published_or_final_version / Biological Sciences / Doctoral / Doctor of Philosophy

Not available

Protein-protein interactions are central to their biological functions in cells. Many approaches have been applied to study protein-protein interactions in a genomic-scale. In an attempt to develop new strategies to study protein-protein interactions, FRET by using ECFP and EYFP as the donor and receptor was evaluated for possible application in protein-protein interaction study in a high-throughput fashion. Due to the intrinsic properties of ECFP and EYFP, FRET-based protein-protein interaction assay is not suitable for large-scale studies. Instead, tandem affinity purification coupled with mass spectrometry approach proved to be a useful strategy to identify protein interacting partners. Several transcription factor complexes in yeast were successfully purified and novel components in the complexes were identified by combining a shotgun mass spectrometry approach and a differential analysis of the mass spectrometry data. In particular, a negative regulator of G1 to S phase transition during cell cycle, Whi5p, was identified to be a component of SBF complex; a regulator of nitrogen metabolism, Gln3p, was identified to be a component of Hap2/3/5 complex that regulates carbon metabolism, suggesting a crosstalk between nitrogen and carbon metabolism. Additionally, one-step purification coupled with shotgun mass spectrometry analysis was applied to simplify and improve the affinity purification approach used for protein-protein interaction studies. In order to map protein complexes in their native state, a sucrose density gradient was used to separate protein complexes in cells. The proteins within each fraction from the sucrose density gradient were analyzed and quantified with mass spectrometry to obtain the protein abundance profiles across the gradient. The known protein complexes were identified by clustering the protein abundance profiles. This method could possibly be improved to become a generic approach to mapping protein complexes. The goal of protein-protein interaction studies is to determine the protein functions. In an effort to identify ribosome biogenesis genes from a yeast gene network reconstructed from diverse large-scale interaction data sets, at least 25 new ribosome biogenesis genes were confirmed by extensive experimental validations, underscoring the value of proteinprotein interaction studies and gene interaction network.

Genetic mapping of complex traits has been successful over the last decade, with over 2,000 regions in the genome associated to disease. Yet, the translation of these findings into a better understanding of disease biology is not straightforward. The true promise of human genetics lies in its ability to explain disease etiology, and the need to translate genetic findings into a better understanding of biological processes is of great relevance to the community. We hypothesized that integrating genetics and protein- protein interaction (PPI) networks would shed light on the relationship among genes associated to complex traits, ultimately to help guide understanding of disease biology. First, we discuss the design, testing and implementation of a novel in silico approach (“DAPPLE”) to rigorously ask whether loci associated to complex traits code for proteins that form significantly connected networks. Using a high-confidence set of publically available physical interactions, we show that loci associated to autoimmune diseases code for proteins that assemble into significantly connected networks and that these networks are predictive of new genetic variants associated to the phenotypes in question. Next, we study variation in the electrocardiographic QT-interval, a heritable phenotype that when prolonged is a risk factor for cardiac arrhythmia and sudden cardiac death. We show that a large proportion of QT-associated loci encode proteins that are members of complexes identified by immunoprecipitations in mouse cardiac tissue of proteins known to be causal of Mendelian long-QT syndrome. For several of the identified proteins, we show they affect cardiac ion channel currents in model organisms. Using replication genotyping in 17,500 individuals, we use the complexes to identify genome-wide significant loci that would have otherwise been missed. Finally, we consider whether PPIs can be used to interpret rare and de novo variation discovered through recent technological advances in exome-sequencing. We report a highly connected network underlying de novo variants discovered in an autism trio exome-sequencing effort, and we design, test and implement a novel statistical framework (“DAPPLE/SEQ”) to analyze rare inherited variants in the context of PPIs in a way that significantly boosts power to detect association.

