Ion binding and selectivity of the rotor ring of the Na⁺-transporting V-ATPase

Abstract

The vacuole-type ATPases (V-ATPases) are proton pumps in various intracellular compartments of eukaryotic cells. Prokaryotic V-ATPase of Enterococcus hirae, closely related to the eukaryotic enzymes, provides a unique opportunity to study ion translocation by V-ATPases because it transports Na⁺ ions, which are easier to detect by x-ray crystallography and radioisotope experiments. The purified rotor ring (K-ring) of the E. hirae V-ATPase binds one Na⁺ ion per K-monomer with high affinity, which is competitively inhibited by Li⁺ or H⁺, suggesting that the K-ring can also bind these ions. This finding is also supported by the K-ring structure at 2.8 Å in the presence of Li⁺. Association and dissociation rates of the Na⁺ to and from the purified K-ring were extremely slow compared with the Na⁺ translocation rate estimated from the enzymatic activity, strongly suggesting that interaction with the stator subunit (I-subunit) is essential for Na⁺ binding to /release from the K-ring.

In eukaryotic cells, vital processes such as protein trafficking, endocytosis, neurotransmitter release, and intracellular pH regulation depend on the movement of H⁺ ions across membranes. The movement of these ions is catalyzed by vacuole-type ATPases (V-ATPases) (1), multisubunit complexes related in structure and mechanism to the F-type ATPases (F-ATPases) found in eubacteria, mitochondria, and chloroplasts (2). Like F-ATPases, V-ATPases have a globular catalytic moiety, V₁ (equivalent to F₁) where ATP is hydrolyzed. This moiety is attached by central and peripheral stalks to an intrinsic membrane moiety, V_o (equivalent to F_o), which pumps ions across the membrane. In both F- and V-ATPases, ATP hydrolysis causes rotation of the central stalk and the attached membrane ring comprised of hydrophobic subunits. It is generally accepted that the H⁺ ion is pumped at the interface between the rotating ring and a static membrane component (a-subunit), although the precise mechanism of ion translocation by V (or F)-ATPase is still not clear (1).

Several organisms living in high salinity or high pH have evolved a variety of primary sodium pumps. In Propionigenum modestum, the sodium ion potential generated by decarboxylase is converted to ATP by Na⁺-transporting F-ATPase (3). The purified Na⁺–ATPase uses the hydrolysis of ATP to transport Na⁺, Li⁺, or H⁺ (in the absence of Na⁺ or Li⁺) (4). We have identified a variant of V-ATPase in a fermentative bacterium Enterococcus hirae, which physiologically transports Na⁺ as well as Li⁺ (5). The enzyme is encoded by nine ntp genes (ntpFIKECGABD) organized in an ntp operon (6). Amino acid sequences of NtpF, NtpI, NtpK, NtpE, NtpC, NtpG, NtpA, NtpB, and NtpD are homologous to those of subunit G, a, c, E, d, F, A, B, and D, respectively of eukaryotic V-ATPases (7). The V₁ moiety responsible for ATP-driven rotation is composed of the NtpA, NtpB, NtpC, NtpD, NtpE, and NtpG subunits (Fig. 1). In V₁, the three A subunits and the three B subunits are arranged alternately around a central D subunit. The V_o moiety that uses the rotation of V₁ for Na⁺ (or Li⁺) transportation is composed of oligomers of the 16-kDa NtpK that form a membrane rotor ring (K-ring) and a single copy of the NtpI subunit. The NtpA₃B₃D complex and V_o moiety are connected by central subunits NtpG and NtpC of V₁ and peripheral stalks, composed of subunits NtpF and NtpE (8). Recently, we solved a crystal structure of the Na⁺-bound K-ring as a high-resolution ring structure and proposed an ion transport mechanism based on the structure (9) (Fig. 1). We have obtained the crystal structure of the Li⁺-bound K-ring and examined the ion binding/release properties of the purified K-ring. The velocities of Na⁺ binding/release from the rotor suggested that the interaction with the I-subunit is essential for the Na⁺ binding to /release from the K-ring. We discuss ion selectivity of the rotor and the ion transport mechanism of E. hirae V-ATPases compared with F-ATPases.

Fig. 1.

