Effect of Phosphorylation on Interactions between Transmembrane Domains of SERCA and Phospholamban

Abstract

We have used site-directed spin labeling and electron paramagnetic resonance (EPR) to map interactions between the transmembrane (TM) domains of the sarcoplasmic reticulum Ca²⁺-ATPase (SERCA) and phospholamban (PLB) as affected by PLB phosphorylation. In the cardiac sarcoplasmic reticulum, PLB binding to SERCA results in Ca-dependent enzyme inhibition, which is reversed by PLB phosphorylation at Ser16. Previous spectroscopic studies on SERCA-PLB have largely focused on the cytoplasmic domain of PLB, showing that phosphorylation induces a structural shift in this domain relative to SERCA. However, SERCA inhibition is due entirely to TM domain interactions. Therefore, we focus here on PLB’s TM domain, attaching Cys-reactive spin labels at five different positions. In each case, continuous-wave EPR indicated moderate spin-label mobility, with the addition of SERCA revealing two populations, one indistinguishable from PLB alone and another with more restricted rotational mobility, presumably due to SERCA-binding. Phosphorylation had no effect on the rotational mobility of either component but significantly decreased the mole fraction of the restricted component. Solvent-accessibility experiments using power-saturation EPR and saturation-recovery EPR confirmed that these two spectral components were SERCA-bound and unbound PLB and showed that phosphorylation increased the overall lipid accessibility of the TM domain by increasing the fraction of unbound PLB. However—based on these results—at physiological levels of SERCA and PLB, most SERCA would have bound PLB even after phosphorylation. Additionally, no structural shift in the TM domain of SERCA-bound PLB was detected, as there were no significant changes in membrane insertion depth or its accessibility. Therefore, we conclude that under physiological conditions, the phosphorylation of PLB induces little or no change in the interaction of the TM domain with SERCA, so relief of inhibition is predominantly due to the previously observed structural shift in the cytoplasmic domain.

Introduction

The sarcoplasmic reticulum Ca²⁺-ATPase (SERCA, Fig. 1, gray) is responsible for initiating muscle relaxation by pumping Ca²⁺ from the cytoplasm into the sarcoplasmic reticulum lumen. In cardiac muscle, SERCA is regulated by phospholamban (PLB, Fig. 1, red), a 52-residue protein comprised of a highly dynamic N-terminal cytosolic helix, followed by a short loop and a C-terminal transmembrane (TM) helix (1, 2, 3). Upon binding, PLB inhibits SERCA pump activity by decreasing its apparent Ca²⁺ affinity, whereas high (μM) Ca²⁺ concentrations restore full activity (3, 4, 5, 6, 7). In addition, phosphorylation of PLB at Ser16 by protein kinase A (PKA) partially restores apparent Ca²⁺ affinity while also linking SERCA regulation to β-adrenergic stimulation. The mechanistic basis for this effect is not yet clear.

Figure 1.

An illustration of the monomeric mutant of PLB (red) bound to SERCA (gray) in a lipid bilayer adapted from x-ray crystallography (49) (Protein Data Bank: 4Y3U) and hybrid NMR (2) (Protein Data Bank: 2KB7) structures. Spin-label locations across the TM domain of PLB are marked with green beads, from top to bottom, as follows: Q29, F32, L39, L43, and M50. Cysteines were labeled with either MSL or MTSSL (top right). This image was rendered using visual molecular dynamics (51). To see this figure in color, go online.

Previous spectroscopic studies have shown that Ser16 phosphorylation (5) and phosphomimetic mutations (S16E, S16D) (8) promote an order-to-disorder transition in the cytoplasmic (CP) domain of PLB. Mutagenesis studies have demonstrated that structural dynamics of the CP domain can be directly correlated with inhibitory potency (9), though it is unknown how these dynamics changes translate to the restoration of SERCA activity. Thus, it is clear that the CP domain is essential for SERCA regulation, but there is considerable evidence that the TM domain alone is responsible for inhibition of SERCA; experiments in which truncated TM fragments of PLB were coexpressed with SERCA in HEK-293 cells (10) and synthesized and coreconstituted with SERCA (11, 12) showed that the apparent Ca²⁺ affinity of SERCA was reduced.

Because of its sensitivity to both dynamics and orientation, electron paramagnetic resonance (EPR) spectroscopy is well-suited for probing structural changes within biological membranes. When combined with site-directed spin labeling (13), the experimenter can systematically investigate selected domains of proteins within membranes by introducing nitroxide spin labels. In particular, measurement of T₁ and T₂ spin-label-relaxation times in the presence of paramagnetic relaxation agents (PRAs) (14) reveals important properties, such as the surrounding environment and membrane insertion depth of labeling sites. Both power-saturation EPR (PS-EPR) (14, 15, 16, 17, 18, 19, 20, 21, 22, 23) and saturation-recovery EPR (SR-EPR) (24, 25, 26, 27) have been used extensively to characterize the solvent accessibility of specific sites in both cytosolic and membrane proteins.

A major question in the SERCA-PLB field is whether phosphorylation relieves SERCA inhibition by causing dissociation of PLB from SERCA or by perturbing the structural dynamics of the SERCA-PLB complex (5, 28, 29, 30, 31). Most fluorescence resonance energy transfer (FRET) and EPR studies of PLB to date have labeled the CP domain and focused on phosphorylation-induced changes in the CP domain (4, 5, 32, 33). EPR showed clearly that the CP domain of PLB is partially unfolded by phosphorylation (5). FRET from PLB to SERCA showed that increased [Ca²⁺] (4, 33) and phosphorylation (33) induce structural changes in the CP domain, and only the latter partially dissociates PLB from SERCA, though the CP domain was labeled in both of these studies (33). A recent saturation-transfer EPR (ST-EPR) study with a spin label on PLB’s TM domain showed that there is a significant amount of SERCA-bound PLB after phosphorylation, but the ST-EPR technique does not resolve populations, so it could not discern the fraction that remained bound (7). Additionally, an NMR study (6) did monitor the binding between PLB’s TM domain and SERCA as a function of phosphorylation, though those measurements were performed in DPC micelles, not lipid membranes.