Presented in this dissertation are studies of protein dynamics and protein/protein interactions using solution phase hydrogen/deuterium exchange in combination with mass spectrometry (HXMS). In addition, gas phase fragmentation behaviors of deuterated peptides are investigated, with the purpose of increasing resolution of the HXMS. In the area of single protein dynamics, two protein systems are studied. Studies on the cytochrome c2 from Rhodobacter capsulatus indicate its domain stability to be similar to that of the horse heart cytochrome c. Further comparison of the exchange kinetics of the cytochrome c2 in its reduced and oxidized state reveals that the so-called hinge region is destabilized upon oxidation. We also applied a similar approach to investigate the conformational changes of photoactive yellow protein when it is transiently converted from the resting state to the signaling state. The central β-sheet of the protein is shown to be destabilized upon photoisomerization of the double bond in the chromophore. Another equally important question when it comes to understanding how proteins work is the interactions between proteins. To this end, two protein complexes are subjected to studies by solution phase hydrogen deuterium exchange and mass spectrometry. In the case of LexA/RecA interaction, both proteins show decreases in their extents of exchange upon complex formation. The potential binding site in LexA was further mapped to the same region that the protein uses to cleave itself upon interacting with RecA. In the sHSP/MDH system, hydrogen/deuterium exchange experiments revealed regions within sHSP-bound MDH that were significantly protected against exchange under heat denaturing condition, indicative of a partially unfolded state. Hydrogen/deuterium exchange therefore provides a way of probing low resolution protein structure within protein complexes that have a high level of heterogeneity. Finally, the feasibility of increasing resolution of HXMS by gas phase peptide fragmentation is investigated by using a peptide with three prolines near the C-terminus. Our data show that deuterium migration indeed occurs during the collision activated dissociation process. Caution is required when interpreting the MS/MS spectra as a way of pinpointing the exact deuterium distribution within peptides.

Many protein-protein interactions, especially those involved in eukaryotic signalling, are mediated by PDZ domains through the recognition of hydrophobic C-termini. The availability of experimental PDZ interaction data sets have led to the construction of computational methods to predict PDZ domain-peptide interactions. Such predictors are ideally suited to predict interactions in single organisms or for limited subsets of PDZ domains. As a result, the goal of my thesis has been to build general predictors that can be used to scan the proteomes of multiple organisms for ligands for almost all PDZ domains from select model organisms. A framework consisting of four steps: data collection, feature encoding, predictor training and evaluation was developed and applied for all predictors built in this thesis. The first predictor utilized PDZ domain-peptide sequence information from two interaction data sets obtained from high throughput protein microarray and phage display experiments in mouse and human, respectively. The second predictor used PDZ domain structure and peptide sequence information. I showed that these predictors are complementary to each other, are capable of predicting unseen interactions and can be used for the purposes of proteome scanning in human, worm and fly. As both positive and negative interactions are required for building a successful predictor, a major obstacle I addressed was the generation of artificial negative interactions for training. In particular, I used position weight matrices to generate such negatives for the positive only phage display data and used a semi-supervised learning approach to overcome the problem of over-prediction (i.e. prediction of too many positives). These predictors are available as a community web resource: http://webservice.baderlab.org/domains/POW. Finally, a Bayesian integration method combining information from different biological evidence sources was used to filter the human proteome scanning predictions from both predictors. This resulted in the construction of a comprehensive physiologically relevant high confidence PDZ mediated protein-protein interaction network in human.

Many protein-protein interactions, especially those involved in eukaryotic signalling, are mediated by PDZ domains through the recognition of hydrophobic C-termini. The availability of experimental PDZ interaction data sets have led to the construction of computational methods to predict PDZ domain-peptide interactions. Such predictors are ideally suited to predict interactions in single organisms or for limited subsets of PDZ domains. As a result, the goal of my thesis has been to build general predictors that can be used to scan the proteomes of multiple organisms for ligands for almost all PDZ domains from select model organisms. A framework consisting of four steps: data collection, feature encoding, predictor training and evaluation was developed and applied for all predictors built in this thesis. The first predictor utilized PDZ domain-peptide sequence information from two interaction data sets obtained from high throughput protein microarray and phage display experiments in mouse and human, respectively. The second predictor used PDZ domain structure and peptide sequence information. I showed that these predictors are complementary to each other, are capable of predicting unseen interactions and can be used for the purposes of proteome scanning in human, worm and fly. As both positive and negative interactions are required for building a successful predictor, a major obstacle I addressed was the generation of artificial negative interactions for training. In particular, I used position weight matrices to generate such negatives for the positive only phage display data and used a semi-supervised learning approach to overcome the problem of over-prediction (i.e. prediction of too many positives). These predictors are available as a community web resource: http://webservice.baderlab.org/domains/POW. Finally, a Bayesian integration method combining information from different biological evidence sources was used to filter the human proteome scanning predictions from both predictors. This resulted in the construction of a comprehensive physiologically relevant high confidence PDZ mediated protein-protein interaction network in human.

