High-Throughput Determination of Exchange Rates of Unmodified and PTM-Containing Peptides Using HX-MS

Abstract

Despite the widespread use of MS for hydrogen/deuterium exchange measurements, no systematic, large-scale study has been conducted to compare the observed exchange rates in protein-derived, unstructured peptides measured by MS to the predicted exchange rates calculated from NMR-derived values and how neighboring residues and post-translational modifications influence those exchange rates. In this study, we sought to test the accuracy of predicted values by performing hydrogen exchange measurements on whole cell digests to generate an unbiased dataset of 563 unique peptides derived from naturally occurring protein sequences. A remarkable 97% of observed exchange rates of peptides are within two-fold of predicted values. Using fully deuterated controls, we found that for approximately 50% of the peptides, the amino acid sequence and, consequently, the intrinsic exchange rate, are the primary contributors to back exchange. A meta-analysis of the remaining physicochemical properties of peptides revealed multiple features that contribute either positively or negatively to back exchange discrepancies. Employing our workflow for comparable measurements on synthetic peptide mixtures containing post-translational modifications, and their unmodified counterparts, we show that lysine acetylation has a strong effect on the observed exchange rate, whereas serine/threonine phosphorylation does not. Our automated workflow enables high-throughput determination of exchange rates in complex biological peptide mixtures with diverse properties.

Keywords: acetylation; cell digest; exchange rates; hydrogen exchange; mass spectrometry; phosphorylation.

Graphical abstract

Fig. 1

Experimental and analysis workflow. A, sample generation workflow B, intensities of proteins identified with a latest generation nanoLC-MS/MS using a 2-h gradient (grey, 6000 proteins) and the HX-UPLC-MS/MS using a 6-min gradient (red, 235 proteins). Intensities are ranked from highest to lowest. C, Labeling workflow of the Jurkat digest D, FASTA generation workflow using non-deuterated and fully deuterated samples.

Fig. 2

Comparisons between calculated and measured exchange rates. A, example D-uptake plot of DSYVGDEAQSK (+2) and heatmap containing all peptides that passed quality control in our dataset, ranked by the median value per peptide. Confidence intervals are drawn by calculating Welch’s unequal variances t-test using all of the replicates available for the best charge state of the peptide. B, distributions of the different methods (absolute values log10 transformed protection factors and root sum of square deviations, RSSD) used to compare calculated and observed curves and respective scatter plots (see Experimental Procedures for further details). Strong positive correlations between the absolute log10 PFs and RSSDs were calculated using Pearson (R = 0.92) and Spearman (ρ = 0.82) correlations. Sequences of outlier peptides with PF and RSSD values outside the 97.5 percentile are shown. Sequences of outlier peptides detected with both methods are shown in black. Sequences of outliers detected only using PF and RSSD values are shown in grey and light blue respectively. Carbamidomethylated cysteines are colored red.

Fig. 3

Back-exchange correction workflow.A, sequence cartoon depicting the amide bonds in a peptide and boxplots showing BE values upon correction for D-loss from the N-term, first amide, last amide and middle amides based on Bai, Connely, Nguyen (see Main Text and Experimental Procedures for calculation details). B, BE corrections for example peptides (top) and proline-rich peptides (bottom) C, boruta result plots for N-term and Fully corrected BE values. Yellow and green boxplots represent z-scores of tentative and confirmed attributes, respectively. Blue boxplots correspond to the maximum z-score of a shadow attribute.

Fig. 4

Effects of K acetylation and S,T phosphorylation on exchange rates. A, example peptide sequences of the non-modified and acetylated peptide pairs B, example D-uptake plots for peptides belonging to groups (A–C), before and after FD corrections; statistically significant differences are marked with an asterisk (Welch’s test). Confidence intervals are drawn by calculating a Welch’s unequal variances t test using all the replicates available for the best charge state of the peptide. C, Bird’s eye view of D-uptake differences before and after FD corrections. D, calculated PF (log2) for the non-modified and acetylated or phosphorylated peptide pairs following FD corrections. Statistically significant differences (paired t test) are detected for the acetylated peptide pairs only (p < 0.0001). E, Site-specific determination of amide exchange kinetics using ETD for GFALVGVGSEASSKK (+3).