RESULTS & DISCUSSION

Data Description

Data 1: Data produced by Konstantinopoulos et al who conducted microarray experiment using AffyMetrix GeneChip U133 Plus 2.0 measuring gene expression of 36M2 ovarian cancer cells with or without 100 μM carboplatin in vitro for 24 hours and harvested and processed the cells for RNA isolation at 24 hours, 30 hours, and 36 hours in drug free media. (GEO: GSE13525)

Data 2: Data produced by Almeida et al who conducted microarray experiment using HG-Focus AffyMetrix GeneChip measuring gene expression of A549 non-small lung cancer cell line (NSCLC) treated or untreated with 50 μM of cisplatin for 1 hour and harvested at 10 hours in drug free media. (GEO: GSE6410)

Figure 1

Figure 1. Volcano plot visualization of differential gene expression at (A) 24 hours, (B) 30 hours, and (C) 36 hours in Data 1 (p-value cutoff: 0.05).

In the research article by Konstantinopoulos et al., the authors performed a time-course microarray experiment to identify genes and pathways that were regulated by carboplatin treatment. They treated 36M2 ovarian cancer cells with 100 μM carboplatin in vitro for 24 hours and harvested and processed the cells for RNA isolation at 24, 30, and 36 hours after treatment. With their data published on GEO (Gene Expression Omnibus), we first normalized and annotated the raw data using Robust Multi-Array Average expression measure (RMA) and AnnotateDbi papckages in R. Then we calculated the differential gene expression by limma package in R and visualized by volcano plots (Figure 1). Among more than 50,000 genes analyzed, at 24 hour time point, there were 35 and 86 genes that were significantly down- or up-regulated (p-value < 0.05) compared to the control group, respectively (Figure 1A). At 30 hours after treatment, the points (genes) became more divergent and genes that were significantly down- or up-regulated became 439 and 1108, respectively (Figure 1B). At 36 hour time point, as expected, the number of significantly up-regulated genes increased to 1737, and 2054 genes were significantly down-regulated (Figure 1C). Therefore, the platinum-base drug carboplatin was able to continuously induce ovarian cancer cell response and alter the expression of an increasing number of genes with respect to time within at least 36 hours after 24-hour treatment.

Figure 2

Fanconi Anemia Pathway

Base Excision Repair

Homologous Recombination

P53 Signaling Pathway

Mismatch Repair

Nucleotide Excision Repair

Figure 2. GSEA analysis results of significantly enriched pathways in Data 1. (A) Normalized enrichment scores (NESs) of top 15 enriched KEGG pathways in the treated groups. (B) Heat maps of gene expression values of DNA repair pathways. (C) Enrichment plots of DNA repair pathways.

The horizontal bar graph shows the 15 most enriched pathways out of a total of 187 pathways assayed from Data 1 (Figure 2A). The NES (Normalized enrichment score) of each pathway was calculated by dividing actual ES (enrichment score) with the mean of ESs against all permutations of the dataset to account for the size of the set. The large positive NES indicates higher expression of members in the gene set in treatment groups compared with the control groups. In the bar graph, we observed several DNA repair pathways have been highly enriched. As we are interested in characterizing the endogenous DNA damage repair mechanism and damage response in cells treated with carboplatin, we then looked specifically into 6 representative biological pathways (Fanconi Anemia pathway, base excision repair pathway, homologous recombination pathway, P53 signaling pathway, mismatch repair pathway, and nucleotide excision repair pathway) of DNA repair/stress response and performed GSEA (Gene Set Enrichment Analysis) accordingly.

The gene lists of each DNA repair pathways were downloaded from KEGG (Kyoto Encyclopedia of Genes and Genomes) using KEGGREST API. In the heat maps (Figure 2B), expression values were represented as colors, where the range of colors (red, pink, light blue, dark blue) showed the range of expression values (high, moderate, low, lowest). In each pathway, the general patterns of the heat maps suggested that the selected genes were distinctly differentially expressed in the treatment groups compared to the control groups. In other words, these DNA repair pathways were significantly promoted by carboplatin treatment in different time series. The following enrichment plots (Figure 2C) showed that members of our gene set tend to occur toward the top of the gene lists of the selected pathways, in which case our gene set was enriched in the carboplatin treated cells rather than the control cells. The ES on the y-axis reflects the degree to which a gene set is overrepresented at the extremes (top or bottom) of the entire ranked list of genes.