Therefore, in this study, we use more flexible spin labels attached to PLB’s TM domain, using EPR techniques that can resolve the SERCA-bound and free components and assess the movement of the TM domain relative to SERCA and the membrane. We used continuous-wave EPR (CW-EPR) to measure changes in side-chain mobility, resolving the populations of SERCA-bound and free PLB. Additionally, we used PS-EPR and time-resolved SR-EPR to measure accessibility to lipid and aqueous PRAs, thus determining whether PLB’s TM domain moves vertically or horizontally upon phosphorylation. The PRAs used in this study were oxygen (zero-grade air) and nickel(II)-ethylenediamine-N,N'-diacetic acid (NiEDDA), which provide accessibility profiles inside the lipid bilayer and in the surrounding solvent, respectively (14).

Materials and Methods

Protein purification and spin labeling

All PLB constructs were based on a triple-mutant background (C36A, C41F, C46A) that renders PLB Cys-free and monomeric (34). A plasmid encoding rabbit PLB as a Tobacco Etch Virus-cleavable fusion to maltose-binding protein was mutated by QuikChange mutagenesis (Stratagene, San Diego, CA) to introduce Cys residues at sites within the TM domain. Single-Cys constructs were then transformed into BL21(DE3) cells (Lucigen, Middleton, WI) for protein expression (33, 35). Maltose-binding protein-PLB fusion proteins were purified from cell lysates by amylose resin chromatography (New England Biolabs, Ipswich, MA), dialyzed extensively against H₂O, and digested with S219V-Tobacco Etch Virus (36) to release PLB as a precipitate. Crude PLB was then collected by centrifugation and dissolved in 10% (w/v) SDS, 50 mM DTT for further purification by high-performance liquid chromatography and subsequent lyophilization as done previously (33, 37).

For labeling, either 4-maleimido-2,2,6,6-tetramethyl-1-piperidinyloxy (MSL; Sigma-Aldrich, St. Louis, MO) or 1-oxyl-2,2,5,5- tetramethyl-3-pyrroline-3-methylmethanethiosulfonate (MTSSL; Toronto Research Chemicals, North York, Ontario, Canada) (Fig. 1, top right) were reacted with each single-Cys variant. Labeling locations were chosen on the side of the PLB TM domain opposite the residues that interact with SERCA (Fig. 1) so as not to perturb binding behavior or SERCA inhibition. Previous spectroscopic studies have shown that PLB, when labeled with EPR (3, 7, 32) or FRET (4, 33) probes, still retains wild-type functional behavior, as assessed by the inhibition of SERCA ATPase activity. Lyophilized PLB powders were dissolved to 100 μM in 50 mM HEPES, 1% (w/v) SDS (pH 7.5), to which 1 mM MSL or 2 mM MTSSL was added from a 50 mM stock in dimethylformamide or 20 mM acetonitrile, respectively. The sample was mixed overnight at room temperature before high-performance liquid chromatography purification. Spin-labeled, purified, lyophilized proteins were dissolved to ∼2 mg/mL in PLB storage buffer (20 mM MOPS, 1% (w/v) C₁₂E₈ (pH 7.0)), and labeling efficiency was determined by electrospray ionization mass spectrometry and spin counting. For all EPR samples used in this work, spin-counting results showed that ≥90% of MTSSL- and MSL-PLB were labeled, and the protein concentrations were between 200 and 500 μM. When applicable, before spin labeling, lyophilized PLB was phosphorylated using PKA catalytic subunit C in 2 mL of 20 mM MOPS, 5 mM DTT, 10 mM MgCl₂, 5 mM ATP (pH 7.0) and incubating overnight at 30°C.

SERCA was purified by reactive red chromatography as described previously (38), omitting reducing agents, and concentrated to ∼6 mg/mL using an Amicon ultrafiltration cell (Sigma-Aldrich).

Sample preparation and EPR spectroscopy

For all EPR experiments, spin-labeled PLB or its phosphorylated form (referred to as PLB and pPLB, respectively) were reconstituted into 4:1 (mol/mol) DOPC/DOPE lipid bilayers in the absence and presence of SERCA (PLB + SERCA and pPLB + SERCA). Because 700:1 lipids/SERCA (mol/mol) has been shown to minimize effects because of oligomers (7), we used 700:1:0.5 lipids/SERCA/PLB for SERCA + PLB samples and 700:0.5 lipids/PLB for PLB-only samples. DOPC and DOPE (Avanti Polar Lipids, Alabaster, AL), dissolved in organic solvent, were added to a glass culture tube, dried using N₂ gas, and desiccated overnight. Vesicles were formed by vortexing with buffer (100 mM MOPS, 500 mM KCl, 5 mM MgCl₂ (pH 7.0)) and subsequently solubilized using C₁₂E₈ (Sigma-Aldrich). Detergent-solubilized PLB and SERCA were then added, and the solution was mixed for 30 min using a mechanical rotator. Reconstitution of the proteins was performed by removing the detergent using Bio-Beads SM-2 (Bio-Rad, Hercules, CA) at a 4:1 (w/w) ratio relative to C₁₂E₈ over a period of 3 h, adding 1/3 of the total Bio-Beads every hour. The resulting vesicles were collected by centrifugation for 1 h at 200,000 × g and 4°C, then loaded into 0.6-mm-inner-diameter quartz capillaries for CW-EPR measurements or gas-permeable TPX capillaries (L&M EPR Supplies, Milwaukee, WI) for EPR accessibility experiments. For relevant samples, 200 mM NiEDDA was added before the addition of proteins. NiEDDA was synthesized as previously reported (14).

