Genetic interactions describe how the presence of one gene affects the function of a second gene. Negative genetic interactions arise when two mutations together cause a stronger growth defect than expected based on the individual single mutations, while positive genetic interactions correspond to cases where the double mutant is either no sicker (epistatic) or healthier (suppressive) than the sickest single mutant.

Our lab has developed a high-throughput technology in budding yeast for quantitative analysis of the entire spectrum of genetic interactions, termed epistatic miniarray profile (E-MAP). The genetic profiles generated by this method provide highly specific readouts that can be used to identify genes that are functionally related. In this dissertation I describe my work related to the employment and extension of the E-MAP technology. With collaborators, I have interrogated several biological processes using E-MAPs, including the phosphorylation network, RNA processing, lipid metabolism, and the plasma membrane. We further extended the technology to facilitate analysis of higher-order genetic interactions by examination of triple mutants, and we developed an adaptation of E-MAP to function in E. coli.

My main focus was on the development of an important advance of the E-MAP technology, which allows us to address higher levels of complexity by examining genetic interactions of point mutant alleles of multi-functional genes. The technique, termed point mutant E-MAP (or pE-MAP), greatly increases the achievable resolution as it allows assignment of genetic relationships to individual residues and domains. We applied this system to genetically dissect RNA polymerase II (RNAPII) in budding yeast, and generated a pE-MAP comprising ~60,000 quantitative genetic interactions. Using these data, we assigned functions to RNAPII sub-domains and uncovered connections to protein complexes. The pE-MAP further allowed us to characterize connections related to RNAPII activity. These include an inverse relationship between in vitro transcription rate and in vivo splicing efficiency, classification of fast and slow mutants that shift transcription start upstream and downstream, respectively, and identification of Sub1 as a positive transcription factor that regulates start site selection and influences splicing. The pE-MAP approach provides a powerful strategy to understand other multi-functional machines at amino acid resolution.

Viruses are obligate parasites, unable to replicate outside of the host to which they are adapted. The adaptation of viruses to their accustomed host cell milieu is exquisite, contacting hundreds or thousands of host proteins in order to hijack host machinery and avoid antiviral defenses. Identifying the key functional interactions between virus and host is a critical step towards interfering with viral replication, as implicated host proteins are attractive and often highly-tractable therapeutic targets. This identification remains challenging, especially as it is best done directly in the primary cells or tissues in which the virus typically replicates. We have built on recent developments using CRISPR-Cas9 ribonucleoproteins that allowed perturbation of genomic sequences in primary human CD4+ T cells to functionally interrogate HIV-human interactions, identifying 86 that significantly alter HIV infection, including 44 not previously reported and 24 that harbor restrictive activity. We sequenced each knockout locus to illuminate the cell-type-specific DNA repair processes in T cells and built an algorithm for enhanced prediction of their CRISPR editing outcomes. We then adapted the CRISPR-Cas9 ribonucleoprotein editing for use in primary human myeloid cells, allowing for interrogation of host factors for many additional pathogens. Finally, faced with a viral pandemic, we identified questions we were well-positioned to answer, first assessing the performance of commercial SARS-CoV-2 antibody assays before returning to host-pathogen interaction mapping. We carried out comparative viral-human protein-protein interaction and viral protein localization analysis for all three pathogenic coronaviruses SARS-CoV-1, MERS-CoV and SARS-CoV-2. Subsequent functional genetic screening identified host factors that functionally impinge on coronavirus proliferation, including Tom70, a mitochondrial chaperone protein that interacts with both SARS-CoV-1 and SARS-CoV-2 Orf9b, an interaction we structurally characterized using cryo-EM. Combining genetically-validated host factors with both COVID-19 patient genetic data and medical billing records identified important molecular mechanisms and potential drug treatments with effectiveness against COVID-19 that merit further molecular and clinical study.

Multi-subunit Cullin-RING ligases (CRLs) comprise the largest family of ubiquitin E3 ligases in humans. Although some CRLs attenuates substrate degradation, it remains unclear whether their modularity can be targeted to regulate CRL subfamilies. With proteomics, we discovered cross talk between CRL4AMBRA1 and CRL5VIF. We also found that CRL4AMBRA1 targets Elongin C (ELOC), the essential adapter protein of CRL5 and CRL2 complexes, for polyubiquitination and proteasomal degradation. Depletion of AMBRA1 promotes accumulation of ELOC, which enhances ligase activity and autocatalysis of CRL5 and CRL2 substrate receptors. We also found that CRL4AMBRA1 reduces the assembly and ligase activity of both CRL5SOCS3 and HIV-1 CRL5VIF complexes. Furthermore, AMBRA1 depletion blunts IL-6-induced STAT3 phosphorylation and enhances the degradation of the antiviral APOBEC3G protein, which are regulated by CRL5SOCS3 and CRL5VIF, respectively. Thus, we revealed that CRL4AMBRA1 negatively regulates CRL5 ubiquitin E3 ligases by targeting their shared ELOC for degradation.

The retrovirus genus Lentivirus, which includes the human pandemic virus human immunodeficiency virus 1 (HIV-1), infect a wide range of mammals and are characterized by slow infections of host immune systems. HIV-1 encodes the accessory protein Vif, which hijacks a host Cullin-RING ubiquitin ligase (CRL) complex as well as the non-canonical cofactor CBFβ, to antagonize host APOBEC3 anti-viral proteins. To date, non-canonical cofactor recruitment to CRL complexes by viral factors has only been attributed to HIV-1 Vif. To further study this phenomenon, we employed a comparative approach combining proteomic, biochemical, structural, and virological techniques to investigate Vif complexes across representative clades within the Lentivirus genus, including primate (HIV-1, SIVmac) and non-primate (FIV, BIV, and MVV) viruses. We find that CBFβ is completely dispensable for the activity of non-primate lentiviral Vif proteins. Furthermore, we find that BIV Vif does not require any non-canonical cofactor MVV Vif requires a novel cofactor, Cyclophilin A (CYPA), for stable CRL complex formation and anti-APOBEC3 activity. Lastly, we present evidence testing the idea of novel Vif functions associated with non-canonical cofactor use. We propose a model of modular conservation of Vif complexes that allows for exaptation of novel functions through the acquisition of non-CRL associated host cofactors while preserving anti-APOBEC3 activity.

Glycoproteomics is a rapidly developing field, and data analysis has been stimulated by several technological innovations. As a result, there are many software tools from which to choose; and each comes with unique features that can be difficult to compare. This work presents a head-to-head comparison of five modern analytical software: Byonic, Protein Prospector, MSFraggerGlyco, pGlyco3, and GlycoDecipher. To enable a meaningful comparison, parameter variables were minimized. One potential confounding variable is the glycan database that informs glycoproteomic searches. We performed glycomic profiling of the samples and used the output to construct matched glycan databases for each software. Up to 19,000 glycopeptide spectra were identified across three replicates of wild-type SH-SY5Y cells. There was substantial overlap among most software for glycoproteins identified, locations of glycosites, and glycans, although Byonic reported a suspiciously large number of glycoproteins and glycosites of questionable reliability. We show that Protein Prospector identified the most glycopeptide spectrum matches with high agreement to known glycosites in UniProt. Overall, our results indicate that glycoproteomic searches should involve more than one software to generate confidence. It may be useful to consider software with peptide-first approaches and with glycan-first approaches.