Recalibrating Protection Factors Using Millisecond Hydrogen/Deuterium Exchange Mass Spectrometry

Abstract

Hydrogen/deuterium exchange mass spectrometry (HDX-MS) is a powerful technique to interrogate protein structure and dynamics. With the ability to study almost any protein without a size limit, including intrinsically disordered ones, HDX-MS has shown fast growing importance as a complement to structural elucidation techniques. Current experiments compare two or more related conditions (sequences, interaction partners, excipients, conformational states, etc.) to determine statistically significant differences at a number of fixed time points and highlight areas of changed structural dynamics in the protein. The work presented here builds on the fundamental research performed in the early days of the technique and re-examines exchange rate calculations with the aim of establishing HDX-MS as an absolute and quantitative, rather than relative and qualitative, measurement. We performed millisecond HDX-MS experiments on a mixture of three unstructured peptides (angiotensin, bradykinin, and atrial natriuretic peptide amide rat) and compared experimental deuterium uptake curves with theoretical ones predicted using established exchange rate calculations. With poly-dl-alanine (PDLA) commonly used as a reference,^, we find that experimental rates are sometimes faster than theoretically possible, while they agree much better, and are never faster, with the fully unstructured trialanine peptide (3-Ala). Molecular dynamics (MD) simulations confirm the high helical propensity of the longer and partially structured PDLA peptides, which need as few as 15 residues to form a stable helix and are therefore not suitable as an unstructured reference. Reanalysis of previously published data by Weis et al. at 100 mM NaCl however still shows a discrepancy with predictions based on 3-Ala in the absence of salt, highlighting the need for a better understanding of salt effects on exchange rates. Such currently unquantifiable salt effects prevent us from proposing a comprehensive, universal calibration framework at the moment. Nevertheless, an accurate recalibration of intrinsic exchange rate calculations is crucial to enable kinetic modeling of the exchange process and to ultimately allow HDX-MS to move toward a direct link with atomistic structural models.

Figure 1

Hydrogen–deuterium exchange of angiotensin, bradykinin, and ANP. The experimental fractional uptake data (black) are compared with theoretical deuterium uptake calculated using the intrinsic exchange rate calculations from Englander using 3-Ala (green) or PDLA (red) as reference. The error associated with the experimental measurements is the pooled standard deviation.

Figure 2

Structural propensity of PDLA peptides of increasing length from molecular dynamics simulations. (A) The helical propensity, calculated by using the DSSP algorithm and averaged over the amino acids of the peptide, is shown as a function of the peptide length. The error bars associated with the helical propensity are the standard deviations. Snapshots were taken every 100 ps of the simulation. (B, C) Snapshots of a double helical bundle from the simulations of PDLA with 24 residues (B) and a triple helical bundle for PDLA with 32 residues (C).

Figure 3

Experimental HD exchange data of unstructured peptides (black) previously published by Al-Naqshabandi and Weis is compared with the deuterium uptake calculated by us, using PDLA (red) or 3-Ala (green) as reference and accounting for the corrections published in ref (31). Data points were extracted from figures published in ref (23) using a plot digitizer, and a default error of ±0.1 Da was assigned to experimental measurements. Mass increase is shown on a linear time scale instead of fractional uptake on a logarithmic time scale to facilitate direct comparison with the original paper. The cis-Pro assumption is required for the peptide FKPGI.

Figure 4

Fractional uptake of angiotensin, bradykinin, and ANP in the absence (black) or presence of 150 mM NaCl (blue). The error associated with experimental measurements is the pooled standard deviation. Experimental data are fitted with a stretched exponential model, and the observed rates k_obs are reported.