Spectroscopic validation of the pentameric structure of phospholamban

Results

Solid-State NMR in Oriented Lipid Bilayers.

Solid-state NMR is the method of choice for the simultaneous identification of membrane protein structure, topology, and dynamics (24–26). Unlike solution NMR, the ^{N chemical shift in solid-state spectra depends on its orientation with respect to the magnetic field. This basic principle allows for the measurement of orientation (topology) information with respect to the lipid membrane. Here, the [H/N] polarization inversion spin exchange at the magic angle (PISEMA) experiment was used. This experiment correlates N chemical shift with H–N dipolar coupling (}27). For α-helices within oriented lipid bilayers (B₀ parallel to the membrane normal), PISEMA spectra show wheel-like patterns called polarity index slant angle (PISA) wheels (28, 29). From the PISA wheels it is possible to identify the tilt and rotation angles of the helix within the bilayer. Specifically, helices that are approximately parallel with respect to the bilayer normal give ^{N chemical shifts ≈200 ppm and H–N dipolar couplings ≈6–8 kHz, whereas those with an orientation perpendicular to the normal resonate at ≈75 ppm and ≈3–4 kHz, respectively. Because there are large orientation differences in the cytoplasmic domain within the four structural models, PISA wheel patterns measured in fully fluid lipid bilayers are able to distinguish between these models with high sensitivity.}

Fig. 2 shows an overlay of six different PISEMA experiments (high- and low-field regions) acquired with selectively labeled [^{N-Leu], [N-Ala], [N-Ile], [N-Cys], [N-Thr], and [N-Asn] wt-PLN samples. The full PISEMA spectra are shown in} SI Fig. 9, indicating the degree of alignment for the samples. The overlaid spectra clearly show the presence of two PISA wheels for the transmembrane (Fig. 2B) and cytoplasmic domains (Fig. 2C). Residue assignments were carried out by using an in-house software that exploits the periodic nature of the chemical shift and dipolar coupling for regular secondary structures (30, 31). Analogous to the approach by Nevzorov and Opella (32), this program uses an exhaustive search algorithm to find the assignment that best matches the observed resonances with those calculated from an ideal α-helix for given tilt (θ) and rotation (ρ) angles with respect to the lipid bilayer. The PISA wheel for the transmembrane domain of wt-PLN (Fig. 2D) comprises 21 assigned resonances and reveals an ≈15° tilt angle with respect to the bilayer normal. This tilt angle differs from that of the PLN monomer by ≈6° (18), showing that formation of the leucine/isoleucine zipper imposes a smaller tilt to each monomer with respect to the depolymerized species. Two previous studies reported tilt angles of 28 ± 6° (8) and 23° by using computation and mutagenesis (33). The former study used dimyristoylphosphatidylcholine (DMPC) lipid bilayers. The difference in the tilt angle between our study and that of Arkin et al. (8) can be attributed to hydrophobic mismatch.

Fig. 2.

Solid-state NMR spectra of PLN pentamer in lipid bilayers. (A) 1D cross-polarization spectrum of [U-^{N] wt-PLN in DOPC/DOPE oriented lipid bilayers. (}B and C) An overlay of selectively labeled PISEMA spectra for the transmembrane and cytoplasmic helices, respectively. The resonances are color-coded: [^{N-Ala], green; [N-Cys], purple; [N-Leu], orange; [N-Ile], red; [N-Asn], gray; and [N-Thr], blue. (}D and E) Simulated PISA wheels (dashed lines) for both transmembrane (θ = 15°) and cytoplasmic (θ = 92°) domains. The PISEMA simulations assumed a regular α-helical geometry for both helical domains.

Unlike the difference in tilt angle between the monomer and pentamer, the rotation angle of the transmembrane domain helix is similar, indicating that pentamer formation requires only a change in tilt angle to pack. Moreover, the regular PISA wheel obtained for the transmembrane domain shows an unbent helix, excluding the pronounced transmembrane helix curvature reported in the bellflower model as well as the proposed antiparallel β-sheet configuration for residues 22–32 in the extended helix/sheet model. The simulated PISA wheel in Fig. 3E shows the expected pattern for the transmembrane domain within the bellflower structure. Note that the availability of Protein Data Bank (PDB) coordinates allows us to simulate PISEMA spectra for the pinwheel and bellflower models only.

Fig. 3.

Simulated PISEMA spectra obtained for the pinwheel and bellflower models. The simulations for the cytoplasmic domain (A and B) assumed ideal α-helices with helix tilt angles of 92° (A) and 20° (B) with a rotation angle (defined in ref. 18) of 15°. PISEMA spectra for the transmembrane domains (tm) are calculated directly from the PDB coordinates (D and E). Unlike the pinwheel model, the bellflower model does not show any high-field resonances. Experimental PISEMA spectra show the remarkable agreement with the pinwheel model (C and F).