The baculovirus chitinase (CHIA) and cathepsin protease (V-CATH) enzymes cause terminal host insect liquefaction, thereby enhancing dissemination of progeny virions in nature. Regulated and delayed cellular release of these host tissue-degrading enzymes ensures liquefaction starts only after optimal viral replication has occurred. Baculoviral CHIA remains intracellular due to its C-terminal KDEL endoplasmic reticulum (ER) retention motif. However, the intracellular processing and trafficking of the baculovirus v-cath expressed cathepsin (V-CATH) is poorly understood and a mechanism for cellular retention of the inactive V-CATH progenitor (proV-CATH) has not been determined. The cathepsins of Autographa californica multiple nucleoplyhedrovirus (AcMNPV) and most other group I alphabaculoviruses have well-conserved chymotrypsin cleavage (Y11) and myristoylation sites (G12) suggestive of proteolytic cleavage to generate proV-CATH, and subsequent acylation which could promote membrane anchoring in order to foster cellular retention of the protein. Proteolytic iii N-terminal processing of baculoviral procathepsin was determined by fusing HA epitope-coding tags to the 5’ and/or 3’ ends of v-cath, indicating that the gene is expressed as a pre-proenzyme. However no evidence for myristoylation of proV-CATH was found, suggesting that another mechanism is responsible for retaining proV-CATH in cells. Prior evidence suggested that CHIA is a proV-CATH folding chaperone and that lack of chiA expression causes proV-CATH to become insoluble and unable to mature into V-CATH enzyme. A putative CHIA chaperone activity for assisting in proV-CATH folding implies that proV-CATH and CHIA interact in the ER of infected cells. Fluorescence microscopy demonstrated co-localization of CHIA-GFP and proV-CATH-RFP fusion proteins in the ER. An mRFP-based bimolecular fluorescence complementation (BiFC) assay helped to determine not only that AcMNPV proV-CATH interacts directly with the full-length viral CHIA, but also that it independently binds to the N-terminal chitin-binding domain (CBD) and C-terminal active site domain (ASD) of CHIA, in the ER during virus replication. Moreover, reciprocal Ni/HIS pull-downs of HIS-tagged proteins confirmed the proV-CATH interactions with CHIA, or with the CBD and ASD biochemically. The reciprocal co-purification of proV-CATH with all three polypeptides (CHIA, CBD, ASD) suggests proV-CATH specifically interacts with each of them, and corroborates the BiFC data. Furthermore, CHIA KDEL deletion allowed for premature secretion of not only CHIA but also of proV-CATH, suggesting that the CHIA/proV-CATH interaction in the ER aids cellular retention of proV-CATH. In contrast to prior reports, it was also determined that CHIA is iv dispensable for correct folding of proV-CATH since proV-CATH produced by a chiA-deficient virus was soluble, prematurely secreted from cells and could mature into V-CATH enzyme. Taken together, these data indicate that the viral chitinase plays a major role in ensuring that proV-CATH is neither prematurely secreted nor activated to V-CATH enzyme.

