Ca²⁺-dependent structural rearrangements within Na⁺–Ca²⁺ exchanger dimers

Abstract

Cytoplasmic Ca²⁺ is known to regulate Na⁺–Ca²⁺ exchanger (NCX) activity by binding to two adjacent Ca²⁺-binding domains (CBD1 and CBD2) located in the large intracellular loop between transmembrane segments 5 and 6. We investigated Ca²⁺-dependent movements as changes in FRET between exchanger proteins tagged with CFP or YFP at position 266 within the large cytoplasmic loop. Data indicate that the exchanger assembles as a dimer in the plasma membrane. Addition of Ca²⁺ decreases the distance between the cytoplasmic loops of NCX pairs. The Ca²⁺-dependent movements detected between paired NCXs were abolished by mutating the Ca²⁺ coordination sites in CBD1 (D421A, E451A, and D500V), whereas disruption of the primary Ca²⁺ coordination site in CBD2 (E516L) had no effect. Thus, the Ca²⁺-induced conformational changes of NCX dimers arise from the movement of CBD1. FRET studies of CBD1, CBD2, and CBD1–CBD2 peptides displayed Ca²⁺-dependent movements with different apparent affinities. CBD1–CBD2 showed a Ca²⁺-dependent phenotype mirroring full-length NCX but distinct from both CBD1 and CBD2.

The Na⁺–Ca²⁺ exchanger (NCX) is a plasma membrane protein that uses the electrochemical gradient of Na⁺ to extrude Ca²⁺ from cells. NCX is organized into nine transmembrane segments (TMSs) with a large cytoplasmic loop between TMSs 5 and 6 composed of two adjacent Ca²⁺-binding domains, CBD1 and CBD2, flanked by linkers that connect the CBDs to the TMSs (1).

Regulatory Ca²⁺ binds to the CBDs, stimulates exchanger activity, but is not transported. It also alleviates Na⁺-dependent inactivation (2, 3). CBD1 and CBD2 bind four and two Ca²⁺ ions, respectively. Recent studies have identified the important role of Ca²⁺ sites 3 and 4 of CBD1 and site 1 of CBD2 in regulating Ca²⁺ transport (4–6). How the binding of Ca²⁺ to these sites is translated into exchanger activation is unknown.

CBD1 and CBD2 share a similar structure consisting of seven antiparallel β-strands arranged as a β-sandwich (4, 6–12) with the Ca²⁺-binding sites at one end. The CBDs are only separated by 5–7 aa and therefore are organized in a head-to-tail fashion (Fig. 1A).

Fig. 1.

NCX undergoes Ca²⁺-dependent conformational changes. (A) Secondary structure of the fluorescent-tagged exchanger illustrating the relative position of the two Ca²⁺–binding domains (CBD1 and CBD2) and fluorophores (FP). Small spheres indicate bound Ca²⁺. (B) Effect of changes in intracellular Ca²⁺ on CFP (□) and YFP (■) emissions (Left) and corresponding FRET_R (Right) after coexpression of NCX^C and NCX^Y. (C) Changes in normalized FRET_R versus free [Ca²⁺] for NCX^C + NCX^Y. FRET_R values were normalized to their maximum and fitted to a single Hill function (K_h = 40 nM and Hill coefficient =1.8).

To advance our understanding on the molecular mechanism leading to NCX activation by cytoplasmic Ca²⁺, we inserted either YFP or CFP at position 266 within the large cytoplasmic loop of NCX. Additionally, we expressed fusion proteins composed of CBD1, CBD2, or CBD1 plus CBD2 (CBD12) flanked by CFP and YFP to detect Ca²⁺-induced movements in real time by FRET. FRET efficiency (FRET_E) depends strongly on the distance between two fluorophores, and the technique can be used for monitoring protein–protein interactions and conformational changes (13).