Fig. 2C shows the overlay of the upfield region (cytoplasmic domain) of the PISEMA spectrum (50–100 ppm). All of the expected resonances (Leu-7, Thr-8, Ala-11, Ile-12, Ala-15, Thr-17, and Ile-18) are present for each selectively labeled sample, showing unambiguously that the cytoplasmic helix is oriented parallel to the surface of the lipid bilayer, supporting the pinwheel model. Structural fitting with an ideal PISA wheel gives an ≈92° tilt angle of the cytoplasmic domain with respect to the bilayer normal, essentially the same as that previously described for the monomer (18). A bellflower structure would give rise to cytoplasmic domain resonances within the range 160–220 ppm (Fig. 3B). Likewise, the continuous helix and extended helix/sheet models would show the majority of cytoplasmic domain resonances in the same region as those expected for the bellflower model.

Solution NMR in DPC Micelles.

Is the pentamer structure in micelles different from that in lipid bilayers? Following the sample preparation used by Oxenoid and Chou (2), we compared the transverse relaxation-optimized spectroscopy/heteronuclear sequential quantum correlation (TROSY-HSQC) spectrum of monomer and pentamer and tested the topology of the pentamer by using three different paramagnetic probes: gadopentetate dimeglumine (Gd ^{for simplicity), 5′-doxyl stearic acid, and 16′-doxyl stearic acid, which probe solvent-exposed residues, hydrophobic residues underneath the phosphate head group, and those residues deeply embedded in the micellar core, respectively.}

Remarkably, the resonance positions of the protein fingerprint in the [^{H/N]-TROSY-HSQC spectrum for residues 2–31 within wt-PLN and monomeric mutant AFA-PLN are nearly identical as shown in} Fig. 4. The largest differences between monomer and pentamer are located in domain II, resulting from the mutated residues and the quaternary interactions within the pentamer. The similarity within the cytoplasmic region (residues 2–31) suggests that the local chemical environment and the conformations for both monomer and pentamer are identical.

Fig. 4.

Solution NMR studies of PLN pentamer in DPC micelles. (A) Overlay of [^{H/N]-TROSY-HSQC spectra of AFA-PLN monomer (black) and wt-PLN pentamer (red). (}B) Difference plot of the combined ^{H and N chemical-shift variations between the monomeric and pentameric species (Δδ = [(Δδ}_H ^{+ Δδ}_N^{/25)/2)]). The three mutations in AFA-PLN (C36A, C41F, and C46A) are indicated with asterisks.}

Fig. 5A shows the intensity retention in the TROSY-HSQC spectrum of wt-PLN after adding Gd^{. The quenching of residues Glu-2, Lys-3, Ser-10, Gln-22, Gln-23, and Ala-24 indicates that these residues are the most solvent-exposed, whereas the absence or minimal perturbation of the remaining residues of the transmembrane and cytoplasmic domains suggests that these are somewhat solvent-protected.}

Fig. 5.

Paramagnetic mapping of PLN topology in DPC micelles. Intensity retention upon addition of Gd ⁽A), 5′-doxyl stearic acid (B), and 16′-doxyl stearic acid (C). The red bars in B and C highlight those residues quenched that face the micelle interior (Val-4, Leu-7, and Ala-11 in B and Leu-7 in C). The asterisks indicate overlapped resonances.

To investigate what residues were associated with or embedded within the DPC micelle, we used 5′- and 16′-doxyl stearic acid. Val-4, Leu-7, and Ala-11 amide resonances are among the most quenched when exposed to 5′-doxyl stearic acid, whereas Ser-10 (the most quenched in Gd ^{experiments) retains ≈80% intensity (}Fig. 5B). Similarly, 16′-doxyl stearic acid shows Leu-7 to be the most quenched in the cytoplasmic domain with residues 41–45 embedded in the core of the micelle. These results support the conclusion that Val-4 and Leu-7 are buried in the micellar hydrophobic core, placing the cytoplasmic domain helix in contact with the surface of the micelle. If domain Ia (residues 1–16) were fully exposed to solvent, as in the bellflower, continuous helix, and extended helix/sheet models, one would expect to observe uniform quenching of these domains upon addition of Gd ^{and no differential quenching upon addition of 5′- or 16′-doxyl stearic acid. In the pinwheel model, one would expect to see differential quenching in the cytoplasmic domain from addition of Gd and 5′-doxyl stearic acid, indicating a preferential surface for the cytoplasmic domain helix to interact with the lipid surface.}

In summary, the similarity of spectra between the pentamer and monomer, and the topological mapping by paramagnetic agents, all establish that in DPC micelles both the monomer and pentamer have cytoplasmic domains that interact with the surface of the micelle.

EPR in DPC Micelles.

To further test the structural models, we performed complementary EPR experiments. Although a single EPR experiment lacks the residue-by-residue view offered by NMR, it has several advantages stemming from the larger magnetic susceptibility of the electron, including increased sensitivity, resolution of distinct conformations based on dynamics (35), and the ability to probe distances up to 8 nm in length (36). To measure intermonomer distances in the wt-PLN pentamer, a four-pulse double electron-electron resonance (DEER) experiment was used (36). wt-PLN was spin-labeled at Lys-3 with 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO)-succinimide.

