Displacement of the Na⁺/K⁺ pump’s transmembrane domains demonstrates conserved conformational changes in P-type 2 ATPases

Significance

The P-type 2 ATPase subfamily includes several essential primary active ion transporters like the Na⁺/K⁺, H⁺/K⁺, and SERCA pumps. These transporters are thought to act by a conserved mechanism. However, major conformational changes seen in the transmembrane domains in crystal structures of SERCA and also proposed to occur in H⁺/K⁺ pumps are absent in available Na⁺/K⁺-pump structures. Here we use cross-linking of engineered cysteines and voltage clamp fluorometry (VCF) to show that the movements of transmembrane segments conserved in other pumps, which are absent in Na⁺/K⁺ pump crystal structures, occur in functional Na⁺/K⁺ pumps. In addition, we illustrate a twist to the usual VCF approach that may help in future studies using this technique with other membrane proteins.

Abstract

Cellular survival requires the ion gradients built by the Na⁺/K⁺ pump, an ATPase that alternates between two major conformations (E1 and E2). Here we use state-specific engineered-disulfide cross-linking to demonstrate that transmembrane segment 2 (M2) of the pump’s α-subunit moves in directions that are inconsistent with distances observed in existing crystal structures of the Na⁺/K⁺ pump in E1 and E2. We characterize this movement with voltage-clamp fluorometry in single-cysteine mutants. Most mutants in the M1–M2 loop produced state-dependent fluorescence changes upon labeling with tetramethylrhodamine-6-maleimide (TMRM), which were due to quenching by multiple endogenous tryptophans. To avoid complications arising from multiple potential quenchers, we analyzed quenching of TMRM conjugated to R977C (in the static M9–M10 loop) by tryptophans introduced, one at a time, in M1–M2. This approach showed that tryptophans introduced in M2 quench TMRM only in E2, with D126W and L130W on the same helix producing the largest fluorescence changes. These observations indicate that M2 moves outward as Na⁺ is deoccluded from the E1 conformation, a mechanism consistent with cross-linking results and with proposals for other P-type 2 ATPases.

P-type ATPases are enzymes that catalyze formation of ion or lipid gradients across membranes in all phyla. These membrane proteins use homologous E1–E2 mechanisms (Fig. 1A) in which uphill transport of a subtype-specific substrate is powered by ATP hydrolysis (1). In nearly all animal cells the Na⁺/K⁺ pump, a P-type 2C ATPase, exports three Na⁺ and imports two K⁺ into the cell, against their electrochemical gradients, thereby building the ion gradients that energize other essential membrane processes, including electric signaling and secondary-active transport. This ion pump is formed by a catalytic α-subunit, homologous to the catalytic subunits of all other P-type ATPases, and an auxiliary β-subunit. The α-subunit contains the ion-transport sites within its 10 transmembrane α-helices (M1–M10) and the nucleotide (N), phosphorylation (P), and actuator (A) domains formed by its intracellular loops, while the β-subunit, with a single transmembrane segment and a large extracellular globular domain, is necessary for ion occlusion (2) and membrane stability (3). In each catalytic cycle, the pump transits through a set of partial reactions while occupying two major conformations E1 and E2, in what is also known as the Post-Albers mechanism (Fig. 1A).

Fig. 1.

(A) Post-Albers kinetic scheme of the Na⁺/K⁺ pump (clockwise forward direction). The pump alternates between two major conformations, E1 and E2, which can be phosphorylated (P) or dephosphorylated. Parentheses indicate ions occluded within the protein. Transitions within the red box E1P(3Na⁺) ↔ E2P + 3 Na⁺_o produce transient charge movement. (B and C) Overall change in structure between Na⁺/K⁺ pump structures E1(3Na⁺) (PDB ID 3WGV; cyan) and E2P-ouabain (PDB ID 4XE5; pink) shown from two angles after aligning through the C-terminal end of the Na⁺/K⁺ pump α-subunit, (B) A lateral view, approximately parallel to the membrane plane (indicated by gray dotted lines), and (C) a view from the extracellular side, approximately perpendicular to the membrane (rotated 90° from the view in B). Residues and distances addressed in the discussion are labeled.

Utilizing a combination of generic P-type ATPase as well as pump-specific inhibitors, all intermediate states of the sarcoplasmic/endoplasmic reticulum ATPase (SERCA, another P-type 2 ATPase) have been determined by X-ray crystallography (4–14), but only a handful of states have been resolved for the Na⁺/K⁺ pump (15–20). Considering the high sequence similarity and conserved catalytic cycle, it was expected that the main conformational changes should also be conserved. However, comparison of the available crystal structures in equivalent states reveals significant differences, particularly in the movement of the M1–M2 stalk. In SERCA, transitions between E1P(2Ca²⁺) and E2P states involve kinking as well as movement of the luminal part of M1–M2 perpendicular to the membrane plane (9), a movement that seems largely absent in the available Na⁺/K⁺ pump crystal structures (20) (SI Appendix, Fig. S1). Fig. 1B compares two states of such Na/K pump structures: the aluminum fluoride-inhibited E1 with 3 Na⁺ bound (ref. 20; cyan carbons) and the ouabain- and Mg²⁺-bound beryllium fluoride-inhibited structure (ref. 21; pink carbons). The closest physiologic states to these inhibitor-locked structures are enclosed with a dotted box in the catalytic cycle (Fig. 1A). Displacements in the membrane plane are absent in the side view (Fig. 1B and SI Appendix, Fig. S1), despite large movements in other directions, which are obvious in the external view perpendicular to the membrane (Fig. 1C).