All CW-EPR experiments were performed using a Bruker EleXsys E500 spectrometer equipped with an ER4123D resonator (Bruker, Billerica, MA) and a gas-flow-temperature controller to maintain sample temperatures at 298 ± 0.2 K. PS-EPR data were acquired after 15-min equilibration under 100% N₂ gas (NiEDDA and control) or 100% zero-grade air (O₂), with the appropriate atmosphere maintained throughout the experiment.

SR-EPR experiments were performed at the National Biomedical EPR Center at the Medical College of Wisconsin using an X-band spectrometer, a loop-gap resonator, and N₂/air gas-flow-control gauges to differentially vary the amount of O₂ (as described in (39)). SR-EPR decay curves for PLB and pPLB samples (single exponential decays) were measured under 100% N₂ gas, 30% air, and 50% air. SR-EPR decay curves for PLB + SERCA and pPLB + SERCA samples (double exponential decays) were measured under the same conditions, but with two additional conditions (15 and 40% air) for accurate fitting. A 140-mW, 300-ns pump pulse was used to saturate the spectrum at the $m_I=0$ peak, followed by acquiring 2048 data points of each decay curve at a rate of 50 MHz (under N₂), 100 MHz (15 and 30% air), or 1 GHz (40 and 50% air) using a CW observe power of 31.5 μW. Resonator artifacts and background signals were subtracted by acquiring decay curves both on- and off-resonance using a field step of −33 G at a rate of 40 kHz. Samples were loaded into TPX capillaries and equilibrated with the appropriate gas content as in the CW-EPR experiments.

Analysis of CW-EPR data

Nanosecond rotational dynamics was determined from CW-EPR spectra. PLB samples labeled with MSL were analyzed using the outer splitting $2T_{\left|\right|}^{\prime }$ of the spectrum, which decreases with submicrosecond rotational motion, defined as the difference between the magnetic field locations of the $m_I=1$ maximum and the $m_I=-1$ minimum. MTSSL-labeled samples with and without SERCA were analyzed by spectral simulation and fitting performed using EasySpin (40) to quantitate the bound and free spectral components. Both PLB + SERCA and pPLB + SERCA spectra were best fit globally in a least-squares sense using a weighted sum of the corresponding empirically measured PLB-only spectrum (component α), with a small amount of Lorentzian broadening added by convolution (0.5 G) to account for the presence of SERCA, and a simulated restricted component (component β). EPR spectra acquired for PLB-only and pPLB-only showed no detectable differences. The mole fraction $x_{\alpha }$ was allowed to vary between PLB + SERCA and pPLB + SERCA spectral fits. For the β component, values for the electron g-tensor elements were fixed as g = [2.0078, 2.0055, 2.0023], whereas the starting values for the hyperfine tensor elements were $A_{\text{par}}=6\text{and}{A}_{\text{prp}}=36$ G and were allowed to vary. All other parameters were held constant between the spectra, as varying the other parameters did not improve the fits.

Analysis of PS-EPR data

Analysis of PS-EPR data was performed as done previously (15), in which the amplitude A of the central resonance line $\left(m_I=0\right)$ was plotted as a function of microwave power P and fitted to the equation

$$A=c{P}^{1/2}{\left[1+\left(2^{1/\mathit{ϵ}}-1\right)P/P_{1/2}\right]}^{-\mathit{ϵ}},$$ (1)

where $P_{1/2}\propto {\left(T_1{T}_2\right)}^{-1}$ is the power at which the amplitude is reduced by half of its value in the absence of saturation; T₁ and T₂ are the longitudinal and transverse relaxation times, respectively; c is a constant depending on properties of the resonator; and $\mathit{\epsilon }$ can vary between 1/2 and 3/2, depending on the amount of inhomogeneous and homogeneous linewidth broadening, respectively. To characterize the accessibility to a particular PRA, the change in $P_{1/2}$ was normalized by the peak-to-peak linewidth, $\mathit{\Delta }{H}_{\text{pp}}$, to account for differences in spin-label mobility (different labeling sites present different local spin-label environments, e.g., steric hindrance, protein secondary structure, etc.):

$${\mathit{\Pi }}_{\text{PRA}}=\left(P_{1/2,\text{PRA}}-P_{1/2,{\text{N}}_2}\right)/\mathit{\Delta }{H}_{\text{pp}}.$$ (2)

The position of a spin label relative to the lipid bilayer (membrane insertion depth) was characterized by comparing its accessibility to polar and nonpolar PRAs as (15, 23),

$$\mathit{\Phi }=\ln \left(\frac{{\mathit{\Pi }}_{{\text{O}}_2}}{{\mathit{\Pi }}_{\text{NiEDDA}}}\right).$$ (3)