Assuming a rigid model, the shortest intermonomer distance for Lys-3 is 3.8 nm in the pinwheel model and 2.0 nm in the bellflower model, as illustrated in Fig. 6. The interaction between the spin labels is a dipole–dipole interaction (r ^{distance dependence). The closer the two spin labels are in space, the stronger the interaction between the dipoles and, subsequently, the faster the DEER decay curve relaxes. The results in DPC micelles are shown in} Fig. 6C, with the simulated decay curves based on the two models shown as solid lines. It is clear that the distribution of distances between the spin labels is in close agreement with those predicted by using the pinwheel model. If the major population in detergent micelles were the bellflower arrangement, a much steeper decay would be expected, as illustrated in Fig. 6C. Similar to NMR data, these EPR results indicate that, in DPC micelles, the pinwheel model is in much better agreement with the data than the bellflower model.

Fig. 6.

Structural models (A, pinwheel; B, bellflower) showing the sites of spin-label attachment (Lys-3) and the predicted interprotomer distances, measured from β-carbons. Pulsed EPR (DEER) decays (black squares) observed for wt-PLN, spin-labeled at Lys-3, in DPC micelles (C) and lipid bilayers (D). Solid curves in C and D show simulated decays predicted by the bellflower model (red), pinwheel model (blue), and the best-fit Gaussian distribution of distances (green).

EPR in Lipid Bilayers.

DEER measurements also were performed on wt-PLN labeled at Lys-3 with TEMPO-succinimide in DOPC/DOPE lipid bilayers. The results (Fig. 6D) are only slightly different from those found in detergent micelles (Fig. 6C), showing a decay profile that agrees well with the pinwheel model (Fig. 6C and D, blue curve). The decay in lipids is slightly faster than in micelles, indicating a slightly shorter distance that is in even better agreement with the pinwheel model, perhaps because of the curvature of the DPC micelle (37). In both micelles and bilayers, at least 1 nm [half-width at half maximum (HWHM)] of disorder in the pinwheel model (see SI Figs. 10 and 11) is necessary to dampen the oscillations in the decay and fit the data. However, the bellflower model did not fit the data regardless of the amount of disorder modeled. The best fit of the data to a Gaussian distribution of distances (green curves in Fig. 6C and D) is centered at 5.4 nm (HWHM 2.3 nm) in micelles (SI Fig. 10 B and D) and 4.7 nm (HWHM 1.9 nm) in bilayers (Fig. 11 B and D), which is in good agreement with the mean distance calculated by using the pinwheel model (4.8 nm). In contrast, the best Gaussian fit to the bellflower model is a distribution of distances centered at 2.8 nm.

As shown previously (13, 15), EPR spectra of AFA-PLN, labeled in the cytoplasmic domain at either Lys-3 with TEMPO-succinimide or at position 11 with 2,2,6,6-tetramethyl-piperidine-1-oxyl-4-amino-4-carboxyl (TOAC), have two dynamically dissimilar components within DPC micelles and DOPC/DOPE mixed lipid bilayers (Fig. 7A and B). These two components are especially well resolved in the case of the 11-TOAC spin label (Fig. 7B), which is attached rigidly to the α-carbon and thus accurately reflects the dynamics of the peptide backbone. The two components correspond to the T and R states of the PLN cytoplasmic domain (15), which are detected in DPC micelles by both EPR and solution NMR (13). Also, a lipid anchor attached to the N terminus of AFA-PLN stabilizes the membrane-bound conformation of the cytoplasmic domain, eliminating the EPR peak attributed to the R state (34), thus confirming that the R state corresponds to the more dynamic and extended conformation (Fig. 7).

Fig. 7.

EPR dynamics (A and B) and Ni ^{accessibility (}C–E) data comparing AFA-PLN (data shown in blue) with wt-PLN (data shown in red). EPR spectra were obtained from membrane-reconstituted PLN with TEMPO-succinimide (SUCSL) attached to Lys-3 (A and C) or with TOAC substituted for Ala-11 (B and D). In B, the two low-field peaks correspond to the resolved T and R states. C and D are progressive saturation curves, showing the enhanced relaxation caused by collisions with membrane-surface-bound Ni^{. Relative accessibilities derived from} C and D are shown in E.

Fig. 7A and B shows that the spectra of AFA-PLN and wt-PLN in the same reconstituted DOPC/DOPE lipid bilayer system are nearly identical, indicating that the T and R states have similar dynamics and mole fractions in the PLN pentamer and monomer. Two small but significant differences are seen: (i) in wt-PLN, the mole fraction of the R state is slightly greater (9%) than in AFA-PLN (7%) and (ii) the resonance corresponding to the T state (or bent population) of wt-PLN has a larger splitting, indicating a slightly greater restriction of rotational motion than the corresponding T state in the monomer. Accessibility measurements using Ni ^{chelated to the lipid surface also were performed with both spin labels on AFA-PLN and wt-PLN (}Fig. 7C–E). These results show that both cytoplasmic spin labels interact strongly with the membrane surface, with little or no difference between AFA-PLN and wt-PLN. These results are in agreement with the solution NMR topological mapping with 5′- and 16′-doxyl stearic acids (Fig. 5B and C).

In summary, all EPR results strongly support the conclusion that the predominant population of wt-PLN is one in which the overall geometry is a pinwheel with the cytoplasmic domain in direct contact with the surface of the bilayer.

Characterization of Oligomerization State.

