Active-Site Structure of the Thermophilic F_oc-Subunit Ring in Membranes Elucidated by Solid-State NMR

Abstract

F_oF₁-ATP synthase uses the electrochemical potential across membranes or ATP hydrolysis to rotate the F_oc-subunit ring. To elucidate the underlying mechanism, we carried out a structural analysis focused on the active site of the thermophilic c-subunit (TF_oc) ring in membranes with a solid-state NMR method developed for this purpose. We used stereo-array isotope labeling (SAIL) with a cell-free system to highlight the target. TF_oc oligomers were purified using a virtual ring His tag. The membrane-reconstituted TF_oc oligomer was confirmed to be a ring indistinguishable from that expressed in E. coli on the basis of the H⁺-translocation activity and high-speed atomic force microscopic images. For the analysis of the active site, 2D ¹³C-¹³C correlation spectra of TF_oc rings labeled with SAIL-Glu and -Asn were recorded. Complete signal assignment could be performed with the aid of the C^α_i+1-C^α_i correlation spectrum of specifically ¹³C,¹⁵N-labeled TF_oc rings. The C^δ chemical shift of Glu-56, which is essential for H⁺ translocation, and related crosspeaks revealed that its carboxyl group is protonated in the membrane, forming the H⁺-locked conformation with Asn-23. The chemical shift of Asp-61 C^γ of the E. coli c ring indicated an involvement of a water molecule in the H⁺ locking, in contrast to the involvement of Asn-23 in the TF_oc ring, suggesting two different means of proton storage in the c rings.

Introduction

H⁺-driven F_oF₁-ATP synthase plays a major role in energy production in most organisms. It consists of a water-soluble F₁ part and a membrane-integrated F_o part (Fig. 1 A). It converts the electrochemical potential generated by the H⁺ gradient across membranes into the rotation of the c-subunit ring (red in the figure) in F_o and then into that of the γ subunit in F₁, or vice versa (1). The c ring comprises 8–15 c subunits depending on the biological species (2–7). The mechanism underlying the energy conversion at F_o is one of the major unresolved issues regarding F_oF₁-ATP synthase. The proton translocation across F_o is the key process in the energy conversion and consists of five steps, namely, H⁺ transfer through the first H⁺ channel, H⁺ transfer between the a and c subunits, the rotation of the c ring, H⁺ transfer between the c and a subunits, and H⁺ transfer through the second H⁺ channel. The conserved acidic amino acid residue in the c subunit is known to be essential for this process. Therefore, the active-site structure of the c ring, including the conserved acidic residue, plays an important role in the second and fourth steps. The crystal structures of the c rings of H⁺-driven F_oF₁-ATP synthases in detergent micelles have been reported for chloroplasts (8), a cyanobacterium (9), Bacillus pseudofirmus (10), and yeast mitochondria (11). Those of F-type (5) and V-type (12,13) Na⁺-driven ATPases from bacteria have also been reported. In the case of H⁺-driven F_oF₁, the structure of the active site, including the conserved acidic amino acid, was found to be different in the abovementioned crystals. This may be due to the different detergent conditions. Therefore, investigation of a membrane-embedded c ring is required to clarify the structure and function of the active site.

Figure 1.

F_oF₁-ATP synthase and characterization of the TF_oc-subunit oligomers. (A) A model of bacterial F_oF₁-ATP synthase. (B) Blue native PAGE images of the molecular-weight markers and TF_oc oligomers after anion-exchange column chromatography. The sample solution included 0.04% Coomassie G-250, 0.15% DM, and 0.02% DDM. (C) ATP-driven H⁺-translocating activity of TF_oF₁ reconstituted from the isolated TF_oc oligomer, TF_oab₂, and TF₁. H⁺-translocating activity of proteoliposomes of TF_oF₁ was analyzed at 42°C by monitoring ACMA fluorescence at 480 nm and excitation at 410 nm. (a) c-oligomer (cell-free) + ab₂, (b) c oligomer (cell-free) + ab₂ + F₁, and (c) c ring (TF_oF₁) + ab₂ + F₁. Here, (cell-free) and (TF_oF₁) indicate the sources of the c subunits. % ACMA represents the level of quenching of ACMA fluorescence.