The dislodgment of M1–M2 is thought to guide opening of the SERCA ion-binding sites as Ca²⁺ is deoccluded and released toward the luminal side prior to occlusion of the countertransported H⁺. Because a similar mechanism has been proposed for the gastric H⁺/K⁺ pump (22), the absence of such conserved conformational change in the closely related Na⁺/K⁺ pump is perplexing, and this movement has never been evaluated in a functional Na⁺/K⁺ pump in the membrane environment. Here we address the movement of the M1–M2 region in functional Na⁺/K⁺ pumps utilizing cross-linking of engineered disulfides under voltage clamp, as well as voltage-clamp fluorometry (VCF), under conditions in which we can control the occupancy of phosphorylated states in the pump cycle. Our results indicate that the movement of the externalmost part of the α-subunit’s M1–M2 is required for Na⁺/K⁺ pump to function and, thus, conserved through the P-type 2 ATPase family. Furthermore, we illustrate a slight modification to the usual VCF approach, where TMRM is attached to a static residue while relative distance changes are evaluated by scanning with quenching tryptophans in the moving region, allowing one to study both the movement and the structural characteristics of transmembrane segments and their loops in a functional Na⁺/K⁺ pump, a powerful modification that can be exploited to investigate other transport proteins as well.

Results

Cross-Linking of R977C in the Fifth Loop with M1–M2 Residues.

An early homology model of the Na⁺/K⁺ pump based on the luminal-open SERCA structure suggested that two residues conserved in all Na⁺/K⁺ pump α-subunits, R977 and D126 (throughout this article, we use sheep α1 residue numbering, two lower than pig α1, and four lower than Xenopus α1), formed a salt bridge in the externally open E2P Na⁺/K⁺ pump when ouabain is bound (16). We used a Xenopus α1-subunit with the mutation C109Y, to simultaneously reduce ouabain affinity (23) and avoid functional effects of cysteine reactivity on the extracellular surface (24, 25) (Methods), as the template to introduce R977C and, subsequently, one cysteine at a time in the first external loop of the α-subunit (from Q116 in M1 to L130 in M2). We expressed these mutants in Xenopus oocytes and tested for functional effects of spontaneous cross-linking between each engineered double-cysteine with two-electrode voltage clamp (TEVC) (Fig. 2).

Fig. 2.

Cross-linking of R977C with residues in M1–M2. (A and B) Currents from oocytes expressing C109Y/R977C (A) or C109Y/D126C/R977C (B). Application of 3 mM K⁺ (3K) in N-methyl D-glucamine (NMG⁺) solution activated outward Na⁺/K⁺ pump current in both oocytes. The maneuver was repeated after a 10-min incubation in the reducing agent TCEP (applied in Na⁺ solution). Note that only the double-cysteine mutant in B showed increased currents after TCEP. Na⁺/K⁺ pump current disappeared following application of Cu:Phen (100 μM Cu:200 μM Phen). (C) K⁺-induced current before TCEP divided by the current after TCEP. Indicated mutants were introduced in the C109Y/R977C background. Gray bars indicate significantly smaller than 1. TCEP did not change current in the template (C109Y) and control (C109Y/R977C). Mean of 4 to 12 experiments. (D and E) Current from two oocytes expressing C109Y/D126C/R977C where repeated 30-s-long applications of 1 μM Cu:Phen are made at −50 mV in 125 mM NMG⁺ (D) or at −90 mV in 125 mM Na⁺ (E). Following Cu:Phen, each K⁺ (3 or 10 mM) application induced a smaller pump current. (F) Remnant Na⁺/K⁺ pump current vs. the cumulative time in Cu:Phen from experiments like those in D and E. The point at 5 min was obtained independently, with a single Cu:Phen application. Each point is the average of three to five experiments. Error bars are SEM in all panels.

Outward Na⁺/K⁺ pump currents were always observed when we applied 3 mM K⁺ to oocytes expressing the single-cysteine mutant R977C (Fig. 2A). Subsequently, the reducing agent TCEP (10 mM) was applied for 10 min. K⁺ was then reapplied, inducing currents identical to those observed before TCEP. In contrast, when a similar experiment was performed on an oocyte expressing the double-cysteine mutant D126C/R977C, the Na⁺/K⁺ pump currents after TCEP application were significantly larger than before TCEP (Fig. 2B). Furthermore, subsequent application of the oxidizing reagent copper phenanthroline (Cu:Phen, 100 μM) toward the end of the experiment nearly eliminated the subsequent Na⁺/K⁺ pump current. These results indicate either that the disulfide bond between D126C and R977C blocked function in all cross-linked pumps, with the remnant un–cross-linked pumps functioning normally, or that cross-linking in all the pumps leaves them with clearly impaired pump function. Considering our experiments measuring transient currents reporting the E1P(3Na⁺) ↔ E2P transition (SI Appendix, Fig. S2) that were nearly abolished after oxidation, we favor the hypothesis that cross-linking blocks the E1P(3Na⁺)↔E2P transition.

Fig. 2C summarizes the effect of TCEP when similar experiments were performed in oocytes expressing all double-cysteine mutants that displayed spontaneous cross-linking between R977C and a second cysteine at M1–M2 positions (E120C, E122C, P123C, Q124C, and D126C). We also tested the effect of 100 μM Cu:Phen (as in Fig. 2B) on these residues. The fractional current after Cu:Phen was 0.96 ± 0.12 for the C109Y template (n = 4), 0.64 ± 0.11 for E120C (n = 5), 0.56 ± 0.06 for E122C (n = 3), 0.71 ± 0.10 (n = 4) for P123C, 0.62 ± 0.16 (n = 4) for Q124C, and 0.24 ± 0.09 (n = 8) for D126C (errors are SD).

