Fast and Accurate Disulfide Bridge Detection

Abstract

Recombinant expression of proteins, propelled by therapeutic antibodies, has evolved into a multibillion dollar industry. Essential here is the quality control assessment of critical attributes, such as sequence fidelity, proper folding, and posttranslational modifications. Errors can lead to diminished bioactivity and, in the context of therapeutic proteins, an elevated risk for immunogenicity. Over the years, many techniques were developed and applied to validate proteins in a standardized and high-throughput fashion. One parameter has, however, so far been challenging to assess. Disulfide bridges, covalent bonds linking two cysteine residues, assist in the correct folding and stability of proteins and thus have a major influence on their efficacy. Mass spectrometry promises to be an optimal technique to uncover them in a fast and accurate fashion. In this work, we present a unique combination of sample preparation, data acquisition, and analysis facilitating the rapid and accurate assessment of disulfide bridges in purified proteins. Through microwave-assisted acid hydrolysis, the proteins are digested rapidly and artifact-free into peptides, with a substantial degree of overlap over the sequence. The nonspecific nature of this procedure, however, introduces chemical background, which is efficiently removed by integrating ion mobility preceding the mass spectrometric measurement. The nonspecific nature of the digestion step additionally necessitates new developments in data analysis, for which we extended the XlinkX node in Proteome Discoverer to efficiently process the data and ensure correctness through effective false discovery rate correction. The entire workflow can be completed within 1 h, allowing for high-throughput, high-accuracy disulfide mapping.

Keywords: EThcD; FAIMS; MAAH; XlinkX/PD; disulfide bridge.

Graphical abstract

Fig. 1

MAAH digestion properties.A, cleavage settings denoted as [NAME= MAAH], [MISSED CLEAVAGES= 20], and [SPECIFICITY= all amino acids] in ExPASy PeptideCutter notation. Below are the examples of theoretical peptides. B, sequence coverage obtained with peptides of lysozyme C grouped on first position only. Heat colors denote the number of peptide-spectrum-matches at each peptide (ranging from black (n = 1) → red (n = 100) → yellow (n = 250)). Positions of Cys residues involved in canonical disulfide bridges highlighted in yellow. Signal peptide in gray letters inside a gray box. C, peptide lengths for lysozyme C for different hydrolysis times. D, effect of protease specificity on the number of theoretical peaks (or search space) generated by the XlinkX/PD search engine. MAAH, microwave-assisted acid hydrolysis; XlinkX/PD, XlinkX node embedded in Proteome Discoverer.

Fig. 2

Mass spectrometry optimizations to support disulfide bridge detection.A, ETD-driven gas-phase disulfide bridge reduction. B, EThcD MS/MS spectrum of disulfide-bridged peptide TPEVTCVVVDVSHE-GKEYKCKVSN originating from acid hydrolysis of trastuzumab with diagnostic peaks (yellow) and backbone fragmentation (blue and red). In the annotations, α denotes peptide A and β peptide B. C, effect of FAIMS integration into the acquisition investigated for trastuzumab as a function of signal-to-noise on signal-to-noise ratios of precursor ions and numbers of MS/MS spectra, peptide-spectrum-matches, and CSMs. CSM, crosslink spectrum match; ETD, electron transfer dissociation; EThcD, electron transfer higher energy dissociation; FAIMS, field asymmetric ion mobility spectrometry.

Fig. 3

Data processing steps and their results in XlinkX/PD. All data were originating from acid hydrolysis of trastuzumab. A, workflow for processing RAW data with the challenging conditions imposed by MAAH digestion and producing the final crosslink spectrum match (CSM) and crosslink tables. Linear peptide search in yellow; cross-linked peptide search in brown. B, mass accuracy of precursor m/z values after recalibration in Proteome Discoverer. The boundaries (dotted red line) are estimated with interquartile range fences. C, sequence coverage of peptide RTPEVTCVVVDVSHED with c-ions obtained during the step “search peptide A.” Each line is a single fragmentation spectrum, 218 identifications in total. D, the rank of the correct peptide A after sorting on score with or without considering the diagnostic peaks in EThcD MS/MS. E, effectivity of the implemented FDR approach on CSM and crosslink level for trastuzumab. FDR, false discovery rate; EThcD, electron transfer higher energy dissociation; MAAH, microwave-assisted acid hydrolysis.

Fig. 4

Detection of disulfide bridge scrambling in MAAH digested chicken lysozyme C.A, detected disulfide bridges as a function of the used conditions (top time, bottom temperature). B, summed intensities of the detected disulfide bridges and linear peptides. MAAH, microwave-assisted acid hydrolysis.

Fig. 5

Disulfide mapping of relevant proteins.A, schematic of trastuzumab with the canonical disulfide bridges indicated. B, all canonical disulfide bridges were detected (numbering in UniProt, for EU numbering subtract 3) and found to have near 100% occupancy rates, except for the disulfide bridges 214-223 and 370-428 (middle panel). C, structure of trastuzumab with the disulfide bridges highlighted (inset: zoom-in on the structure around 370–428). D, schematic of integrin alpha-IIb with the disulfides indicated in the domains. E, all canonical disulfide bridges were detected and found to have high occupancy rates. F, structure of integrin alpha-IIb with the disulfide bridges highlighted. Top right, comparison of the occupancy rates of this study and those obtained by Pijning et al (52). Bottom right, zoom-in on structure around Cys504-Cys515, which showed a low occupancy rate).