A schematic structure model for ion translocation by the V-ATPase of E. hirae. The V-ATPase is composed of a hydrophilic catalytic moiety (V₁) and an integral membrane moiety (V_o) for Na⁺ translocation. ATP hydrolysis in A₃B₃ generates rotation of the central stalk (D, G, and C subunits) and an attached membrane ring (K-ring) as indicated by the red arrow, where Na⁺ ions are pumped across the membrane at the interface between the K-ring and I-subunit. This ion translocation model is based on the Na⁺-bound K-ring structure that supports the two-channel model of F-ATPase (9). The K-ring is shown in space-filling representation with essential Glu-139 (yellow and red for oxygen atoms). Sodium ions are shown as blue spheres. I-subunit (structure unknown) possesses two half-channel and an essential Arg-573, which are required for Na⁺ translocation across the membrane. Clockwise rotation of the ring (red arrow) driven by ATP hydrolysis and electrostatic Arg-573–Glu-139 interaction enable the release of bound Na⁺ into a half-channel connected to the periplasm as indicated by the blue arrow and the binding of Na⁺ from a second half-channel connected to the cytoplasm as shown by the purple arrow.

Results

²²Na⁺ Binding (Release) Properties of the Purified K-Ring.

The K-ring of V-type Na⁺-ATPase from E. hirae was purified as described (10). The purified K-ring exhibited a similar level of ²²Na⁺ binding in detergent micelles as the purified whole complex (V₁V_o-ATPase). The ²²Na⁺ binding to the purified K-ring increased over time, reaching a maximum after 60 min (Fig. 2A). This level of binding was maintained for several days. The observed association rate constant (k_ob) was calculated to be 2 × 10⁻³± 2.3 × 10⁻⁴ s⁻¹ by using a one-phase exponential association equation: no significant difference was found between calculated constants according to models fitted for one and two association constants. The bound ²²Na⁺ ions on the K-ring were exchangeable with excess (10 mM; 1,000-fold) of nonradioactive Na⁺ (Fig. 2B). These findings suggest that the Na⁺ binding site of the K-ring in detergent is accessible to the aqueous environment. The slope of the semilogarithmic plots of the data indicated a dissociation rate constant (k_off) of 1 × 10⁻³± 2 × 10⁻⁴ s⁻¹. The association rate constant (k_on) of the Na⁺ binding calculated from the k_ob and k_off values was 1 × 10² M⁻¹·s⁻¹. The dissociation constant [K_D(Na+)] of 10 μM was obtained from the k_on and k_off.

Fig. 2.

²²Na⁺ binding to or release from the purified K-ring. θ is defined as the number of moles of bound Na⁺ per mole of the K-subunit but not the K-ring. (A) Binding experiments were initiated by the addition of 10 μM ²²Na⁺. (B) Dissociation experiments were initiated at time 0 by the addition of 10 mM NaCl into the mixture containing purified K-ring incubated with 10 μM ²²Na⁺. (Inset) The dissociation rate constant (k_off) was estimated from a semilogarithmic plot. (C) Na⁺ concentration dependence of Na⁺ binding to purified K-ring. (Inset) The Scatchard plot of the specific binding of Na⁺ to the purified K-ring.

Fig. 2C shows that ²²Na⁺ binding depends on the NaCl concentration. The Scatchard plot (Fig. 2C Inset) of this data has an intercept on the abscissa around one, indicating that one K-subunit binds one sodium ion (SD 0.95 ± 0.06). This finding corresponds well to the crystal structure of the Na⁺-bound K-ring that had one Na⁺ ion per K-monomer (9). The slope of the Scatchard plot indicates that the dissociation constant [K_D(Na+)] is 12 μM (SD 12 ± 3 μM), similar to the K_D(Na+) value obtained from the k_on and k_off as described above. These values are also similar to that of the purified whole complex [K_D(Na+) = 15 ± 5 μM] (11), suggesting that the K-ring is responsible for Na⁺ binding by the V-ATPase.

H⁺ Binding to the K-Ring.

