Search

Objective Nuclear Factor One X (NFIX) is a transcription factor expressed by neural stem cells within the developing mouse brain and spinal cord. In order to characterise the pathways by which NFIX may regulate neural stem cell biology within the developing mouse spinal cord, we performed an microarray-based transcriptomic analysis of the spinal cord of embryonic day (E)14.5 Nfix−/− mice in comparison to wild-type controls. Data description Using microarray and differential gene expression analyses, we were able to identify differentially expressed genes in the spinal cords of E14.5 Nfix−/− mice compared to wild-type controls. We performed microarray-based sequencing on spinal cords from n = 3 E14.5 Nfix−/− mice and n = 3 E14.5 Nfix+/+ mice. Differential gene expression analysis, using a false discovery rate (FDR) p-value of p ± 1.5, revealed 1351 differentially regulated genes in the spinal cord of Nfix−/− mice. Of these, 828 were upregulated, and 523 were downregulated. This resource provides a tool to interrogate the role of this transcription factor in spinal cord development.

Chromatin immunoprecipitation followed by sequencing (ChIP-seq) is the primary protocol for detecting genome-wide DNA-protein interactions, and therefore a key tool for understanding transcriptional regulation. A number of factors, including low specificity of antibody and cellular heterogeneity of sample, may cause “peak” callers to output noise and experimental artefacts. Statistically combining multiple experimental replicates from the same condition could significantly enhance our ability to distinguish actual transcription factor binding events, even when peak caller accuracy and consistency of detection are compromised.We adapted the rank-product test to statistically evaluate the reproducibility from any number of ChIP-seq experimental replicates. We demonstrate over a number of benchmarks that our adaptation “ChIP-R” (pronounced ‘chipper’) performs as well as or better than comparable approaches on recovering transcription factor binding sites in ChIP-seq peak data. We also show ChIP-R extends to evaluate ATAC-seq peaks, finding reproducible peak sets even at low sequencing depth. ChIP-R decomposes peaks across replicates into “fragments” which either form part of a peak in a replicate, or not. We show that by re-analysing existing data sets, ChIP-R reconstructs reproducible peaks from fragments with enhanced biological enrichment relative to current strategies.

Transcription factors are a set of proteins that bind to DNA and regulate transcription. They play a key role in regulating genes through the developmental process and disruptions to transcription factor binding have been linked to cancer and other human diseases. This thesis develops a bioinformatic workflow to identify high confidence transcription factor binding sites and characterise the functional role of a transcription factor by integrating available genetic and epigenetic data sets.Chromatin immunoprecipitation followed by sequencing (ChIP-seq) is the assay of choice to identify where a transcription factor is binding in vivo. The ChIP-seq protocol is known to generate false positive signals and requires a large cell population resulting in significant biological variance across experiments. Our workflow addresses these weaknesses in two ways. First, DNase I hypersensitivity (DHS) accessibility data is used to filter out false positive binding events that are inaccessible as transcription factor binding is dependent on access to DNA. Second, we built a new tool to identify reproducible peaks across biological replicates improving our ability to accurately report transcription factor binding. This tool had marked improvement over available approaches at identifying true binding sites from ChIP-seq data.ChIP-seq cannot be used to assay binding across all factors and cell types due to lack of a specific antibody, inability to extract cells and general infeasibility. To combat this, we built a Bayesian network model that predicts in vivo binding sites allowing a transcription factor to be characterised without the need for ChIP-seq data. Our model did not achieve the performance required to be included in our workflow but it did allow identification of chromatin accessibility, DNA footprints and sequence patterns associated with binding across different factors and conditions. We also indicate where improvements can be made in future models and ultimately the need for new biological information to accurately predict binding events within and across cell types.After identifying high confidence binding events, our workflow determines where binding events occur relative to genes and integrates available histone modification data to build epigenetic profiles that provide insight into how a transcription factor operates. Using these profiles, we demonstrated key differences in the operation of two transcription factors; showing one to preferentially bind far from the nearest gene at locations with strong regulatory activity while the other binds closer to genes at locations with epigenetic patterns linked to weaker or repressive activity. Therefore, including such profile generation in our workflow successfully provides new insight into the regulatory role a transcription factor plays.RNA-sequencing (RNA-seq) experiments provide insight into gene expression patterns and, through comparison of wild type and transcription factor knock out experiments, can be used to determine which genes are being regulated by a transcription factor. This information is invaluable when attempting to characterise the functional role of a transcription factor. However, transcription factors are capable of regulating multiple genes over potentially long distances making target gene identification challenging. Our workflow used a combination of existing approaches to predict transcription factor target genes followed by functional annotation to characterise the regulatory role of a transcription factor and identify novel target genes that were subsequently experimentally validated. This new knowledge could not have been identified without the application of our integrative bioinformatic workflow.While valuable biological insights were gained using available approaches, the task of identifying transcription factor target genes accurately is yet to be resolved. Approaches to predict long distance target genes are based on chromosome conformation capture experiments designed to detect long distance interactions driven by DNA-looping. These experiments have high false negative rates, low resolution, are difficult to reproduce and only a handful measure interactions genome wide making them a poor gold standard against which predictive approaches are measured. We reviewed available approaches identifying numerous limitations, flaws and oversights and conclude that no approach can predict target genes of transcription factor binding events accurately when learning from low quality validation data.In summary, this thesis presents a bioinformatic workflow as a set of methods and techniques focused on data integration and machine learning that can be adapted to a range of different experiments and types of data. Specifically, we demonstrate the application of our workflow to regulation across the genome with a focus on transcription factor biology. Our bioinformatic workflow clearly outlines a number of approaches to characterise a transcription factor’s regulatory role. This includes an approach to filter ChIP-seq binding sites using accessibility data, a robust method for combining biological replicates, a means to build epigenetic profiles of binding events and construction of a transcription factor targetome linking high confidence binding sites to biologically relevant or misregulated target genes leading to a better understanding of transcription factor driven regulation in the genome. We applied this workflow to a number of transcription factors successfully identifying and validating novel target genes and providing insight into the functional role each factor plays: knowledge that was inaccessible without the application of our bioinformatic workflow.