To characterize the oligomeric states under our NMR and EPR conditions, we visualized PLN species by using denaturing (SDS/PAGE) and nondenaturing (native) gel-shift assays [supporting information (SI) Fig. 8]. Both gels clearly show that in 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC)/1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE) lipid bilayers and DPC micelles, wt-PLN is pentameric. Note that all of these preparations were tested for SERCA inhibition, and under both NMR and EPR conditions PLN causes a shift in SERCA affinity for calcium as previously reported (14, 23). These functionally identical preparations were subjected to structural studies by both NMR and EPR spectroscopy.

Solid-State NMR in Oriented Lipid Bilayers.

Solid-state NMR is the method of choice for the simultaneous identification of membrane protein structure, topology, and dynamics (24–26). Unlike solution NMR, the ^{N chemical shift in solid-state spectra depends on its orientation with respect to the magnetic field. This basic principle allows for the measurement of orientation (topology) information with respect to the lipid membrane. Here, the [H/N] polarization inversion spin exchange at the magic angle (PISEMA) experiment was used. This experiment correlates N chemical shift with H–N dipolar coupling (}27). For α-helices within oriented lipid bilayers (B₀ parallel to the membrane normal), PISEMA spectra show wheel-like patterns called polarity index slant angle (PISA) wheels (28, 29). From the PISA wheels it is possible to identify the tilt and rotation angles of the helix within the bilayer. Specifically, helices that are approximately parallel with respect to the bilayer normal give ^{N chemical shifts ≈200 ppm and H–N dipolar couplings ≈6–8 kHz, whereas those with an orientation perpendicular to the normal resonate at ≈75 ppm and ≈3–4 kHz, respectively. Because there are large orientation differences in the cytoplasmic domain within the four structural models, PISA wheel patterns measured in fully fluid lipid bilayers are able to distinguish between these models with high sensitivity.}

Fig. 2 shows an overlay of six different PISEMA experiments (high- and low-field regions) acquired with selectively labeled [^{N-Leu], [N-Ala], [N-Ile], [N-Cys], [N-Thr], and [N-Asn] wt-PLN samples. The full PISEMA spectra are shown in} SI Fig. 9, indicating the degree of alignment for the samples. The overlaid spectra clearly show the presence of two PISA wheels for the transmembrane (Fig. 2B) and cytoplasmic domains (Fig. 2C). Residue assignments were carried out by using an in-house software that exploits the periodic nature of the chemical shift and dipolar coupling for regular secondary structures (30, 31). Analogous to the approach by Nevzorov and Opella (32), this program uses an exhaustive search algorithm to find the assignment that best matches the observed resonances with those calculated from an ideal α-helix for given tilt (θ) and rotation (ρ) angles with respect to the lipid bilayer. The PISA wheel for the transmembrane domain of wt-PLN (Fig. 2D) comprises 21 assigned resonances and reveals an ≈15° tilt angle with respect to the bilayer normal. This tilt angle differs from that of the PLN monomer by ≈6° (18), showing that formation of the leucine/isoleucine zipper imposes a smaller tilt to each monomer with respect to the depolymerized species. Two previous studies reported tilt angles of 28 ± 6° (8) and 23° by using computation and mutagenesis (33). The former study used dimyristoylphosphatidylcholine (DMPC) lipid bilayers. The difference in the tilt angle between our study and that of Arkin et al. (8) can be attributed to hydrophobic mismatch.

Fig. 2.

Solid-state NMR spectra of PLN pentamer in lipid bilayers. (A) 1D cross-polarization spectrum of [U-^{N] wt-PLN in DOPC/DOPE oriented lipid bilayers. (}B and C) An overlay of selectively labeled PISEMA spectra for the transmembrane and cytoplasmic helices, respectively. The resonances are color-coded: [^{N-Ala], green; [N-Cys], purple; [N-Leu], orange; [N-Ile], red; [N-Asn], gray; and [N-Thr], blue. (}D and E) Simulated PISA wheels (dashed lines) for both transmembrane (θ = 15°) and cytoplasmic (θ = 92°) domains. The PISEMA simulations assumed a regular α-helical geometry for both helical domains.

Unlike the difference in tilt angle between the monomer and pentamer, the rotation angle of the transmembrane domain helix is similar, indicating that pentamer formation requires only a change in tilt angle to pack. Moreover, the regular PISA wheel obtained for the transmembrane domain shows an unbent helix, excluding the pronounced transmembrane helix curvature reported in the bellflower model as well as the proposed antiparallel β-sheet configuration for residues 22–32 in the extended helix/sheet model. The simulated PISA wheel in Fig. 3E shows the expected pattern for the transmembrane domain within the bellflower structure. Note that the availability of Protein Data Bank (PDB) coordinates allows us to simulate PISEMA spectra for the pinwheel and bellflower models only.

Fig. 3.

Simulated PISEMA spectra obtained for the pinwheel and bellflower models. The simulations for the cytoplasmic domain (A and B) assumed ideal α-helices with helix tilt angles of 92° (A) and 20° (B) with a rotation angle (defined in ref. 18) of 15°. PISEMA spectra for the transmembrane domains (tm) are calculated directly from the PDB coordinates (D and E). Unlike the pinwheel model, the bellflower model does not show any high-field resonances. Experimental PISEMA spectra show the remarkable agreement with the pinwheel model (C and F).