We previously reported a solution structure of the c subunit of thermophilic Bacillus PS3 H⁺-driven F_oF₁-ATP synthase (TF_oF₁) (14). However, the structure of its c-subunit ring (TF_o c ring) has not yet been determined. The number of subunits in the ring was determined to be 10 using a combination of gene engineering and a biochemical method (4). The amino acid sequence of TF_o c in one-letter codes is as follows:

• MHLGV LAAAI¹⁰ AVGLG ALGAG²⁰ IGNGL IVSRT³⁰ IEGIA RQPEL⁴⁰ RPVLQ TTMFI⁵⁰

• GVALV EALPI⁶⁰ IGVVF SFIYL⁷⁰ GR

Here, E56 is the conserved glutamate. The solution structure comprised two α helices folded in a hairpin style, as found in every crystal c-ring structure. The conserved Glu is thought to accept/release a proton from/to a proton channel upon interaction with the a subunit. The purpose of this work is to elucidate the structure of the active site of the TF_o c ring in lipid membranes using solid-state NMR (ssNMR) in combination with the stereo-array isotope specific labeling. The TF_oc ring is suitable for such extensive work because of its stability.

The structural analysis of membrane proteins by ssNMR has been extensively carried out (15–18). However, it is still challenging. We performed ssNMR analysis of ¹³C, ¹⁵N uniformly and specifically labeled F_oc subunit rings from Escherichia coli (19,20). Although site-specific isotope labeling by chemical synthesis provided well-defined information, chemical synthesis of multiple samples is a risky and tedious task. In contrast, the fully labeled sample suffered from severe signal overlapping. Therefore, we developed a more efficient, specific labeling method in this work. Cell-free expression can provide a variety of methods for isotope labeling and reconstitution of F_oF₁. We used stereo-array isotope-labeled amino acids (SAIL-AA) (21) and a wheat-germ extract (WGE) system for specific labeling of TF_oc rings. The most critical point in cell-free production is the formation of active rings from the synthesized c-subunit monomers. We examined this by referring to the c rings prepared from TF_oF₁-ATP synthase. Since the active-site structures in crystals in micelles were different from one another, we focused our attention on this difference. We successfully prepared specifically isotope-labeled TF_oc rings and obtained structural information about the active site in lipid membranes. Our results provide insights into the chemical mechanism underlying the proton locking in F_oc rings and H⁺ transfer at the interface of the c and a subunits.

Materials and Methods

The experimental details are presented in the Supporting Material.

Synthesis of TF_oc in a WGE system

A mutated protein, S2H-TF_oc, was synthesized from messenger RNA (mRNA) in a WGE HPOWG translation mixture at 26°C for ∼24 hr with a supply of amino acids from the dialysis buffer according to a previously described method (22). Because only S2H-TF_oc was used in this work, the term “TF_oc” will be employed instead of S2H-TF_oc hereafter. For isotope labeling of Ala, Gly, Val, Glu, and Asn, ¹³C,¹⁵N-labeled ones were used in the amino acid mixture. SAIL-Asn and -Glu were used at 0.12 mM with the other amino acids deuterated. After the reaction, the collected product was solubilized and applied to Ni-NTA chromatography (23). The column was eluted with an imidazole gradient (20–200 mM) in 100 mM KCl, 0.5% sodium deoxycholate (DOC), 0.15% decyl maltoside (DM), and 20 mM HEPES-KOH buffer (pH 7.5), and the TF_oc fraction was collected. This fraction was used for reconstitution into membranes in some cases. Otherwise, this fraction was further applied to an anion exchange HiTrap Q (HQ) column. The HQ column was eluted with a NaCl gradient (0–50 mM) in 0.2 mM EDTA, 10 mM Tris-HCl buffer (pH 7.5). The purity of TF_oc was confirmed by Tricine-SDS-PAGE and western blotting.

Reconstitution of TF_oc oligomers into lipid bilayers

We solubilized 1,2-diperdeuteriomyristoyl-sn-glycero-3-phosphocholine (DMPC-d₅₄, 98% ²H, 95% chemical purity) in a 0.5% DM solution. This solution was slowly titrated into the protein solution (protein concentration: 0.5 mg/mL; lipid/protein (molar ratio) = 24) under mixing at room temperature. The detergent was removed with Bio-Beads SM-2 (20–50 mesh) followed by dialysis against 100 mM KCl, 0.4 mM NaN₃, 20 mM HEPES-KOH/Tris-HCl buffer (pH 7.5), and then against 0.4 mM NaN₃, 10 mM HEPES-KOH/Tris-HCl buffer (pH 7.5). The collected precipitate was subjected to freeze-and-thaw treatment 10 times. In the case of SAIL-AA-labeled samples, the buffer was replaced with 0.4 mM NaN₃, 10 mM deuterated Tris-²HCl buffer (in ²H₂O, p²H 7.5).