To evaluate if the cross-link between D126C and R977C forms in E2P as predicted from the early SERCA models, we tested whether conditions promoting this state accelerate cross-linking (Fig. 2 D–F). We performed brief applications of Cu:Phen (1 μM) either at −50 mV in the absence of external Na⁺ (replaced with external N-methyl D-glucamine, NMG⁺) to favor E2P occupancy (Fig. 2D) or in the presence of 125 mM Na⁺ at −90 mV to favor E1P(3Na⁺) occupancy (Fig. 2E). The effect of each Cu:Phen application was tested in the subsequent activation of Na⁺/K⁺ pump current by K⁺. Each oxidizing insult in NMG⁺ induced larger reductions in pump current than in the presence of Na⁺ at −90 mV. Plotting the remnant Na⁺/K⁺ pump current against the cumulative time in Cu:Phen (Fig. 2F) demonstrates that disulfide formation was faster in NMG⁺ at −50 mV (E2P) than in Na⁺ solutions at −90 mV [favoring E1P(3Na⁺)]. Since a 5-min application reduced current by 90% even without reaching steady state after this long application, this is also consistent with a complete block of function in the D126C/R977C cross-link. The slow modification rate in E2P may reflect a combination of slow cross-linking with 1 μM Cu:Phen with a relative high freedom for M2 to move while in E2P. Such movement would be consistent with the high B factors of the available structures and the current lack of a high-resolution E2P open (ion-free and ouabain-free) analog structure. The slower cross-linking under E1 promoting conditions can still happen, probably because the pumps can transit to E2P at a rate of ∼50 s⁻¹ (SI Appendix, Fig. S2A), and although many pumps will come back quickly to E1P at −90 mV (where ∼2% of the pumps are in E2P; SI Appendix, Fig. S2B), the cross-linked state is absorbent, producing the slower current decay as pumps cannot return to E1.

Overall, cross-linking results indicate that D126, located in the rigid M2 helix, comes in close proximity to R977 in E2P and separates in E1, in contrast to expectations from the distances seen in crystal structures (Fig. 1C). Smaller reductions in other cross-linking residues Q124C, P123C, E122C, and E120C with R977C probably reflect a combination of a more flexible region toward the M1–M2 loop with incomplete cross-linking due to larger distances and/or orientation factors. Interestingly, the published crystal structure models report some of the highest B factors of the transmembrane domain around the loop 1–2, and additionally, within E1P(3Na⁺) crystal structure the helix has been modeled as starting at Q124 while the E2P models suggest it starts at E122. Our cross-linking data are consistent with the E122 assignment.

VCF Demonstrates Movement of M1–M2.

The results above suggest that during the pump cycle the M1–M2 stalk must move with respect to R977. However, interpretation of the functional effects of cross-linking can be difficult (26). We used VCF to gain insight into the movement of M1–M2, utilizing a sheep α1 template, which has been previously evaluated by VCF (27–30). This template has reduced affinity for ouabain thanks to a double substitution Q116R/N127D and extracellular cysteines removed by the double substitution (C911S and C964A). We introduced single-cysteine mutants from R116 to L130, one at a time, and labeled them with the fluorophore tetramethylrhodamine-6-maleimide (TMRM; Methods) to simultaneously measure the current and fluorescence signals reporting conformational changes. TMRM labeled all positions from R116C to D127C but failed to label from L128C to L130C, three residues deeper into M2. Initially, we studied conformational changes induced by voltage in the presence of Na⁺, without K⁺, to force the pump to transit between E1P(3Na⁺) at negative voltages and the Na⁺-free E2P at positive voltages. (Fig. 3 and Table 1).

Fig. 3.

Current and fluorescence changes in oocytes expressing E120C. (A) Ouabain-sensitive currents elicited by 500-ms-long pulses to the indicated voltages. The dashed line is a single-exponential fit to the data starting 5 ms from the voltage step. (B) Fluorescence signal, expressed as percentile value of the level at −50 mV, at the same voltages measured in Na⁺ solution. Dashed lines are double-exponential fits to the data, with shared slow time constant. (C) Fluorescence signal in Na⁺ solution with 10 mM ouabain. (D) Mean rate (1/τ) as a function of voltage for the ouabain-sensitive current (black squares, n = 7) and fast component of fluorescence in Na⁺ (orange circles, n = 7). (E) Overlap of the mean Q–V (right axis, black symbols, n = 9) and the mean ΔF/F–V (left axis, orange symbols) measured in the absence (circles, n = 20) or presence (triangles, n = 15) of ouabain. Line plots are Boltzmann distributions with parameters V_1/2 = −70 mV, kT/ez_q = 47 mV, F/F_max = 18.1% for ΔF/F and V_1/2 = −65 mV, kT/ez_q = 47 mV for Q–V. Error bars are SEM.

Table 1.

Mean values of several quantities from current and fluorescence measurements for M1–M2 mutants

	R116C	A117C	A118C	T119C	E120C	E121C	E122C	P123C	Q124C	N125C	D126C	D127C
IK,max (nA)	256 ± 155	171 ± 68	265 ± 49	155 ± 48	220 ± 111	380 ± 162	411 ± 496	453 ± 406	519 ± 570	265 ± 138	81 ± 36	32 ± 21
Qtot (nC)	20 ± 17	37 ± 12	25 ± 7	11 ± 4	19 ± 27	30 ± 16	24 ± 12	24 ± 15	20 ± 6	26 ± 8	14 ± 9	17 ± 10
V1/2,Q (mV)	−45 ± 12	−155 ± 30	−88 ± 4	−51 ± 10	−68 ± 11	−77 ± 14	−78 ± 14	−63 ± 19	−81 ± 10	−71 ± 8	−20 ± 23	−13 ± 7
kT/ezq,Q (mV)	34 ± 13	37 ± 8	39 ± 22	37 ± 6	39 ± 17	44 ± 10	34 ± 11	31 ± 6	41 ± 11	32 ± 5	41 ± 11	33 ± 10
ΔF/Fmax (%)	1 ± 0.4*	ND	0.5 ± 0.5*	17 ± 15	19 ± 5	5 ± 3	2 ± 0.8	2 ± 0.9	6 ± 2	7 ± 3	4 ± 2	ND
V1/2,F (mV)	ND	ND	ND	−0.5 ± 1	−59 ± 27	−102 ± 14	−103 ± 25	−48 ± 4	−90 ± 3	−141 ± 6	−47 ± 6	ND
kT/ezq,F (mV)	ND	ND	ND	47 ± 0.5	48 ± 14	44 ± 8	45 ± 23	34 ± 2	54 ± 3	40 ± 1	38 ± 5	ND
ΔF/Fouab (%)	0.6 ± 0.6	0.8 ± 0.6	1.7 ± 1	5.5 ± 3	0.3 ± 3†	−0.5 ± 2.4†	−2.2 ± 1	−0.7 ± 1†	1.4 ± 1.9†	2.2 ± 1.9	3.6 ± 2	0.8 ± 0.4
ΔF/FK (%)	0.3 ± 0.5†	0.1 ± 0.9†	0.7 ± 0.5	−0.6 ± 2†	−2.6 ± 2.6†	−2.4 ± 1.5	0.3 ± 0.7†	0.1 ± 1†	−1.7 ± 2†	−1.9 ± 1.6	−0.5 ± 0.04	0.7 ± 0.3