Fig. 2C shows the overlay of the upfield region (cytoplasmic domain) of the PISEMA spectrum (50–100 ppm). All of the expected resonances (Leu-7, Thr-8, Ala-11, Ile-12, Ala-15, Thr-17, and Ile-18) are present for each selectively labeled sample, showing unambiguously that the cytoplasmic helix is oriented parallel to the surface of the lipid bilayer, supporting the pinwheel model. Structural fitting with an ideal PISA wheel gives an ≈92° tilt angle of the cytoplasmic domain with respect to the bilayer normal, essentially the same as that previously described for the monomer (18). A bellflower structure would give rise to cytoplasmic domain resonances within the range 160–220 ppm (Fig. 3B). Likewise, the continuous helix and extended helix/sheet models would show the majority of cytoplasmic domain resonances in the same region as those expected for the bellflower model.

Solution NMR in DPC Micelles.

Is the pentamer structure in micelles different from that in lipid bilayers? Following the sample preparation used by Oxenoid and Chou (2), we compared the transverse relaxation-optimized spectroscopy/heteronuclear sequential quantum correlation (TROSY-HSQC) spectrum of monomer and pentamer and tested the topology of the pentamer by using three different paramagnetic probes: gadopentetate dimeglumine (Gd ^{for simplicity), 5′-doxyl stearic acid, and 16′-doxyl stearic acid, which probe solvent-exposed residues, hydrophobic residues underneath the phosphate head group, and those residues deeply embedded in the micellar core, respectively.}

Remarkably, the resonance positions of the protein fingerprint in the [^{H/N]-TROSY-HSQC spectrum for residues 2–31 within wt-PLN and monomeric mutant AFA-PLN are nearly identical as shown in} Fig. 4. The largest differences between monomer and pentamer are located in domain II, resulting from the mutated residues and the quaternary interactions within the pentamer. The similarity within the cytoplasmic region (residues 2–31) suggests that the local chemical environment and the conformations for both monomer and pentamer are identical.

Fig. 4.

Solution NMR studies of PLN pentamer in DPC micelles. (A) Overlay of [^{H/N]-TROSY-HSQC spectra of AFA-PLN monomer (black) and wt-PLN pentamer (red). (}B) Difference plot of the combined ^{H and N chemical-shift variations between the monomeric and pentameric species (Δδ = [(Δδ}_H ^{+ Δδ}_N^{/25)/2)]). The three mutations in AFA-PLN (C36A, C41F, and C46A) are indicated with asterisks.}

Fig. 5A shows the intensity retention in the TROSY-HSQC spectrum of wt-PLN after adding Gd^{. The quenching of residues Glu-2, Lys-3, Ser-10, Gln-22, Gln-23, and Ala-24 indicates that these residues are the most solvent-exposed, whereas the absence or minimal perturbation of the remaining residues of the transmembrane and cytoplasmic domains suggests that these are somewhat solvent-protected.}

Fig. 5.

Paramagnetic mapping of PLN topology in DPC micelles. Intensity retention upon addition of Gd ⁽A), 5′-doxyl stearic acid (B), and 16′-doxyl stearic acid (C). The red bars in B and C highlight those residues quenched that face the micelle interior (Val-4, Leu-7, and Ala-11 in B and Leu-7 in C). The asterisks indicate overlapped resonances.

To investigate what residues were associated with or embedded within the DPC micelle, we used 5′- and 16′-doxyl stearic acid. Val-4, Leu-7, and Ala-11 amide resonances are among the most quenched when exposed to 5′-doxyl stearic acid, whereas Ser-10 (the most quenched in Gd ^{experiments) retains ≈80% intensity (}Fig. 5B). Similarly, 16′-doxyl stearic acid shows Leu-7 to be the most quenched in the cytoplasmic domain with residues 41–45 embedded in the core of the micelle. These results support the conclusion that Val-4 and Leu-7 are buried in the micellar hydrophobic core, placing the cytoplasmic domain helix in contact with the surface of the micelle. If domain Ia (residues 1–16) were fully exposed to solvent, as in the bellflower, continuous helix, and extended helix/sheet models, one would expect to observe uniform quenching of these domains upon addition of Gd ^{and no differential quenching upon addition of 5′- or 16′-doxyl stearic acid. In the pinwheel model, one would expect to see differential quenching in the cytoplasmic domain from addition of Gd and 5′-doxyl stearic acid, indicating a preferential surface for the cytoplasmic domain helix to interact with the lipid surface.}

In summary, the similarity of spectra between the pentamer and monomer, and the topological mapping by paramagnetic agents, all establish that in DPC micelles both the monomer and pentamer have cytoplasmic domains that interact with the surface of the micelle.

EPR in DPC Micelles.

To further test the structural models, we performed complementary EPR experiments. Although a single EPR experiment lacks the residue-by-residue view offered by NMR, it has several advantages stemming from the larger magnetic susceptibility of the electron, including increased sensitivity, resolution of distinct conformations based on dynamics (35), and the ability to probe distances up to 8 nm in length (36). To measure intermonomer distances in the wt-PLN pentamer, a four-pulse double electron-electron resonance (DEER) experiment was used (36). wt-PLN was spin-labeled at Lys-3 with 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO)-succinimide.