Activity analysis and high-speed atomic force microscopy imaging

Active TF_oF₁ was reconstituted into liposomes. Then, the proton translocation activity was measured according to a previously described method (23,24). A laboratory-built, high-speed (HS) atomic force microscope was used (25). Atomic force microscopy (AFM) images were obtained in the tapping mode. A droplet of TF_oc/DMPC-d₅₄ liposome solution was deposited on a freshly cleaved mica surface. After incubation for 5 min, the surface was washed with a solution of 10 mM MgCl₂, 50 mM KCl, 20 mM Tris-HCl (pH 7.5). The liposomes retained on the surface were subjected to HS-AFM in the same buffer.

ssNMR measurements

NMR measurements were performed with Varian Infinity-plus 600 and 700 spectrometers operating at 14.09 and 16.44 T, respectively. Broadband double- and triple-resonance magic-angle-spinning (MAS) probes for 3.2 and 2.5 mmϕ rotors were used. The rotor size and MAS rate used for measurements of two-dimensional (2D) spectra were 3.2 mm and 12.5 kHz unless otherwise specified. The amount of protein used was ∼3 mg/3.2 mmϕ rotor. The probe temperature was set at 233 K. The sample temperature was approximately −25°C, because the water signal disappeared in the region of the set temperature from −15 to −20°C due to freezing. 2D dipole-assisted rotational resonance (DARR) (26) and 2D C^α_i+1-(C′C^α)_i correlation (27) spectra under MAS were recorded with a 3 s repetition time. The ¹³C chemical shift (CS) was referenced to 2,2-dimethylsilapentane-1-sulfonic acid (DSS) by using the methine carbon signal of adamantane at 40.5 ppm relative to DSS.

Results

Cell-free production of TF_oc with the WGE system and isolation of TF_oc oligomers

In the absence of a detergent and liposomes, a 1.5–2 mg TF_oc/mL reaction mixture was obtained as precipitate. However, the precipitate could not be efficiently solubilized by 2% DOC. To improve the recovery, we examined the effect of an additive such as detergent micelles (Tween 20, NP-40, Triton-X, sucrose monolaurate, or dodecyl maltoside (DDM)) or soybean phosphatidylcholine (sPC) liposomes on expression. The best recovery was achieved in the presence of sPC liposomes. Thus, we decided to synthesize TF_oc with WGE in the presence of 0.44% sPC liposomes. We also used 2% DOC to promote ring formation in the solubilization step because it has been used to isolate TF_oc rings from E. coli membranes (23). To isolate the TF_oc rings, we used a virtual His tag. Although TF_oc monomer has only one His at the second position, a decamer ring, for example, should have a noncovalent ring His₁₀ tag. The tricine-SDS-PAGE images in Fig. S1 A indicate the feasibility of this His tag in Ni-NTA chromatography. The obtained TF_oc oligomer amounted to a 0.8–1.6 mg/ml reaction mixture. Then, this fraction was applied to an HQ column. The Tricine-SDS-PAGE image at each purification step is presented in Fig. S1 A. Blue native PAGE revealed that the obtained oligomers are homogeneous (Fig. 1 B). To determine whether the isolated oligomers are functional TF_oc decamers, we measured the H⁺-translocation activity of F_oF₁ reconstituted from the isolated TF_oc oligomers, the TF_oab₂ complexes, and TF₁ in sPC liposomes (23). The results are presented in Fig. 1 C. An active F_oF₁ will translocate H⁺ from the outside to the inside of the liposome at the expense of ATP added to the outside. The acidification of the inside quenches the fluorescence of 9-amino-6-chloro-2-methoxyacridine (ACMA). The quench rate of the reconstituted TF_oF₁ including cell-free c oligomers was 69.5%, whereas that of reconstituted TF_oF₁ including the intact TF_oc ring isolated from TF_oF₁ expressed in E. coli membranes was 71%, i.e., the activity of the former was almost the same as that of the latter. The analysis of the activity of mutant TF_oF₁ incorporating genetically fused TF_oc oligomer (from 2-mer to 14-mer) revealed that only TF_oF₁ incorporating the decamer ring was active (4). Therefore, we can conclude that the TF_oc oligomers obtained from the cell-free system are decamer rings as in the case of the intact ones.

