In this study we used an integrated bioinformatics approach () to analyze and interpret the proteomics and gene expression data from radiation treated cells with mutant or wild type ATM genotypes. The iProXpress system provides a protein-centric data integration for functional analysis and allows direct comparison of different molecules (mRNA vs. protein) from same samples under study. As functional understanding of the omics data is underpinned by the current knowledge annotated in databases for given lists of genes or proteins from high throughput experiments, it is crucial to maximize the use of known knowledge from heterogeneous databases and resources. The iProXpress system uses both iProClass and UniProtKB databases for data mapping, data analysis and interpretation, and also takes advantage of the extensive informatics infrastructure at PIR, e.g. the Text Search engine for data browsing and searching. One of the most useful features of the iProXpress system is to allow comparison of functional profiles across multiple data sets or groups obtained from different issue/cell types and time points, or from different omics experiments. In particular, while differential profiles with GO slim or pathway terms may not be evident when generated from combined data groups, profiles from more specific groups may reveal clearer differences. For example, purine metabolism became evident when examining individual time points from both AT5BIVA and ATCL8 cells.
While existing annotations from databases are critical for the omics analysis, the knowledge base is still limited. GO has become a common standard for annotation and functional analysis, but currently only about half of all human genes/proteins are annotated with GO terms, and even less with experimentally validated and manually annotated GO functions. Compared to GO profiling, pathway and network mapping provide more biological insight, however, an estimated <10% of human genes/proteins have been annotated with biological pathways in databases. Therefore, as part of the integrated bioinformatics approach, expert-guided analysis should be coupled with review of scientific literature for functional interpretation of the large scale omics data and for formulation of scientific hypothesis.
The expression profiling and pathway/network analyses have shown that enzymes of purine metabolisms, especially surrounding steps of the ADP/ATP and GDP/GTP synthesis, were differentially affected in irradiated AT5BIVA and ATCL8 cells. RRM2 is a small subunit of the RR complex that is well known for its role in DNA synthesis. RR is the only enzyme responsible for the reduction of ribonucleotides to their corresponding deoxyribonucleotides, providing a balanced supply of precursors for DNA synthesis and repair. It has been shown that an increase in RRM2 protein levels and RR activity in human nasopharyngeal cancer cells results in ionizing radiation resistance, which appears mediated by enhanced ionizing radiation damage repair during G2 phase of the cell cycle. However, overexpression of the large subunit, RRM1, of RR in these cells did not affect RR activity or ionizing radiation response (Kuo et al., 2003
). RRM2 overexpression is also associated with gemcitabine chemoresistance in pancreatic adenocarcinoma cells, and that suppression of RRM2 expression using RNA interference enhances gemcitabine-induced cytotoxicity in vitro (Duxbury et al., 2004
). Human RRM2 has been shown to be a target of p53 through direct protein-protein interaction that leads to the nuclear accumulation of RR subunits after UV exposure (Xue et al., 2003
), and inhibition of RRM2 by hydroxyurea results in increased sensitivity to UV irradiation in prostate cancer (PC3) cells (Zhou et al., 2003
). Our results suggest that RRM2 is involved in the ATM and p53-mediated signaling pathway leading to DNA repair in response to radiation in ATCL8 cells, while the ATM-mutated AT5BIVA cells became more sensitive to radiation possibly due to the impaired activation of RRM2 expression.
Most of proteins in this study were derived from the 2D-gel/MS experiment, and not all identified proteins from given 2D-gel spots were responsible for the observed changes. We used this integrated bioinformatics to help rational selection of candidate proteins for validation. Based on common pathways (e.g. purine metabolism) and their differential expression patterns, we can preferentially select those proteins commonly associated with a pathway over those not associated with the pathway for validation. Indeed, the enzyme RRM2, identified from a spot with 40 identified proteins at 3hr and a spot with 12 identifications at 1hr in ATCl8 cells (not shown), was actually one that was most likely to have changed, also consistent with the finding that RRM2 mRNA was up-regulated at 1hr in the same cells.
It was noted that the intersection between changed proteins and genes from the proteomics and gene expression data in this study was small. The lack of direct correlation between changes in proteins and genes from gene expression and proteomics experiments has been previously observed (Jansen et al., 2002
; Hewick et al., 2003
). This is due in part to the experimental artifacts and in part to differential post-transcriptional or post-translational regulation of genes or proteins. For example, an increased or new 2D-gel spot may result from increased protein phosphorylation without corresponding mRNA changes. Constructing gene regulatory networks may potentially help identify correlations between proteomics and gene expression data when direct correlation between the two is not apparent (Perco et al., 2005
Besides identifying RRM2 as a potential downstream target of the ATM-p53-mediated pathway for DNA repair in response to radiation, other enzymes in purine metabolism and several other metabolic pathways, such as AK2, IMPDH, and NDK, were differentially expressed as well in the two cell lines. Interestingly, three forms of NDKs (nucleoside diphosphate kinase) were observed to be down-regulated in AT5BIVA cells. NDKs have recently been found to have DNA binding and exonuclease activities (Yoon et al., 2005
). It is not clear however whether this is related to the reduced DNA damage repair in ATM mutated cells. Their roles and significance of these metabolic enzymes in the ATM-mediated pathways and in radiation responses remain to be further examined. Currently we are extending this study by applying metabolomics measurement to the two cell lines after irradiation, aiming to identify changes in metabolites in response to irradiation and the anticipated differential patterns in wild-type ATM vs. mutant ATM-expressing cells. Our current proteomics and gene expression data will provide a valuable reference for future analysis and interpretation of radiation damage-induced metabolites. We envision that integration and correlation of proteomics, functional genomics and metabolomics data generated from the same experimental system will provide new biological insight.
In conclusion, we have demonstrated an integrated bioinformatics approach that includes expert-guided examination of data to define radiation-induced and ATM-mediated pathways in cell models with wild-type or mutant ATM genotype. We have shown that purine metabolic pathways were differentially affected in response to radiation, and that RRM 2 was up-regulated only in ATM-wild type but not in ATM-mutated cells. We hypothesize that in this cell model, ionizing radiation activates ATM-p53-mediated pathway that directly targets RRM2 and leads to DNA damage repair, thus increasing radiation resistance in the ATCL8 cells.