Our results indicate that FRET occurs between adjacent exchangers, revealing that NCX exists as a dimer in the plasma membrane. Previous studies have demonstrated the multimeric nature of NCX (14, 15) without, however, resolving its subunit composition. We also show that regulatory Ca²⁺ decreases the distance between the paired NCXs by triggering conformational changes within the cytoplasmic portion of the protein linking CBD1 to TMS 5. Our results show the detection of regulatory Ca²⁺-dependent movements in the full-length functional exchanger.

Results

Ca²⁺-Dependent Conformational Changes Between NCX Monomers.

To investigate the Ca²⁺-dependent movements of NCX during Ca²⁺ regulation, YFP or CFP proteins were inserted at amino acid 266 of NCX to give NCX^C or NCX^Y (Fig. 1A). Introduction of fluorophores at amino acid 266 does not alter the biophysical properties of NCX, including regulation by cytoplasmic Na⁺ or Ca²⁺, as shown previously (16). Because NCX can form oligomers (14, 15), we hypothesized that FRET would occur between adjacent exchangers.

To have fast access to the cytoplasmic face of NCX, we used fluorescent plasma membrane sheets (FluoPMS) derived from Xenopus laevis oocytes (17) (SI Materials and Methods). FluoPMS were prepared from oocytes coexpressing NCX^C and NCX^Y to monitor Ca²⁺-dependent movement as changes in the YFP/CFP ratio (FRET_R). As shown in Fig. 1B, addition of 13 μM Ca²⁺ to the bath triggered an increase in FRET_R, which was reversed by 10 mM EGTA. Thus, regulatory Ca²⁺ promotes movements in the cytoplasmic domains of NCX, decreasing the distance between the large intracellular loops of adjacent exchangers. To determine the range of Ca²⁺ needed to trigger movements in the full-length exchanger, FRET_R was measured at different free Ca²⁺ concentrations. These values were fitted to a Hill function. A 50% change in FRET_R was induced by ∼40 nM Ca²⁺ (Fig. 1C).

NCX Exists as Dimer.

The detection of FRET between NCX molecules suggests that NCX exists in a multimeric complex. Alternatively, FRET may arise from a nonspecific encounter of the two fluorophores because of a high density of NCX molecules within the membrane. Based on theoretical models (13, 18, 19), two conditions need to be satisfied to determine true oligomerization: (i) the level of energy transfer (FRET_E) between units within a true oligomer is expected to rise with increasing concentration of acceptor protein (donor protein concentration is constant) until a plateau is reached, which is when donors are all coupled to acceptors; and (ii) the FRET_E should be independent of the absolute protein concentration. Fig. 2A shows the effects of increasing the level of NCX^Y acceptors while keeping the NCX^C donors constant. An increase in FRET_E [calculated by using the gradual acceptor photobleaching protocol (20)] is observed until a plateau is reached (FRET_E = 0.2). This hyperbolic curve satisfies the first condition for true oligomerization (18, 19), in contrast to the almost linear relationship between FRET_E and acceptor concentration expected for random encounters. Although FRET generated by protein overexpression (E_random) is also predicted to plateau when the cell surface is packed with proteins, this occurs at extremely high protein concentrations, not attainable for overexpressed NCX. We estimate that at least 200,000 exchangers per μm² (18, 21) are required to saturate E_random. However, the maximal density of NCX in Xenopus laevis oocytes is 300–400 exchangers per μm² (22), generating a FRET_E of about 0.01 by random associations, well below our measurement of 0.1.

Fig. 2.

