Huntington’s disease (HD) is a fatal neurodegenerative disease caused by a CAG repeat expansion in exon 1 of the huntingtin gene, resulting in an expanded polyglutamine (polyQ) repeat in the huntingtin protein. Patients receive symptomatic treatment for motor, emotional, and cognitive impairments; however, there is no treatment to slow the progression of the disease, with death occurring 15-20 years after diagnosis. Mutant huntingtin protein interferes with multiple cellular processes leading to cellular dysfunction and neuronal loss. Due to the complexity of mutant huntingtin toxicity, many approaches to treating each effect are being investigated. Unfortunately, addressing one cause of toxicity might not result in protection from other toxic insults, necessitating a combination of treatments for HD patients. Ideally, single therapy targeting the mutant mRNA or protein could prevent all downstream toxicities caused by mutant huntingtin. In this work, I used animal models to investigate a potential therapeutic target for decreasing mutant huntingtin protein, and I apply bioluminescent imaging to investigate RNA interference to silence mutant huntingtin target sites. The enzyme puromycin sensitive aminopeptidase (PSA) has the unique property of degrading polyQ peptides and been implicated in the degradation of huntingtin. In this study, we looked for an effect of decreased PSA on the pathology and behavior in a mouse model of Huntington’s disease. To achieve this, we crossed HD mice with mice with one functional PSA allele and one inactivated PSA allele. We found that PSA heterozygous HD mice develop a greater number of pathological inclusion bodies, representing an accumulation of mutant huntingtin in neurons. PSA heterozygous HD mice also exhibit worsened performance on the raised-beam test, a test for balance and coordination indicating that the PSA heterozygosity impairs the function of neurons with mutant huntingtin. In order to test whether increasing PSA expression ameliorates the HD phenotype in mice we created an adeno-associated virus (AAV) expressing the human form of PSA (AAV-hPSA). Unexpectedly, testing of AAV-hPSA in non-HD mice resulted in widespread toxicity at high doses. These findings suggest that overexpression of PSA is toxic to neurons in the conditions tested. In the second part of my dissertation work, I designed a model for following the silencing of huntingtin sequences in the brain. Firefly luciferase is a bioluminescent enzyme that is extensively used as a reporter molecule to follow biological processes in vivo using bioluminescent imaging (BLI). I created an AAV expressing the luciferase gene containing huntingtin sequences in the 3'-untranslated region (AAV-Luc-Htt). After co-injection of AAV-Luc-Htt with RNA-silencing molecules (RNAi) into the brain, we followed luciferase activity. Using this method, we tested cholesterol-conjugated siRNA, un-conjugated siRNA, and hairpin RNA targeting both luciferase and huntingtin sequences. Despite being able to detect silencing on isolated days, we were unable to detect sustained silencing, which had been reported in similar studies in tissues other than the brain. We observed an interesting finding that co-injection of cholesterol-conjugated siRNA with AAV-Luc-Htt increased luminescence, findings that were verified in cell culture to be independent of serotype, siRNA sequence, and cell type. That cc-siRNA affects the expression of AAV-Luc-Htt reveals an interesting interaction possibly resulting in increased delivery of AAV into cells or an increase in luciferase expression within the cell. My work presents a method to follow gene silencing of huntingtin targets in the brain, which needs further optimization in order to detect sustained silencing. Finally, in this dissertation I continue the study of bioluminescent imaging in the brain. We use mice that have been injected in the brain with AAV-Luciferase (AAV-Luc) to screen 34 luciferase substrate solutions to identify the greatest light-emitting substrate in the brain. We identify two substrates, CycLuc1 and iPr-amide as substrates with enhanced light-emitting properties compared with D-luciferin, the standard, commercially available substrate. CycLuc1 and iPr-amide were tested in transgenic mice expressing luciferase in dopaminergic neurons. These novel substrates produced luminescence unlike the standard substrate, D-luciferin which was undetectable. This demonstrates that CycLuc1 and iPr-amide improve the sensitivity of BLI in low expression models. We then used CycLuc1 to test silencing of luciferase in the brain using AAV-shRNA (AAV-shLuc). We were unable to detect silencing in treated mice, despite a 50% reduction of luciferase mRNA. The results from this experiment identify luciferase substrates that can be used to image transgenic mice expressing luciferase in dopaminergic neurons. My work contributes new data on the study of PSA as a modifier of Huntington’s disease in a knock-in mouse model of Huntington’s disease. My work also makes contributions to the field of bioluminescent imaging by identifying and testing luciferase substrates in the brain to detect low level of luciferase expression.