The top four rows are maximal K⁺-induced current (I_max,K), and parameters from Boltzmann fits to Q–V (Q_tot, V_1/2,Q, and kTez_q,Q). The bottom five rows are parameters from the Boltzmann fits to ΔF/F–V (ΔF/F_max, V_1/2,F, and kT/ez_{q, F}), followed by the steady-state changes ΔF/F induced by ouabain and K⁺ at −50 mV. Errors are SD from 4 to 13 determinations. ΔF/F_max includes data from the first set of pulses in each oocyte to avoid bleaching effects.

^*ΔF/F_max was obtained by adding the ΔF/F value measured at −180 mV and at +60 mV because a Boltzmann function could not be reliably fitted.

^†The ΔF/F value is not significantly different from zero at P < 0.05. Note that positions 116 and 127 occupied by a Gln and an Asn in the Xenopus clone are an Arg and Asp, respectively, in the ouabain-resistant sheep α1 clone (Methods).

Experiments from oocytes expressing E120C, a position with large fluorescence changes, are illustrated in Fig. 3. After labeling with TMRM, we applied 0.5-s-long square pulses to voltages ranging from −180 to +40 mV to an oocyte held at −50 mV, in Na⁺ solution. The pulse protocol was repeated, first in the absence and then in the presence of the specific inhibitor ouabain. The ouabain-sensitive transient current (current in the absence of inhibitor minus current in the presence of 10 mM ouabain) represents the occlusion/deocclusion of external Na⁺, rate limited by the E1–E2 conformational change of the protein (31). The current time course was adequately described by a monoexponential decay fit (Fig. 3A, dashed lines). Transient currents were slower at positive voltages and became faster at negative voltage due to the voltage-dependent rebinding of Na⁺, as it is observed in Na⁺/K⁺ pumps of diverse origin (31–37). The fluorescence signal measured simultaneously in the same oocyte (Fig. 3B) illustrates that fluorescence increased (unquenched) at positive voltages, which augment E2P occupancy, while it decreased (quenched) at negative voltages, which promote E1P(3Na⁺). The fluorescence changes were blocked by ouabain (Fig. 3C). The time courses were described by a biexponential function (Fig. 3B, dashed lines), with a voltage-independent slow component (τ = 109 ± 14 ms, SD) and a voltage-dependent fast component with a fractional amplitude of 72 ± 5% for negative pulses and 54 ± 8% for positive ones. Fig. 3D shows the mean voltage dependence of the rate constants (1/τ) for the fast component of ΔF/F (orange circles) and the transient current (black squares). Both signals had comparable voltage dependencies (faster at negative voltages and slower at positive voltages), but the fluorescence changes were twofold to fourfold slower than the transient current. Fig. 3E overlaps the Q–V (right axis, black squares) and the steady state ΔF/F–V in 125 mM Na⁺ (left axis, orange circles) and in 125 mM Na⁺ with 10 mM ouabain (orange triangles). The Q–V and ΔF/F–V in Na⁺ solution were fitted with a Boltzmann distribution (line plots) with comparable voltage-dependent parameters. Furthermore, when the Na⁺ concentration was reduced to 30 mM, both the ouabain-sensitive Q–V and the ΔF/F–V were significantly shifted to the left by identical amounts when compared to the curves in 125 mM Na⁺ (ΔV_1/2 = 37 ± 9 for Q/V and ΔV_1/2 = 36 ± 22 for ΔF/F; n = 4, SD), the expected behavior for the parameters reflecting the voltage-dependent binding of external Na⁺.

Results from similar experiments with each TMRM-labeled M1–M2 mutant are summarized in Table 1, and examples are illustrated in SI Appendix, Fig. S3. We also evaluated the effect of adding 10 mM ouabain or 10 mM K⁺ on the steady-state fluorescence at −50 mV (Table 1, bottom five rows). Because ouabain binds to E2P (38), we expected that its application would either increase steady-state fluorescence or be without effect, but it quenched the fluorescence of E122C and P123C. In addition, ouabain application induced a statistically significant increase in fluorescence (∼1%) of A117C, a mutant that lacked voltage- and K⁺-dependent fluorescence responses. This result suggests that ouabain binding induces a conformational change, as previously reported using lanthanide resonance energy transfer (39). Thus, these fluorescence changes at each position may include both a component due to ouabain-specific conformational changes and the change in the E1–E2 equilibrium.

On the other hand, K⁺, which activated pump current in all constructs (first line in Table 1), is expected to favor E1 occupancy and therefore reduce fluorescence, particularly in mutants whose Q–V distribution is centered close to or below −50 mV, although the pump will cycle through all states in the presence of external K⁺. This was not the case; for instance, A118C and D127C all increased fluorescence (∼1%) in the presence of K⁺. Importantly, all mutations in Table 1 were introduced on a template whose Q–V is centered at –61 ± 15 mV (n = 8). It is clear that introducing the mutation and the TMRM adduct in M1–M2 has a wide range of effects on the Q–V, inducing shifts of the Boltzmann distributions in every direction. This result again suggests that the movement and the interactions of M1–M2 residues are important determinants of Na⁺/K⁺ pump function. In general, the results in Table 1 indicate that large displacements in M1–M2 lead to a fluorescence increase in E2 and a reduction in E1.