Most V-ATPases transport H⁺ ions. In addition it has been shown that Na⁺-translocating F-ATPase from P. modestum can transport H⁺ in the absence of Na⁺ (4). Here, we examined whether the K-ring of E. hirae V-ATPase can also bind H⁺ ions, although we have not observed ATP-driven H⁺ uptake activity into proteoliposomes reconstituted with the whole V-ATPase complex so far (data not shown). Na⁺ binding was shown to decrease at acidic pHs (Fig. 3A). Fig. 3B shows double reciprocal plots of Na⁺ concentration dependence of Na⁺ binding to the K-ring at various pHs (pH 5.2–8.0; 0.01–6 μM H⁺). Maximal number of binding sites (BS_max) values shown at the y axis intercept are constant (≈1) at different H⁺ concentrations, whereas the apparent K_D(Na+) values varied (Fig. 3B), suggesting competitive binding between Na⁺ and H⁺. The apparent K_D(Na+) values were replotted against concentrations of H⁺ (Fig. 3C). The K_i of H⁺ was calculated from the negative intercept on the ordinate as 3.4 μM (pH 5.5). These results suggest that the K-ring binds H⁺ to the same binding pocket as Na⁺.

Fig. 3.

Inhibition by other cations of ²²Na⁺ binding of the K-ring. θ is defined as the number of moles of bound Na⁺ per mole of K subunit. (A) The pH profile of ²²Na⁺ binding. (B and E) Double reciprocal plots of Na⁺ concentration dependence of Na⁺ binding to purified K-ring at various H⁺ concentrations (B) or various Li⁺ concentrations (E). (C and F) The apparent K_D(Na+) values obtained from each of the double reciprocal plots were replotted against concentrations of H⁺ (C) or Li⁺ (F). K_i values were calculated from the negative intercept of the replots. (D) Dissociation experiments were started at time 0 by adding 10 mM LiCl (■) or other cation (KCl, RbCl, CsCl, TlNO₃, CaCl₂, or MgSO₄) into the mixture containing purified K-ring incubated with 10 μM ²²Na⁺.

Li⁺ Binding to the K-Ring.

E. hirae V-ATPase transports Li⁺ as well as Na⁺ (5). The K_m values for Na⁺ and Li⁺ for the ATP hydrolysis by the purified V-ATPase complex have previously been shown to be similar (12). Bound ²²Na⁺ ions on the K-ring were exchangeable with an excess (10 mM) of Li⁺ (Fig. 3D) in a similar manner to Na⁺ (Fig. 2B). When 10 mM K⁺, Rb⁺, Tl⁺, Cs⁺, Ca²⁺, or Mg²⁺ was used in place of Na⁺ or Li⁺, no release of bound ²²Na⁺ ions was observed (Fig. 3D). Similar results were obtained for the Na⁺ concentration dependence of Na⁺ binding at various Li⁺ concentrations (Fig. 3E) as for H⁺ (Fig. 3 B and C). The K_i of Li⁺ was calculated from the negative intercept on the ordinate as 48 μM, indicating a 5-fold lower binding affinity for Li⁺ compared with Na⁺ (Figs. 2C and 3F). These findings suggest that the K-ring competitively binds Li⁺, Na⁺, or H⁺ to the binding pocket, but does not bind K⁺, Rb⁺, Tl⁺, Cs⁺, Mg²⁺, or Ca²⁺.

Crystal Structure of the Li⁺ Bound K-Ring.

The isolated K-ring was crystallized in the presence of 50 mM Li⁺ (in place of Na⁺) by vapor diffusion. The biochemical findings of the exchange reaction (Fig. 3D) and competitive inhibition (Fig. 3 E and F) of ²²Na⁺ binding with Li⁺ strongly suggested that the K-ring should bind Li⁺ under the crystallization conditions. The structural data are summarized in Table 1. The structure was solved by molecular replacement using the Na⁺-bound K-ring structure (9) as a search model. The model has good stereochemistry (rmsd of bond lengths, 0.009 Å; rmsd of bond angles, 1.13°) with an R factor of 21% (R_free, 22%) for all data to 2.8 Å. There are no outliers on the Ramachandran plot. The final model contains 11,541 protein atoms, 322 water molecules, 10 lithium ions, 2 detergent molecules (n-undecyl-β-maltopyranoside), and 20 lipid molecules (10 1,2-dipalmitoyl-phosphatidylglycerol; 10 1,2-dipalmitoyl-glycerol).

Table 1.