Reconstitution of TF_oc oligomers into lipid membranes and HS-AFM imaging

Purified TF_oc was reconstituted into DMPC-d₅₄ membranes. The detergent was removed by a combination of Bio-Beads treatment and dialysis as described above. The obtained membrane preparation was characterized by sucrose-density gradient (10–44%) ultracentrifugation as in the case of the EF_oc ring (19). The membrane fraction gave a sharp single band that was different from the band for pure DMPC liposomes. To elucidate the macroscopic architecture of the TF_oc oligomers in DMPC-d₅₄ membranes, we obtained successive HS-AFM images of the preparation on the mica surface in the presence of buffer at room temperature. As a reference, we observed images of the intact TF_oc rings in DMPC-d₅₄ membranes. They were isolated from TF_oF₁ expressed in E. coli membranes and well characterized in our previous study (23). The images are presented in Fig. 2 A. A torus-shaped substance with a pore at the center could be clearly observed. Since the DMPC-d₅₄ membranes are attached to the mica surface, the TF_oc rings should be oriented perpendicular to it. However, it was difficult to discriminate the two different bottom ends (the loop and termini sides) of the TF_oc cylinder. The ring diameter was measured using the top position of well-defined rings (see Supporting Material for Materials and Methods). The diameter of the TF_oc rings expressed in E. coli was 3.8 ± 0.2 nm (average of 35 images). In Fig. 2 B, we can directly confirm the ring structure of the oligomeric TF_oc prepared with the WGE cell-free system. Its diameter was 3.9 ± 0.2 nm (average of 40). Therefore, the oligomeric TF_oc is indistinguishable from the intact TF_oc decamer ring in terms of the macroscopic structure under physiological conditions.

Figure 2.

HS-AFM images of the TF_oc-subunit oligomers in DMPC-d₅₄ membranes. The images were obtained in 10 mM MgCl₂, 50 mM KCl, and 20 mM Tris-HCl buffer (pH 7.5) at room temperature. (A) TF_oc rings isolated from TF_oF₁ expressed in E. coli membranes. The imaging rates were 1.72 and 0.95 frames per second (f/s), and the scan ranges were 100 × 100 and 50 × 50 nm on the left and right, respectively. (B) TF_oc oligomers isolated from the WGE reaction mixture. The rates were 1.03, 1.63, 1.90, and 1.42 f/s on the left, top right, and left and right of the bottom right, respectively. The scan ranges were 200 × 200 and 50 × 50 nm on the left and right, respectively. (C) Successive images of a TF_oc oligomer from the cell-free system disrupted by a cantilever (in the blue circle). The time intervals were 0.6, 0.6, 10.8, and 6.4 s, from a to e, respectively. The imaging rate was 2.95 f/s, except for a (1.63 f/s). The scan range was 20 × 20 nm. The length of the curvature of interest in each shot and its difference are presented.

The advantage of HS-AFM is that it allows one to not only obtain images of a c ring under physiological conditions but also follow a change in the same ring as a function of time. As can be seen in Fig. 2 C, a nonring structure was generated from time to time during scanning. Its curvature changed as a function of the scanning time. This could be ascribed to the degradation caused by the cantilever. The change in the curvature length presented in Fig. 2 C can be summarized as 1.2 × N nm (N, integer), suggesting that regular units are removed from the ring by the cantilever. Although there is no direct evidence, the c subunit is a possible candidate for the regular unit. If this is the case, the circumference of the ring should be described as 1.2 × N^ring. Here, N^ring is the subunit number in the ring. Since the observed circumference is 12.2 nm, 12.2 ≅ 1.2 × N^ring. This leads to the most possible N^ring = 10. The most possible number is consistent with the TF_oc decamer ring, strongly suggesting that the removed regular unit is a c subunit.

ssNMR measurements

Taking advantage of the WGE cell-free system, we prepared specifically labeled TF_oc rings reconstituted in DMPC-d₅₄ membranes to analyze its active site. We focused on Glu-56 (E56) and Asn-23 (N23) because we had previously suggested that the side chain of E56 in the C-terminal helix would interact with that of N23 in the N-terminal helix in the absence of the a subunit in the membrane (14). Because of the low yield and low impurity, the preparation obtained from the Ni-NTA chromatography was used for NMR analysis unless otherwise specified. To obtain information on the TF_oc structure in general, and site-specific assignment of E56 and N23, Ala, Gly, Val, Glu, and Asn in TF_oc were specifically ¹³C- and ¹⁵N-labeled ([¹³C,¹⁵N]AGVEN-TF_oc). The other amino acid residues were nonlabeled ones. Because Ala (nine residues), Gly (11 residues), and Val (eight residues) are abundant, they can provide information about the general structure. In contrast, there are only one Asn and three Glus that will provide specific information about the active site. A 2D DARR spectrum of [¹³C,¹⁵N]AGVEN-TF_oc under MAS with a mixing time of 15 ms is presented in Fig. 3 A. The spectrum in the aliphatic region is relatively simple and can be readily assigned in terms of amino acids. Because N23 is unique, the assignment is clear on the basis of its spin connectivity (solid lines). The crosspeaks of Ala and Asn can be used to check the intactness of the TF_oc-ring structure obtained from the cell-free system, because they were also isolated in the DARR spectrum of the uniformly ¹³C-labeled TF_oc ring obtained from TF_oF₁ (23). The CSs of N23 C^α, C^β, and C^γ were identical for these two spectra. The overlapped crosspeaks of Ala C^β/C^α of the uniformly labeled TF_oc ring obtained from TF_oF₁ are presented in the inset of the figure. The peculiar pattern is identical to that of this spectrum. Although the crosspeaks of Val C^α/C^β of the uniformly labeled TF_oc ring partially overlap those of Pro, they are also similar to those in this spectra. Since most of Ala, Val, and Asn are distributed in the transmembrane region, it is strongly suggested that the ring structure composed of the TF_oc monomers produced with the cell-free system is indistinguishable from that of the intact TF_oc ring also at atomic resolution.