Ancestral sequence reconstruction is a technique that is gaining widespread use in molecular evolution studies and protein engineering. Accurate reconstruction requires the ability to handle appropriately large numbers of sequences, as well as insertion and deletion (“indel”) events, but available approaches exhibit limitations. To address these limitations, we developed Graphical Representation of Ancestral Sequence Predictions (GRASP), which efficiently implements maximum likelihood methods to enable the inference of ancestors of families with more than 10,000 members. GRASP implements partial order graphs (POGs) to represent and infer insertion and deletion events across ancestors, enabling the identification of building blocks for protein engineering.To validate the capacity to engineer novel proteins from realistic data, we predicted ancestor sequences across three distinct enzyme families: glucose-methanol-choline (GMC) oxidoreductases, cytochromes P450, and dihydroxy/sugar acid dehydratases (DHAD). All tested ancestors demonstrated enzymatic activity. Our study demonstrates the ability of GRASP (1) to support large data sets over 10,000 sequences and (2) to employ insertions and deletions to identify building blocks for engineering biologically active ancestors, by exploring variation over evolutionary time.Author summaryMassive sequencing projects expose the extent of natural, genetic diversity. Here, we describe a method with capacity to perform ancestor sequence reconstruction from data sets in excess of 10,000 sequences, poised to recover ancestral diversity, including the evolutionary events that determine present-time biological function and structure.We introduce a novel strategy for suggesting “indel variants” that are distinct from, but can be explored alongside, substitution variants for creating ancestral libraries. We demonstrate how indels can be used as building blocks to form “hybrid ancestors”; based on this strategy, we synthesise ancestor variants, with varying enzymatic activities, for wide-ranging applications in the biotechnology sector.

Transcriptional regulation plays a central role in controlling neural stem and progenitor cell proliferation and differentiation during neurogenesis. For instance, transcription factors from the nuclear factor I (NFI) family have been shown to co-ordinate neural stem and progenitor cell differentiation within multiple regions of the embryonic nervous system, including the neocortex, hippocampus, spinal cord and cerebellum. Knockout of individual Nfi genes culminates in similar phenotypes, suggestive of common target genes for these transcription factors. However, whether or not the NFI family regulates common suites of genes remains poorly defined. Here, we use granule neuron precursors (GNPs) of the postnatal murine cerebellum as a model system to analyse regulatory targets of three members of the NFI family: NFIA, NFIB and NFIX. By integrating transcriptomic profiling (RNA-seq) of Nfia- and Nfix-deficient GNPs with epigenomic profiling (ChIP-seq against NFIA, NFIB and NFIX, and DNase I hypersensitivity assays), we reveal that these transcription factors share a large set of potential transcriptional targets, suggestive of complementary roles for these NFI family members in promoting neural development.

Transcription factors regulate gene expression and play an essential role in development by maintaining proliferative states, driving cellular differentiation and determining cell fate. Transcription factors are capable of regulating multiple genes over potentially long distances making target gene identification challenging. Currently available experimental approaches to detect distal interactions have multiple weaknesses that have motivated the development of computational approaches. Although an improvement over experimental approaches, existing computational approaches are still limited in their application, with different weaknesses depending on the approach. Here, we review computational approaches with a focus on data dependency, cell type specificity and usability. With the aim of identifying transcription factor target genes, we apply available approaches to typical transcription factor experimental datasets. We show that approaches are not always capable of annotating all transcription factor binding sites; binding sites should be treated disparately; and a combination of approaches can increase the biological relevance of the set of genes identified as targets.