Data collection and crystallographic analysis

Wavelength, Å	1.0000
Space group	P212121
Unit cell dimensions, Å	a = 119.6, b = 125.8, c = 210.2
Resolution, Å	107.83–2.80
No. of reflections	73,353
Rmerge*	0.094 (0.706)
Completeness, %	98.1 (94.2)
Redundancy	4.1 (3.6)
I/σ	11.2 (1.9)
Wilson B factor, Å2	56.0
R factor, %	20.8 (28.6)
Free R factor, %	21.8 (32.6)
rmsd bonds, Å	0.009
rmsd angles, °	1.134

Statistics for the highest-resolution bin (2.95–2.80 Å) are shown in parentheses.

*R_merge = Σ_hklΣ_I _l_i(hkl) − <l(hkl)>_/Σ_hklΣ_il_i(hkl).

The overall structure of the Li⁺-bound K-ring, comprised of 10 K-subunits, is almost identical (rmsd for all atoms, 0.09 Å) to that of Na⁺-bound K-ring except for the ion binding pocket of each K-subunit (Fig. 4 A and B). A 7-fold ring symmetry has previously been obtained by averaging single particle images (10). It is unclear why there is a discrepancy between the observed symmetries but it is likely that the true symmetry is 10-fold as suggested by these crystal structures. Fig. 4 C and D shows the ion binding pocket viewed from outside of the K-ring structures bound to Li⁺ and Na⁺, respectively. The 2 F_o − F_c electron density maps were calculated in the absence of the Li⁺ or Na⁺ at the binding pocket. Clear electron density for Li⁺ was not observed at the ion binding pocket of the obtained K-ring structure (Fig. 4C), in contrast to a strong peak of density for Na⁺ observed in the Na⁺-bound K-ring structure obtained previously (9) (Fig. 4D). A Li⁺ ion with two electrons is more difficult to detect than a Na⁺ ion with 10 electrons because of the Fourier truncation error in the 2.8-Å resolution electron density map. From the density at the binding pocket, the biochemical data described above and the crystallization condition, we concluded that we had indeed obtained the Li⁺ bound K-ring structure. Li⁺ placed in the cavity of the binding pocket in the structure fitted well with no negative peak after refinement. Each Li⁺ is surrounded by five oxygen atoms, four of them in the side chains of Thr-64, Gln-65, Gln-110, and Glu-139 and the fifth in the main-chain carbonyl of Leu-61; residues in helices 2, 3, and 4 all contribute to the Li⁺-binding pocket as seen in the Na⁺-bound K-ring structure (Fig. 4 E and F). The distances between Li⁺ and the oxygen atoms are slightly shorter than those between Na⁺ and the oxygen atoms. Thus, the ion binding pocket of the K-ring specifically binds Na⁺ or Li⁺ by using identical binding pockets with slightly different metal–oxygen distances.

Fig. 4.

Structure of Li⁺-bound K-ring. Helix 0 (residues 1–8), helix 1 (residues 11–46), helix 2 (residues 51–79), helix 3 (residues 85–124), helix 4 (residues 127–156), and loops (residues 9–10, 47–50, 80–84, and 125–126) are colored in violet, blue, green, orange, yellow, and red in ribbon representations, respectively. Residues involved in ion binding are shown in stick representation. Oxygen and nitrogen atoms are in red and dark blue, respectively. (A) A side view of the K-ring. The red box indicates the location of the ion-binding pocket. (B) A view from the cytoplasmic side of the membrane. (C and E) The ion-binding pocket viewed from outside of the Li⁺-bound K-ring. The Li⁺ ion is in violet. (D and F) The ion-binding pocket of the Na⁺-bound K-ring as reported (9). The Na⁺ ion is in light blue. The 2 F_o − F_c maps are shown in C and D. Black lines are Li⁺ (or Na⁺)–O bonds with distances (E and F).

Discussion

Ion Binding of the K-Ring.