Assuming a rigid model, the shortest intermonomer distance for Lys-3 is 3.8 nm in the pinwheel model and 2.0 nm in the bellflower model, as illustrated in Fig. 6. The interaction between the spin labels is a dipole–dipole interaction (r ^{distance dependence). The closer the two spin labels are in space, the stronger the interaction between the dipoles and, subsequently, the faster the DEER decay curve relaxes. The results in DPC micelles are shown in} Fig. 6C, with the simulated decay curves based on the two models shown as solid lines. It is clear that the distribution of distances between the spin labels is in close agreement with those predicted by using the pinwheel model. If the major population in detergent micelles were the bellflower arrangement, a much steeper decay would be expected, as illustrated in Fig. 6C. Similar to NMR data, these EPR results indicate that, in DPC micelles, the pinwheel model is in much better agreement with the data than the bellflower model.

Fig. 6.

Structural models (A, pinwheel; B, bellflower) showing the sites of spin-label attachment (Lys-3) and the predicted interprotomer distances, measured from β-carbons. Pulsed EPR (DEER) decays (black squares) observed for wt-PLN, spin-labeled at Lys-3, in DPC micelles (C) and lipid bilayers (D). Solid curves in C and D show simulated decays predicted by the bellflower model (red), pinwheel model (blue), and the best-fit Gaussian distribution of distances (green).

EPR in Lipid Bilayers.

DEER measurements also were performed on wt-PLN labeled at Lys-3 with TEMPO-succinimide in DOPC/DOPE lipid bilayers. The results (Fig. 6D) are only slightly different from those found in detergent micelles (Fig. 6C), showing a decay profile that agrees well with the pinwheel model (Fig. 6C and D, blue curve). The decay in lipids is slightly faster than in micelles, indicating a slightly shorter distance that is in even better agreement with the pinwheel model, perhaps because of the curvature of the DPC micelle (37). In both micelles and bilayers, at least 1 nm [half-width at half maximum (HWHM)] of disorder in the pinwheel model (see SI Figs. 10 and 11) is necessary to dampen the oscillations in the decay and fit the data. However, the bellflower model did not fit the data regardless of the amount of disorder modeled. The best fit of the data to a Gaussian distribution of distances (green curves in Fig. 6C and D) is centered at 5.4 nm (HWHM 2.3 nm) in micelles (SI Fig. 10 B and D) and 4.7 nm (HWHM 1.9 nm) in bilayers (Fig. 11 B and D), which is in good agreement with the mean distance calculated by using the pinwheel model (4.8 nm). In contrast, the best Gaussian fit to the bellflower model is a distribution of distances centered at 2.8 nm.

As shown previously (13, 15), EPR spectra of AFA-PLN, labeled in the cytoplasmic domain at either Lys-3 with TEMPO-succinimide or at position 11 with 2,2,6,6-tetramethyl-piperidine-1-oxyl-4-amino-4-carboxyl (TOAC), have two dynamically dissimilar components within DPC micelles and DOPC/DOPE mixed lipid bilayers (Fig. 7A and B). These two components are especially well resolved in the case of the 11-TOAC spin label (Fig. 7B), which is attached rigidly to the α-carbon and thus accurately reflects the dynamics of the peptide backbone. The two components correspond to the T and R states of the PLN cytoplasmic domain (15), which are detected in DPC micelles by both EPR and solution NMR (13). Also, a lipid anchor attached to the N terminus of AFA-PLN stabilizes the membrane-bound conformation of the cytoplasmic domain, eliminating the EPR peak attributed to the R state (34), thus confirming that the R state corresponds to the more dynamic and extended conformation (Fig. 7).

Fig. 7.

EPR dynamics (A and B) and Ni ^{accessibility (}C–E) data comparing AFA-PLN (data shown in blue) with wt-PLN (data shown in red). EPR spectra were obtained from membrane-reconstituted PLN with TEMPO-succinimide (SUCSL) attached to Lys-3 (A and C) or with TOAC substituted for Ala-11 (B and D). In B, the two low-field peaks correspond to the resolved T and R states. C and D are progressive saturation curves, showing the enhanced relaxation caused by collisions with membrane-surface-bound Ni^{. Relative accessibilities derived from} C and D are shown in E.

Fig. 7A and B shows that the spectra of AFA-PLN and wt-PLN in the same reconstituted DOPC/DOPE lipid bilayer system are nearly identical, indicating that the T and R states have similar dynamics and mole fractions in the PLN pentamer and monomer. Two small but significant differences are seen: (i) in wt-PLN, the mole fraction of the R state is slightly greater (9%) than in AFA-PLN (7%) and (ii) the resonance corresponding to the T state (or bent population) of wt-PLN has a larger splitting, indicating a slightly greater restriction of rotational motion than the corresponding T state in the monomer. Accessibility measurements using Ni ^{chelated to the lipid surface also were performed with both spin labels on AFA-PLN and wt-PLN (}Fig. 7C–E). These results show that both cytoplasmic spin labels interact strongly with the membrane surface, with little or no difference between AFA-PLN and wt-PLN. These results are in agreement with the solution NMR topological mapping with 5′- and 16′-doxyl stearic acids (Fig. 5B and C).