E2A-PBX1 is expressed as a consequence of a recurring chromosomal translocation seen in 5% of acute lymphoblastic leukemia cases. We recently reported that substitution of a leucine residue (L20A) within the N-terminal transcriptional activation domain (AD1) of E2A-PBX1 markedly impairs binding to the KIX domain of CBP/p300 and, importantly, leukemia induction in a mouse bone marrow transplantation model. Since both the protein-protein interaction and consequent leukemogenesis rely on a focal contact point and might therefore be susceptible to antagonism by small molecules, we devised a cell-free assay based on fluorescence anisotropy (FA) to detect binding of a fluorescently labeled peptide derived from AD1 of E2A-PBX1 (FITC-E2A) with recombinantly expressed KIX domain. The optimized FA assay reveals a dissociation constant of 2 µM for the wild-type interaction and correctly detects disruption of the complex by naphthol AS-E phosphate, a compound previously shown to antagonize KIX binding. The optimized FA assay was used to screen the Prestwick, Spectrum and Chembridge libraries containing 12400 compounds in total. Of the initial 43 positive hits from the libraries, 10 caused a reproducible decrease in FA. Since intrinsic small molecule fluorescence can produce false positive results in the FA-based screen, intrinsically fluorescent compounds were excluded from further analysis unless they could be shown to bind to KIX. Two hits, L1 and C2, were intrinsically fluorescent but demonstrated KIX interactions and one hit, P9, was not intrinsically fluorescent. These three compounds were tested for their ability to inhibit binding of a larger portion of E2A (residues 1 to 483) to full length CBP in a pull down assay with only compound P9 demonstrating efficacy. Further characterization of P9 by NMR showed no binding to KIX, however evaluation by FA showed binding to FITC-E2A with a 20 µM affinity. A cell-based cytotoxicity assay demonstrated that compound P9 was slightly more toxic on leukemic cells that express E2A-PBX1, compared to leukemic cells lacking E2A-PBX1 expression. Mammalian two-hybrid analysis did not provide details of the effects of P9 on the E2A:KIX interaction. We expect the identification of a novel compound, P9, capable of disrupting the oncogenic E2A-PBX1:CBP interaction, to guide the development of effective, less toxic leukemia drugs and provide new tools for elucidating the molecular mechanisms of leukemia induction by E2A-PBX1. / Thesis (Master, Pathology & Molecular Medicine) -- Queen's University, 2009-08-31 11:13:19.517

The E2A proteins are transcription factors critical for B-lymphopoiesis. A chromosomal translocation involving the E2A gene promotes acute lymphoblastic leukemia (ALL) through expression of the oncoprotein E2A-PBX1. Two activation domains of E2A-PBX1, AD1 and AD2, have been implicated in transcription mediated by recruitment of the transcriptional co-activator CBP/p300. A motif has been identified within AD1 that is important for recruiting CBP/p300, known as PCET. This recruitment requires an interaction between the activation domains of E2A-PBX1 and the KIX domain of CBP/p300. The KIX domain recognizes a generic ΦXXΦΦ sequence (Φ corresponds to a hydrophobic residue) found in the activation domains of numerous transcription factors. Mutation of leucine 20 in PCET has been shown to abrogate ex vivo immortalization of murine bone marrow and oncogenesis in a murine bone marrow transplantation model. A similar sequence is also found in AD2 and is implicated in E2A transcriptional activity and recruitment of CBP/p300. The structural details of these interactions remain largely unknown. NMR spectroscopy was used to determine the solution structure of the PCET:KIX complex, and the functional consequences of the Leu20Ala mutation were structurally rationalized. Other PCET mutations informed by this structure were tested and correlations were found between in vitro binding affinities and both transcriptional activation and immortalization. The binding site of the ΦXXΦΦ-containing E2A AD2 peptide was mapped to the same site on the KIX domain used by the PCET motif. A model of this complex was generated and mutations were tested using a similar approach as was used for PCET. E2A AD2 binds the KIX domain with lower affinity than the PCET motif and is not required for immortalizing bone marrow. A mutation that increases the affinity of E2A AD2 for the KIX domain to levels approaching that seen for the PCET:KIX interaction restores transcriptional activation and immortalization, demonstrating that immortalization by E2A-PBX1 is an affinity dependent process involving the KIX domain of CBP/p300. These studies indicate that the activation domains of E2A-PBX1 serve to support the in vivo function of the oncoprotein and that the PCET:KIX complex is a potential target for novel therapeutics in E2A-PBX1+ leukemia. / Thesis (Ph.D, Biochemistry) -- Queen's University, 2012-09-13 13:30:48.848