Bound ²²Na⁺ in the purified K-ring was exchangeable with bulk cold Na⁺ or Li⁺ (Figs. 2B and 3D). How are these ions bound to and released from the binding pocket? Fig. 5 A and C shows solvent-accessible surface representations of the ion binding pocket of the Na⁺- and Li⁺-bound structures, respectively. The ions are exposed only partially to the external milieu and accessibility to the ion from other directions is negligible by the presence of protein entity. In the model, the bound ion (Na⁺ or Li⁺) is occluded by Glu-139, and the position of its γ-carboxylate is stabilized by hydrogen bonds with side chains of Gln-110, Tyr-68, and Thr-64 (closed form). Accessibility to the bound ion allowing its exchange with the bulk phase free ions can be achieved by changing the torsion angles of the side chain of Glu-139, eliminating any steric hindrance (open form; Fig. 5B) as described (9). Release of the bound ion is then easily achieved through the open space. Ion binding to the ring can be brought about by the electrostatic interaction of the hydrated ion with Glu-139, resulting in charge neutralization causing dehydration of the ion. The dehydrated ion can then bind to the pocket. Stabilization of ion binding may require the side chain of Glu-139 to resume its position as in the closed form (Fig. 5 A and C). Thus, the Na⁺ exchange reaction of the purified K-ring observed experimentally can be accomplished by only torsion angle changes of the side chain of Glu-139.

Fig. 5.

A model for ion binding/releasing processes at the ion binding pocket in the K-ring. Residues involved in ion binding are shown in space-filling representation. The side chain of E¹³⁹ is colored in yellow, other residues are in gray. Oxygen and nitrogen atoms are in red and dark blue, respectively. Residues not involved in binding are represented as the solvent-accessible surface. (A–D) The ion binding pocket is viewed from outside. (A) The structure of the Na⁺ (light blue)-bound K-ring. (B) To open the Na⁺-binding pocket, the torsion angles of the side chain of E¹³⁹ have been changed from (−170, 72) to (70, 170), and the Na⁺ ion has been released. (C) The structure of the Li⁺ (purple)-bound K-ring. (D) The predicted structure of the H⁺-bound K-ring. The protonated oxygen of E¹³⁹ is in green. (E) Spheres representing the sizes (ion radii) of the following cations for comparison. The ion sizes are drawn proportionally to the atom sizes represented in A–D. The ion radii (r; in Å) were derived from the sizes of ions in crystal lattices; H⁺, black, r = 0.012; Li⁺, purple, r = 0.76; Na⁺, light blue, r = 1.02; K⁺, gray, r = 1.38; Tl⁺, brown, r = 1.5; Rb⁺, orange, r = 1.52; Cs⁺, light orange, r = 1.67; Mg²⁺, yellow, r = 0.72; Ca²⁺, pink, r = 0.99.

Ion Selectivity of K-Ring.

Bound ²²Na⁺ in the purified K-ring was exchangeable with Li⁺ but not with K⁺, Rb⁺, Tl⁺, Cs⁺, Mg²⁺, or Ca²⁺ (Fig. 3D). Li⁺ and H⁺ competitively inhibited Na⁺ binding of the K-ring (Fig. 3 B, C, E, and F), indicating that Na⁺, Li⁺, and H⁺ bind to the same pocket. How does the ion binding pocket of the rotor ring distinguish the kind of ions to bind?

Na⁺ and Li⁺ binding.

Na⁺-coupled enzymes usually recognize Li⁺ with different affinities. Na⁺-translocating F-ATPase from P. modestum transports Li⁺ with a 10-fold lower affinity than Na⁺ (4). In the case of the Na⁺-coupled aspartate transporter (Glt_ph) from Pyrococcus horikoshii, the binding of Li⁺ is ≈10- to 30-fold lower than Na⁺ (13). In contrast the Na⁺/H⁺ antiporter from Helicobacter pylori and Escherichia coli exhibits a higher affinity for Li⁺ than Na⁺ (14, 15). In this study, we reported that the K-ring of E. hirae V-ATPase binds Li⁺ with ≈5-fold lower affinity than Na⁺ (Figs. 2C and 3F). According to the model of the Na⁺- (or Li⁺) binding site in the K-ring described above, an ion (Na⁺ or Li⁺) is bound at the center of five residues in the binding pocket. The cavity size of the binding site formed by the oxygen atoms accommodates the Na⁺ ion (diameter of Na⁺ = 2.04 Å; mean Na⁺–O distance = 2.3 Å) better than the Li⁺ ion (diameter of Li⁺ = 1.52 Å; mean Li⁺–O distance = 2.1 Å). Indeed, as demonstrated by the binding studies the Na⁺ ion binds with a 5-fold higher affinity than the Li⁺.

