Toward protein NMR at physiological concentrations by hyperpolarized water-Finding and mapping uncharted conformational spaces

Abstract

Nuclear magnetic resonance (NMR) spectroscopy is a key method for determining the structural dynamics of proteins in their native solution state. However, the low sensitivity of NMR typically necessitates nonphysiologically high sample concentrations, which often limit the relevance of the recorded data. We show how to use hyperpolarized water by dissolution dynamic nuclear polarization (DDNP) to acquire protein spectra at concentrations of 1 μM within seconds and with a high signal-to-noise ratio. The importance of approaching physiological concentrations is demonstrated for the vital MYC-associated factor X, which we show to switch conformations when diluted. While in vitro conditions lead to a population of the well-documented dimer, concentrations lowered by more than two orders of magnitude entail dimer dissociation and formation of a globularly folded monomer. We identified this structure by integrating DDNP with computational techniques to overcome the often-encountered constraint of DDNP of limited structural information provided by the typically detected one-dimensional spectra.

Fig. 1.. NMR and DDNP of MAX.

(A) Sketch of the experimental concept. Upon injection of hyperpolarized water, the target protein is diluted, while at the same time, the signal loss is (over)compensated by exchange between the water and the target, which introduces hyperpolarized protons into the latter. (B) Signal decay (bulk integrals) of the ¹H^N resonance of MAX after dilution. Directly after injection, turbulences obscure the spectra (gray bar). After 1 s, the turbulences settle, and signal-amplified spectra become detectable. After 10 s, the signal amplitude has decayed below the noise level. The experiment was repeated two times. The error bars indicate the maximum deviation between the repetitions. (C) Comparison of skyline projections on the ¹H dimension of an HSQC (in purple) and a TROSY (in blue) experiment detecting ¹⁵N-labeled MAX (0.3 mM) with a hyperpolarized MAX spectrum (first detected spectrum) at near-physiological concentration (1 μM). The latter was recorded using an adapted BEST-HMQC pulse sequence. The spectra at higher concentrations can be considered fingerprints of the folded dimer (25°C) and the unfolded state (37°C). The spectrum at low concentrations matches neither of these cases, indicating the presence of a third conformational state. (D) ¹⁵N-edited ¹H spectra of ¹⁵N-labeled MAX at different dilutions detected on an 11.7-T NMR spectrometer in thermal equilibrium (purple and blue) compared to the hyperpolarized ¹⁵N-edited ¹H spectrum (magenta; average over the first 20 detected spectra). The black straight lines trace the position of significative peaks from the thermal equilibrium to the hyperpolarized spectrum. The dashed boxes indicate significantly boosted signals in the NMR experiment that remained undetected in the thermal equilibrium spectrum. The DDNP experiment highlights almost invisible signals in the spectral envelope obtained in the standard experiments. a.u., arbitrary units.

Fig. 2.. Integration of MD simulations with the DDNP spectra.

(A) Structural transition observed for the MAX monomer in a 550-ns MD simulation in a 15 nm³ large box. Stripping the well-documented dimer of one subunit provided the starting structure. After 150 ns, a transition into a globular, tightly folded conformation could be observed. (B) Development of the hydration radius R_h of MAX in the MD simulations over time. After ~150 ns, MAX converts from the initial elongated shape into a compact conformation. The dashed box indicates the part of the simulation that we considered equilibrated and used for chemical shift predictions. (C) Zoom onto the R_h trajectory and the last 30 ns of the simulation. All conformations used for chemical shift prediction stem from the last 20 ns of simulation, highlighted in pink. (D to F) Comparison of the hyperpolarized spectrum (S) (pink line; average over the first 20 detected spectra) and the spectrum predicted on the basis of the MD simulations (blue bars) of the MAX monomer for three independent runs. The gray shade indicates side-chain resonances not included in the chemical shift calculations. The good match between the experimental and the calculated spectrum supports our conclusion that the monomer dominates the conformational ensemble at low concentrations. (G) “Negative control” comparing the experimental hyperpolarized spectrum with a spectrum predicted from the elongated MAX dimer structure. The mismatch shows that the experiment does not reflect this structure. (H) Negative control comparing the experimental hyperpolarized spectrum with a spectrum predicted for a random-coil state. The mismatch shows that the experiment does not reflect this structure either.