The development of methods and techniques is the driving force of scientific research. In this work, we described two large-scale screens in studying transcriptional activation and tumor suppression. In Part I, we studied transcriptional activation mechanisms by deriving and characterizing activation defective mutants. Promoter-specific transcriptional activators stimulate transcription through direct interactions with one or more components of the transcription machinery, termed the “target.” The identification of direct in vivo targets of activators has been a major challenge. We perform a large-scale genetic screen to derive and characterize tra1 alleles that are selectively defective for interaction with Gal4 in vivo. Utilizing these mutants, we demonstrated that Tra is an essential target for Gal4 activation, Gal4 and Tra1 bind cooperatively at the promoter and the Gal4–Tra1 interaction occurs predominantly on the promoter. In addition, we demonstrated that the Gal4-interaction site on Tra1 is highly selective. In Part II, we described a functional genomics approach to discover new tumor suppressor genes. A goal of contemporary cancer research is to identify the genes responsible for neoplastic transformation. Cells that are immortalized but non-tumorigenic were stably transduced with pools of short hairpin RNAs (shRNAs) and tested for their ability to form tumors in mice. ShRNAs in any resulting tumors were identified by sequencing to reveal candidate TSGs, which were then validated both experimentally and clinically by analysis of human tumor samples. Using this approach, we identified and validated 33 candidate TSGs. We found that most candidate TSGs were down-regulated in >70% of human lung squamous cell carcinoma (hLSCC) samples, and 17 candidate TSGs negatively regulate FGFR signalling pathway, and their ectopic expression inhibited growth of hLSCC xenografts. Furthermore, we suggest that by examining at the expression level of TSGs in lung cancer patients, we can predict their drug responsiveness to FGFR inhibitors. In conclusion, we have identified many new lung squamous cell cancer TSGs, using an experimental strategy that can be broadly applied to find TSGs in other tumor types.

Mobile genetic elements represent a large portion of the genome in many species. Posing a danger to the integrity of genetic information, silencing and structural machinery has evolved to suppress the mobility of foreign and transposable elements within the genome. Condensin proteins – which regulate chromosome structure to promote chromosome segregation – have been demonstrated to function in repetitive gene regulation and transposon silencing in several species. In model system Caenorhabditis elegans, microarray analysis studies have implicated Condensin II subunit HCP-6 in the silencing of multiple loci, including DNA transposon MIRAGE. To address the hypothesis that HCP-6 has a direct function in transcriptional gene silencing of the MIRAGE transposon, we queried MIRAGE expression and chromatin profiles in wild-type and hcp-6 mutant animals. Our evidence confirms that HCP-6 does indeed function during silencing of MIRAGE. However, we found no significant indication that HCP-6 binds to MIRAGE, nor that HCP-6 mediates MIRAGE enrichment of H3K9me3, the repressive heterochromatin mark observed at regions undergoing transcriptional silencing. We suggest that the silencing of MIRAGE, a newly evolved transposon and the only tested mobile element considerably derepressed upon loss of HCP-6, is managed by HCP-6 indirectly.