H⁺ binding.

Of the five residues involved in Na⁺ binding to the K-ring only one, an essential glutamate (E¹³⁹), is conserved in all H⁺-transporting V-ATPases. The other four residues (L⁶¹, T⁶⁴, Q⁶⁵, and Q¹¹⁰) are not conserved (9). It is widely accepted that the H⁺ binding/release in H⁺ transport of V-ATPase occurs via protonation/deprotonation of the carboxyl group of the glutamate residue (1). The K-ring whose carboxyl group should be protonated at acidic pH does not bind ²²Na⁺ with the same high affinity at alkali pH (Fig. 3A). Analysis of inhibition by H⁺ ions suggested that the K-ring binds Na⁺ and H⁺ competitively (Fig. 3 B and C). The K_i value of H⁺ corresponds to the acid dissociation constant (K_a) of the E, the pK_a of which is 5.5 (Fig. 3C). However, we have not observed ATP-driven H⁺ uptake activity into proteoliposomes reconstituted with the whole V-ATPase complex at pH 5.5 in the effective absence of Na⁺ (data not shown). It proved impossible to obtain crystals of H⁺-bound K-ring under acidic conditions, as the K-ring was unstable at <pH 4.5. Fig. 5D shows the predicted structure of the H⁺-bound K-ring. The protonated E¹³⁹ should be stabilized by hydrogen bonds to side chains of Gln-110, Tyr-68, and Thr-64 as seen in both the Na⁺- and Li⁺-bound K-ring structures.

K⁺, Tl⁺, Rb⁺, and Cs⁺.

It is generally accepted that the key factors involved in ion selectivity are the cavity size and the coordination number of oxygen atoms forming the binding pocket (16–18). The cavity size of K-ring binding pocket matches well to the ion size of Na⁺ (r = 1.02), but not to K⁺ (r = 1.38). The coordination number in the K-ring binding pocket is five. The average coordination numbers for Na⁺ and K⁺ from an analysis of Na⁺- and K⁺- containing structures in the Protein Data Bank are 5.4 and 6.4, respectively (17), suggesting that the coordination number in the K-ring binding pocket is more suited to Na⁺ binding. Tl⁺ is typically used as a heavy-ion analogue for K⁺, because of the similarity of the ion radii of Tl⁺ (r = 1.5) and K⁺ (r = 1.38). However, Tl⁺ is a highly polarizable ion and can bind to the Na⁺-binding site of the Na⁺-coupled aspartate transporter (Glt_ph) from P. horikoshii, although K⁺ does not bind (13). In this study, Tl⁺ did not bind to the K-ring (Fig. 3D). The cavity size of the K-ring is likely to have a role in distinguishing K⁺ and Tl⁺ from Na⁺. Rb⁺ (r = 1.52) and Cs⁺ (r = 1.67), which have larger radii than that of K⁺, are also unable to bind to the K-ring because of the small cavity size (Fig. 5).

Mg²⁺ and Ca²⁺.

The radii of Mg²⁺ and Ca²⁺ are 0.72 and 0.99 Å, respectively. Average coordination numbers for Mg²⁺ and Ca²⁺ estimated from the analyses of Mg²⁺- and Ca²⁺-containing structures in the Protein Data Bank are 5.8 and 6.7, respectively (18). Therefore, theoretically these ions can bind to the ion binding pocket of the K-ring. Previous work has suggested that Mg²⁺- and Ca²⁺-binding sites require a high charge density necessary to compensate for dehydration of the divalent cation (16, 18). Thus, it is conceivable that these divalent cations are unable to bind to the K-ring because of the presence of only one fully charged oxygen in the binding site.

Ion-Transporting Mechanism.