In summary, all EPR results strongly support the conclusion that the predominant population of wt-PLN is one in which the overall geometry is a pinwheel with the cytoplasmic domain in direct contact with the surface of the bilayer.

Discussion

The fully functional monomeric AFA-PLN mutant has been well characterized by the G.V. and D.D.T. groups using solution NMR (1, 11–14) and EPR in detergent micelles (13), EPR in lipid vesicles (13, 15, 34), and solid-state NMR in mechanically oriented lipid bilayers (18). In all of these experiments, the results are unambiguous: the major population of AFA-PLN is L-shaped. Our studies had focused on the monomeric mutant to study the complex with SERCA, given that the monomer is believed to be responsible for inhibition, as well as to eliminate potential complex equilibria between various oligomers of wt-PLN. However, recent studies suggested that the pentameric form of PLN could play a larger role in SERCA regulation than previously was thought (5). Therefore, we embarked in the elucidation of the structure and topology of the pentamer in lipid bilayers.

Four distinct structural models proposed for pentameric wt-PLN (Fig. 1) differ primarily in the orientation of the cytoplasmic domain. In the present study, we investigated the topology of wt-PLN in detergent micelles, lipid vesicles, and oriented lipid bilayers by using EPR and solution/solid-state NMR spectroscopies, representing the most exhaustive structural investigation of wt-PLN pentamer to date. Our data in detergent micelles and lipid bilayers all support the same conclusion: the cytoplasmic domain of wt-PLN, like that of the AFA-PLN monomer, lies on the surface of the membrane, supporting an overall pinwheel geometry for the pentamer.

The pinwheel model qualitatively describes the NMR and EPR data yet is not a definitive high-resolution structure of pentameric PLN. It was constructed based on the AFA-PLN monomer structure with the addition of three distance restraints measured by FRET (10). Because the data in DPC micelles and lipid bilayers are consistent, the inclusion of restraints from a number of different techniques—including NOEs from solution NMR, DEER distance measurements from EPR, rotational-echo double-resonance (REDOR) distances from solid-state NMR, and PISEMA solid-state NMR restraints—should be used. This inclusion will reduce the ambiguity associated with data interpretation from one technique, further validating a structural model.

Although there are differences between lipid bilayers and detergent micelles (surface curvature, dynamics, etc.), the careful choice of a micellar system has been shown to be a good mimic for a physiological lipid environment (38, 39). Furthermore, for our system, biological function can be tested directly under the conditions used for NMR experiments (14, 31). In our case, DPC preserves both SERCA function and the PLN inhibitory activity under NMR conditions (13, 14). In the present study, our results in DPC micelles indicate that the monomer tree and within the pentamer adopt a similar structural and topological arrangement. Using both solid-state NMR and EPR, we also show that these structures are essentially the same in lipid bilayers.

A weakness of the previous studies is the complete reliance on single data sets provided by each technique employed. For instance, the structure proposed by Oxenoid and Chou is heavily based on the use of residual dipolar couplings (RDCs). In fact, there are no distance restraints that keep the cytoplasmic domains perpendicular to the plane of the lipid bilayers. Although RDC measurements are a well established method for obtaining long-range restraints and domain orientations (40–42), protein dynamics modulates the observed dipolar coupling (41, 43), complicating the interpretation of RDCs (43, 44). Our analysis of RDC values for PLN monomer shows that multiple orientations of the cytoplasmic domains are compatible with the RDC data, including the L-shaped and bellflower conformations (30). However, we selected those solutions that satisfied structural restraints and paramagnetic quenching experiments simultaneously (1, 30).

In this article, we show that only a multitechnique approach is capable of converging to a unique solution. Specifically, anisotropic PISEMA solid-state NMR (Figs. 2 and ​and3),3), paramagnetic quenching data obtained by both solution NMR (Fig. 5) and EPR (Fig. 7E), and the long-range distance restraints provided by EPR (DEER) data (Fig. 6) all clearly demonstrate that only the pinwheel model is supported in both DPC micelles and lipid bilayers. Finally, as illustrated in Fig. 7, the EPR dynamics data (Fig. 7B) show that the cytoplasmic domain of PLN is just as dynamic in the pentamer as in the monomer. The extended R state is dynamically disordered (order parameter, S < 0.1) and is a minor component (<10%), whereas the bent T state (pinwheel structure) is the predominant population in both monomeric and pentameric PLN. Given the highly structured and tight ensemble of PDB ID code 1ZLL (bellflower structure), it is implausible that the dynamically disordered R state of pentameric wt-PLN detected in our experiments resembles the bellflower structure. One possibility, strongly indicated by the two-component EPR spectra of Fig. 7B, is that the two states (T and R) represent the conformational equilibrium between a more stable, predominantly α-helical ensemble and a more dynamic, less populated unfolded state (15). Interestingly, phosphorylation at Ser-16 by protein kinase A and single-site mutations at position 21 show the R state to become more populated, indicating the importance of preexisting equilibria necessary for SERCA recognition (12, 14, 34). Similar conformational interconversions also have been detected for fd coat protein, in which lipids may induce different topologies consistent with either the arrangement in the virus coat or in the interaction with the lipid membrane (45–47).