NCX exists as dimer in the plasma membrane. (A) Dependence of FRET_E on NCX^Y expression (measured as fluorescence emission intensity, a.u.) at a constant NCX^C expression level and in presence of 13 μM Ca²⁺. Points were fitted to FRET_E = (FRET_max * F_YFP)/F_YFP + K, where FRET_max is the maximal FRET_E, F_YFP is YFP emission intensity, and K is analogous to a dissociation constant (19). F test comparison between hyperbolic and linear regressions indicates that a nonlinear model best fits the points (P < 0.0001). (B) FRET_E versus protein concentration measured as the sum of YFP (before bleaching) and CFP (after YFP bleaching) fluorescence intensity. FRET_E was measured in the absence of Ca²⁺ from FluoPMS expressing a constant ratio of NCX^Y to NCX^C of 0.50 ± 0.03 (n = 20) or 4.3 ± 0.2 (n = 18) in absence of Ca²⁺ and 4.3 ± 0.1 (n = 28) in presence of 13 μM Ca²⁺. The R² values are 0.086, 0.00002, and 0.01, respectively. (C) Plot of FRET_E as function of the NCX^Y/NCX^C concentration ratio in absence or presence of 13 μM Ca²⁺. Points were fitted to a hyperbole as described in the Materials and Methods and in ref. 23. Each point is the average of 4–12 measurements. (D) Comparison between the experimental data shown in Fig. 2C (normalized to their maxima) and modeled FRET_E curves for dimer, trimer, and tetramer according to [(a + d)ⁿ − aⁿ − dⁿ]/[(a + d)ⁿ − aⁿ − dⁿ + ndⁿ] (24), where n is the number of molecule(s) in the oligomers, a is the number of acceptors, and d is the number of donors. The SDs of the vertical distances of the points from the lines describing the dimer, trimer, or tetramer models are 0.08, 0.18, and 0.24 (0 Ca²⁺) and 0.08, 0.17, and 0.24 (13 μM Ca²⁺).

To verify the second condition, we investigated the relationship between FRET_E and protein expression levels. Because FRET_E depends on the fraction of donor–acceptor complexes with respect to total donor molecules, we measured FRET_E from patches expressing a constant ratio of acceptors, NCX^Y to donors, NCX^C, but with different total expression levels. As shown in Fig. 2B, our data indicate that FRET_E remained constant despite different total NCX expression levels both in the absence or presence of Ca²⁺. The combined data are consistent with NCX existing as a functional oligomer in the plasma membrane.

To determine the maximal FRET_E between tagged exchangers, we measured FRET_E from FluoPMS expressing different ratios of NCX^Y and NCX^C in the presence and absence of Ca²⁺ and generated FRET saturation curves as shown in Fig. 2C. The measured FRET_E increases with the probability of interaction of CFP with YFP and therefore with the relative amounts of free acceptors and donors. The FRET titration curves could be fitted to a hyperbolic function (23). Ca²⁺ significantly increased the extrapolated maximal FRET_E value, indicating movements between the large cytoplasmic loops of neighboring NCX proteins.

Since FRET_E versus NCX^Y/NCX^C depends on the stoichiometry of the oligomer, we compared our experimental values to mathematical models describing the expected FRET_E saturation curves for dimers, trimers, and tetramers (24). Fig. 2D shows that the theoretical dimer curve best approximates the experimental data, suggesting that oligomerized NCX exists predominately as a dimer in the membrane.

CBD1 Confers Ca²⁺-Dependent Movement to the Full-Length NCX.

To determine the contributions of the individual CBDs to the Ca²⁺-induced movements of NCX, we mutated residues in NCX that disrupt Ca²⁺ binding to either CBD1 or CBD2. Two separate constructs were engineered. The first consisted of NCX^C and NCX^Y exchangers with three mutations in CBD1: D421A, E451A, and D500V (NCX^C-CBD1 and NCX^Y-CBD1), which disrupt all four Ca²⁺-binding sites of CBD1 (6, 8, 10), and the second construct had the mutation E516L (site 1 in CBD2, NCX^C-CBD2, and NCX^Y-CBD2) (4, 5). Mutations in CBD1 drastically decreased the sensitivity of NCX peak current for regulatory Ca²⁺ whereas mutant E516L in CBD2 abolished Ca²⁺ regulation of NCX (Fig. S1). Analogous results have previously been reported in the untagged NCX with the same mutation(s) (4), emphasizing the fact that the tagged NCX behaves in a manner similar to that of untagged NCX.

