Research - Kim lab at McGill U

Improved personalized medicine by drug partner discovery via a novel proteome-wide gene overexpression perturbation platform

Throughout my career, I have focused on the generation and integration of high-throughput data to propose novel hypotheses for experimental validation or new clinical biomarkers. In the Kim lab at Roswell Park Comprehensive Cancer Center, we have a broad research interest in developing and utilizing novel platforms to generate various omics data to solve clinical problems, such as drug resistance in cancer chemotherapy.

Our lab adapts two state-of-the-art tools to perform high-throughput overexpression screens to look for novel players in human disease. The ORFeome library, constructed by Drs. David E. Hill, Marc Vidal, and Frederick P. Roth’s groups, is a collection of barcoded ORFs that are cloned into a Gateway compatible system. The newest version hOFReome 9.1 (Collaboration and The ORFeome Collaboration, 2016; Luck et al., 2020) includes ~90% of all human protein-coding genes, allowing for an unbiased proteome-wide screening for searching candidate biomarkers and therapeutic targets.

The second system, the Bxb-landing pad system (Matreyek et al., 2017, 2020), was constructed by Dr. Doug Fowler’s lab (Matreyek et al., 2017, 2020) and improved by Dr. Frederick P. Roth’s group to deliver exogenous DNA to a designated safe harbor locus in transfectable mammalian cell lines. Bxb landing pad system enables us to select gene-inserted cells with FACS for making every cell in the recombined pool express a single gene. Therefore, when we do an en masse transfection of our ORFeome collection, we are able to express “one gene per cell” with an artificial promoter to achieve “homogenous gene expression”.

Current projects at the Kim lab

Using gene overexpression phenotypes to discover drug partners that eliminate therapeutic resistance

Mechanisms for therapeutic resistance often depend on the regulation of gene overexpression. The current "One Treatment Fits All" approach has limitations due to cancer heterogeneity and does not address an individual's distinct response to therapy (Hamburg and Collins, 2010). Combinatorial drug use is limited in current personalized medicine approaches, but the combinatorial space is much larger and each patient may be treated with a ‘unique fingerprint' of drug combinations to overcome resistance.

We use our overexpression screening platform to discover novel pathways whose upregulation induces anti-cancer drug resistance in various cancer types. By identifying the ORFeome gene whose overexpression alters sensitivity to a particular drug, we can find causal genes for therapeutic resistance. We can propose a novel combinatorial therapy using a second drug that can suppress resistance, or induce toxicity. Moreover, we are interested in the application of an overexpression perturbation system to find novel developmental drivers or regulators of extracellular vesicles’ biogenesis.

Exploring disease mutants that cause a gain of protein-protein-interactions

Mutations rewiring PPI networks are known to be important in cancer(Bowler et al., 2015)and other diseases. Comprehensively investigating mutations in PPI-mediating domains may be a new approach to find treatments for cancers and rare diseases. We are interested in applying proteomics approaches to detect changes in PPIs of selected human disease variants to address how these variants alter the interactome of the protein. We seek to explore phenotypic effects caused by novel gain-of-interactions, and ultimately discover drugs that can inhibit the interaction.

DK’s biography

Systematic analysis of extracellular vesicle

After developing a protein sequence comparison program (Yang et al., 2011) for my undergraduate thesis using my expertise in bioinformatics, I turned my interest to extracellular vesicles (EVs) during my Ph.D.protein–protein training. I profiled the contents of extracellular vesicles via mass spectrometry-based proteomics (Choi et al., 2018, 2020, 2011, 2012, 2015a; Lee et al., 2009, 2013, 2015; Luck et al., 2020; Minciacchi et al., 2015), lipidomics (Kim et al., 2013, 2015a, 2015b), and microarray/next-generation sequencing-based transcriptomics(Chelakkot et al., 2018; Choi et al., 2014, 2015b; Kang et al., 2013; Lässer et al., 2017; Lunavat et al., 2015; Yoon et al., 2014). Using proteome profiles of various EVs, through collaboration with experimental biologists, we found that EVs can work as angiogenic factors(Yoon et al., 2014), transporters for beta-lactamase to share antibiotic resistance(Lee et al., 2013), protectors against colitis(Kang et al., 2013), and providers of insulin resistance(Choi et al., 2015b). Furthermore, I built an integrative database of extracellular vesicles, EVpedia, which contains i) the extracellular vesicle-related publications and researchers, ii) the data repository of extracellular vesicle components (protein, mRNA, miRNA, lipid, metabolite, and glycan), and iii) the bioinformatics tools to analyze the data in the database(Kim et al., 2013, 2015a, 2015b).