Analysis of SR-EPR data

Analysis of SR-EPR data was performed as done previously (41). When using SR-EPR to monitor the equilibrium conformational exchange between two states α and β, the relevant physical timescales are the average exchange rate $k=\left(k_{\alpha \beta }+k_{\beta \alpha }\right)/2$ and the effective electron relaxation rates of the individual states,

$$W_{1\alpha }=W_{1\alpha ,0}+W_{\text{ex}\alpha ,\text{PRA}}$$

and

$$W_{1\beta }=W_{1\beta ,0}+W_{\text{ex}\beta ,\text{PRA}},$$

where $W_{1\alpha \left(\beta \right)}=1/\left(2T_{1\alpha \left(\beta \right),0}\right)$ is the longitudinal relaxation rate of state α (β) in the absence of a PRA, and $W_{\text{ex}\alpha \left(\beta \right),\text{PRA}}$ is the PRA interaction exchange rate for state α (β). For interactions between PLB and SERCA, the exchange rate of PLB is estimated to be $k\approx 1$ Hz (42), placing the system in the slow exchange limit, where $k\ll \left|W_{1\alpha }-W_{1\beta }\right|$. In this regime, the SR-EPR signal reduces to the form (41)

$$I\left(t\right)=A_{\alpha }{\text{e}}^{-W_{1\alpha }t}+A_{\beta }{\text{e}}^{-W_{1\beta }t}.$$ (4)

Here, A_α and A_β are the SR-EPR signal amplitudes of the α and β states, respectively, and in general depend on the electron-relaxation rates, signal-component intensities, and other quantities (we refer the reader to (41) for the full expressions).

To accurately determine the individual relaxation rates $W_{1\alpha }$ and $W_{1\beta }$, we differentially varied their values by adding O₂ (air) (39, 41), yielding

$$W_{1\alpha }=W_{1\alpha ,0}+j_{\alpha ,{\text{O}}_2}\left[\text{\% air}\right]$$

and

$$W_{1\beta }=W_{1\beta ,0}+j_{\beta ,{\text{O}}_2}\left[\text{\% air}\right],$$ (5)

where $j_{\alpha ,{\text{O}}_2}$ is the “accessibility constant” of state α with respect to air because of Heisenberg exchange.

As a starting point, each set of SR-EPR data was independently fitted to Eq. 4, treating A_α and A_β as independent constants, to determine the dependence of the relaxation rates on air content and show that there were two resolved spin populations, one of which only appeared in the presence of SERCA (Fig. 5 a, inset, and Fig. 6). Once it was established that there were two spin populations with different accessibility constants $j_{\alpha ,{\text{O}}_2}$ and $j_{\beta ,{\text{O}}_2}$, the independent fitting results were then used as starting values for global fitting of all SR-EPR data to Eqs. 4 and 5, using the full expressions of the amplitudes A_α and A_β from (41). For the global fitting step, the mole fractions for the states $x_{\alpha }$ and $x_{\beta }=1-x_{\alpha }$ were fixed using the values obtained from CW-EPR fitting, as done previously (41).

Figure 5.

(a) SR-EPR data for PLB-only (black), PLB + SERCA (red), and pPLB + SERCA (blue) acquired under 50% air (1:1 air/N₂) for MTSSL-L43. The inset shows the electron relaxation rates W₁ as a function of air content; these were extracted from fitting the SR-EPR data to Eq. 4 using single (PLB and pPLB) or double (PLB + SERCA and pPLB + SERCA) exponential decays. The group of faster relaxation rates were assigned to the state α, and the slower relaxation rates to β. Each value is given as the mean and standard error from multiple experiments, including error due to the fitting process. (b) The residuals of least-squares fits to SR-EPR data under 50% air are given, indicating that a single exponential works well for PLB and pPLB, whereas two exponentials are required when SERCA is present. To see this figure in color, go online.

Figure 6.

Oxygen exchange rates $j_{{\text{O}}_2}$ for α (green) and β (purple) obtained from independent least-squares fits to electron relaxation rates W₁ versus air content using Eq. 5. The values of $j_{{\text{O}}_2,\alpha }$ are very similar for all conditions, suggesting that they correspond to the same state of PLB. Additionally, there is no significant change in $j_{{\text{O}}_2,\beta }$ after phosphorylation. To see this figure in color, go online.

Results

To investigate the behavior of PLB’s TM domain relative to SERCA and the effects of phosphorylation, we performed CW-EPR spectroscopy. A recent EPR study (7) that labeled PLB’s TM domain in the presence of SERCA employed 2,2,6,6-tetramethylpiperidine-1-oxyl-4-amino-4-carboxylic acid (TOAC), a nonnatural amino acid spin label rigidly coupled to the protein backbone. This allowed for determination of changes in the effective rotational motion of the TM domain’s backbone via ST-EPR after SERCA-binding, but ST-EPR does not provide resolution of different (e.g., unbound and bound) rotational mobilities, and CW-EPR was unable to resolve the SERCA-bound and unbound populations because of the very slow motion of TOAC on the microsecond timescale. Here, we use the more flexible spin labels MSL and MTSSL, whose mobilities are more representative of side-chain dynamics on timescales of ∼1–10 ns. The goal was to resolve the SERCA-bound and unbound populations because of changes in 1) side-chain dynamics via CW-EPR and 2) accessible surface area to the lipids via PS-EPR and SR-EPR.

Adding SERCA decreases rotational mobility of MSL on PLB’s TM domain, and phosphorylation partially reverses this effect

The rotational motion of MSL, relative to the peptide backbone, is more restricted than that of MTSSL because of its larger functional group (Fig. 1, right), making it useful for measuring changes in side-chain mobility on slower timescales corresponding to rotational correlation times $\left({\tau }_{\text{R}}\right)$ of ∼10 ns (43). For MSL-labeled sites, EPR spectra (Fig. 2, left) of PLB + SERCA (red) showed increased broadening relative to PLB alone (black), whereas pPLB + SERCA (blue) showed a level of broadening between PLB and PLB + SERCA. To quantitatively measure changes in these spectra, the outer splitting $2T_{\left|\right|}^{\prime }$ was determined for each condition (Fig. 2, right). As with spectral broadening, $2T_{\left|\right|}^{\prime }$ was larger for PLB + SERCA relative to PLB alone, whereas pPLB + SERCA reported a $2T_{\left|\right|}^{\prime }$ somewhat less than PLB + SERCA for all labeling sites.