We next investigated the Ca²⁺-dependent conformational changes of these mutants. As shown in Fig. 3A, the CBD1 mutant abolished Ca²⁺-dependent movements of NCX, although changes in FRET_R of ∼0.02, at the limit of our detection capability, were observed upon addition of ∼40 μM Ca²⁺. FRET_E was also unaffected by Ca²⁺ (Fig. 3B). In contrast, the CBD2 mutant did not significantly modify either the FRET_R (Fig. 3C) or FRET_E (Fig. 3D) response to Ca²⁺. These results indicate that the Ca²⁺-dependent conformational changes of NCX are associated with the occupancy of Ca²⁺-binding sites within CBD1. The conformational changes associated with CBD1, however, are not sufficient to regulate NCX if CBD2 is nonfunctional.

Fig. 3.

Mutations within CBD1 abolish Ca²⁺-dependent conformational changes. (A and B) Effect of intracellular Ca²⁺ on changes in FRET_R and FRET_E between CFP- and YFP-tagged exchangers carrying the mutations D421A, E451A, and D500V (NCX^C–CBD1 + NCX^Y-CBD1). Mutations within CBD1 abolished Ca²⁺-dependent FRET changes. (C and D) FRET_R and FRET_E measurements from fluorescent-tagged exchangers with E516L in CBD2 (NCX^C–CBD2 + NCX^Y-CBD2). Note the increases in FRET_R and FRET_E upon addition of cytoplasmic Ca²⁺.

Ca²⁺-Dependent Conformational Changes in Isolated CBD1, CBD2, and CBD12.

To further explore the effects of the binding of Ca²⁺ to the cardiac CBDs (NCX1.1 isoform), we studied their Ca²⁺-dependent properties as single Ca²⁺-binding domains. Based on crystal structures (4, 6–12). CBD1 encompasses residues 371–501, CBD2 encompasses residues 501–689, and CBD12 encompasses residues 371–689. We linked these constructs to CFP and YFP at their N and C termini (^CCBD1^Y, ^CCBD2^Y, and ^CCBD12^Y), respectively, and investigated their responses to Ca²⁺. All constructs were targeted to Xenopus laevis oocyte plasma membranes by fusing the Ki-Ras sequence to the C termini of YFP (25). The Ca²⁺-dependent movements were first monitored as changes in FRET_R. Examples of changes in the FRET_R are shown in Fig. 4 A–C. Rapid switches between solutions with different Ca²⁺ concentrations caused a decrease in FRET_R, indicating that the distance between the fluorophores increases upon binding of Ca²⁺ to all three peptides.

Fig. 4.

Cytoplasmic Ca²⁺ triggers movement in the isolated Ca²⁺-binding domains. (A–C) Effect of Ca²⁺ on FRET_R signals generated by ^CCBD1^Y (A), ^CCBD2^Y (B), and ^CCBD12^Y (C) in FluoPMS. Note the lower Ca²⁺ concentrations required to activate movements in CBD12. Images were acquired every 10 s. (D) Changes in normalized FRET_R value versus free [Ca²⁺]. Points, which are averaged between four and six experiments, were fitted to a Hill function. K_h values are 0.20 μM, 12 μM, and 0.010 μM for CBD1, CBD2, and CBD12. Average K_h values obtained from the single experiments are (in μM) 0.19 ± 0.01 (n = 4), 12 ± 0.01 (n = 6), and 0.014 ± 0.01 (n = 4), respectively. For CBD12, only the high Ca²⁺ affinity component was analyzed because the movements associated with higher Ca²⁺ concentrations were too small for accurate analysis. (E) Average FRET_E values in the absence and presence of 13 μM Ca²⁺ for CBD1 and CBD12 and 139 μM Ca²⁺ for CBD2. Values are 0.35 ± 0.01 and 0.23 ± 0.01 for CBD1, 0.184 ± 0.003 and 0.162 ± 0.003 for CBD2, and 0.114 ± 0.003 and 0.090 ± 0.004 for CBD12, respectively. ^CCBD1^Y-L is a lengthened construct consisting of residues 371–508, and ^CCBD12^Y-CBD1 is ^CCBD12^Y with mutations D421A, E451A, and D500V in CBD1. Values are 0.28 ± 0.01 and 0.23 ± 0.01 for ^CCBD1^Y-L and 0.14 ± 0.01 and 0.13 ± 0.01 for ^CCBD12^Y-CBD1. Number of experiments is indicated within each bar.