For most positions, positive pulses produced slower signals than negative ones (SI Appendix, Fig. S3) for both fluorescence and current, as expected if both signals report the E1P(3Na⁺)↔E2P conformational change. However, the fluorescence change was always noticeably slower than the transient current (Fig. 3 A, B, and D and SI Appendix, Fig. S3), and as evident in Table 1, the center of the fluorescence and Q–V did not coincide for most mutants. The discrepancies between fluorescence and current probably reflect incomplete labeling with TMRM. In this case, depending on the labeling efficiency and the actual kinetics of the labeled pumps, the slower transient current signals from labeled pumps may go undetected (lost in the baseline of the traces), with the current signal reflecting the behavior of unlabeled pumps; obviously, only labeled pumps contribute to the fluorescence signal. Experiments measuring Na⁺/K⁺ pump function, before and after tagging with TMRM for 5 min at room temperature, revealed that TMRM labeling slows down transient currents and reduced both the pump current and the transient currents but failed to demonstrate if all the differences are solely due to inefficient labeling (SI Appendix, Fig. S4).

Identification of Quenching Residues.

Tetramethylrhodamine is quenched by tryptophan and to a lesser extent by tyrosine (40). The quenching mechanisms require the collision between the quencher and the fluorophore (40), instead of energy transfer mechanisms at longer distances, like FRET. Three putative quenching tryptophans were identified in the crystal structures: W315 (in M3) as well as W888 and W892 (in the long loop between M7 and M8). To evaluate if there are position-specific tryptophan quenchers, each tryptophan was substituted, one at a time, with alanine in E120C and D126C (two mutants at different places within the loop with strong fluorescence signals) and ΔF/F–V was measured (Fig. 4). Compared to the E120C template, W315A and W892A showed largely reduced ΔF/F amplitudes, while W888A showed a somewhat smaller reduction in ΔF/F (Fig. 4A). On the other hand, compared to the D126C template, W888A and W892A had similar ΔF/F amplitude, and W315A had a slightly increased ΔF/F amplitude (Fig. 4B). These experiments demonstrate that different quenching tryptophans are involved at different positions and illustrate the difficulties in assigning specific displacements to the fluorescent signals at each position.

Fig. 4.

Distinct tryptophans quench TMRM at two positions in M1–M2. (A) Mean ΔF/F–V curves for E120C (orange circles), E120C/W315A (red stars), E120C/W888A (green triangles), or E120C/W892A (blue diamonds). (B) Mean ΔF/F–V curves for D126C (orange circles), D126C/W315A (red stars), D126C/W888A (green triangles), or D126C/W892A (blue diamonds). Error bars are SEM. Number of experiments are in parentheses. To avoid bleaching effects each curve is the average of the first ΔF/F from each oocyte.

Engineered Quenchers to Evaluate Conformational Changes.

Due to the uncertainties of the cysteine-TMRM scan, we tried a different approach to further define M1–M2 movement. Instead of attaching TMRM to the moving region, we attached it to the putatively static region while engineering tryptophan quenchers into the putative moving parts. Since the loop between M9 and M10 appears static in the crystal structures, as they pertain to the transmembrane region that supports transport in the catalytic subunit (41), we reasoned that R977C might lack voltage-dependent fluorescent changes when tagged with TMRM. SI Appendix, Fig. S5, shows that this is indeed the case. Then, we evaluated the fluorescence response of tryptophan residues introduced, one at a time, at each position from R116 to L134. Robust state-dependent fluorescence signals were observed with D126W/R977C (Fig. 5). Representative traces show the ouabain-sensitive current (Fig. 5A) and fluorescence change (Fig. 5B) induced by pulses from −50 to −160 or +40 mV in the absence or presence of 10 mM ouabain. Because D126 is an important determinant of ouabain binding (42), D126W reduces ouabain affinity, and fluorescence signals remain in the presence of the high concentration of ouabain. Furthermore, ouabain unbinding can be seen at pulses that are more negative. This observation happens even though D126W was introduced in the D127N background, which has higher ouabain affinity than the D127 background used for the other tryptophan mutants (Methods).

Fig. 5.

Voltage-dependent reactions in D126W/R977C. (A) Ouabain-sensitive transient current and (B) fluorescent signal elicited by pulses to +40 mV (black and blue) or −160 mV (gray and red) in a TMRM-tagged oocyte expressing D126W/R977C-TMRM. Note incomplete inhibition by ouabain and unbinding of ouabain during negative voltage pulses (magenta line). (C) Rate (from exponential fits to traces like those in A and B as a function of voltage. (D) Ouabain sensitive Q–V (left axis, black squares) and ΔF/F–V (right axis) curves in Na⁺ (open circles) or in the presence of 10 mM ouabain (orange stars), fitted to Boltzmann distribution (line plots) with parameters Q_max= 21.2 nC, V_1/2 = 26.3 mV, kT/ez_q = 25.8 mV, ΔF/F_max= 11.5%, V_1/2 = −28.3 mV, and kT/ez_q = 32 mV in the presence of Na⁺ and ΔF/F_max = 4.3%, V_1/2 = −33 mV, and kT/ez_q = 40 mV in the presence of 10 mM ouabain.

The directionality of the fluorescence changes is congruent with the cross-linking results presented earlier. Quenching occurs at positive voltages (E2P, Fig. 5B, blue trace) and unquenching at negative voltages [E1P(3Na⁺), Fig. 5B, red trace]. The time courses of the fluorescence and current signals are much closer together at all voltages (Fig. 5C) than when TMRM was attached to D126C (SI Appendix, Fig. S3). The voltage dependencies for ΔF/F (Fig. 5D, orange circles, right axis) and charge movement (Fig. 5D, black squares, left axis) were described by Boltzmann distributions (line plots). The voltage dependencies of both the rate and steady-state fluorescence are shifted ∼50 mV to more negative voltages than the current signal. As shown below, this discrepancy appears to reflect incomplete labeling with TMRM and the altered function of TMRM tagged pumps as a similar discrepancy between the fluorescence and current curves was observed for all tryptophan mutants.

