Of Men and Water Fleas: Gene regulatory network discovery in human cancer and in an ecological context

Gene expression is a central process in the biology of a cell, influencing the phenotype at the cellular, tissue and organism level. It underlies fundamental biological processes ranging from development - including the development of diseases - to responses to environmental conditions. The complexity of those processes is mirrored in the complexity of the regulation of gene expression. Genes and their products form intricate networks that are tightly regulated at many levels, including the sequence-specific binding of transcription factors to promoter and enhancer regions, and histone modifications that steer chromatin accessibility. The advent of high-throughput sequencing methods has greatly facilitated whole-genome sequencing and genome-wide profiling of gene expression, transcription factor binding and histone modifications. In conjunction with bioinformatics methods to integrate those data, it is now possible to unravel gene regulatory networks (GRNs) and thereby to understand how cells and organisms react to internal and external stimuli by modulating gene expression in a genotype-dependent manner. Last but not least, the recent development of single-cell sequencing technologies allows the mapping of transcriptional interactions precisely where they happen, namely in the individual cells. In this thesis, I first used those techniques to explore GRNs in human, specifically in melanoma skin cancer. Next, I asked whether similar methods can be applied to species that are not traditional genetic model organisms, namely water fleas, to identify the GRNs involved in ecological responses and adaptation. In Part I, my colleagues and I investigated heterogeneity in melanoma cell cultures using single-cell RNA-Seq and chromatin accessibility data. A long-standing question in cancer research is, how tumour cells switch from a proliferative state that leads to tumour growth, to an invasive state that is marked by a higher migratory potential and therapy resistance, and can result in tumour dissemination. Understanding this phenotypic switch and which genes control the different cell states can help to define targets for therapy and prognostic markers. We profiled single cells by their transcriptional phenotype and migratory behaviour, and found that the patient-derived cell cultures reside either in two previously-described extreme states, namely (1) the melanocytic state, which maintains characteristics of lineage-specificity and pigmentation genes, and (2) the mesenchymal-like state which has switched towards a more undifferentiated and invasive phenotype; or they (3) exhibit an intermediate melanocytic state, which shares properties with both extremes, and also exhibits an intermediate migratory phenotype. With the GRN inference method SCENIC, we identified distinct and stable regulatory profiles that characterise the three states and that are also present in vivo. To study the transcriptional changes during the phenotypic switch from the intermediate to the mesenchymal-like state, we knocked down an important melanocytic lineage-factor, as well as two regulators of the intermediate state. All three knock-downs initiated an increase in signatures of migration, invasion and therapy resistance, and demonstrate the importance of the three factors for the maintenance of the intermediate cell state. In Part II, I focused on GRN recovery in the aquatic microcrustacean Daphnia spp. Daphnia, or water fleas, are keystone grazers in lake and pond ecosystems that have been studied for more than 200 years in ecology, evolutionary biology and ecotoxicology. Since the genomes of two species, Daphnia magna and Daphnia pulex, have been sequenced, they have become the first crustacean model organisms in ecological genomics. One of the biggest challenges in ecological genomics remains the identification of genes that determine responses to complex ecological gradients in nature and the subsequent step of identifying their functions. High-throughput experiments that yield associations between environmental stressors and gene regulation, combined with computational analyses, are thus becoming a bottleneck in environmental genomic studies. Currently, many transcriptome sequencing studies are conducted on ecological model species and data analysis remains typically restricted to genes with functional annotation through homologies in other, genetically well-studied species. Despite those difficulties, enormous potential lies in the investigation of genetic non-model species. Discoveries can have direct implications for human health (e.g. new disease models, drug development, nutrition), or they contribute to informed policy decisions directed towards the maintenance of ecosystem services. The era of non-model species has already begun, and bioinformatics tools to analyse the plethora of new data are needed. Therefore, in chapter 3, I tested whether an ontology-free gene prioritisation method that has initially been developed for genetic model species, can be applied to emerging model species, in particular Daphnia. This method, called Daphnia-cisTarget, facilitates the analysis of gene expression data by combining genomic and transcriptomic data to infer GRNs. I validated the approach using a heat shock data set, and could show that Daphnia-cisTarget indeed produces biologically meaningful results and recovers the highly conserved heat shock factor as master regulator. Since it clusters genes into transcriptional units only based on conserved genomic sequences, the method can be used to approximate the function of so far unannotated genes. In chapter 4, I applied this motif discovery method to two RNA-Seq data sets obtained by exposing D. magna to cyanobacteria. Cyanobacteria can severely impact freshwater ecosystems by producing toxic algae blooms. However, Daphnia have the potential to adapt and become less susceptible and even suppress blooms. Previous studies have identified individual genes that might confer tolerance to cyanobacteria. I re-analysed those data sets from a networks biology perspective to pinpoint the master regulators and to gain a better insight into the system-level responses. With Daphnia-cisTarget, I indeed found multiple GRNs that could be attributed to cuticle formation, the moulting cycle and nutritional control. Finally, in chapter 5, I analysed the transcriptional response of D. magna to fish kairomones. Fish predation is one of the most important Daphnia stressors. Phenotypic responses of Daphnia to fish are well-studied and include morphological, life-history and behavioural traits. However, the genes underlying those traits are yet unknown. For this study, I used a natural, resurrected D. magna population that has experienced two shifts in fish predation pressure, first from no to high fish presence, and then to reduced fish presence. This population has been characterised phenotypically and genotypically and has been shown to rapidly adapt to the presence of fish. Using a GRN inference approach adopted from the single-cell method SCENIC, I found transcriptional networks that react to fish smell and that are reminiscent of the moulting- and midgut-related response described earlier in the cyanobacteria study. The GRNs that are constitutively differentially expressed between the no- and the high-fish subpopulations likely respond to juvenile hormone and control reproductive processes. The results of the last two chapters show that research in species that are genomically underexplored can benefit from a network-level analysis approach that does not rely solely on gene function conservation, but also incorporates genomic sequence information.

Jaar van publicatie:2022

Toegankelijkheid:Open