As shown in Fig. 4D, the K_h values for Ca²⁺ of ^CCBD1^Y and ^CCBD2^Y are 0.20 and 12 μM, respectively. The response of ^CCBD12^Y to Ca²⁺ appears biphasic with apparent Ca²⁺ affinities of ∼0.01 and 16 μM with relative contributions of ∼80% and 20%, respectively. We lack confidence in the quantification of the low-affinity component of ^CCBD12^Y because of the small changes in FRET_R. The finding that CBD12 has a higher affinity for Ca²⁺ than either CBD1 or CBD2 contrasts with an earlier study (26) but is consistent with the result that CBD12 has substantially altered Ca²⁺ dissociation kinetics (see below). Our data indicate that ^CCBD12^Y behaves differently than the sum of the separate ^CCBD1^Y and ^CCBD2^Y moieties, indicating an interaction between the two domains.

Consistent with the ratiometric data, FRET_E values decreased after Ca²⁺ application (Fig. 4E), showing the most prominent decrease in CBD1. As observed for NCX, mutation of residues 421, 451, and 500 abolished the Ca²⁺-dependent movements of ^CCBD12^Y (^CCBD12^Y-CBD1 in Fig. 4E): FRET_E values are 0.14 ± 0.01 and 0.13 ± 0.01 in the absence and presence of 13 μM Ca²⁺, stressing the role of CBD1 in triggering these movements.

We then resolved the temporal changes of FRET_R upon addition or removal of Ca²⁺ to estimate the speed of the Ca²⁺-dependent movements. Examples of FRET_R changes upon binding of Ca²⁺ are shown in Fig. S2. Generally, we find that higher concentrations of Ca²⁺ result in faster changes in FRET_R. The changes in FRET follow a monoexponential time course. ^CCBD1^Y had slow kinetics at 140 nM free Ca²⁺ (τ = 0.38 ± 0.03 s, n = 6), which accelerated to the limit of our temporal resolution (∼150 ms) in response to micromolar Ca²⁺ concentrations. Fast kinetics have been previously described (26, 27), which would not be resolved because of our image sampling rate. Similarly, ^CCBD2^Y and ^CCBD12^Y showed slow movements upon addition of low levels of Ca²⁺ and faster movements when Ca²⁺ was raised to higher levels (Fig. S2).

Removal of Ca²⁺ triggered different responses among the CBDs (Fig. 5). In the case of ^CCBD1^Y and ^CCBD2^Y, FRET_R changed too rapidly for an accurate measurement (<0.15 s), whereas CBD12 showed a substantially slower time course (τ = 16 ± 2 s, n = 8). Similar kinetics were measured for NCX dimers (τ = 19 ± 4 s, n = 7), suggesting that the movements observed in NCX originate from the two CBDs acting in concert.

Fig. 5.

Removal of Ca²⁺ causes a slow conformational change in CBD12 and in the full-length NCX. (A–C) FRET_R changes upon removal of regulatory Ca²⁺ for the indicated constructs. Images were acquired every 200 or 500 ms. (D) Summary of time constants for FRET_R changes upon removal of Ca²⁺ for CBD12 and NCX. ^#Values for the individual CBDs represent an upper limit and cannot be accurately measured because of our limited frame rate (5 Hz). Faster kinetics are likely to occur as previously reported (26, 27).