Proteome-wide screening of protein-protein interaction

In parallel, during my Ph.D. training, I carried out proteome-scale screens for interactors of SIRT6/7(Lee et al., 2014), REST/NRSF(Lee et al., 2016), and VRK1/3(Lee et al., 2017) by affinity-purification mass spectrometry (AP-MS) and explored the relatedness of interaction partners to aging, neuronal differentiation, and the cell cycle of liver cancer, respectively.

Subsequently, during my postdoctoral training, I have been expanding my expertise by constructing and analyzing an all-by-all protein-protein interaction map. First, I co-led the Human Reference Interactome (HuRI) project published at Nature (Luck et al., 2020), yielding ~53K biophysical protein-protein interactions (PPIs) identified by yeast-two hybrid screening covering most of the human protein-coding genes (~18K out of 20K). By integrating transcriptome and proteome data, we exploited several cellular context-specific PPI networks to generate and support several hypotheses: i) a new cell-death protein, OTU6DA, validated by a viability assay in HeLa cells, ii) candidate extracellular vesicle recruiters mediate trafficking of interaction partners into EVs, confirmed with CRISPR knockout in U373vIII cells, and iii) dominant-negative function of an uncharacterized short isoform of NCK2 during brain development, confirmed in zebrafish.

To overcome the limitation of HuRI and other Y2H maps as ‘static’ networks, assessed under a single environment, we engineered the fluorescence-Barcode Fusion Genetics-Yeast Two-Hybrid (fBFG-Y2H) method. In this approach, we can simultaneously query >2 million protein pairs in a quantitative way using fluorescence-activated cell sorting (FACS) and next-generation sequencing. Applying fBFG-Y2H to nearly all possible yeast protein pairs, we generated high-quality interaction maps under four conditions: baseline, poor carbon source, oxidative stress, and DNA damage, and have called this the Conditional Yeast Reference Interactome (CYRI). Using stringent thresholds, we discovered ~2,500 total interactions, of which ~1K are environment-specific. Our network uncovered ‘contextual hubs’ which gain or lose many interaction partners exclusively in specific environments. Contextual networks showed enriched connectivity between proteins with relevant functions while revealing dense subnetworks that enable functional predictions. To further investigate these environment-specific PPIs, we combined phosphoproteomics and RNA-seq to investigate potential mechanisms of environment-dependent regulation. For example, the analysis revealed that homodimerization of URA7 could be regulated by phosphorylation on Ser424 in a poor carbon source environment. Thus, we provide the first proteome-wide maps of PPIs across multiple environments, enabling understanding of principles that underlie global cellular response to changing contexts.

Recently, we succeeded in cloning all coding sequences of SARS-CoV-2 (Kim et al., 2020) by gene synthesis, and facilitated the sharing of these clones (>5K times). A consortium of ~10 labs together started protein and genetic interaction screening for SARS-CoV-2, of which my part was to lead the “working hand” level of trainees and staff(Laurent et al.; Samavarchi-Tehrani et al.; St-Germain et al., 2020). By fBFG-Y2H and classical Y2H, we identified ~200 virus-host protein interactions, which provided us with a novel biological hypothesis of viral ORF6’s function as a protein degrader (Kim et al., 2021).