To test the effect of more complete labeling with TMRM, we incubated the R977C template for 2 h in 100 μM TMRM. This treatment reduces viability of the oocytes, increases background fluorescence, and reduces total charge and also altered the center of the Q–V. The center was V_1/2 = −124 ± 33 mV (Q_tot = 7.6 ± 6.7 nC, kT/z_qe = 43 ± 14 mV, n = 4) after 2 h incubation with 100 μM TMRM and V_1/2 = −81 ± 12 mV (Q_tot = 32.4 ± 9.7 nC, kT/z_qe = 46 ± 6 mV, n = 5) after incubation with 20 μM TMRM for 30 min. The approximately −40 mV shift in the Q–V suggests that labeling of only a fraction of the pumps with TMRM can explain the left shift of the ΔF/F–V (reflecting only transitions in labeled pumps) with respect to the Q–V curves (reflecting both labeled and unlabeled pumps, with a larger contribution of the latter).

To evaluate which positions reach close proximity to the fluorophore attached to R977C, tryptophan residues were introduced from Q116W to L134W. Clear voltage-dependent fluorescence changes were observed only at positions on M2, from P123W to L130W (Fig. 6A). The signals that became evident with P123 reached a maximum for D126W, in keeping with cross-linking results. As the tryptophan was moved inward through M2, the signal first decreases with D127W and then increases again reaching a second maximum between Y129W and L130W, exactly one helical turn below D126. This helical pattern is more obvious when fluorescence was normalized to the total charge moved (ΔF/F_max/Q_tot) to correct for expression level (Fig. 6B); such normalization was impossible for L130W because it was insensitive to ouabain, impeding charge movement determination. This helical pattern reflects that the first two residues after D126 face in suboptimal directions for quenching, while the 10-fold reduced amplitude for L130W compared to D126W indicates that the former, although on the same helix side, is farther away from the fluorophore than the latter. The result demonstrates that position 130 reaches sufficient proximity for quenching TMRM on the extracellular side, with the signals quickly fading away after L130W.

Fig. 6.

Helical pattern of ΔF/F_max. (A) ΔF/F_max from Boltzmann fits to the fluorescence signals of the M1-M2 tryptophan mutants showing voltage-dependent signals when TMRM was attached to R977C. (B) ΔF/F_max normalized to the maximal charge measured in the same oocyte when both determinations were possible. To correct for the reduced ouabain sensitivity of D126W, the fluorescence used for normalization was the ouabain-sensitive one (∼70% of the maximum; cf. D). Note the y-axes breaks and the presence of two maxima, one in D126W and the other between Y129W and L130W. Number of experiments is in parentheses. Error bars are SEM. No signals were observed with mutants from Q116W to E122W or from V132W to L134W (n = 5 to 14 for each position without signal).

Discussion

Our disulfide scanning results present compelling evidence that the M1–M2 region of the α-subunit is highly dynamic during the pump cycle and its movement with respect to M9–M10 loop is required for the Na⁺/K⁺ pump to complete a catalytic cycle. Cross-linking of D126C and R977C in E2P abolished both Na⁺/K⁺-pump current due to pump cycling (Fig. 2) and transient currents due to the E1P(3Na⁺) ↔ E2P + Na⁺_o conformational transition (SI Appendix, Fig. S2). These effects demonstrate that displacement of M2 is required for the Na⁺/K⁺ pump to undergo its major conformational change. The partial reduction of Na⁺/K⁺ pump current observed by spontaneous cross-linking between R977C and other residues located toward the middle of the first loop indicates that these residues reach close proximity with lower frequency than D126C but may also reflect that the middle of the first loop is a more flexible structure than the rigid M2 helix, meaning that some function could remain after cross-linking in the loop, as long as there is enough flexibility to displace the transmembrane α-helices.

We used VCF to further evaluate the movement of the M1–M2 region that we proposed based on the cross-linking results. VCF has been used to identify moving parts in voltage-dependent ion channels (43–46), the M5–M6 region of Na⁺/K⁺ and H⁺/K⁺ pumps (29, 47), and the Na⁺/K⁺ pump β-subunit (27). The presence of clear fluorescence changes when TMRM was attached to every position within M1–M2 (Table 1 and SI Appendix, Fig. S3) reveals the highly dynamic nature of this region. Overall, the VCF results are consistent with our cross-linking data and mechanistic models based on the closely related SERCA (48) and the gastric H⁺/K⁺ pumps (22). However, the finding that residues at positions 126 and 977 are in close proximity in E2P and move away in E1P(3Na⁺) (Fig. 2 D–F) contrasts with predictions based on available Na⁺/K⁺ pump structures (15, 17, 18, 20) since in these structures the α-carbons of D126 and R977 are 12 Å away in E2P and 8 Å away in E1P(3Na⁺) (Fig. 1C).

Tryptophan reduces fluorescence from tetramethylrhodamine and other fluorophores by collisional quenching. Therefore, a tryptophan residue must be at atomic proximity of the TMRM fluorophore in order to quench its fluorescence. Upon labeling with TMRM, E120C gives the largest fluorescence change reported so far with Na⁺/K⁺ pumps. Because we identified the main quenching tryptophans, we can use the available structures to rationalize our E120C data. TMRM adds ∼12.5 Å to the cysteine’s sulfhydryl (i.e., the TMRM molecule’s edge would be ∼15 Å away from the Cα). The large movement of the membrane transport core formed by M1–M6 (41) is obvious when the E1P(3Na⁺) and E2P structures are aligned at the so-called “support membrane domain” formed by M7–M10 (49) (Fig. 1 B and C). For W315, the distance from E120’s Cα to the tryptophan rings is 16 Å in E1 and 21 Å in E2, consistent with the observed quenching and with the strong reduction of the E120C-TMRM signal by the W315A mutation. In contrast, for W892, such distance is 14 Å in E1P(3Na⁺) and 10 Å in E2P. This apparent discrepancy may reflect the same crystallization issues making cross-linking between D126 and R977 incompatible with crystal structures’ distances, although this could also reflect movement between a near-optimal (14 Å) and a suboptimal (10 Å) quenching distance.