To rule out a possible contribution of intermolecular FRET due to oligomerization of the CBDs, we fused either CFP or YFP to the N or C terminus of CBD12 to generate ^YCBD12, ^CCBD12, and CBD12^C. HEK cells were then transfected with the cytosolic ^YCBD12 in the presence of either ^CCBD12 or CBD12^C. None of the combinations were able to generate FRET (a FRET_E of 0.01 was calculated in both cases), indicating that free CBDs did not oligomerize and that FRET only occurs intramolecularly. Finally, FRET_E measured from the CBDs targeted to the plasma membrane was independent of the expressed protein concentration, indicating that FRET was not caused by crowding of CBDs in the membrane (Fig. S3).

Discussion

The Na⁺-Ca²⁺ exchanger is regulated by cytoplasmic Ca²⁺, which antagonizes Na⁺-dependent inactivation and directly activates NCX (2, 3). This Ca²⁺ binds to two domains, CBD1 and CBD2, the structures of which have been determined (4, 6–12). The mechanisms involved in Ca²⁺ regulation are not well understood. To gain insight into these molecular processes, we used FRET measurements to characterize the Ca²⁺-dependent conformational changes both in the full-length exchanger and in isolated CBD peptides.

First, we assessed FRET between full-length NCXs, each labeled with a single fluorophore. Our data indicate that adjacent exchangers are close enough to generate FRET, confirming previous crosslinking studies that NCX exists as oligomers (15). By analyzing the changes in FRET as a function of fluorophore concentration, we determined that the exchanger likely exists as a dimer in the plasma membrane, similarly to the NCKX transporter (28). In some cases, oligomerization of transporters regulates either function (29–31) or trafficking (32). We have previously shown that coexpression of a tagged WT NCX with a nonfunctional mutant caused retention of the tagged NCX within the endoplasmic reticulum (14). Thus, dimerization may be important for trafficking NCX to the plasma membrane. Whether interactions between neighboring NCX subunits influence transport function remains to be determined.

Next, we determined whether cytoplasmic Ca²⁺ influences the spatial arrangement of the dimer. Upon addition of Ca²⁺, the pair NCX^C + NCX^Y shows an increase in FRET, indicating that the binding of Ca²⁺ decreases the distance between the N-terminal portions of the adjacent large cytoplasmic loops. These results provide a physical measurement of Ca²⁺-dependent movements within NCX.

We demonstrated by mutagenesis that the dimer movements arise mainly from the binding of Ca²⁺ to CBD1. Mutations that eliminate Ca²⁺ binding to CBD1 severely decrease the apparent affinity for Ca²⁺ to regulate NCX activity (5) and abolish the Ca²⁺-dependent changes in FRET_E, ruling out the Ca²⁺ transport site as the source of FRET changes. The residual low-affinity Ca²⁺ regulation is attributable solely to CBD2 (5). In contrast, a mutation within CBD2 that abolishes Ca²⁺ regulation, left Ca²⁺-dependent changes in FRET fully intact. Possibly, fluorophores at position 266 fail to track changes in conformation induced by Ca²⁺ binding to CBD2.

Overall, these results are consistent with previous findings that Ca²⁺ binding to CBD1 both induces a substantial conformational change (8, 33–35) and controls the affinity of Ca²⁺ regulation (5, 34, 36). Nevertheless, the conformational changes evoked by CBD1 are not sufficient to account for all of the steps involved in Ca²⁺ regulation because a functional CBD2 is also required.