Figure 2.

(a) CW-EPR spectra of MSL-Q29 (left) and outer splitting $2T_{\left|\right|}^{\prime }$ and rotational correlation time ${\tau }_{\text{R}}$ for each spectrum (right) acquired from PLB alone (black), PLB + SERCA (red), and pPLB + SERCA (blue). Each spectrum is normalized by its double integral value. (b and c) show the same data as (a), but for MSL-F32 and MSL-M50, respectively. To see this figure in color, go online.

Increases in broadening and $2T_{\left|\right|}^{\prime }$ can indicate greater motional restriction of the spin label. For a model of isotropic motion of a single spin population, $2T_{\left|\right|}^{\prime }$ can be used to estimate the rotational correlation time ${\tau }_R$ for the spin label (44) (Fig. 2, right). Because protein-protein interactions probably reduce the mobility of PLB’s side chains after SERCA-binding, this could explain the observed spectral changes in the presence of SERCA. With regard to phosphorylation, neither the broadening nor $2T_{\left|\right|}^{\prime }$ returned to the same values as PLB alone. One possibility is that the mobility of MTSSL in the SERCA-bound state of PLB increases after phosphorylation, whereas the mole fractions corresponding to each state remain constant. Another conclusion would be that there is a shift in equilibrium toward the unbound state of PLB, which would cause a shift in the apparent splitting shown in the spectrum. Presumably, the PLB + SERCA and pPLB + SERCA samples each contain both SERCA-bound and unbound populations of PLB and pPLB, respectively. However, for each spin-labeling location, these samples yielded what appear to be single-component EPR spectra that are similar to those of PLB-only samples, and thus MSL is unable to fully resolve the multiple states of PLB in the presence of SERCA. There may indeed be two spin populations present in the spectra for PLB + SERCA and pPLB + SERCA whose mole fractions shift, which could cause changes in apparent splitting, but if they do exist, they are not clearly resolvable. To gain more insight into whether these effects were due to a change in the properties of the bound state itself or a change in equilibrium mole fractions of multiple spin populations, we also performed CW-EPR using a more mobile spin label.

Adding SERCA yields two EPR spectral components for MTSSL, and phosphorylation shifts their relative populations

We also acquired CW-EPR spectra using MTSSL (Fig. 3 a). MTSSL has significant mobility relative to the host protein backbone (${\tau }_{\text{R}}\approx 1-10$ ns) because of its disulfide linker (Fig. 1, upper right), making it sensitive to changes in side-chain mobility and the surrounding environment. The PLB-only spectrum (black) at L43 showed a single spectral component with relatively fast motion (${\tau }_{\text{R}}\approx 2$ ns), indicating high side-chain mobility for the unbound state of PLB. For PLB + SERCA (red), there are two spectral components, one that appears to be identical to that of PLB (mobile component, α) and another with larger outer splitting (restricted component, β). Upon phosphorylation, pPLB + SERCA also shows these two spectral components, but the equilibrium mole fraction of α is increased with respect to β. Additionally, EPR spectra of PLB-only and pPLB showed no detectable differences (data not shown), indicating that the presence of SERCA is required to observe these spectral changes. We interpret the presence of β as the restriction of MTSSL’s dynamics by interactions with one of SERCA’s TM helices in the bound state. Similar results were obtained from MTSSL-L39 (Fig. S1).

Figure 3.

(a) CW-EPR spectra of MTSSL-L43 acquired from PLB (black), PLB + SERCA (red), and pPLB + SERCA (blue). Each spectrum is normalized by its double integral value. The PLB + SERCA and pPLB + SERCA spectra are resolved into two spectral components, designated α and β, which are labeled and highlighted in green and purple, respectively. (b) A global least-squares fitting of MTSSL-L43 PLB + SERCA (red) is given. A scaled version of the PLB-only spectrum (with a 0.5 G Lorentzian broadening) was used in place of α (green), whereas component β was simulated, and the total fit (gray) is a weighted sum of α and β. The mole fraction $x_{\alpha }$ obtained from fitting is also displayed. (c) shows the same as (b), but for pPLB + SERCA (blue). Spectral simulations and fitting were performed using EasySpin (40) and a custom script written in MATLAB (The MathWorks, Natick, MA). To see this figure in color, go online.

To investigate whether the α component observed in PLB + SERCA and pPLB + SERCA represented the same state as the PLB-only spectra, we performed spectral decomposition through spectral simulation and fitting using EasySpin, which can simulate CW-EPR spectra in the slow-motion regime (40, 45). For the α component, we used the empirically measured PLB-only spectrum with some amount of Lorentzian broadening, whereas the β component was simulated. The results for MTSSL-L43 (Fig. 3, b and c, fit parameters in Table S1) show that this model agrees very well with the observed EPR spectra, indicating that the α component in PLB + SERCA and pPLB + SERCA are probably caused by the presence of unbound PLB, whose mole fraction shifted from $x_{\alpha }=0.42\pm 0.08$ to 0.65 ± 0.05 after phosphorylation. Similar results were obtained from MTSSL-L39, in which the mole-fraction shift was $x_{\alpha }=0.48\pm 0.04$ to 0.84 ± 0.06 (Fig. S1, fit parameters in Table S1).

