Abstract

Cellular function is closely tied to protein-protein interactions. Mapping these on a large scale, therefore, provides fundamental knowledge about the regulation and structure of biological systems. With the onset of proteomics, the use of affinity purification coupled to mass spectrometry (MS) has become the major tool to map protein interactions. Already twenty years ago, researchers endeavored to build interaction maps of model organisms such as yeast. However, previous large-scale interaction studies in Saccharomyces cerevisiae date back more than ten years, covered only about half of all genes, and made use of non-quantitative MS and tandem-affinity purification strategies. These approaches were limited by harsh purification protocols and required large amounts of cell lysate. Additionally large false positive and negative rates hampered their use as a fully reliable source for network studies. Building on recent improvements in sensitivity and speed of MS technology and the introduction of the concept of ‘affinity enrichment coupled to MS,’ I developed a fast, robust, and highly reproducible workflow for proteome-wide interaction studies. I applied and optimized the approach for a first full screen in S. cerevisiae. The workflow starts from only a few hundred µg of proteins per pull-down and is performed entirely in 96-well format, including cell growth, lysis, and affinity enrichment of GFP-tagged proteins. To increase sample throughput and minimize MS idle time between injections, I turned to the high throughput Evosep One liquid chromatography system. This allowed me to obtain data on 60 baits per day. The system is coupled online to a timsTOF Pro mass spectrometer capable of fragmenting over 100 peptides per second using the parallel accumulation – serial fragmentation (PASEF) technology. This combination of miniaturization and standardization ensured high sample throughput, sensitivity, and robustness. Altogether, I successfully performed over 4150 pull-downs and completed more than 8300 measurements for the yeast interactome using this next-generation workflow, all in less than 20 weeks of mass spectrometer running time. The dataset has a very high success rate for pull-downs. The near-complete coverage of expressed proteins in our study enabled a novel two-dimensional analysis strategy that efficiently scores interactions. We examined well-known protein complexes, which confirmed very high data quality. Although the yeast interactome has been studied by large-scale methods for decades, the majority of interactions were novel compared to known high-quality interaction databases. Among many striking novel discoveries - I found compelling evidence for interactions between the conserved chromatin remodeler SWI/SNF and SPX-domain-containing plasma-transporters. Using the common GFP-tag for quantification of protein abundance confirmed that our workflow covers a wide range of cellular protein abundances down to a few copies per cell. Redefining the yeast interactome with very high data quality and completeness enabled the study of its fundamental network properties that have been controversially discussed over many years. In total, our protein-protein interaction network encompasses about 4,000 proteins connected via about 30,000 interactions. A full browsable web application is accessible at yeast-interactome.org and allows (sub-) network exploration, interactor validation via volcano plots and correlation maps, and sample quality control. In a collaboration with the CZ Biohub, we set out to implement the mass spectrometry pipeline developed here to an interaction screen with CRISPR GFP-tagged human HEK293T cells. The reduced sample amount allowed us to screen cell cultures grown in 12-well plates for high throughput. The interaction and localization results of 1,311 processed interactomes in biological triplicates can be accessed at opencell.czbiohub.org.