An important step for understanding the physiological role of Ca²⁺ regulation of NCX is to determine its affinity for Ca²⁺, a nontrivial task because apparent affinities are typically measured under different conditions by using diverse techniques. For example, electrophysiological measurements of exchanger activity are generally carried out in the presence of high intracellular Na⁺ with varying Ca²⁺ concentrations. Increasing cytoplasmic Ca²⁺ activates NCX but also competes with Na⁺ at the transport site, thereby decreasing exchanger activity (2, 3). As a result, estimates of NCX affinity for regulatory Ca²⁺ have ranged from 22 nM in whole-cell measurements (37) to 800 nM in giant excised patches (5). In this work, we find that the concentration of Ca²⁺ needed to increase FRET_R by 50% between two exchangers is ∼40 nM, which suggests that, at resting levels of myocardial Ca²⁺ (∼100 nM), NCX has already undergone large changes in conformation, limiting the effects of further increases of Ca²⁺ during excitation–contraction coupling. However, two issues must be considered: First, our experiments were conducted in the absence of both Na⁺ (to prevent Na⁺-dependent inactivation) and Mg²⁺. Although cytoplasmic Na⁺ does not alter the Ca²⁺ binding properties of the isolated CBDs (27), its effects on the Ca²⁺ regulation of the full-length exchanger have not been investigated. Intracellular Mg²⁺ decreases both the affinities of NCX and the isolated CBDs for Ca²⁺ (27, 33, 38). Thus, Ca²⁺-dependent conformational changes of NCX associated with CBD1 may indeed take place in a cardiac cell. Second, the lower affinities measured for regulation may involve binding of Ca²⁺ to CBD2, after CBD1 is saturated. As we have shown, CBD2 is essential for regulation, making this possibility feasible. Binding of Ca²⁺ to CBD2 produces only minor FRET changes that are not readily detectable in NCX experiments. We are currently investigating these issues.

To understand how the individual Ca²⁺-binding domains behave during regulation of NCX, we investigated the movements of membrane-targeted CBDs. We first analyzed the response of CBD1 to Ca²⁺. We had previously investigated the properties of the soluble CBD1 flanked by CFP and YFP encompassing residues 371–508 and shown that binding of Ca²⁺ to this domain decreases FRET (33). However, after consideration of the crystal structure (10), we removed residues 502–508 from CBD1 because they form part of the first β-strand of CBD2. Truncation of residues 502–508 did not affect the Ca²⁺ affinity of CBD1, although the change in the FRET_E signal was somewhat larger (Fig. 4E with the longer form of CBD1 designated as ^CCBD1^Y-L). The result has two important implications: residues 502–508 are not essential for the Ca²⁺-dependent movements in CBD1 and targeting the construct to the plasma membrane did not affect Ca²⁺ binding properties.

CBD2 behaved similarly to CBD1: Ca²⁺ binding resulted in an increase in the distance between the fluorophores, although the movements were smaller and less sensitive to Ca²⁺ when compared with CBD1 (K_h = 12 μM vs. K_h = 200 nM for CBD1). The results are consistent with previous reports showing smaller structural effects of Ca²⁺ on CBD2 than CBD1 and a reported lower affinity of CBD2 for Ca²⁺ (4, 7, 26, 27). The greater FRET_E changes observed for CBD1 might be accounted for by stabilizing or rigidifying effects of Ca²⁺ binding, as determined by other structural approaches (8–11).

The Ca²⁺-binding domains of NCX are adjacent to each other and are aligned in a head-to-tail arrangement (Fig. 1) with the residues coordinating Ca²⁺ ions in CBD1 in proximity to the end of CBD2 that does not bind Ca²⁺. The arrangement suggests that interactions between these domains may play a role in Ca²⁺ regulation (7). This possibility is strongly supported by the data that CBD12 movements are slower and occur at lower Ca²⁺ levels than for either CBD1 or CBD2. Synergistic interactions between the two domains have been documented in time-resolved experiments (26) in which a slow dissociation of Ca²⁺ was observed from the brain splice variant of CBD12 but not in CBD1 or CBD2. These two independent lines of evidence strongly suggest that the slow kinetics of Ca²⁺ dissociation (26) and the slow conformational transitions in CBD12 shown here are closely related events, even across splice variants.