The existence of the α and β components in CW-EPR spectra indicates the presence of multiple spin populations that are experiencing significantly different environments. Regarding the spectra acquired from PLB + SERCA and pPLB + SERCA, the components α and β reveal different amounts of motional restriction. This behavior could be caused by different spin-label states (often referred to as rotameric states) or different protein states. In addition to spectral decomposition, the problem of distinguishing between these two cases for EPR spectra of spin-labeled proteins can be addressed using SR-EPR (26, 27, 41), which was also performed in this study and will be discussed later.

SERCA-binding and phosphorylation affect PLB lipid accessibility but not membrane insertion depth

A more systematic investigation of the TM domain’s proximity to the lipid membrane environment was performed by measuring spin-label accessibility. Using PS-EPR, the EPR signal can be progressively saturated with increasing microwave power P in the presence and absence of a PRA (examples of power-saturation curves are given in Fig. S2). The lipid accessibility of each spin-label location for MSL (at residues 29, 32, and 50; CW-EPR spectra given in Fig. 2) and MTSSL (at residues 39 and 43; CW-EPR spectra given in Fig. 3) was measured using O₂ (zero grade air) (Fig. 4 a, ${\mathit{\Pi }}_{{\text{O}}_2}$), and solvent accessibility was measured using NiEDDA (Fig. 1 b, ${\mathit{\Pi }}_{\text{NiEDDA}}$), for PLB, PLB + SERCA, and pPLB + SERCA. Overall, ${\mathit{\Pi }}_{{\text{O}}_2}$ decreased significantly at all sites for PLB + SERCA relative to PLB, which is consistent with PLB binding to SERCA, in which the surface area at which O₂ molecules can collide with spin labels is greatly reduced. For pPLB + SERCA, ${\mathit{\Pi }}_{{\text{O}}_2}$ increased by a small amount at all sites, indicating either a structural shift of the TM domain outward relative to SERCA or an increase in the fraction of unbound PLB. Regarding solvent accessibility, much smaller changes were observed in the values of ${\mathit{\Pi }}_{\text{NiEDDA}}$, though these changes were mostly consistent with those of ${\mathit{\Pi }}_{{\text{O}}_2}$.

Figure 4.

(a) Oxygen accessibility ${\mathit{\Pi }}_{{\text{O}}_2}$ versus residue number of PLB-only (black), PLB + SERCA (red), and pPLB + SERCA (blue). (b) NiEDDA accessibility ${\mathit{\Pi }}_{\text{NiEDDA}}$ versus residue number is shown. (c) The depth parameter Φ versus residue number is shown. For a point of reference, the dashed line indicates the Φ-value measured from the spin-labeled fatty acid 5-DSA, which has been shown to associate near the charged phospholipid headgroups. Each value is given as the mean and standard error from multiple experiments, including error propagation through the expression for Φ. To see this figure in color, go online.

We also extracted the membrane-insertion-depth parameter Φ from the accessibility results (Fig. 4). For PLB, PLB + SERCA, and pPLB + SERCA, there is a maximum in ${\mathit{\Pi }}_{{\text{O}}_2}$ at L39, whereas ${\mathit{\Pi }}_{\text{NiEDDA}}$ fluctuates at a minimal value between L39 and the nearest measured residue L43. This indicates that L39 is probably located near the center of the lipid bilayer. As one moves away from L39, ${\mathit{\Pi }}_{{\text{O}}_2}$ decreases, whereas ${\mathit{\Pi }}_{\text{NiEDDA}}$ increases and reaches maximal values at F32 and M50, indicating that these residues are located near the surfaces of the outer and inner leaflets of the membrane. Correspondingly, the depth parameter Φ has a maximum at L39 and monotonically decreases at other residues, regardless of the presence of SERCA or phosphorylation (Fig. 4 c). These observations indicate that the insertion depth of PLB does not change significantly after binding to SERCA or phosphorylation.

For reference, Φ was also measured using the spin-labeled fatty acid 5-DOXYL-stearic acid (5-DSA), and its value is indicated using a dashed line (Fig. 4 c). The insertion depth of 5-DSA is known to fluctuate just below charged phospholipid headgroups. The fact that Φ-values for Q29, F32, and M50 are less than that of 5-DSA confirms that these residues are probably located closer to the lipid headgroups in the inner and outer leaflets.

The fact that the pattern in Φ-values across the TM domain was unaffected by SERCA-binding or phosphorylation, whereas ${\mathit{\Pi }}_{{\text{O}}_2}$-values changed significantly, led us to conclude that PLB’s accessible surface area to the lipid bilayer was probably changing via shifts in equilibrium mole fractions between SERCA-bound and unbound states (β and α, respectively). This interpretation is consistent with the observed changes in CW-EPR spectra acquired using MSL (Fig. 2) and MTSSL (Fig. 3). However, measuring changes in the oxygen accessibility via PS-EPR experiments alone is insufficient to distinguish between a structural shift within the SERCA-PLB complex or a change in binding. Therefore, additional oxygen-accessibility measurements were performed using SR-EPR, which can resolve multiple states with different oxygen accessibilities. These results are discussed below after the membrane insertion depth results.

SR-EPR confirms the existence of two spin populations in the presence of SERCA

To investigate the nature of the states α and β in the CW-EPR spectra, as well as the reason behind the observed changes in lipid accessibility, SR-EPR experiments were performed. Because PS-EPR is a steady-state measurement, only the average accessibility of an ensemble can be measured. As such, it is difficult to determine the accessibility of different spectral components using this technique. SR-EPR is a time-resolved, pulsed version of this experiment that can independently measure the accessibility of multiple spin populations (24, 26, 27, 39, 41, 46, 47).