Figure 3

Nucelotide Excision Repair

P53 Signaling Pathway

Figure 3. GSEA analysis results of significantly enriched pathways in Data2. (A) NESs of top 15 enriched KEGG pathways in the treated groups. (B) Heat maps of gene expression values of NER and P53 signaling pathway. (C) Enrichment plot of NER and P53 signaling pathway.

Similarly, Figure 3 shows the GSEA analysis results of the microarray Data 2 generated by Almeida et al. The authors had A549 non-small lung cancer cell line (NSCLC) treated or untreated with 50 μM of platinum-based drug cisplatin. However, the treatment time in this experiment was only 1 hour followed by 10 hour incubation in drug free media before the cisplatin-induced gene expression changes were investigated. As a result, as shown in Figure 3, only 2 endogenous DNA repair/stress responese pathways (nucleotide excision repair and P53 signaling pathway) were identified to be significantly enriched in the cisplatin-treated groups. However, considering the lower drug concentration and much shorter treatment time, the results do not rule out the possibility that pathways shown in Figure 2 would be significantly regulated in cisplatin treatment groups with longer treatment time. Additionally, because the two experiments used two different cancer cell lines, we cannot safely compare the efficacy of carboplatin and cisplatin despite the fact that stronger nucleophile cisplatin is probably able to induce more DNA damage and trigger apoptosis in cancer cells. In general, our heat maps and enrichment plots in Figure 2 and Figure 3 did show the strong enrichment of DNA repair pathways in platinum-based drug treated groups, which could be accountable for one of the drug resistance of cancer cells upon platinum-base chemotherapy by up-regulating endogenous DNA repair/stress response pathways.

Figure 4

Figure 4. Mapping and visualization of gene expression in enriched pathways from (A) Data 1 HR at 36 hour time point and (B) Data 2 P53 signaling pathway.

In Figure 4, using KEGG PATHWAY database and pathview package, KEGG pathway mapping analyses were performed on the log2 fold change values to provide an intuitive overview of the roles of regulated genes in the pathways. Here we showed homologous recombination repair and P53 signaling pathways as two of the top regulated pathways from Data 1 and Data 2. Green and red background color of the gene blocks highlight differentially expressed genes in the pathways. We can clearly identify that several genes encoding positive regulators in the pathways were up-regulated, such as Nbs1 and Rad54 in homologous recombination pathway from Data 1 shown in Figure 4A, while negative regulators such as cyclin D and cyclin B in P53 signaling pathway were down-regulated from Data 2 shown in Figure 4B.

Figure 5

Figure 5. Network and clustering of enriched pathways of Data 1.

Figure 5 shows the network of GO (Gene Ontology) pathway enrichment from carboplatin vs. control data. Cytoscape software was used to generate and cluster the GO network. Red and blue color of each dot indicates positive and negative NES, and the sizes of the dots represent the number of genes in each pathway. The thickness of the edges represent the similarity of shared genes between pathways. The network was grouped into several clusters in different sizes by the degree of similarity and shared genes, which helps us to intuitively visualize the degree of enrichment in DNA repair and communication with other biological processes in a general picture. We did not show the network analyses of GO enrichment network of the Data 2 because the lower drug concentration and shorter treatment time of the experiment did not provide significant and representative results with less than 25% FDR (false discovery rate).

In conclusion, by analyzing the microarray data generated from carboplatin and cisplatin control and treated cancer cells, platinum-based drugs are able to trigger endogenous DNA repair pathways to fight against DNA damage and serve as one of the major drug-resistance responses in cancer cells. The actual function relevance of the response still requires future investigation.

Page updated

Report abuse