Figure 3.

2D ¹³C-¹³C correlation DARR (A) and 2D C^α_i+1-C^α_i correlation (B) spectra of [¹³C, ¹⁵N] AGVEN TF_oc-rings in DMPC-d₅₄ membranes at 14.09 T. (A) Mixing time = 15 ms. The amino-acid-specific assignment is presented with a one-letter code. The data size was 1024(d1) × 408(d2) for 60 × 60 kHz spectral widths, and n = 232 scans. The inset is the Ala C^β/C^α region of the DARR spectrum in Yumen et al. (23). (B) Sequential walking (solid and broken lines) and unambiguous assignment of the crosspeaks are presented; assignment of overlapping peaks is in parentheses. The data size was 1024(d1) × 100(d2) for 60 × 50 kHz spectral widths, and n = 840 scans. A squared sine bell with −80° shift and an exponential with 50 Hz broadening factor and linear prediction were used as window functions in Fourier transformation for the d1 and d2 axes, respectively, in A, and the same window functions with 150 Hz broadening factor were used in B.

To obtain sequential information, we recorded a 2D C^α_i+1-(C′C^α)_i correlation spectrum under MAS of the same sample (Fig. 3 B). Here, the magnetization was transferred from ¹³C^α_i+1 to ¹³C^α through ¹⁵N_i+1 and ¹³C = O_i. Now, well-resolved crosspeaks could be obtained because of the uniqueness of successive pairs (for example, A11V12) in the sequence. Sequential walking could be performed, for example, from A11V12 to V12G13 through broken lines. The horizontal line crossed the diagonal line at the CS of V12 at the Ci axis. A vertical line starting from this point found the crosspeak of V12G13. There were two crosspeaks, which might be ascribed to V55E56. However, we could identify the correct one because of the sequential connectivity of V55E56 and E56A57 (indicated by solid lines). The other crosspeak was assigned to V52A53, which was the only possible pair in the region. Thus, the C^α CSs of A11, V12, G13, E32, G33, G51, V52, A53, V55, E56, A57, V63, and V64 were obtained unequivocally. From this result, the crosspeak of Glu observed in both ω1/ω2 and ω2/ω1 (aliphatic areas in the DARR spectrum) turned out to be due to E56 C^α/C^β. The obtained CSs are summarized in Table S1. These CSs and the CS confinement based on the C^α/C^β crosspeaks in the DARR spectrum suggest that the backbone structure represented by these residues is composed of α-helical conformations as in the cases of the intact TF_oc ring (23) and the solution structure (28).

Then, we focused our attention on E56 and N23. To obtain simple and well-resolved spectra with analyzable intensities, we labeled TF_oc with SAIL-Glu and -Asn using the cell-free system (SAIL-EN TF_oc). The stereochemical labeling sites are available in Kohno and Endo (22). We used deuterated amino acid residues, except for Glu and Asn. DARR spectra of the SAIL-EN TF_oc rings in DMPC-d₅₄ membranes, with mixing times of 100 and 50 ms, are presented in Fig. 4, A and B, respectively. Those with mixing times of 15, 200, and 400 ms are presented in Fig. S2, A–C, respectively. In Fig. 4 A, the C^α/C^β crosspeaks of N23 and E56 can be identified on the basis of the assignment made for the [¹³C,¹⁵N]AGVEN-TF_oc-ring signals. The resolution in the carbonyl carbon region was better than that in the aliphatic region. Typical spin connectivity for C^α, C^β, C^γ, C^δ, and C′ of E56 is represented by solid lines. The CSs of E56 were determined from those crosspeaks. For the assignment of the other two Glu residues, an expansion of the upper half of the carbonyl region is presented in the inset of Fig. 4 A. The two Glus are designated as E_a and E_b for convenience. They are indicated by solid and broken arrows, respectively. Because the two C^β/C′ crosspeaks indicated by the two arrows in the inset are separated from each other, the assignment (except for E_b C^α) could be performed using the connectivity with them (see Supporting Material for Results). The E_b C^α CS was determined from the isolated C^α/C′ crosspeak in Fig. 4 B. The CS differences of (E_a-E56) and (E_b-E56) for C^α were 1.1 and 0.3 ppm, respectively. Because the CS difference between E32 and E56 in Fig. 3 B was larger than 1.0 ppm, E_a and E_b should be assigned to E32 and E39, respectively. The assignment was also confirmed with the connectivities in the spectra at 15, 200, and 400 ms mixing times (Fig. S2). The assigned CSs are summarized in Table S2. CSs obtained from SAIL signals were corrected using a calibration table kindly provided by Prof. Masatsune Kainosho.