SR-EPR experiments were performed on PLB, pPLB, PLB + SERCA, and pPLB + SERCA for MTSSL-L43, which had the highest measured value of oxygen accessibility with varying air content (Fig. 5 a shows SR-EPR data acquired using 50% air). Least-squares fitting was performed using Eq. 4 to model the data and obtain the electron relaxation rates (residuals shown in Fig. 2 b; a table of fit parameters is given in Table S2). For PLB and pPLB, only a single exponential decay function was required (the first term on the right-hand side of Eq. 4, yielding $W_{1\alpha }$), as expected based on the presence of a single component in each of their CW-EPR spectra. However, adequately fitting PLB + SERCA and pPLB + SERCA recovery curves required the full sum of two exponential decay functions (yielding both $W_{1\alpha }$ and $W_{1\beta })$, which was expected based on the presence of two components in each of their CW-EPR spectra. This indicates the presence of two spin populations with different electron relaxation rates, $W_{1\alpha }$ and $W_{1\beta }$, undergoing exchange on a timescale that is slow relative to SR-EPR, i.e., $k_{\text{ex}}\mathrm{\lesssim }100$ KHz. This observation is consistent with a previous study that used Förster transfer recovery to measure the binding and unbinding rates of SERCA-PLB (42).

Plotting the electron-spin-relaxation rates obtained from each configuration of PLB as a function of air content was further revealing (Fig. 5 a, inset). Linear fits to this data yielded $j_{\alpha ,{\text{O}}_2}$ and $j_{\beta ,{\text{O}}_2}$ (a time-resolved equivalent to ${\mathit{\Pi }}_{{\text{O}}_2}$ from PS-EPR) from the slopes (Fig. 6) and the bare electron relaxation rates $W_{1\alpha ,0}$ and $W_{1\beta ,0}$ from the intercepts, as in Eq. 5. The results showed that 1) values for $W_{1\alpha }$ are nearly identical between PLB and pFPLB; 2) the values of $W_{1\alpha }$, and hence $j_{\alpha ,{\text{O}}_2}$, for PLB and pPLB are very similar to those of PLB + SERCA and pPLB + SERCA; and 3) the values of $j_{\beta ,{\text{O}}_2}$ are very similar between PLB + SERCA and pPLB + SERCA. From these findings, we determined that the state α is present for all conditions, whereas the state β is present only in the presence of SERCA, in agreement with the results of CW-EPR global spectral fitting on the same samples.

As a further test for consistency between the results of CW-EPR and SR-EPR, all SR-EPR data were globally fitted using Eq. 5 with the full expressions from (41) and the mole-fraction values $x_{\alpha }$ obtained from CW-EPR fitting (Fig. 7; fit parameters given in Table S3). Here, the values of $W_{1\alpha ,0}$ and $j_{\alpha ,{\text{O}}_2}$ were linked across all SR-EPR spectra, and the values of $W_{1\beta ,0}$ and $j_{\beta ,{\text{O}}_2}$ were linked across PLB + SERCA and pPLB + SERCA. The starting values for other variables were chosen using the results from independent fitting (Fig. 5). The residuals from global fitting indicate that our interpretation of the CW-EPR spectra, namely that only the mole fractions of the states α and β change after phosphorylation, is in excellent agreement with all SR-EPR data.

Figure 7.

Global fitting to SR-EPR data for PLB and PLB + SERCA (red and blue, top), and pPLB and pPLB + SERCA (black and blue, bottom) acquired under various mixtures of air and N₂ for L43. The results are offset for clarity. Residuals of the least-squares global fits are shown at the top of each graph using the same color schemes, indicating that a model that uses the mole fractions obtained from CW-EPR fits the data well. To see this figure in color, go online.

Discussion

Using site-directed spin labeling and EPR spectroscopy, we have investigated the effects of SERCA coreconstitution and phosphorylation on PLB’s TM domain. By spin-labeling six different locations on the side opposite the SERCA-interacting residues, we systematically measured both O₂ (Fig. 4, top) and NiEDDA (Fig. 4, middle) accessibility and showed that, for all labeling sites, adding SERCA significantly reduced lipid accessibility, whereas phosphorylation partially reversed it toward PLB-only levels. Additionally, no changes in the membrane insertion depth were detected (Fig. 4, bottom). At two of these labeling locations, L39 and L43, we used the flexible probe MTSSL, whose CW-EPR spectra revealed two spin populations: a mobile component (α) and a restricted component (β). The α component was present for all conditions, whereas the β component only existed in the presence of SERCA. To investigate this behavior further, we then performed time-resolved accessibility measurements on L43 using SR-EPR. The SR-EPR data confirmed that PLB showed only one spin population, whereas PLB + SERCA and pPLB + SERCA both showed two spin populations, one of which (α) had identical lipid accessibility to that shown by PLB alone.

Binding of MTSSL-PLB to SERCA is detectable by CW-EPR and SR-EPR

We conclude that α represents unbound PLB and β represents SERCA-bound PLB based on the following: 1) CW-EPR spectra of PLB + SERCA and pPLB + SERCA were successfully fit in a global sense using the respective PLB-only spectra in place of component α for MTSSL-L39 (Fig. S1) and MTSSL-L43 (Fig. 3); 2) the T₁ relaxation rates $\left(W_{1\alpha ,0}\right)$ and lipid accessibilities $\left(j_{\text{ex},\alpha }\right)$ shown by SR-EPR of PLB-only and pPLB-only were very similar to those shown by the α components from both PLB + SERCA and pPLB + SERCA data (Fig. 6); 3) $W_{1\beta ,0}$ and $j_{\text{ex},\beta }$ were very similar between PLB + SERCA and pPLB + SERCA; and 4) the SR-EPR data can be globally fit using the mole fractions obtained from the CW-EPR spectral fitting results. These results are consistent with previous FRET studies (33, 48) and an STEPR study (7) performed on the same system.