When TMRM is attached to cysteines introduced in highly dynamic regions, as it is commonly done, TMRM modification frequently alters the behavior of the protein (sometimes drastically like in D126C; SI Appendix, Fig. S4). Furthermore, comparison of fluorescence data obtained following TMRM treatment of mutants with cysteines at different positions is complicated by differences in labeling efficiency and distinct fluorescence quenchers (which have to be determined for each residue). Positions with inefficient labeling would have fluorescent signals from labeled pumps and currents arising mostly from unlabeled pumps, and positions with more efficient labeling would have current signals from two pump populations. These caveats provide a logical explanation for the high variability of voltage-dependent parameters (Table 1) and the discrepant rates between the fluorescence and current signals (SI Appendix, Fig. S3).

The uncertainties with the standard VCF approach described above were mitigated when we performed tryptophan scanning in the putative moving region, to evaluate quenching of TMRM attached to the static part of the protein (Fig. 6). The small twist to the usual VCF approach provides more reliable structural and dynamic information, and we think it will be useful in studies with other membrane proteins. Specifically, a cysteine or an azidohomoalanine (50) engineered on a putatively static region of the protein can be used to attach a fluorophore. The dynamics and structural patterns of moving regions can then be followed through introduction of tryptophan quenchers there. This strategy has at least three main advantages with respect to attaching TMRM to the moving part of the protein: First, the structural and kinetic effects of introducing the tryptophan in the dynamic region may be less drastic than those resulting from adding the larger TMRM (as we observed; e.g., compare Fig. 5 with SI Appendix, Fig. S3). Second, tryptophans can be added deeper into the transmembrane domains where a cysteine would be inaccessible for labeling with the fluorophore (e.g., L128–L134 in M2), providing information about regions that could not be probed by the standard method. Third, the TMRM labeling is expected to be the same for all tryptophan mutants because it is performed on the same target, which would facilitate comparison of the results. In keeping with the advantages described above, tryptophan scanning results of R977C labeled with TMRM (Fig. 6) are fully consistent with the cross-linking kinetics in the two major conformations (Fig. 2). In particular, the large fluorescence change in D126W, showing quenching in E2P and unquenching in E1P(3Na⁺) (Fig. 5), is the expected result given promotion of cross-linking between D126C and R977C under conditions where E2P occupancy is favored.

A stunning result revealed by tryptophan scanning is that every residue in the externalmost part of M2 induces conformation-dependent collisional quenching, only in E2P, also displaying the α-helical pattern of M2 expected from the structure (Fig. 6). Aside from being consistent with M2’s helical structure known from the crystal structure, this result suggests that the direction of the M2 movement cannot be explained by a simple rocking tilt of the whole protein, of the kind that has also been demonstrated for SERCA structures embedded on a lipid bilayer (51), or by rotation of M2 (in which cases some residues should quench in E2P and others in E1P, depending on the reorientation of a helix). Rather, these observations indicate a movement in and out of the membrane plane, in a direction perpendicular to the membrane. Thus, taken together, our data indicate that the externalmost part of M2 moves inwardly in E1P(3Na⁺) (as indicated by the tryptophan scan of TMRM tagged R977C) while changing its tilt (suggested by the results with TMRM attached to E120C but difficult to define due to the multiplicity and mobility of quenchers), in a way similar to proposals of E1–E2 conformational changes in SERCA (48) (SI Appendix, Fig. S1) and H⁺/K⁺ pumps (22). It should be noted that the directionality of M1–M2 movements perpendicular to the membrane plane suggested by the SERCA crystal structures is not obvious, and the crystals in the same state from different laboratories show differences in the position of the M1–M2 stalk, probably reflecting the difficulties generated by this highly dynamic region for crystal-structure determination [for instance, compare the E2BeF4 structures Protein Data Bank (PDB) ID 2ZBE (9) and PDB ID 3B9B (6) and the MgF₄ inhibited pump with the external/luminal gate closed PDB ID 1WPG (10) and PDB ID 1T5T (7)]. Although our approach provides directionality to these movements in a functional pump, our technique is unable to provide absolute distances due to uncertainties regarding the exact position of the TMRM fluorophore.

In conclusion, the results presented here, which study the dynamics of the M1–M2 stalk in a protein in its physiologic environment, demonstrate that the mechanism opening the external gate to release ions to the extracellular (or luminal) side, which requires significant movement of the M1–M2 stalk, is conserved across the P-type 2 ATPase subfamily. The contrast between our observations and the available crystal structures reemphasizes the need to complement studies from static crystal structures, obtained in the presence of inhibitors and after removal of lipids, with studies that provide structural and dynamic information of functional membrane proteins in their native environment.

Methods

Oocyte Preparation and Molecular Biology.

Oocytes were enzymatically isolated and maintained for up to a week at 16 °C until recording, as previously described (52). Oocytes were maintained in SOS solution (100 mM NaCl, 2 mM KCl, 1.8 mM CaCl₂, 1 mM MgCl₂, and 5 mM HEPES) supplemented with horse serum and antimycotic-antibiotic solution (Antibiotic-Antimycotic, Gibco) at 16 °C, for up to a week, until recording. Robust and stable Na⁺/K⁺ pump currents were observed in oocytes injected with functional Na⁺/K⁺ pumps, as previously described (30, 52, 53), indicating that long-term incubation of oocytes without nutrients has no effect on its ATP levels (∼5 mM; ref. 54). A total of 50 nL of cRNA mixture were injected (containing 1 to 2 μg/μL of α-cRNA and an equimolar amount of β-cRNA). For cross-linking experiments the injected cRNA was in vitro transcribed from cDNA of Xenopus Na⁺/K⁺ pump isoforms: the ouabain-resistant α1-mutant C113Y-α1 (C109Y in sheep α1 numbering; ref. 25) and the β3-subunit. For VCF we expressed Na⁺/K⁺ pump formed by the rat β1 and a sheep α1 isoform template. Although it has been reported that a fully cysteineless Na⁺/K⁺ pump can be expressed (55), the template used here (a generous gift from Thomas Friedrich, Technical University of Berlin, Berlin, Germany) has been previously studied with VCF (28, 30) and is only cysteineless in the extracellular surface by mutations C911S and C964A (56).