The decrease in FRET_E upon Ca²⁺ binding to CBD12 indicates an increase in distance between the fluorophores. In contrast, a study based on small-angle X-ray scattering analysis (7) suggests that within CBD12 a decrease in the maximal distance (D_max; 120 ± 10 Å to 135 ± 10 Å) occurs upon binding of Ca²⁺. D_max provides information on the maximum size of the protein without specifying the location at which these distances are obtained. Thus, FRET and small-angle X-ray scattering measurements reflect movements with different reference points and there is not necessarily a discrepancy in the two measurements.

The slow conformational changes of CBD12 observed upon removal of Ca²⁺ mirrors that of the full-length NCX and correlates with the deactivation times (∼5–10 s) of NCX activity measured electrophysiologically from patches from cardiac myocytes (2). The results suggest that NCX remains in an activated state for several seconds upon the removal of Ca²⁺, indicative of possible regulation over several beats. A persistent activation of NCX after reduction in intracellular Ca²⁺ in CHO cells has also been reported (39).

Overall, the results establish that interactions between the two CBDs take place within both the isolated conjunct CBDs and NCX. Possibly, as previously suggested (7), CBD2 exerts a stabilizing effect on the Ca²⁺-binding sites of CBD1 conferring high Ca²⁺ sensitivity. This hypothesis is supported by evidence that insertion of a linker between the CBDs decreases NCX sensitivity to Ca²⁺ (40) and abolishes the slow Ca²⁺ kinetics from the isolated CBD12 (26). Therefore, data obtained with isolated CBD1 and CBD2 are not representative of in vivo responses, and CBD12 is a more realistic model for in vitro studies.

In summary, our data provide evidence that NCX is a dimer in the plasma membrane, and binding of Ca²⁺ to CBD1 triggers a substantial decrease in the distance between the N-terminal portions of the large cytoplasmic loops of adjacent exchangers. These movements may be transduced to the adjoining TMS 5 to modulate NCX activity. Possibly other movements are transmitted between CBD2 and TMS 6. Interactions between CBD1 and CBD2 are clearly important in regulation of NCX as shown here and elsewhere (26, 40). The use of a fluorescent probe adjacent to TMS 6 for FRET measurements may be informative, and such experiments are being initiated.

Materials and Methods

Molecular Biology.

The generation of canine cardiac NCX-266CFP (NCX^C) and NCX-266YFP (NCX^Y) has been previously described (16). Construction of the plasma membrane targeted CBDs is as described in the SI Materials and Methods.

FRET Measurements.

Xenopus laevis oocytes FluoPMS isolation and changes in FRET_R measurements were conducted as described previously (17). FRET_E was calculated using the gradual acceptor photobleaching technique (20). Additional information is presented in SI Materials and Methods. Solutions used for FRET measurements consist of (mM): 100 CsCl, 10 Hepes, 10 EGTA or N-(2-hydroxyethyl) ethylenediamine-N,N,N-triacetic acid (HEDTA), pH = 7 (using MES) and different free CaCl₂ concentrations (calculated according to the WEBMAXC program (41) and validated with a Ca²⁺ electrode).

The relative expression of acceptor to donor as NCX^Y/NCX^C ratio was calculated as fluorescence intensities of the acceptor and donor before and after acceptor bleaching, respectively. The ratio values were then converted into concentration ratios by normalizing to the value obtained from a construct with a fixed CFP/YFP stoichiometry, imaged with the same settings. For this purpose, the full-length exchanger with the fluorophores inserted at positions 266 and 467 was used, and an average FRET_R of 8 was determined (7.97 ± 0.23, n = 64). The maximal FRET_E (FRET_max) was deduced by fitting FRET_E versus the NCX^Y/NCX^C ratio with the equation FRET_E = (FRET_max * r)/(1 + r), where r is the NCX^Y/NCX^C concentration ratio (modified from ref. 23).

FRET_E calculation for randomly distributed donors and acceptors were conducted using the Wolber and Hudson model (18, 21) as described in SI Materials and Methods.

Electrophysiology.

Outward NCX currents were recorded from oocytes in the inside-out configuration by using the giant patch technique, as previously described (5). Additional information is presented in SI Materials and Methods.