Small silencing RNAs function in almost every aspect of cellular biology. Argonaute proteins bind small RNA and execute gene silencing. The number of Argonaute paralogs range from 5 in Drosophila melanogaster , 8 in Homo sapiens to an astounding 27 in Caenorhabditis elegans. This begs several questions: Do Argonaute proteins have different small RNA repertoires? Do Argonaute proteins behave differently? And if so, how are they functionally and mechanistically distinct? To address these questions, we examined the thermodynamic, kinetic and functional properties of fly Argonaute1 (dAgo1), fly Argonaute2 (dAgo2) and mouse Argonaute2 (mAGO2). Our studies reveal that in fly, small RNA duplexes sort into Argonaute proteins based on their intrinsic structures: extensively paired siRNA duplex is preferentially sorted into dAgo2 while imperfectly paired miRNA duplex is channeled into dAgo1. The sorting of small RNA is uncoupled from its biogenesis. This is exemplified by mir-277, which is born a miRNA but its extensive duplex structure licenses its entry into dAgo2. In the Argonaute protein, the small RNA guide partitions into functional domains: anchor, seed, central, 3' supplementary and tail. Of these domains, the seed initiates binding to target. Both dAgo2 and mAGO2 (more closely related to and a surrogate for dAgo1 in our studies) bind targets at astonishing diffusion-limited rates (~107–108 M−1s−1). The dissociation kinetics between dAgo2 and mAGO2 from their targets, however, are different. For a fully paired target, dAgo2 dissociates slowly (t½ ~2 hr) but for a seed-matched target, dAgo2 dissociates rapidly (t½ ~20 s). In comparison, mAGO2 does not discriminate between either targets and demonstrates an equivalent dissociation rate (t½ ~20 min). Regardless, both dAgo2 and mAGO2 demonstrate high binding affinity to perfect targets with equilibrium dissociation constants, KD ~4–20 pM. Functionally, we also showed that dAgo1 but not dAgo2 silence a centrally bulged target. By contrast, dAgo2 cleaved and destroyed perfectly paired targets 43-fold faster than dAgo1. In target cleavage, dAgo2 can tolerate mismatches, bulged and internal loop in the target but at the expense of reduced target binding affinities and cleavage rates. Taken together, our studies indicate that small RNAs are actively sorted into different Argonaute proteins with distinct thermodynamic, kinetic and functional behaviors. Our quantitative biochemical analysis also allows us to model how Argonaute proteins find, bind and regulate their targets.

There are two copies of each core histone in a nucleosome, however, it is unclear whether post-translational modifications on each molecule function redundantly or if symmetrical modifications are required to properly regulate gene expression. We tried to address this question by breaking nucleosomal symmetry and measuring its impact on gene expression. Our strategy includes re-engineering specific residues at the H3-H3 interface, generating pairs of mutant proteins, which were predicted by computational methods to form obligate heterodimers. Using S. cerevisiae as a model system, we tested the viability of strains with mutant histones, and analyzed the interaction between by co-immunoprecipitation from mononucleosome preparations. We also measured the changes of gene expression in the strains bearing single-tailed or tailless H3 heterodimers. The data suggested that the best computationally-derived H3 pair was frequently, but not exclusively heterodimeric in vivo. In order to obtain a more stringent H3 heterodimer, random mutagenesis was performed on four codons in the original computational design, and then genetic screening of the mutant libraries was performed.