The ion-transporting mechanism of F-ATPases has been suggested to be either a two-channel model or a one-channel model. The two-channel model of H⁺ translocation proposed by Junge et al. (19) is based on an occluded position of the H⁺-binding pockets in the ring, embedded deeply in the lipid bilayer. In the model, H⁺ binding/release of the c-ring occurs through the two half channels at the interface between the a-subunit and the c-ring. The one-channel model of Na⁺ translocation proposed by Dimroth et al. (20) is based on biochemical evidence showing the free cytoplasmic access of ions to the ion-binding sites in the c-ring. In the model, Na⁺ binding/release of ions to the c-ring occurs through a half channel at the a-subunit/c-ring interface and an intrinsic channel in the c-ring connecting the ion-binding sites with the cytoplasm. Recently, Meier et al. (21) solved the crystal structure of the Na⁺-bound c-ring of F-type Na⁺-ATPase from Ilyobacter tartaricus and revealed that there is no intrinsic ion channel in the c-ring. Thus, it is likely that Na⁺ binding/release of the c-ring in the F-ATPase from I. tartaricus and P. modestum also occurs at the interface between the a-subunit and the c-ring as outlined in the recently proposed push-and-pull model (22).

Previously, we reported the accessibility of ions to the Na⁺-binding site of E. hirae V-ATPase by using the ²²Na⁺-binding assay system (11). About two-thirds of the total amount of bound ²²Na⁺ to purified V₁V_o complex was released within 30 s of the first sampling point of measurement (k_off > 2.8 × 10 ⁻² s⁻¹). This result suggested that the multiple Na⁺-binding sites are accessible from the aqueous phase, being explained by the one-channel model (5, 11, 23). However, the crystal structure of the K-ring revealed that the binding pocket is located close to the middle of the lipid bilayer, with no intrinsic ion channel in the ring, supporting the two-channel model (8, 9). In this study, we examined the ²²Na⁺ binding/release properties of purified K-ring. Although the data suggested that the Na⁺-binding pocket in the K-ring was accessible to the aqueous phase in detergent (Fig. 2), the obtained rate constants (k_ob = 2 × 10⁻³ s⁻¹, k_on = 1 × 10² M⁻¹·s⁻¹, k_off = 1 × 10⁻³ s⁻¹) were extremely slow. The purified whole complex showed high ATP hydrolysis activity (4 units/mg protein) stimulated by 10 μM Na⁺ ion in detergent as well as in lipid (12), which corresponds to the speed that the enzyme rotates >10 times per s, transporting (binding/release) >100 ions per s. These estimations suggest that the observed slow rates of binding/release of the ²²Na⁺ of the purified K-ring in detergent are far from the parameters of the actual ion-transporting reaction of the Na⁺-ATPase. These findings strongly suggest that the interaction with the stator subunit (I-subunit) is essential for the Na⁺ binding to/release from the K-ring. The faster release (k_off > 2.8 × 10⁻² s⁻¹) of ²²Na⁺ from purified V₁V_o–ATPase may have required an idling rotatory motion (without ATP) of the rotor at the K-ring/I-subunit interface. All of the kinetic data for ion binding to the K-ring presented here are consistent with the two-channel model proposed for ion translocation of V-ATPase based on the Na⁺-bound K-ring structure (9) (Fig. 1). To elucidate whether the direct access of Na⁺ to the isolated K-ring observed in detergent occur in the lipid bilayer, a comparison of ion exchange rates of the K-ring in detergent versus those in the membrane is necessary. We have tried to measure ²²Na⁺ binding to/release from the purified K-ring reconstituted into liposomes, but have been unable to obtain reliable results so far because of problems depleting contaminating Na⁺ in lipid and difficulties in distinguishing between ion exchange rates of the K-ring in lipid bilayer and the leakage rate of free ²²Na⁺ trapped in the proteoliposomes. Establishment of a more suitable assay system will be valuable for further investigation of the ion-coupling mechanism of V-ATPase.

Materials and Methods

Protein Preparation.

The V-ATPase was purified from cells of E. hirae as described (24). The K-ring was released from the isolated V-ATPase by treatment with 10% isopropanol and then purified by anion exchange and gel filtration chromatography as described (10).

Measurement of ²²Na⁺ Binding to Purified Samples.

Binding of ²²Na⁺ to the K-ring.

The K-ring is comprised of 10 16-kDa K-subunits. One milligram of the purified K-ring corresponds to 60 nmol K-subunit (6 nmol K-ring). Binding experiments were initiated by the addition of 10 μM ²²NaCl in the reaction mixture that contained 6 μM K-subunit (0.6 μM K-ring) in A buffer (20 mM Tris·HCl, 20% glycerol, 0.05% n-dodecyl β-d-maltoside, pH 7.5). Free ²²Na⁺ was rapidly separated by using a Dowex-50 method at various time intervals as described (11). The observed association rate constant (k_ob) was determined by using a one-phase exponential association equation with Prism software (GraphPad). SD was calculated from three independent experiments. The association rate constant (k_on) was calculated according to the following equation: k_on = (k_ob − k_off) / [²²Na⁺].