During forebrain development, radial glia generate neurons through the production of intermediate progenitor cells (IPCs). The production of IPCs is a central tenet underlying the generation of the appropriate number of cortical neurons, but the transcriptional logic underpinning this process remains poorly defined. Here, we examined IPC production using mice lacking the transcription factor nuclear factor I/X (Nfix). We show that Nfix deficiency delays IPC production and prolongs the neurogenic window, resulting in an increased number of neurons in the postnatal forebrain. Loss of additional Nfi alleles (Nfib) resulted in a severe delay in IPC generation while, conversely, overexpression of NFIX led to precocious IPC generation. Mechanistically, analyses of microarray and ChIP-seq datasets, coupled with the investigation of spindle orientation during radial glial cell division, revealed that NFIX promotes the generation of IPCs via the transcriptional upregulation of inscuteable (Insc). These data thereby provide novel insights into the mechanisms controlling the timely transition of radial glia into IPCs during forebrain development.

Transcription factors from the nuclear factor one (NFI) family have been shown to play a central role in regulating neural progenitor cell differentiation within the embryonic and post-natal brain. NFIA and NFIB, for instance, promote the differentiation and functional maturation of granule neurons within the cerebellum. Mice lacking Nfix exhibit delays in the development of neuronal and glial lineages within the cerebellum, but the cell-type-specific expression of this transcription factor remains undefined. Here, we examined the expression of NFIX, together with various cell-type-specific markers, within the developing and adult cerebellum using both chromogenic immunohistochemistry and co-immunofluorescence labelling and confocal microscopy. In embryos, NFIX was expressed by progenitor cells within the rhombic lip and ventricular zone. After birth, progenitor cells within the external granule layer, as well as migrating and mature granule neurons, expressed NFIX. Within the adult cerebellum, NFIX displayed a broad expression profile, and was evident within granule cells, Bergmann glia, and interneurons, but not within Purkinje neurons. Furthermore, transcriptomic profiling of cerebellar granule neuron progenitor cells showed that multiple splice variants of Nfix are expressed within this germinal zone of the post-natal brain. Collectively, these data suggest that NFIX plays a role in regulating progenitor cell biology within the embryonic and post-natal cerebellum, as well as an ongoing role within multiple neuronal and glial populations within the adult cerebellum.

During forebrain development, radial glia generate neurons through the production of intermediate progenitor cells (IPCs). The production of IPCs is a central tenet underlying the generation of the appropriate number of cortical neurons, but the transcriptional logic underpinning this process remains poorly defined. Here, we examined IPC production using mice lacking the transcription factor nuclear factor I/X (Nfix). We show that Nfix deficiency delays IPC production and prolongs the neurogenic window, resulting in an increased number of neurons in the postnatal forebrain. Loss of additional Nfi alleles (Nfib) resulted in a severe delay in IPC generation while, conversely, overexpression of NFIX led to precocious IPC generation. Mechanistically, analyses of microarray and ChIP-seq datasets, coupled with the investigation of spindle orientation during radial glial cell division, revealed that NFIX promotes the generation of IPCs via the transcriptional upregulation of inscuteable (Insc). These data thereby provide novel insights into the mechanisms controlling the timely transition of radial glia into IPCs during forebrain development.

Cerebellar granule neurons are the most numerous neuronal subtype in the central nervous system. Within the developing cerebellum, these neurons are derived from a population of progenitor cells found within the external granule layer of the cerebellar anlage, namely the cerebellar granule neuron precursors (GNPs). The timely proliferation and differentiation of these precursor cells, which, in rodents occurs predominantly in the postnatal period, is tightly controlled to ensure the normal morphogenesis of the cerebellum. Despite this, our understanding of the factors mediating how GNP differentiation is controlled remains limited. Here, we reveal that the transcription factor nuclear factor I X (NFIX) plays an important role in this process. Mice lacking Nfix exhibit reduced numbers of GNPs during early postnatal development, but elevated numbers of these cells at postnatal day 15. Moreover, Nfix-/- GNPs exhibit increased proliferation when cultured in vitro, suggestive of a role for NFIX in promoting GNP differentiation. At a mechanistic level, profiling analyses using both ChIP-seq and RNA-seq identified the actin-associated factor intersectin 1 as a downstream target of NFIX during cerebellar development. In support of this, mice lacking intersectin 1 also displayed delayed GNP differentiation. Collectively, these findings highlight a key role for NFIX and intersectin 1 in the regulation of cerebellar development.

More than 30 human genetic diseases are linked to tri-nucleotide repeat expansions. There is no known mechanism that explains repeat expansions in full, but changes in the epigenetic state of the associated locus has been implicated in the disease pathology for a growing number of examples. A comprehensive comparative analysis of the genomic features associated with diverse repeat expansions has been lacking. Here, in an effort to decipher the propensity of repeats to undergo expansion and result in a disease state, we determine the genomic coordinates of tri-nucleotide repeat tracts at base pair resolution and computationally establish epigenetic profiles around them. Using three complementary statistical tests, we reveal that several epigenetic states are enriched around repeats that are associated with disease, even in cells that do not harbor expansion, relative to a carefully stratified background. Analysis of over one hundred cell types reveals that epigenetic states generally tend to vary widely between genic regions and cell types. However, there is qualified consistency in the epigenetic signatures of repeats associated with disease suggesting that changes to the chromatin and the DNA around an expanding repeat locus are likely to be similar. These epigenetic signatures may be exploited further to develop models that could explain the propensity of repeats to undergo expansions.