Figure 4.

(A and B) 2D ¹³C-¹³C correlation DARR spectra of SAIL-EN TF_oc rings in DMPC-d₅₄ membranes at 16.44 T with mixing times of 100 (A) and 50 (B) ms. Connectivities for N23 and E56 are presented as broken and solid lines, respectively, and those for E32 and E39 are indicated by solid and broken arrows, respectively. (A) The top half of the carbonyl carbon region is expanded in the inset. The data size was 1024(d1) × 210(d2) for 70 × 70 kHz spectral widths, and n = 80 scans. (B) A 2.5 mmϕ rotor (1.5 mg protein) was used at a 15 kHz spinning rate. The data size was 2048(d1) × 210(d2) for 100 × 100 kHz spectral widths, and n = 280 scans. The window functions were a squared sine bell with −80° shift (d1) and an exponential with 50 Hz broadening factor and linear prediction (d2).

The CS of E56 C^δ is significantly different from those of E32 and E39. It is 174.5 ppm for E56, but 183.1 and 182.6 ppm for E32 and E39, respectively. According to a report by Gu et al. (29), isotropic CSs of E32 and E39 C^δ are in the range for deprotonated carboxyl groups. In contrast, that of E56 C^δ is basically in the range for protonated carboxyl groups. However, the isotropic CS of the deprotonated carboxyl group was reported to go down to 172 ppm in crystals (29). To make the protonation state of E56 convincing, we performed CS tensor analysis of E56 C^δ. There is a strong correlation between the principal CS tensor elements (δ₁₁, δ₂₂, and δ₃₃) and chemical structures in solid state (29,30), namely, δ₁₁ is larger than 250 ppm for the protonated carboxyl groups. Also, δ₂₂ provides information about the strength of a hydrogen bond involving the carboxyl group. To determine the principal CS tensor elements of E56 C^δ, we measured 1D spectra of the SAIL-EN TF_oc ring (purified with NTA-Ni and HQ columns) in membranes at spinning rates of 4.0, 4.3, 4.5, and 5.0 kHz at 16.44 T. The spectra at spinning rates of 5.0 and 4.5 kHz are presented in Fig. 5 A. In spite of the low signal/noise ratio, the intensity change could be followed, as indicated by arrows in the figure. The side-band intensity data for the four spinning rates were analyzed with Herzfeld-Berger analysis software (HBA 1.6.12) (31,32). Details of the analysis are described in the Supporting Material for Results. The obtained principal elements of the CS tensor were (264 ± 11, 139 ± 7, 121 ± 10) ppm. Because δ₁₁ is larger than 250 ppm, we can safely conclude that this carboxyl group is protonated. Furthermore, the value of δ₂₂ suggests that the E56 carboxyl group is involved in weak hydrogen bonding (29).

Figure 5.

Carbonyl regions of 1D and 2D ¹³C-NMR spectra of SAIL-EN TF_oc rings in DMPC-d₅₄ membranes at 16.44 T. (A) 1D spectra at spinning rates of 4.5 (bottom) and 5.0 (top) kHz. The E56 C^δ signals are indicated by arrows. A 2.5 mmϕ rotor was used. The data size was 4096 for 70 kHz spectral width, and n = 2000 scans. The exponential window function with a 100 Hz broadening factor was used. (B) 2D proton-driven, spin-diffusion ¹³C-¹³C correlation spectrum. The relevant crosspeaks are connected by solid lines. The mixing time was 700 ms. Other parameters are given in Fig. S2D.

To obtain structural information around E56, we carried out a ¹H-driven spin-diffusion experiment on the SAIL-EN TF_oc rings with a mixing time of 700 ms. The relevant parts of the spectrum are presented in Fig. 5 B and Fig. S2 D, in which the crosspeaks of N23 C^δ with E56 C^β, C^γ, and C^δ can be seen, revealing that they are close to each other.

