Involvement of F1296 and N1303 of CFTR in induced-fit conformational change in response to ATP binding at NBD2

Abstract

The chloride ion channel cystic fibrosis transmembrane conductance regulator (CFTR) displays a typical adenosine trisphosphate (ATP)-binding cassette (ABC) protein architecture comprising two transmembrane domains, two intracellular nucleotide-binding domains (NBDs), and a unique intracellular regulatory domain. Once phosphorylated in the regulatory domain, CFTR channels can open and close when supplied with cytosolic ATP. Despite the general agreement that formation of a head-to-tail NBD dimer drives the opening of the chloride ion pore, little is known about how ATP binding to individual NBDs promotes subsequent formation of this stable dimer. Structural studies on isolated NBDs suggest that ATP binding induces an intra-domain conformational change termed "induced fit," which is required for subsequent dimerization. We investigated the allosteric interaction between three residues within NBD2 of CFTR, F1296, N1303, and R1358, because statistical coupling analysis suggests coevolution of these positions, and because in crystal structures of ABC domains, interactions between these positions appear to be modulated by ATP binding. We expressed wild-type as well as F1296S, N1303Q, and R1358A mutant CFTR in Xenopus oocytes and studied these channels using macroscopic inside-out patch recordings. Thermodynamic mutant cycles were built on several kinetic parameters that characterize individual steps in the gating cycle, such as apparent affinities for ATP, open probabilities in the absence of ATP, open probabilities in saturating ATP in a mutant background (K1250R), which precludes ATP hydrolysis, as well as the rates of nonhydrolytic closure. Our results suggest state-dependent changes in coupling between two of the three positions (1296 and 1303) and are consistent with a model that assumes a toggle switch-like interaction pattern during the intra-NBD2 induced fit in response to ATP binding. Stabilizing interactions of F1296 and N1303 present before ATP binding are replaced by a single F1296-N1303 contact in ATP-bound states, with similar interaction partner toggling occurring during the much rarer ATP-independent spontaneous openings.

Figure 1.

Statistical coupling analysis predicts coevolution of the triad F1296-N1303-R1358. (A, a and b) Amino acid frequency distributions of sites 1 and 2 (corresponding to positions 1296 and 1303 in CFTR) in a multiple sequence alignment (MSA) of >10,000 ABC NBD sequences (http://pfam.sanger.ac.uk/family?PF00005). Phenylalanine and asparagine are the most frequent amino acids at sites 1 and 2, respectively. (c) Within the subset of sequences (n = 1,023) that contain glutamine at site 2, serine and alanine are prevalent at site 1. (d) In the subset (n = 1,364) that contains serine at site 1, glutamine is the most frequent residue at site 2. (B) Amino acid frequency distributions of site 3 (corresponding to position 1358 in CFTR), calculated for the total sequence alignment (a; arginine is found in 60% of sequences), and for the subsets that contain a glutamine at site 2 (b) or a serine at site 1 (c). (C) Ribbon diagram of model structure encompassing the entire α and part of the β subdomain of CFTRs NBD2 (residues 1250–1371; from Mornon et al., 2008). To illustrate their relative spatial positioning, conserved residues of the Walker A and B motif and the conserved glutamine in the Q loop are shown in blue stick representation, and residues of the triad studied here are highlighted in navy blue. Bound ADP is shown in magenta. (D and E) Ribbon diagrams of the segment of TAP1 corresponding to segment 1296–1358 of CFTR from the crystal structures of TAP1 solved in complex with MgADP (D; Gaudet and Wiley, 2001) and ATP (E; Procko et al., 2006). Residues corresponding to F1296, N1303, and R1358 in CFTR NBD2 are shown in stick representation (navy blue) and labeled site 1, 2, and 3, respectively. Dotted magenta lines are H bonds computed by Swiss Pdb Viewer v.3.7.

Figure 2.

Phosphorylation dependence is little affected by site-1 and site-2 mutations. (A–D) Inward chloride currents recorded in patches excised from resting oocytes expressing WT (A), F1296S (B), N1303Q (C), and F1296S/N1303Q (D) CFTR. In each case, the application of 2 mM ATP (bars) elicits only small currents relative to those activated by subsequent exposure to 300 nM PKA plus 2 mM ATP (bars). Note the rapid partial deactivation after the removal of PKA in A–D and the persistent channel activity after the removal of ATP in D (magnified in inset). Membrane potential was −80 mV in A–C, but −20 mV in D.

Figure 3.

A stabilizing interaction between sites 1 and 2 facilitates channel opening in the absence of ATP. (A) Representative traces of WT, F1296S, N1303Q, and F1296S/N1303Q currents illustrating segments in 0 mM ATP and bracketing segments in 2 mM ATP. Dotted lines show zero current level. (B) Estimation of P_o;max for WT (black), F1296S (red), N1303Q (blue), and F1296S/N1303Q (green) by stationary noise analysis. Each symbol plots the variance of macroscopic current fluctuations divided by the unitary current amplitude for a steady segment of recording in 2 mM ATP, as a function of the mean current. Open probabilities calculated for each individual segment were averaged to obtain final P_o;max estimates (refer to Materials and methods) for each construct. (C) P_o values in 0 mM ATP (P_o;bas), computed as the product of P_o;bas/P_o;max ratios (refer to Materials and methods) and P_o;max values (from B). (D) Thermodynamic mutant cycle built on P_o;bas/(1−P_o;bas) values; each corner is represented by the side chains at sites 1 and 2, respectively. ΔΔG⁰ values (mean ± SEM) on arrows show mutation-induced changes in the stability of the open state with respect to the closed state in the absence of ATP, and were used to calculate (refer to Materials and methods) the coupling energy for the site-1–site-2 interaction (ΔΔG_{int(open–closed in 0 mM ATP)}).