The sheep α1 template has reduced affinity for ouabain due to the double substitution Q116R/N127D. Upon identification of a dramatic reduction of ouabain affinity on mutants of D126 (D126C and D126W), we introduced these mutations on a background with the natural asparagine at 127. For simplicity, all numbering refers to sheep α1, four lower than the Xenopus α1, and three higher than the pig α1 used in several Na⁺/K⁺ pump structures. The reduced ouabain affinity of the templates used here allow us to selectively inhibit the endogenous oocyte pumps by preincubation with 1 μM ouabain, enabling exclusive measurement of signals from exogenous Na⁺/K⁺ pumps (52).

Solutions.

Before recording, oocytes were Na⁺-loaded, to saturate intracellular facing Na⁺-binding sites, by a 1-h incubation in a solution containing (in mM) 90 NaOH, 20 tetraethylammonium-OH, 0.2 EGTA, and 40 HEPES (pH 7.2 with sulfamic acid), supplemented with 1 μM ouabain. Extracellular solutions contained (also in mM): 133 methane sulfonic acid (MS), 10 HEPES, 5 Ba(OH)₂, 1 Mg(OH)₂, 0.5 Ca(OH)₂, titrated with 125 mM NMG⁺ (NMG⁺ solution) or 125 mM NaOH (Na⁺ solution). Intermediate Na⁺ concentrations were achieved by mixing NMG⁺ and Na⁺ solutions keeping osmolality constant. External K⁺ was added from a 450-mM K⁺-MS stock. The oxidizing agent Copper Phenanthroline (Cu:Phen) was freshly prepared each day from a 100-mM CuSO₄ and frozen 100-mM phenanthroline stock solution at a ratio of 1:2.

TMRM Labeling.

A 20-mM stock of TMRM from Molecular Probes or Santa Cruz Biotechnology was made in DMSO and stored at −20 °C in the dark. Oocytes were incubated with 20 μM TMRM in 125 mM Na⁺ for 30 min on ice and rinsed six times to wash free TMRM. Labeled oocytes were subsequently Na⁺-loaded and kept in the dark at 4 °C until the start of the experiment. A 15-min room temperature pretreatment with 10 μM dithiothreitol before TMRM incubation was used to increase labeling efficiency in the mutants that showed inconsistent labeling (A117C, A118C, P123C, Q124C, and D126C). For some experiments, shown in SI Appendix, Fig. S4, the oocytes were labeled in the TEVC chamber for 5 min, at room temperature, to record currents before and after labeling. This treatment indicates that adding TMRM-maleimide at room temperature leads to inhibition of some Na/K pumps, probably due to permeation and modification of functionally important intracellular cysteines in the N domain of the protein (57).

Electrophysiology and Fluorometry.

TEVC was performed at room temperature, with an OC-725C amplifier (Warner Instruments), a Digidata 1440 A/D board, a Minidigi 1A, and pClamp 10 software (Molecular Devices). Signals were filtered at 2 kHz and digitized at 10 kHz. Resistance of both microelectrodes (filled with 3 M KCl) was 0.5 to 1 MΩ. Fluorescence was acquired as described (30). Briefly, the electrophysiological equipment was mounted on an inverted microscope (Olympus IX 71) with a 40× oil-immersion objective (NA 1.3). Excitation was performed with a 530-nm LED controlled by a DC4100 controller (Thorlabs). Excitation light was filtered through a 535-nm excitation filter (Omega Optical). Emission was filtered through a 570-nm long-pass emission filer (Omega Optical) and was detected with a Photomax 200 (DAGAN Corporation).

Data Analysis.

The transient charge moved was measured by integrating the transient current either during the pulse (Q_ON; Table 1) and when the pulse was turned off (Q_OFF; all figures). The steady-state fluorescence signals (ΔF/F, normalized to the value at −50 mV) were measured averaging the last 20 ms of the 500-ms voltage pulse. Q–V and ΔF/F–V curves were fitted with a Boltzmann distribution:

$$\mathrm{Q}\hspace{0.17em}=\hspace{0.17em}{\mathrm{Q}}_{\mathrm{hyp}}+\hspace{0.17em}{\mathrm{Q}}_{\mathrm{tot}}/(1\hspace{0.17em}+\hspace{0.17em}\exp \left[{\mathrm{z}}_{\mathrm{q}}\mathrm{e}\left(\mathrm{V}-{\mathrm{V}}_{1/2}\right)/\mathrm{kT}\right],$$

$$\mathrm{\Delta }\mathrm{F}/\mathrm{F}\hspace{0.17em}=\hspace{0.17em}\mathrm{\Delta }\mathrm{F}/{\mathrm{F}}_{\mathrm{hyp}}+\hspace{0.17em}\mathrm{\Delta }\text{F}/{\mathrm{F}}_{\max }/(1\hspace{0.17em}+\hspace{0.17em}\exp \left[{\mathrm{z}}_{\mathrm{q}}\mathrm{e}\left(\mathrm{V}-{\mathrm{V}}_{1/2}\right)/\mathrm{kT}\right],$$

where Q_hyp and ΔF/F_hyp are the charge or fluorescence moved by hyperpolarizing pulses, Q_tot (|Q_hyp| from negative pulses + |Q_pos| from positive pulses) and ΔF/F_max are the total charge or fluorescence moving over the whole voltage axis, V_1/2 is the center of the distribution, z_q is the valence of a charged particle if it crossed the entire membrane electric field, e is the elementary charge, k is the Boltzmann constant, and T is the temperature. kT/z_qe is also referred to as the slope factor.

Fluorescence and transient-current signals during the pulse were fitted with exponential functions, starting 5 ms after the beginning of the pulse. All analyses were performed with pClamp and Origin (OriginLab). As indicated, error bars in all figures are SEM. Errors in Table 1 and the main text are SD. Statistical significance was assigned whenever t tests yielded t < 0.05.

Data Availability

All analyzed data have been included in the article and SI Appendix. Raw data are available upon request.