Discussion

Strategy for analyzing the active site of the TF_oc ring in membranes

To analyze the active site of the TF_oc ring in membranes, we used specific isotope labeling and CP/MAS NMR. Because this is a membrane protein, we had to develop a method for expression with the WGE cell-free system, formation and purification of the rings, and their reconstitution into membranes. To obtain well-defined structural information about the active site by ssNMR, we used SAIL-Glu and -Asn. This work revealed that the combination of expression with the WGE cell-free system in the presence of liposomes and purification in the presence of 2% DOC could provide functionally active TF_oc-decamer rings. Cotranslational insertion of membrane proteins into liposomes in the WGE cell-free system was also previously reported (33). Although TF_ocs were artificially produced in the cell-free system, they assembled into decamer rings as in the case of TF_oF₁ formation in vivo. This was verified by the proton-translocating activity, and warranted by the macroscopic and microscopic structures based on HS-AFM images and NMR spectra, respectively. This fact strongly suggests that the number of c subunits that form a ring in lipid membranes is determined by their primary sequence.

The SAIL-EN TF_oc rings in the DMPC membranes provided high-resolution ssNMR spectra and structural information in combination with the [¹³C,¹⁵N]AGVEN-TF_oc rings. The sensitivity was better for the SAIL-EN TF_oc ring than for the AGVEN-labeled one in spite of similar intrinsic line widths (see Supporting Material for Discussion). The better sensitivity likely is due to the higher efficiency in CP and DARR for the SAIL-EN TF_oc rings. The longer T_1ρ of protons and carbons would improve CP efficiency, and the longer T₂ and T₁ of carbons may suppress the decay in the evolution and mixing periods of DARR, respectively (26,34,35). The efficient DARR in the presence of only a limited number of protons revealed that the combination of a single ¹³C-¹H group for heterogeneous broadening and a single ¹H in the direct neighbor for homogeneous broadening is the core spin system for generating an effective ¹³C polarization transfer through DARR. The method developed in this work also can be applied to focused analysis of other membrane proteins in lipid membranes.

Structure and function of the active sites of F_oc rings in membranes

The active-site structure of the c subunit involving the conserved acidic amino acid residue is a key factor in elucidating the proton-transfer mechanism at the interface between the a and c subunits in F_o. This is also important for understanding the mechanism of the proton translocation across F_o. Our results provide structural information about the active site of the TF_oc ring embedded in membranes. The CS of E56 C^δ showed that its carboxyl group was protonated in membranes. The presence of the crosspeaks between N23 C^γ and E56 C^δ, C^γ, and C^β in the spin diffusion spectrum revealed that the side chain of E56 points to the inside of the ring and interacts with N23, forming a proton-locked conformation. Furthermore, δ₂₂ of E56 C^δ suggested that the carboxyl group is involved in weak hydrogen bonding. This is also supported by the unique isotropic CS of N23 C^γ (170.6 ppm), which is smaller than the average value found in the BMRB database (176.5 ppm). Actually, the isotropic CS of an amide carbon was reported to be proportional to the length of a hydrogen bond in peptide bonds because of a change in δ₂₂ (36), i.e., the smaller the CS, the longer the hydrogen bond.

Our results can be compared with the active-site structures of the F_oc rings reported thus far. Actually, they were all different from one another. The cyanobacterium c₁₅ assumed a proton-locked conformation with the conserved carboxyl group on the C-terminal (outer) helix, forming hydrogen bonds with a glutamine side chain on the N-terminal (inner) helix and others (9). Here, the protonated carboxyl group is locked by these hydrogen bonds. The chloroplast c₁₄ assumed a modified proton-locked conformation, missing the hydrogen bond with the glutamine in spite of its presence in the N-terminal helix (8), and Bacillus pseudofirmus c₁₃ assumed a water-involved ion-locked conformation, with the carboxyl group hydrogen bonding only with a water molecule (10). In this case, there is no polar amino acid residue in the active-site region of the N-terminal helix. The presence of the water molecule is similar to the Na⁺-coordination structure in F_oc₁₁ from Ilyobactor tartaricus (5). Since the acidic side chains pointed into the rings in all three cases, they were specified as closed conformations in general. In contrast to them, the conserved carboxyl group (E59) of the yeast mitochondrial c₁₀ (YF_oc₁₀) was located outside of the ring in both the protonated and deprotonated states (11). They are called open conformations. The reason for the open-form formation was attributed to the property of detergent by molecular-dynamics simulation. Actually, the YF_oc₁₀ E59 exhibited a closed form in a crystal structure of the YF₁c₁₀ complex (37). Our work reveals that the closed form is the correct conformation in lipid membranes. The open form was assumed to be a conformation at the interface between the a and c subunits (11).

