Solid-state NMR chemical shift analysis for determining the conformation of ATP bound to Na,K-ATPase in its native membrane

Structures of membrane proteins determined by X-ray crystallography and, increasingly, by cryo-electron microscopy often fail to resolve the structural details of unstable or reactive small molecular ligands in their physiological sites. This work demonstrates that 13C chemical shifts measured by magic-angle spinning (MAS) solid-state NMR (SSNMR) provide unique information on the conformation of a labile ligand in the physiological site of a functional protein in its native membrane, by exploiting freeze-trapping to stabilise the complex. We examine the ribose conformation of ATP in a high affinity complex with Na,K-ATPase (NKA), an enzyme that rapidly hydrolyses ATP to ADP and inorganic phosphate under physiological conditions. The 13C SSNMR spectrum of the frozen complex exhibits peaks from all ATP ribose carbon sites and some adenine base carbons. Comparison of experimental chemical shifts with density functional theory (DFT) calculations of ATP in different conformations and protein environments reveals that the ATP ribose ring adopts an C3′-endo (N) conformation when bound with high affinity to NKA in the E1Na state, in contrast to the C2′-endo (S) ribose conformations of ATP bound to the E2P state and AMPPCP in the E1 complex. Additional dipolar coupling-mediated measurements of H–C–C–H torsional angles are used to eliminate possible relative orientations of the ribose and adenine rings. The utilization of chemical shifts to determine membrane protein ligand conformations has been underexploited to date and here we demonstrate this approach to be a powerful tool for resolving the fine details of ligand–protein interactions.


Supplementary Information: the selective HCCH experiment
The usual broadband HCCH experiments on uniformly 13 C molecules excite double quantum (DQ) coherences between bonded pairs of carbons to measure HCCH torsional angles sequentially (Figure S3a).Here, a selective HCCH experiment (Figure S3b) was used to excite only the 13 C DQ coherence between non-bonded C8 and C1' by adjusting the MAS frequency to the n = 1 rotational resonance condition, at which the MAS frequency is set to the difference in the chemical shifts of the two carbons (5226 Hz at an applied magnetic field B 0 of 9.3 T).The DQ coherence is measured after evolution for a period t of up to one cycle of sample rotation, t R , under the influence of the local dipolar field of the bonded protons, H8 and H1'.The evolution of DQ coherence, measured as a difference intensity S(t), is sensitive to the effective torsional angle, H8 -C8 -C1' -H1', and the effective bond angle C8 -C1' -H1', both of which vary according to  (Figure S4).S(t) is also dependent on the effective bond angle C1' -C8 -H8, which does not vary with  and is taken to be 126°.Ideally, a series of S(t) values are measured from t = 0 s to t = t R and the evolution compared with simulated curves to obtain values of .However, the sensitivity of the HCCH experiment is very low and each spectrum required acquisition of 230,000 transients with block averaging to achieve suitable signal-to-noise.Therefore only two spectra were obtained, at t = 0 and at t = 0.5t R to obtain the normalized value of S 0.5 (defined in Figure S5a), which is shown to vary with the value of  according to Figure 5b and is highest for positive values of . Figure 3c of the main text shows HCCH NMR spectra at t = 0 and at t = 0.5t R overlaid with simulated peaks scaled according to Figure 5a.A value of S 0.5 = 0.12 ± 0.10 was determined, the uncertainty reflecting the signal-to-noise.the effective torsional angle H8 -C8 -C1' -H1' (164°), effective bond angle C8 -C1' -H1' (129°) and effective bond angle H8 -C8 -C1' (96°).These values apply to a torsional angle  of -44°.The dipolar coupling between bonded 1 H -13 C pairs was calculated to be -21300 kHz, based on an average internuclear separation of 1.12 Å, and a scaling factor  of 0.45 was applied.The value of  was obtained by calibration with a crystalline leucine sample of known geometry.A MAS frequency of 5226 Hz, corresponding to n = 1 rotational resonance with respect to C8 and C1' was assumed.
(b) Variation of S(0.5) as a function of .The shaded region represents a measured value of S(0.5) of 0.13 ± 0.10.The red portion of the continuous line represents the values of  that are consistent with the measured value of S(0.5).

Figure S2 .
Figure S2.Two structures of the NKA -subunit in E1 conformations complexed with non-hydrolysable ATP analogues.

Figure S3 .
Figure S3.Two HCCH SSNMR pulse sequences.(a) HCCH with broadband excitation of DQ coherence, as described elsewhere.This experiment typically excites DQ coherences between directly bonded pairs of carbons in a uniformly 13 C labeled molecule, which evolves under frequency-switched Lee-Goldberg (FSLG) proton homonuclear decoupling for one rotor period before conversion back into observable magnetization.It is usually run as a two-dimensional experiment to resolve individual DQ coherences.(b) The HCCH with selective excitation of DQ coherence at rotational resonance.This experiment can in principle excite DQ coherence between any pair of carbons in a uniformly 13 C labeled molecule, depending on the difference in their chemical shifts, which is matched to the MAS frequency.This experiment was used to determine the torsional angle .

Figure S4 .
Figure S4.Comparison of experimental 13 C chemical shifts with values calculated with Gaussian using PBE and B3LYP functionals, for AMPPCP bound to NKA in the E1 conformation (left) and in the E2P conformation (right).(a) Calculated shift values.(b) Correlation of experimental values and values calculated with the B3LYP hybrid functional.RMSD values are given for the B3LYP calculations and PBE calculations (in brackets).All calculated values are based on the same binding site clusters taken from PDB 7WYU and PDB 7Y46 as shown in Figure 2b of the main text.

Figure S6 .
Figure S6.Evolution of DQ coherence in the selective HCCH NMR experiment.(a) DQ evolution over one cycle of sample rotation at the magic angle, calculated according to the molecular geometry of ATP in Figure 4a of the main text.The DQ intensity is the difference in the integrals measured from the resonances for C8 and C1', normalized to the value at t = 0 s.The values of S(0.0) and S(0.5) were used to calculate the simulated peak intensities in Figure 5c of the main text.Input values are: the effective torsional angle H8 -C8 -C1' -H1' (164°), effective bond angle C8 -C1' -H1' (129°) and effective bond angle H8 -C8 -C1' (96°).These values apply to a torsional angle  of -44°.The dipolar coupling between bonded 1 H -13 C pairs was calculated to be -21300 kHz, based on an average internuclear separation of 1.12 Å, and a scaling factor  of 0.45 was applied.The value of  was obtained by calibration with a crystalline leucine sample of known geometry.A MAS frequency of 5226 Hz, corresponding to n = 1 rotational resonance with respect to C8 and C1' was assumed.(b)Variation of S(0.5) as a function of .The shaded region represents a measured value of S(0.5) of 0.13 ± 0.10.The red portion of the continuous line represents the values of  that are consistent with the measured value of S(0.5).