- Cervical cancer is the fourth most common cancer in women worldwide with human papillomavirus (HPV) being the main cause of the disease. Currently available treatment methods are limited and emphasize the need for discovery of new therapies that improve patient outcome. Chromosomal amplifications have been identified as a source of upregulation for cervical cancer driver genes but cannot fully explain the increased expression of immune genes in invasive carcinoma. Insight into additional factors that cause a shift from immune tolerance of HPV to the elimination of the virus, such as local microbiota, may improve diagnostic markers. Furthermore, identification of strategies for selection in combinatorial targeted therapies will allow development of efficient methods to combat the disease. In this work we shed the light on both issues.
In our first chapter we examined the currently known roles of microbiota in cancers triggered by viral infections. We provide a model of microbiome contribution to the development of oncogenic viral infections and virus associated cancers, give examples of this process in human tumors, and describe the challenges that prevent progress in the field as well as their potential solutions.
In our second chapter we investigated whether microbiota affect molecular pathways in cervical carcinogenesis by performing microbiome analysis, via sequencing 16S rRNA in tumor biopsies from 121 patients. While we detected many intra-tumor taxa (289 operational taxonomic units (OTUs)), we focused on the 38 most abundantly represented microbes. To search for microbes and host genes potentially involved in the interaction, we reconstructed a transkingdom network by integrating a previously discovered cervical cancer gene expression network with our bacterial co-abundance network and employed the bipartite betweenness centrality metric. The top ranked microbes were represented by the families Bacillaceae, Halobacteriaceae, and Prevotellaceae. While we could not define the first two families to a species level, Prevotellaceae was assigned to Prevotella bivia. By coculturing a cervical cancer cell line with P. bivia, we confirmed that three out of the ten top predicted genes in the transkingdom network (lysosomal associated membrane protein 3 (LAMP3), STAT1, TAP1) -all regulators of immunological pathways- were upregulated by this microorganism. Therefore, we propose that intra-tumor microbiota may contribute to cervical carcinogenesis through the induction of immune response drivers, including the well-known cancer gene LAMP3.
In our third chapter we identified the strategies for combining two cancer drivers as targets for gene therapies to inhibit cell growth. We use a gene co-expression network to model the disease state. Two major theories exist on how to combine regulating genes to gain control over biological network: the first strategy is to take control over as many genes in the network as possible (distantly located regulator nodes that control different parts of the network); in contrast, the second strategy is to manipulate a localized but critical portion of the network (closely located regulator nodes that control same parts of the network). To test which of the two strategies is superior, we first screened 34 potential proliferation drivers using a cervical cancer cell line to identify true regulators of cell growth. In the second step, we reconstructed a gene co-expression network from the union of targets from eight confirmed proliferation regulators (DTL, S100PBP, TPX2, EXO1, CDCA8, NEK2, ITGB3BP, ANP32E); in addition, we identified that 5-38% of driver targeted genes were correlated with cell proliferation. Average shortest path values between each pair of proliferation associated drivers’ targets were chosen as a measure of distance between two drivers in the network. In the next phase of our research, we performed double knock downs for each possible pair of drivers (total – 28 pairs) to see the inhibition effect on cell growth. We found that the average shortest path between drivers and number of unique proliferation associated targets both predict the degree of proliferation inhibition. The best performing driver pairs inhibited cell growth below 50% of the control; these were DTL+S100PBP, DTL+TPX2, and S100PBP+CDCA8. Based on our results, we suggest that gaining control over a large but localized portion of the network responsible for the phenotype of interest provides more control biological processes compared to methods controlling as many nodes in the whole network as possible.