One alternative explanation regarding the nature of α and β is that they are caused by different spin label rotamers, not protein states. However, the observation that both CW-EPR and SR-EPR show two spin populations suggests that these two states are undergoing exchange on a timescale that is slow with respect to both measurements, i.e., $k_{\text{ex}}\mathrm{\lesssim }100$ KHz, which is too slow for detectable spin label conformational exchange. Additionally, spectral decomposition using least-squares fitting was consistent with the α state being present after SERCA coreconstitution. These results are consistent with other studies that have used the same arguments to rule out the possibility of rotameric states in other systems (26, 41).

Phosphorylation causes some dissociation of the SERCA-PLB complex, though a significant amount of PLB remains bound

In the presence of SERCA, CW-EPR spectral fitting (Fig. 3; Fig. S1) showed an increase in $x_{\alpha }$ of 0.23 ± 0.09 at L43 and 0.36 ± 0.07 at L39 after phosphorylation of PLB. These results are consistent with the fractional changes in overall O₂ accessibility relative to PLB alone (Fig. 4) at L43 and L39 (0.24 ± 0.02 and 0.22 ± 0.07, respectively). Because we have shown that α represents unbound PLB, we conclude that phosphorylation causes some dissociation of the SERCA-PLB complex and that there is still a significant amount of bound pPLB. Additionally, no changes in PLB’s relative position within the membrane, based on insertion depth (Fig. 4) or accessibility of the bound state β (Fig. 6), were detected. This suggests that the previously observed equilibrium shift in PLB’s CP domain toward the R-state (33), induced by phosphorylation, is primarily responsible for changes in binding and reduction of SERCA inhibition. Alternatively, it is possible that there exists a noninhibitory PLB binding site on SERCA and/or a noninhibitory conformation of pPLB, though x-ray diffraction studies have only observed one binding site on SERCA (49), and we observed negligible changes in the lipid accessibility of the bound state after phosphorylation.

The observed change in SERCA-binding after phosphorylation is consistent with the main results of a previous time-resolved FRET study with a donor probe on PLB’s CP domain and an acceptor probe on SERCA’s CP domain (33). Under similar reconstitution conditions, that study showed that phosphorylation of PLB completely relieved inhibition of SERCA ATPase activity but eliminated only about half of the PLB molecules bound to SERCA, resulting in a dissociation constant of $K_D\approx 6.0\pm 0.8$ (per 1000 lipids) for SERCA-pPLB interaction. Taking the average between the results for L39 and L43, we found the unbound fraction of pPLB to be $x_{\alpha }=0.75\pm 0.13$, which corresponds to an estimated K_D = 3.6 ± 2.9 (Eq. S3) and is in reasonable agreement with the FRET study. In cardiac sarcoplasmic reticulum, the concentrations of SERCA and pPLB (${\left[\text{SERCA}\right]}_{\text{t}}$ and ${\left[\text{pPLB}\right]}_{\text{t}}$) in the membrane are much higher, with a large molar excess of ${\left[\text{pPLB}\right]}_{\text{t}}/{\left[\text{SERCA}\right]}_{\text{t}}\approx 3.5$. Under those conditions, the K_D from this study implies that the fraction of SERCA with bound pPLB would be 0.83 ± 0.13, indicating that the SERCA-bound state of pPLB would be favored strongly under physiological conditions. Regarding SERCA functional activity in the presence of PLB, it is typically measured under conditions with a significant excess of PLB, and the significant reduction of inhibition because of PLB phosphorylation is well established (5, 7, 8, 33, 50). Therefore, we conclude that the dominant mechanism of phosphorylation-mediated SERCA activity restoration is not due to a change in PLB binding but to the previously observed structural change in the CP domain.

Conclusions

We have investigated the interactions between the TM domains of SERCA and PLB via site-directed PLB spin labeling and EPR spectroscopy in lipid bilayers. The CW-EPR spectra of MTSSL-labeled PLB revealed high rotational mobility. In the presence of SERCA, two components were observed, one indistinguishable from that of PLB-only and a second component that was substantially more restricted, corresponding to SERCA-bound PLB. Phosphorylation of PLB at Ser16 caused no change in the absence of SERCA but in the presence of SERCA caused a shift in mole fraction toward the mobile component and a corresponding increase in lipid accessibility. To confirm that these two spectral components were due to distinct protein states, SERCA-bound and unbound PLB, SR-EPR was performed under varying concentrations of O₂. We found that there indeed was a single spin population for PLB-only and two spin populations with different lipid accessibilities for PLB + SERCA and pPLB + SERCA, with a mole-fraction shift toward the population with increased lipid accessibility, in excellent agreement with the mole-fraction shifts found using CW-EPR. These findings provide a more complete picture of how phosphorylation of PLB affects interactions between PLB and SERCA, and hence SERCA inhibition: phosphorylation does decrease the fraction of SERCA containing bound PLB, but the most important effect for SERCA function is a structural shift in PLB’s CP domain relative to SERCA.

Author Contributions

P.D.M. and Z.M.J. performed CW-EPR and PS-EPR experiments. P.D.M. performed SR-EPR experiments. P.D.M. and D.D.T. analyzed all EPR data. P.D.M. and D.D.T. wrote the manuscript. All authors approved the final version of the manuscript.

Editor: Joseph Falke.