Application of millisecond time-resolved solid state NMR to the kinetics and mechanism of melittin self-assembly

Abstract

Common experimental approaches for characterizing structural conversion processes such as protein folding and self-assembly do not report on all aspects of the evolution from an initial state to the final state. Here, we demonstrate an approach that is based on rapid mixing, freeze-trapping, and low-temperature solid-state NMR (ssNMR) with signal enhancements from dynamic nuclear polarization (DNP). Experiments on the folding and tetramerization of the 26-residue peptide melittin following a rapid pH jump show that multiple aspects of molecular structure can be followed with millisecond time resolution, including secondary structure at specific isotopically labeled sites, intramolecular and intermolecular contacts between specific pairs of labeled residues, and overall structural order. DNP-enhanced ssNMR data reveal that conversion of conformationally disordered melittin monomers at low pH to α-helical conformations at neutral pH occurs on nearly the same timescale as formation of antiparallel melittin dimers, about 6 to 9 ms for 0.3 mM melittin at 24 °C in aqueous solution containing 20% (vol/vol) glycerol and 75 mM sodium phosphate. Although stopped-flow fluorescence data suggest that melittin tetramers form quickly after dimerization, ssNMR spectra show that full structural order within melittin tetramers develops more slowly, in ∼60 ms. Time-resolved ssNMR is likely to find many applications to biomolecular structural conversion processes, including early stages of amyloid formation, viral capsid formation, and protein-protein recognition.

Keywords: melittin; protein folding; self-assembly; solid-state NMR.

Fig. 1.

(A) Conceptual representation of the conversion of 4 unfolded melittin monomers to a melittin tetramer (Protein Data Bank ID 2MLT) after a rapid pH jump. When melittin is ¹³C-labeled at all carbon sites of Gly3, Leu6, and Ile20, helix formation leads to intramolecular ¹³C–¹³C distances less than 4.5 Å between Gly3 and Leu6, while tetramer formation leads to intermolecular ¹³C–¹³C distances less than 4.0 Å between Leu6 and Ile20. (B and C) CD spectra of 300 μM melittin in 20% (vol/vol) glycerol with 75 mM NaHPO₄ as a function of pH at 20 °C and as a function of temperature at pH 7.0.

Fig. 2.

(A) Apparatus for rapid mixing and freeze-trapping. A copper plate (1) is precooled by immersion in a liquid-nitrogen bucket (2), and then rotated continuously by a motor (3). Two solutions are pumped through a mixer (4) that sweeps across the rotating plate under the control of a stepper motor (5). A jet of mixed solution freezes on the rotating plate after traveling a variable distance d_f that determines the structural evolution time τ_e. After rotation stops, the plate is immersed in liquid nitrogen. The frozen solution is scraped from the plate and packed into a MAS rotor (6) for low-temperature ssNMR measurements. (B) DNP-enhanced, DQ-filtered ¹³C ssNMR spectra of frozen melittin solutions with the indicated values of τ_e. The dashed blue and red lines indicate ¹³CO and ¹³C_α chemical shifts for isotopically labeled residues at τ_e = ∞. Spectra were acquired at 30 K, 100.8-MHz ¹³C NMR frequency, and 7.00-kHz MAS frequency.

Fig. 3.

(A) DNP-enhanced 2D ¹³C ssNMR spectra of frozen melittin solutions with the indicated values of τ_e, obtained with a ¹³C–¹³C spin diffusion period τ_SD = 25 ms. The dashed lines indicate the ¹³C_α chemical shifts for Gly3, Leu6, and Ile20 in the fully folded tetrameric state at pH 7.0 (i.e., τ_e = ∞). Contour levels increase by successive factors of 1.3. (B) Dependence on τ_e of the coefficients c_q of the 6 PC spectra, from singular value decomposition of the aliphatic regions of the 6 experimental 2D spectra. (C) Comparison of the dependences of the singular values s_q on q for the experimental 2D spectra, for spectra reconstructed with only the first 2 PCs of the experimental spectra, and for spectra reconstructed with only the first 2 PCs of the experimental spectra plus additional random noise. A logarithmic scale is used in the Inset.

Fig. 4.

(A and B) DNP-enhanced 2D ¹³C ssNMR spectra of frozen melittin solutions with τ_SD = 25 ms and τ_SD = 1.0 s, respectively, and with τ_e = 4.6 ms. Signal amplitudes are displayed on continuous color scales. The numbered regions contain interresidue Gly3–Leu6 (1) and Leu6–Ile20 (3) cross-peak signals or intraresidue Leu6 (2) and Ile20 (4) cross-peak signals. (C) Dependences on τ_e of interresidue cross-peak volume ratios (V₁/V₂ for Gly3–Leu6; V₃/V₄ for Leu6–Ile20, where V_m is the volume within region m). Ratios from spectra with τ_SD = 25 ms are subtracted from those with τ_SD = 1 s to correct for overlap between intraresidue and interresidue cross-peaks. Error bars are derived from the rms noise in the 2D spectra. The color-coded dashed lines are least-squares fits to the form $A-B\exp \left(-{\mathrm{\tau }}_e/{\mathrm{\tau }}_{bu}\right)$, with indicated values of τ_bu. (D) Dependence on τ_e of the coefficient of the second PC for 2D spectra with τ_SD = 25 ms, from Fig. 3B. The dashed line is a least-squares fit to the form $A-B\exp \left[-{\left({\mathrm{\tau }}_e/{\mathrm{\tau }}_{bu}\right)}^{\mathrm{\beta }}\right]$, with indicated values of τ_bu and β.