In this study, we set out to identify gene biomarkers of tumor cell responses to arsenic trioxide-induced cytotoxicity. Using the cytotoxicity data established by the Developmental Therapeutics Program of the NCI, we ranked the tumor cell lines of the NCI-60 panel by their susceptibility to arsenic trioxide-induced killing. Through this ranking we find that there is a general trend of tumor cell susceptibility to arsenic trioxide for different tumor types. For instance, leukemia cell lines are distributed in the range of sensitivity to arsenic trioxide relative to the other tumor types. By associating the baseline gene expression levels of the NCI-60 human tumor cell panel with the arsenic trioxide-specific drug screening results, we identified 209 potential gene biomarkers with baseline expression levels that were significantly associated with tumor cell susceptibility to arsenic trioxide. Of the 209 genes, 169 (80.9%) were associated with arsenic resistance whereas the other 40 (19.1%) were associated with arsenic sensitivity. As expected, there is an association of the gene expression levels of these 209 genes with tumor type whereby many of same types of tumors show similar patterns of gene expression. As an example, in these analyses it is evident that the baseline gene expression levels of leukemia tumor cells with sensitivity to arsenic-induced killing are similar and cluster together. Likewise, colon tumor cells that show resistance to arsenic-induced killing also show baseline gene expression levels that are similar to each other, yet quite distinct from the leukemia tumor cell lines.
We applied a systems biology approach to examine these differentially expressed genes and affiliated networks and pathways, as well as the biological processes underlying tumor cell responses to arsenic-induced cytotoxicity. More specifically, in order to establish the potential biological mechanisms that underlie tumor cell responses to arsenic trioxide, we analyzed the 209 genes for known protein-protein interactions and enriched biological functions. We identified 64 common biological functions that were related to tumor cell responses to arsenic trioxide. Not surprisingly, we found that genes that are associated with arsenic susceptibility in the NCI-60 panel are statistically enriched for biological functions related to tumorigenesis, including cancer, cell death, cell-to-cell signaling and interaction, tumor morphology, and other functions relating to cancer disease.
We were intrigued to find numerous transcription factors with known links to tumorigenesis as well as with known association to arsenic trioxide are among our most significant arsenic-susceptibility gene biomarkers. For example, the transcription factor ID1 is well known for its function in carcinogenesis [25, 28]. Furthermore, a study has shown that the ID1 was induced by inorganic arsenite and may contribute to cell survival after exposure to sodium arsenite . Our findings suggest a potential link between the expression level of this transcription factor and how tumor cells respond when exposed to arsenic trioxide.
By examining canonical pathways in the gene biomarkers, we identified the enrichment of the NRF2-mediated oxidative stress response pathway. Specifically, eight NRF2 target genes were identified as significantly associated and all eight target genes showed high expression in arsenic-resistant tumor cell lines. The NRF2 gene itself did not show an association of its baseline gene expression and arsenic susceptibility. These findings may indicate that the arsenic-resistant tumor cell lines express the same levels of NRF2 mRNA but with higher transcriptional activity compared to the arsenic sensitive cell lines.
NRF2 is a transcription factor that responses to environmental hazardous insults , including reactive oxygen species (ROS) . It has been a promising therapeutic target for various diseases [32–35] and recently linked to chemoprevention as well [14, 36, 37]. NRF2 works as a system with the protein Kelch-like ECH-associated protein 1 (KEAP1) . Under normal conditions, NRF2 is bound by KEAP1 . Exposure to NRF2 inducing agents results in the dissociation of NRF2 from KEAP1 and allows nuclear accumulation of NRF2, which triggers the expression of downstream target genes of NRF2 . The NRF2 signaling pathway has been related to cell survival  and previous studies shown that NRF2 deficiency was associated with decreased rates on cell proliferation and tumor formation . Interestingly, it has also been found that NRF2 and some of its downstream target genes were overexpressed in numerous tumor cell lines and human cancer tissues, which indicates its involvement in tumor formation [41–43]. NRF2 has also been shown to play a role in cellular responses to arsenic. For example, arsenic enhances the cellular expression of NRF2 at the transcript and protein levels and activates the expression of NRF2-related genes in skin cells . In addition, arsenic-induced malignant transformation of human keratinocytes appears to require constitutive NRF2 activation .
To validate our computational prediction that NRF2 may mediate tumor cell survival in response to arsenic, we generated lung carcinoma cells that were deficient for the expression of NRF2. Through the computational analyses we predicted that cells with lower levels of NRF2 would be more sensitive to arsenic trioxide-induced killing. The results of the knock-down experiments support this and show that, as expected, cells that are deficient for NRF2 show increased sensitivity to arsenic-induced cytotoxicity. It should be noted the lung carcinoma cells that were used for these experiments are among the most resistant tumor cells of the NCI-60 panel to arsenic trioxide. It is therefore noteworthy that these highly resistant tumor cells can be altered to show increased cell killing to arsenic trioxide via their expression levels of NRF2.
Several of the NRF2 target genes identified from our study are of interest and support our findings in this work. For example, TXN and TXNRD1 are the key components of the thioredoxin system , which is an anti-oxidant system that has been linked to redoxinduced cell death , cellular growth , and apoptosis . Previous studies shown that the redox status of TXN determines the sensitivity of human liver carcinoma cells (HepG2) to arsenic trioxide-induced cell death . Moreover, research indicates that targeting the thioredoxin system to induce tumor cell apoptosis might underlie the anti-cancer mechanisms of several therapeutic agents, including arsenic trioxide .
ABCC1 is another noteworthy NRF2 target gene, and it is also known as multidrug resistance-associated protein 1 (MRP1). ABCC1 has been associated with chemotherapeutic resistance in several types of cancer , including cancers of the kidney , breast , and prostate [54, 55]. ABCC1, as an ATP binding cassette protein, is believed to participate in chemotherapeutic agents transportation , including arsenic trioxide ; and possibly contributes to the chemoresistance in cancer treatment [51, 57]. Chemotherapy resistance has been a huge obstacle in cancer treatment, and multidrug transporters like ABCC1 provide promising targets in chemotherapy [58–60] and valuable information for drug development. Our results indicate that ABCC1 could be a gene biomarker of arsenic response, as well as a potential chemotherapeutic target when using arsenic trioxide in cancer treatment, for APL and possibly other tumor types.
Another interesting finding is the identification of the transcription factor NFκB as an integrated node in the arsenic-susceptibility sub-network. NFκB is well known for its function in regulating genes for immune response, inflammation and apoptosis [61–63]. Numerous studies have shown that the NFκB signaling pathway is altered in the presence of arsenic trioxide [64–66]. For example, NFκB has been shown to be activated by arsenic at environmentally relevant concentrations [64, 67–71] (reviewed in [72–74]). At higher doses, arsenic represses NF-κB activation . The varied responses of NF-κB upon exposure to arsenic are certainly influenced by arsenic dose, arsenic species, and cell type differences. Similar to NRF2, the baseline expression levels of NFκB were not statistically associated with tumor cell responses to arsenic trioxide. However, its transcriptional targets are. Previous studies have demonstrated the crosstalk between NRF2 and NFκB in biological processes including inflammation and carcinogenesis [76, 77], but the interaction between these two transcription factors under cellular stress is not clearly understood. Our results suggest that NRF2 and NFκB both may contribute to tumor cell resistance upon exposure to arsenic trioxide, and the two transcription factors may work cooperatively in protecting tumor cells from arsenic-induced cytotoxicity.