Fig 1.
Overview of the analysis and validation strategy.
Fig 2.
Co-expression network linking normal and tumour epithelial cells.
The figures represent the modularized NT network with the results of the network connectivity and functional enrichment analysis. (A) The NT network in which genes belonging to the three main modules have been color-coded. The main functional terms identified by the web-based tool G:Profiler are listed in the three panels connected to each module. (B-D) The number of connections of the top 20 most connected hubs in each module. The x axis represents the number of connections, genes are represented on the y axis and color-coded to represent the cell type where they are expressed (blue and red represent normal and tumour epithelial cells respectively). Below each heading two p values are listed. pNT is the p value from a test showing the probability that the proportion of normal/tumour expressed genes in each module is the result of random chance. phub is the p value from the test showing the probability that the proportion of normal/tumour expressed genes in the top 20 hubs is the result of random chance.
Fig 3.
Analysis of the NT network using a novel topological index.
The results of the analysis of the NT network using the polarization index. (A) The distribution of polarization coefficient for 4 different significance thresholds defying the NT network. (B) The same distribution from a multi-variate normal model that generates random datasets with similar distributions of correlation coefficients for the normal and tumour tissue but no constrains on the correlation structure between the two tissues. (C) The number of connections towards normal or tumour expressed genes for each of the selected (rs>|0.75|) polarized genes. The x-axis represent the polarized genes sorted by increasing values of polarization index. The y-axis shows the number of tumour-expressed genes correlated to each polarized gene when expressed in normal tissue (positive values) and the number of normal expressed genes correlated to each polarized gene when expressed in tumour tissue (negative values). (D-E) The number of genes and significance (x axis) for the most enriched functional terms in the negative and positively polarized genes respectively.
Table 1.
Roles of the highly polarised genes.
Fig 4.
In vitro normal and tumour cell co-culture model.
The results of the in vitro cell co-culture experiment used to analyze normal and tumour epithelial cells crosstalk. (A) Table showing the number of overlapping genes between differentially regulated gene lists in normal and tumour cells as a result of co-culture. Rows in the table represent up and down regulated genes in normal cells whereas columns are up and down regulated genes in tumour cells. There is a significant overlap in gene lists changing in opposite directions (p<0.0001, red background). (B) PCA plot representing normal and tumour cells cultured on their own (RWPE1 and DU145) or in co-culture (RWPE1DU145 and DU145RWPE1). (C-D) Scatterplots comparing the number of genes in functional terms represented in the predicted targets of polarized genes and genes differentially expressed in the co-culture model. Panel C represent tumour cells whereas panel D represent normal cells.
Fig 5.
Phenotypic analyses of tumour cells in co-culture experiments.
In each panel is shown the phenotypic characteristic of the tumour cells alone (DU145, no insert), the tumour cells in the presence of normal (RWPE1) or tumour cells (DU145). The presence of normal cells are seen to (A) increase the population doubling time (PDT), (B) decrease the total cell number, (C) have no significant effect on apoptosis, (D) decrease the number and (E) size of the cell clusters and (F) increase the number of single cells. Each of these changes shows a normalisation of the phenotypic characteristics of the tumour cells by the presence of the normal cells. In contrast co-culture with DU145 tumour cells is seen to have the opposite effect and to increase the tumour phenotype of the tumour cells. Panel G shows the results of a clonogenic assay performed on two different tumour cell lines (DU145 and PC-1). The figure shows that the addition of 1:50 dilution of culture media, conditioned by over-expressing Slit-2 induce at least 60% reduction in cell survival respect to control cultures.
Fig 6.
Integration of CNV, polarized genes, mRNA expression and tumour features.
The figure summaries the relationship between gene CNV (blue), mRNA re-expression following hypo-methylating agents (yellow), polarized gene expression (red and green nodes are positively and negatively polarized genes, respectively) and tumour features (cyan). These relationships are the result of the statistical modeling described in the last section of the Results. Genes within the blue open rectangle (a) are under the influence of CNV and are linked to tumour features. Table 1 lists the genes within panel a and their functional classification according to Gene Ontology.
Table 2.
Functional profile of genes linked to CNV and Gleason score.
Fig 7.
Expression of polarized genes in laser-capture micro-dissected tumour and normal tissue.
The differential expression patterns of polarised genes linked to CNV and Gleason score in the Tomlins et al LCM-based gene expression dataset. 36/58 genes were mapped from the network (see Fig 5) to the independent dataset. (A) Expression of all but 1 of the polarised differentially expressed in the LCM dataset (11/36) was downregulated in low and high-grade tumour tissue compared to healthy prostate. (B-D) Expression of several polarised genes linked to CNV and Gleason score in specific prostate tumour compartments isolated by the Tomlins et al study. Nor = normal prostate tissue, BPH = benign prostate hyperplasia, Adj = prostate tissue adjacent to tumour, PIN = intraepithelial neoplasia, PCA-L = low-grade tumour, PCA-H = high-grade tumour, Meta = metastatic tumour tissue. * p < 0.05, ** p < 0.01 *** p < 0.001.