Furthermore, classification of the c subunits as an E/D-only type or mixed type on the basis of their primary sequence was proposed to be useful for understanding the active-site structure (10). In the former, the conserved Glu or Asp is the only residue that can be predicted to be involved in ion coordination, whereas the latter has more polar residues. To make the difference clearer, we use the term “E/D-plus type” instead of mixed type in this discussion. It turned out so far that the closed conformation in the E/D-only-type ring (c₁₃) carries a water molecule coordinated to the carboxyl group, whereas those in the E/D-plus-type rings (c₁₄ and c₁₅) do not in the high-resolution crystal structures.

Although the whole structure of TF_oc₁₀ is not yet known, our result clearly reveals that its conserved carboxyl group is protonated and assumes a proton-locked conformation in lipid membranes. Our result is not consistent with the modified proton-locked conformation of c₁₄, because of the N23/E56 interaction. TF_oc₁₀ is the simplest example of the E/D-plus type, since it has only N23 and E56 as the polar residues in the active-site pocket. The protonated state of E56 in TF_oc₁₀ and the weak nature of its hydrogen bonding to N23 strongly suggest the absence of a water molecule in this pocket, as in the cases of c₁₄ and c₁₅ (Fig. 6 A). In contrast, the E. coli F_oc-subunit ring (EF_oc₁₀) is classified to the E/D-only type, the structure of which is also not yet known. It carries no polar residue other than the conserved D61 in the active-site pocket. We performed an ssNMR analysis of EF_oc₁₀ in lipid membranes (20). The CS of D61 C^γ was 179.6 ppm. Because this was determined using a chemically synthesized EF_oc with specific ¹³C labeling at A24 C^β and E56 C^δ, this value is reliable. This CS is significantly different from the 174.5 ppm of E56 C^δ determined in this work. This fact reveals that the chemical nature of the conserved carboxyl group of EF_oc₁₀ is significantly different from that of TF_oc₁₀. The CS of D61 C^γ suggests that its carboxyl group is either protonated and involved in a strong hydrogen bonding, or is deprotonated and involved in hydrogen bonding (29). Since there is no polar residue to form hydrogen bonds or to coordinate a cation in the active-site pocket, the only possible hydrogen-bonding partner should be a water molecule, as in the case of c₁₃ (Fig. 6 B). The involvement of water is also consistent with the CSs of propionic acid and butyric acid diluted in water, namely, 180.02 and 179.22 ppm for COOH, respectively, and 184.95 and 184.10 ppm for COO⁻, respectively (38). The shorter side chain of D61 may provide a space for the water in the pocket. The presence of water in the pocket may also explain the surviving activity of EF_oc₁₀ on the double mutations A24D/D61G and A24D/D61N (39). Although the essential carboxyl group is located on the inner helix in this case, the hydrogen-bonded water molecule would mediate the proton transfer between D24 and the proton channels, responding to the electrostatic interaction with the conserved Arg in the a subunit. Now, we can conclude that the correlation between the water involvement in proton locking and the E/D-only type also holds for EF_oc₁₀. This strongly suggests that there are two means of proton storage in the c rings in lipid membranes, namely, polar-group-involved proton locking and water-involved proton locking, as presented in Fig. 6. The weak hydrogen bond in TF_oc₁₀ and the presence of the open form in the protonated state in YF_oc₁₀ strongly suggest that the barrier between the closed and open forms in the protonated state is low enough to facilitate the closed/open conversion through thermal fluctuations. The major form will be determined by the hydrophilicity of the environment induced by the interaction with the a subunit or lipids. The deprotonation and protonation probably take place in the open form. In the water-involved type, however, there is a possibility that the carboxyl group leaves the proton in the pocket upon conversion from the closed to the open form under the effect of the positive charge in the a subunit, because the water molecule can stabilize it. The deprotonation in the active site generates an ion pair similar to the Na⁺-locking type. If this is the case, they may have similar mechanisms for cation transfer. There may be also an effect of F₁ on the structure of the active site, which we cannot evaluate at this point.

Figure 6.

Two types of proton locking in c₁₀ rings. Sliced views of the active-site pocket from the cytoplasmic side are presented. (A) Polar-group-involved proton locking (TF_oc₁₀). The inset is a side view of the model. (B) Water-involved proton locking (EF_oc₁₀).

In summary, we have elucidated the chemical nature of the conserved acidic amino acid residue of the TF_oc₁₀ and EF_oc₁₀ rotors in lipid membranes using ssNMR in combination with the stereo-array isotope specific labeling. It was revealed that the conserved E56 of the former assumed a proton-locked conformation through an interaction with N23. However, D61 of the latter was involved in the interaction with a water molecule. In the case of TF_oc₁₀, the fluctuation of the E56 side chain between the closed and open forms could easily take place because of weak hydrogen bonding with N23.