Release of ²²Na⁺ bound to the K-ring.

Dissociation experiments were initiated by the addition of 10 mM nonradioactive NaCl, LiCl, KCl, RbCl, CsCl, TlNO₃, CaCl₂ or MgSO₄ to the incubation mixture (6 μM K-subunit in A buffer) equilibrated with 10 μM ²²Na⁺. Free ²²Na⁺ was rapidly separated by using a Dowex-50 method at various time intervals. The release of ²²Na⁺ bound to the K-ring was caused by the exchange reaction with nonradioactive cations. The exchange rate of ²²Na⁺ from the ring increased depending on the concentrations of added Na⁺ (or Li⁺), and addition of 10 mM (1,000-fold excess) nonradioactive cation is enough to saturate the rate (data not shown). The saturated exchange rate constant should correspond to the dissociation rate constant (k_off) of Na⁺ in the binding reaction.

Na⁺ concentration dependence of ²²Na⁺ binding.

The reaction mixture that contained 60 pmol K-subunit (6 pmol K-ring) and various concentrations of ²²NaCl (30,000 cpm) in 10 μl of A buffer was incubated for 2 h at room temperature, long enough to saturate Na⁺ binding to the K-ring. Na⁺-binding capacity was measured by separating the free ²²Na⁺ by using the Dowex-50 method. To estimate the K_is of H⁺ or Li⁺ inhibition on ²²Na⁺ binding, Na⁺ concentration dependence of ²²Na⁺ binding was measured at various pHs or various Li⁺ concentrations. SD was calculated from three independent experiments.

pH dependence of ²²Na⁺ binding.

The reaction mixture containing 60 pmol K-subunit, 100 pmol ²²NaCl (30,000 cpm), and 1 μmol of various pH buffer (pH 3.5–5.5, citrate-Tris; pH 5.5–7, Bis-Tris·HCl; pH 7–9, Tris·HCl) in 10 μl of A buffer was incubated for 2 h at room temperature. The ²²Na⁺ binding at various pHs was measured as above.

Crystallization of Li⁺-Bound K-Ring.

The purified K-ring (5 μl; 2 mg/ml) in buffer containing 20 mM Tris·HCl (pH 7.5), 100 mM LiCl, 10% glycerol, and 0.32 mM n-dodecyl β-d-maltoside was mixed with reservoir solution (5 μl) consisting of 100 mM Tris·HCl (pH 7.5), 4% glycerol, 240 mM potassium citrate, 1.2 mM undecylmaltoside, and 34% PEG 400. Crystals grew at 20°C in sitting drops by vapor diffusion. They appeared after 3 days and grew to maximum dimensions (250 × 200 × 70 μm) in 3 weeks. Crystals were plunged into liquid nitrogen and stored at 100 K.

Data Collection, Structure Determination, and Refinement.

Diffraction data were collected from a single crystal at cryogenic temperature (100 K) on beam-line BL26B1 at the SPring8 (Harima, Japan). The data were indexed and integrated with MOSFLM (25) and processed further with the CCP4 programs (26). The structure was solved by molecular replacement with AMoRe (27) by using the Na⁺-bound K-ring structure (Protein Data Bank ID code 2BL2). In all subsequent refinement steps, 5% of the data were set aside for calculation of the free R factor. The atomic model was built by using the program O (28) and refined by using REFMAC5 (29). The bound lengths between the Li⁺ ion and its ligands were not restrained. Tight noncrystallographic symmetry restraints (sigma 0.05 Å) were applied to the 10 K-protomers (excluding regions in lattice contacts). TLS refinement, with one TLS group per protomer, was carried out in the final stages. The refined structure was validated with PROCHECK (30). Figures were generated by PYMOL (31).

Analytical Methods.

Protein concentrations were determined by the BCA method (Pierce Chemicals) using BSA as standard. ²²NaCl (1.36 TBq/mmol) was obtained from Dupont/NEN Research Products.