In addition to our study, other spectroscopic data suggest association of the cytoplasmic domains of the pentamer with the lipid membrane. In particular, membrane association has been detected by using ^{H and P solid-state NMR with cytoplasmic domain peptides and full-length versions of wt-PLN by the Middleton (}19, 20) and Lorigan (21, 22) groups. The interactions with the membranes or membrane-mimicking systems induce a stable α-helical conformation. This process is reminiscent of the membrane association mechanisms proposed for small hormone or antimicrobial peptides (48) and of the prefolding states of membrane proteins identified by Engelman and coworkers (49, 50).

Our results strongly support a pinwheel geometry as the predominant arrangement for the pentamer. In addition to this major conformation, we propose that the entire ensemble of conformations sampled by PLN and mediated by lipids is necessary for recognition by the main interacting partners: SERCA, protein kinase A, calmodulin-dependent protein kinase II, protein phosphatase 1, and the more recently found HS-1 associated protein X-1.

Materials and Methods

Protein Preparation.

[U-^{N] wt-PLN (rabbit) was grown and purified in} Escherichia coli bacteria as previously described (23). Selectively labeled wt-PLN was grown in a similar manner with addition of the ^{N-labeled amino acid of interest (0.1 g/liter) and the remaining 19 unlabeled amino acids (0.3 g/liter).}

For EPR experiments, wt-PLN or AFA-PLN samples were prepared by solid-phase peptide synthesis as described previously for AFA-PLN (15, 51). For dynamics measurements, Ala-11 was replaced by the TOAC spin label, which reports directly the dynamics of the peptide backbone (15). For accessibility and distance measurements, wt-PLN was spin-labeled with TEMPO-succinimide at Lys-3. Neither spin label had any effect on PLN function, as determined by inhibition of SERCA (ref. 15 and K. Ha, N.J.T., R.V., J. Zamoon, C.B.K., D.D.T., and G.V., unpublished work) or on pentameric stability, as determined from SDS gels (52). Inhibitory function was identical for expressed and synthetic PLN.

NMR Spectroscopy.

Solution NMR samples were prepared by dissolving PLN in 300 mM DPC, 6 M guanidinium hydrochloride, 50 mM 2-mercaptoethanol, 100 mM NaCl, and 25 mM phosphate buffer (pH 6.0). Samples were thoroughly dialyzed to remove all denaturant. PLN samples were ≈0.4 mM in pentamer and were stable over several weeks at 30°C for NMR measurements.

The [^{H/N]-TROSY-HSQC assignment of wt-PLN is based on the assignment previously published (}2) with selectively ^{N-labeled wt-PLN samples to resolve overlapping resonances. TROSY-HSQC experiments were used for the paramagnetic quenching experiments. Gd (Magnevist; Bayer Schering Pharma, Berlin, Germany) and 5′- and 16′-doxyl stearic acids (Sigma–Aldrich) were titrated into wt-PLN (≈0.4 mM in pentamer) to final concentrations of 3, 2, and 1 mM, respectively.}

Oriented lipid bilayer preparations on glass plate supports have been described previously for a 4/1 molar mixture of DOPC/DOPE (18, 53). Typical preparations used 4–8 mg of [^{N-Leu], [N-Ala], [N-Ile], [N-Cys], [N-Thr], and [N-Asn] wt-PLN. The final molar ratio of lipid/protein for all samples was ≈125/1. All solid-state NMR experiments were acquired at 4°C.}

PISEMA spectra were acquired at a 14.1-T field strength (^{H frequency of 600.1 MHz) equipped with a Bruker DMX spectrometer (National High Magnetic Field Laboratory, Tallahassee, FL). The 2D PISEMA experiments (}27) were performed at an RF field strength of ≈60 kHz for cross-polarization, SEMA, and decoupling, by using a low-E probe (54). Additional details are provided in the SI Methods.

EPR Spectroscopy.

Spin-labeled PLN in lipid bilayer vesicles, containing 4/1 DOPC/DOPE (200/1 lipid/PLN monomer), were prepared as previously described (15). Spin-labeled PLN in DPC micelles were prepared by dissolving the protein in DPC as described for solution NMR experiments (no reducing or denaturing agent) to a final PLN concentration of 0.3 mM.

EPR spectra were acquired with a Bruker EleXsys 500 spectrometer with the SHQ cavity. Samples (20 μl) were maintained at 4°C. The field modulation frequency was 100 kHz, with a peak-to-peak amplitude of 1 G. Accessibility of the spin label to the membrane surface was determined from progressive power saturation measurements, by using Ni ^{ions chelated to the lipid head group as previously described (}15, 17).

Spin-spin distances were measured by DEER with a Bruker E680-pulsed EPR spectrometer (Billerica, MA) at X-band (9.5 GHz) (55). One hundred-microliter samples were contained in 3-mm i.d. quartz tubes and flash-frozen in liquid nitrogen. The static field was set to the low-field resonance of the nitroxide signal. A four-pulse sequence was used with a 16-ns π/2 pulse (36, 55). Temperature was controlled at 65° K during spectral acquisition, which lasted 8–12 h. Data manipulations and other details are provided in the SI Methods.

Supplementary Material