Figure 4.

The stabilizing site-1–site-2 interaction that facilitates channel opening in the absence of ATP is preserved in the ATP hydrolysis–deficient K1250R mutant. (A) Representative traces of K1250R, F1296S/K1250R, N1303Q/K1250R, and F1296S/N1303Q/K1250R currents illustrating segments in 0 mM ATP and bracketing segments in 2 mM ATP. Dotted lines show zero current level (determined for the triple mutant similarly to that in Fig. S2). (B) Estimation of P_o;max for K1250R (black), F1296S/K1250R (red), N1303Q/K1250R (blue), and F1296S/N1303Q/K1250R (green) by stationary noise analysis. (C) P_o values in 0 mM ATP (P_o;bas), computed as in Fig. 3 C. (D) Thermodynamic mutant cycle built on P_o;bas/(1−P_o;bas) values; notation as in Fig. 3 D.

Figure 5.

Energetic coupling between sites 1 and 2 changes between ATP-bound open and ATP-free closed states, but not between ATP-bound closed and open states. (A) Summary of P_o;max values for K1250R (black), F1296S/K1250R (red), N1303Q/K1250R (blue), and F1296S/N1303Q/K1250R (green) obtained from the data presented in Fig. 4 B. (B) Thermodynamic mutant cycle built on P_o;max/(1−P_o;max) values showing changes (mean ± SEM) in the stability of the open state with respect to the closed state in saturating ATP. (C) Time courses of macroscopic current decay upon sudden washout of 2 mM ATP (gray traces), and mono-exponential fit lines (color-coded as in A; the red and black fit lines overlap). The trace for WT (labeled), shown as a comparison, is fitted with a single exponential (magenta) with a time constant of 459 ms. (Inset) Mean (±SEM) closing time constants (τ_relax) obtained from 7–22 similar experiments for each construct. (D) Thermodynamic mutant cycle built on macroscopic relaxation rates (1/τ_relax).

Figure 6.

ATP binding affects energetic coupling between sites 1 and 2 in closed channels. (A) [ATP] dependence of macroscopic currents was assayed for WT (top left), F1296S (top right), N1303Q (bottom left), and F1296S/N1303Q (bottom right) channels by exposure to various test [ATP] bracketed by exposures to 2 mM ATP. (B) ATP-dependent current fractions (I−I₀)/(I_max−I₀) plotted as a function of [ATP] for WT (black), F1296S (red), N1303Q (blue), and F1296S/N1303Q (green). Each plot was fitted by the Michaelis-Menten equation (solid lines); predicted midpoints (K_Po) are shown in the inset. (C) Estimates of K_rCO for each construct, calculated (refer to Materials and methods) using K_Po from B and P_o;max from Fig. 3 B. (D) Thermodynamic mutant cycle built on K_rCO values.

Figure 7.

Removal of the arginine side chain at site 3 affects channel gating regardless of the residue at site 2. (A) Representative traces of R1358A and R1358A/N1303Q currents illustrating segments in 0 mM ATP and bracketing segments in 2 mM ATP. Dotted lines show zero current level. (B) Estimation of P_o;max for WT (black), R1358A (red), N1303Q (blue), and R1358A/N1303Q (green) by stationary noise analysis. Estimated P_o;max was 0.62 ± 0.05 for R1358A and 0.36 ± 0.04 for R1358A/N1303Q. (C) P_o values in 0 mM ATP (P_o;bas), computed as in Fig. 3 C. (D) Thermodynamic mutant cycle built on P_o;bas/(1−P_o;bas) values; notation as in Fig. 3 D. (E) ATP-dependent current fractions (I−I₀)/(I_max−I₀) plotted as a function of [ATP] for WT (black), R1358A (red), N1303Q (blue), and R1358A/N1303Q (green). Each plot was fitted by the Michaelis-Menten equation (solid lines); predicted midpoints (K_Po) are shown in the inset. (F) Estimates of K_rCO for each construct, calculated (refer to Materials and methods) using K_Po from E and P_o;max from B. (G) Thermodynamic mutant cycle built on K_rCO values.

Figure 8.

The intra-NBD2 induced fit upon ATP binding is associated with a toggle switch rearrangement of interactions between sites 1 and 2. (A) Cartoon representation of Scheme 2 with an example set of rates suitable to explain the gating of WT CFTR (black rates on arrows). Green, NBD1; blue, NBD2; cyan, TMD; yellow, ATP. Positions 1 and 2 within NBD2 are denoted by yellow letters, and stabilizing interactions among these and unidentified positions X and Y are represented by yellow connecting lines. The two rates assumed to be changed by the F1296S/N1303Q double mutation, and by the K1250R mutation, are shown in red and magenta, respectively, below the WT rates. (B) Table summarizing parameters P_o;bas and K_Po predicted by Scheme 2 for WT (using the rates in black in A) and F1296S/N1303Q (using the two rates in red in A), as well as P_o;max and τ_relax for K1250R and F1296S/N1303Q/K1250R (using the rates printed in magenta for steps C₄→O₂ and O₂→C₁). Predicted parameters were calculated using standard Q-matrix techniques. For comparison, the corresponding measured parameters are printed underneath in parentheses.