The utility of a model organism for studying biological processes is closely tied to its amenability to genome manipulation. Although tools for targeted genome engineering in mice have been available since 1987, most organisms including zebrafish have lacked efficient reverse genetic tools, which has stymied their broad implementation as a model system to study biological processes. The development of zinc finger nucleases (ZFNs) that can create double-strand breaks at desired sites in a genome has provided a universal platform for targeted genome modification. ZFNs are artificial restriction endonucleases that comprise of an array of 3- to 6-C2H2-zinc finger DNA-binding domains fused with the dimeric cleavage domain of the type IIs endonuclease FokI. C2H2-zinc fingers are the most common, naturally occurring DNA-binding domain, and their specificity can be engineered to recognize a variety of DNA sequences providing a strategy for targeting the appended nuclease domain to desired sites in a genome. The utility of ZFNs for gene editing relies on their activity and precision in vivo both of which depend on the generation of ZFPs that bind desired target sites high specificity and affinity. Although various methods are available that allow construction of ZFPs with novel specificities, ZFNs assembled using existing approaches often display negligible in vivo activity, presumably resulting from ZFPs with either low affinity or suboptimal specificity. A root cause of this deficiency is the presence of interfering interactions at the finger-finger interface upon assembly of multiple fingers. In this study we have employed bacterial-one-hybrid (B1H)-based selections to identify two-finger zinc finger units (2F-modules) containing optimized interface residues that can be combined with published finger archives to rapidly yield ZFNs that can target more than 95% of the zebrafish and human protein-coding genes while maintaining a success rate higher than that of ZFNs constructed using available methods. In addition to genome engineering in model organisms, this advancement in ZFN design will aid in the development of ZFN-based therapeutics. In the process of creating this archive, we have undertaken a broader study of zinc finger specificity to better understand fundamental aspects of DNA recognition. In the process we have created the largest protein-DNA interaction dataset for zinc fingers to be described that will facilitate the development of better predictive models of recognition. Ultimately, these predictive models would enable the rational design of synthetic zinc finger proteins for targeted gene regulation or genomic modification, and the prediction of genomic binding sites for naturally occurring zinc finger proteins for the construction of more accurate gene regulatory networks.

Canavan’s Disease is a fatal Central Nervous System disorder caused by genetic defects in the enzyme – aspartoacylase and currently has no effective treatment options. We report additional phenotypes in a stringent preclinical aspartoacylase knockout mouse model. Using this model, we developed a gene therapy strategy with intravenous injections of the aspartoacylase gene packaged in recombinant adeno associated viruses (rAAVs). We first investigated the CNS gene transfer abilities of rAAV vectors that can cross the blood-brain-barrier in neonatal and adult mice and subsequently used different rAAV serotypes such as rAAV9, rAAVrh.8 and rAAVrh.10 for gene replacement therapy. A single intravenous injection rescued lethality, extended survival and corrected several disease phenotypes including motor dysfunctions. For the first time we demonstrated the existence of a therapeutic time window in the mouse model. In order to limit off-target effects of viral delivery we employed a synthetic strategy using microRNA mediated posttranscriptional detargeting to restrict rAAV expression in the CNS. We followed up with another approach to limit peripheral tissue distribution. Strikingly, we demonstrate that intracerebroventricular administration of a 50-fold lower vectors dose can rescue lethality and extend survival but not motor functions. We also study the contributions of several peripheral tissues in a primarily CNS disorder and examine several molecular attributes behind pathogenesis of Canavan’s disease using primary neural cell cultures. In summary, this thesis describes the potential of novel rAAV-mediated gene replacement therapy in Canavan’s disease and the use of rAAVs as a tool to tease out its pathological mechanism.

Interactions between 3´ untranslated regions (UTRs) and RNA-binding proteins (RBPs) play critical roles in post-transcriptional gene regulation. Metazoan genomes encode hundreds of RBPs and thousands of 3’ UTRs have been experimentally identified, yet the spectrum of interactions between 3´UTRs and RBPs remains largely unknown. Several methods are available to map these interactions, including protein-centered methods such as RBP immunoprecipitation (RIP) and cross-link immunoprecipitation (CLIP), yeast three-hybrid assays and RNAcompete. However, there is a paucity of RNA-centered approaches for assaying an RNA element of interest against multiple RBPs in a parallel, scalable manner. Here, I present a strategy for delineating protein-RNA interaction networks using a gene centered approach. This approach includes annotating RBPs and identifying physical interactions between an RNA of interest and these RBPs using the Protein-RNA Interaction Mapping Assay (PRIMA). Few RBPs have been experimentally determined in most eukaryotic organisms. Therefore I show that existing RBP annotations can be supplemented using computational predictions of RNA binding domains (RBD) from protein sequences. A single RNA of interest can be tested using PRIMA against a library of RBPs constructed from these annotations. PRIMA utilizes the green fluorescent protein (GFP) in yeast as a reporter. PRIMA is based on reconstitution of the interaction between the 5´ and 3´ ends of an mRNA, which increases mRNA stability and enhances translation. PRIMA recapitulates known and uncovers new interactions involving RBPs from human, Caenorhabditis elegans and bacteriophage with short RNA fragments and full-length 3´UTRs. The development of RBP prey libraries will enable the testing of 3´UTRs against the hundreds of RBPs, which is essential to gain broad insights into post-transcriptional gene regulation at a systems level.

