Improving cross-linking mass spectrometry methodologies

Research regarding protein interactions has been a productive area of study over many decades, contributing to our understanding of biology and health and hence to the development of therapies. Since most traditional methods are biased towards purified proteins with a fixed structure or a relatively low molecular mass, novel methods are required to monitor protein interactions in a wider range of situations. Cross-linking mass spectrometry (XL-MS) is emerging as a widely applicable technique that captures dynamic situations in protein complexes, and can readily be applied to endogenous levels of protein due to its high sensitivity. Since digested peptides are measured rather than the intact protein complex, XL-MS can theoretically be applied to proteins of any size. While an XL-MS experiment is relatively straight-forward in terms of experimental design, identification of cross-linked peptides via mass spectrometry remains challenging. Difficulties in optimising the ideal settings during such measurements arise from the variability of biological samples containing cross-linked peptides. We therefore aim to build a library of synthetic cross-linked peptides that will be used for systematic optimisation of MS parameters used in XL-MS studies. A further complication lies in the bioinformatics approaches that are used to identify cross- linked peptides from the MS data because two peptides need to be correctly identified rather than just one, as in other proteomics experiments. Additionally, it is never known exactly which cross-linked peptides are present in the digest of a cross-linked protein, since non- specific cross-links cannot be fully eliminated. We will therefore use the data collected in this study to create a dataset suitable for testing and optimising search engines that have been developed for the identification of cross-links from MS data. The use of synthetic peptides is particularly applicable in this case since the library can be controlled for true positives and true negatives, which is not possible upon analysis of a complex tryptic digestion mixture. We will also develop a standardised cross-linking experiment to track improvements in the full XL-MS workflow upon optimisation of MS parameters. The standardised experiment will be based on cross-linking of the Cas9 protein. Such a standardised is imperative for the cross-linking community to compare the quality of data obtained from different methodologies. Improvements that will be made to the XL-MS technique as a result of this research will enable more protein interaction sites to be discovered and characterised, adding to the biological impact of such experiments and ensuring that cross-linked peptides are correctly identified.

Proteins are molecular machines inside living cells that drive biological processes such as growth, metabolism and homeostasis. Protein function is determined by its three-dimensional shape and by its mode of interaction with other proteins. Solving the structure of important proteins and multi-protein complexes is therefore an important area of research termed 'structural biology', and tools are being constantly developed to address protein systems that are difficult to study. Challenging systems are often tackled with 'integrative' approaches, in which data collected with complementary techniques are combined to generate structural models. Crosslinking mass spectrometry (XLMS) is a key player in integrative methods, and is used to define interacting sites within a protein complex. In a crosslinking experiment a reagent forms a bond between sites in close spatial proximity, the protein complex is broken down into short strings called peptides, and mass spectrometry is used to measure which peptides are bound to each other, thus revealing interacting sites. Many search engines have been developed to convert raw mass spectrometry data into meaningful crosslinking results. However, without the availability of 'ground truth' data to verify results, there was uncertainty in the reliability of such data interpretation strategies. To address this shortcoming, we created libraries of synthetic crosslinked peptides that have the unique advantage of containing crosslinks that are known to be true, and those that are known to be false. We used these libraries as 'ground truth' samples to test the sensitivity and accuracy of several popular crosslinking softwares, and provided the raw data in an online repository as a resource for the scientific community. In our study, we used the mass spectrometry data relating to measurement of the crosslinked peptide library to benchmark the reliability of popular search engines. We filtered the results to an estimated false discovery rate of 5% (i.e. it's predicted by the software that 5% of identified crosslinks will be false). We then counted the number of correct and incorrect crosslinks given by each search engine, which allows us to determine the 'calculated' false discovery rate. The software named pLink identified the highest number of correct crosslinks (217) but with 28 incorrect crosslinks, giving rise to a calculated false discovery rate of 11%, which is higher than the 5% that is estimated. Xi and StavroX identified fewer correct crosslinks (186 and 180, respectively), but have calculated false discovery rates of 5%, as estimated. This research provides an excellent resource for the crosslinking community. The data can be used to assess the many algorithms developed for XL identification. This provides valuable information regarding the performance of crosslink search algorithms, allowing selection of suitable data analysis strategies. The data will also be invaluable in the development of crosslink search engines.