Down Syndrome (DS), the most common aneuploidy seen in live-borns, is caused by trisomy for chromosome 21. DS imposes high risks for multiple health issues involving various systems of the body. The genetic complexity of trisomy 21 and natural variation between all individuals has impeded understanding of the specific cell pathologies and pathways involved. In addition, chromosomal disorders have been considered outside the hopeful progress in gene therapies for single-gene disorders. Here we test the feasibility of correcting imbalanced expression of genes across an extra chromosome by expression of a single gene, XIST, the key player in X chromosome inactivation. We targeted a large XIST transgene into one chromosome 21 in DS iPS cells, and demonstrated XIST RNA spreads and induces heterochromatin and gene silencing across that autosome in cis. By making XIST inducible, this allows direct comparison of effects of trisomy 21 expression on cell function and phenotypes. Importantly, XIST-induction during in vitro hematopoiesis normalized excess production of differentiated blood cell types (megakaryocytes and erythrocytes), known to confer high risk for myeloproliferative disorder and leukemia. In contrast, trisomy silencing enhances production of iPS and neural stem cells, consistent with DS clinical features. Further analysis revealed that trisomy 21 initially impacts the endothelial hematopoietic transition (EHT) to generate excess CD43+ progenitors, and also increases their colony forming potential. Furthermore, results provide evidence for a key role for enhanced IGF signaling, involving over-expression of non-chromosome 21 genes controlled by trisomy 21. Finally, experiments to examine trisomy effects on angiogenesis showed no effect on production of endothelial cells, but it remains unclear whether trisomic cells may differ in ability to form vessels. Collectively, this thesis demonstrates proof-of-principle for XIST-mediated “trisomy silencing”. Phenotypic improvement of hematopoietic and neural stem cells demonstrates the value for research into DS pathogenesis, but also provides a foundation of potential for future development of “chromosome therapy” for DS patients.

In the mammalian genome, 5‟-CpG-3‟ dinucleotides are frequently methylated, correlating with transcriptional silencing. Genome-wide waves of demethylation are thought to occur only twice during development, in primordial germ cells and in the pre-implantation embryo. They are followed by de novo methylation, setting up a pattern that is inherited throughout development. No global methylation changes are thought to occur during further somatic development, although methylation does alter at gene-specific loci, contributing to tissue-specific patterns of gene expression. Here we studied DNA methylation in differentiating mouse erythroblasts in vivo using several approaches including genomic-scale, reduced representation bisulfite sequencing (RRBS). Surprisingly, demethylation at the erythroid-specific β-globin locus was coincident with a wave of global DNA demethylation at most genomic elements, including repetitive elements and genes silenced in erythropoiesis. Over 30% of total methylation is irreversibly lost during erythroid differentiation. Demethylation occurred through a passive mechanism, requiring the rapid DNA replication triggered with the onset of erythroid terminal differentiation. Global loss of DNA methylation was not associated with a global increase in transcription, as determined by GeneChip analysis. We propose that global demethylation is a consequence of cellular mechanisms required for the rapid demethylation and induction of β-globin and other erythroid genes. Our findings demonstrate that, contrary to previously held dogma, DNA demethylation can occur globally during somatic cell differentiation, providing a new experimental model for the study of global demethylation in development and disease.