Tumor mRNA expression profiles predict responses to chemotherapy.


The development of prognostic and diagnostic markers such as tumor staging and more recently molecular factors have greatly contributed to identifying lung cancer patients that can benefit from adjuvant or neoadjuvant chemotherapy. However, the predicting efficiencies vary considerably, and many therapies fail while surviving patients often experience severe toxicities. Moreover, patients displaying similar clinical characteristics often respond differently to therapy, and this is likely a result of heterogeneity of the tumors’ genetic and epigenetic characteristics. The advent of the human genome sequencing project and the concurrent development of many genomic-based technologies, including expression microarrays, have allowed scientists to explore the possibility of using expression profiles to predict tumor drug sensitivity or resistance before treatment, and thereby select the best possible therapies while decreasing the risk of toxicities for the patients. To this end, various gene signatures and sequence alteration in target genes have been obtained for prediction of drug response in patients. Examples include signatures for taxane/doxorubicin sensitivity in breast cancer, for platinum-based therapy in esophageal cancer, for epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (such as gefitinib and erlotinib) in lung cancers with mutated and often amplified EGFR, or for imatinib targeting Abl in chronic myelogenous leukemia. In many (but by no means all), these expression profiles were both “trained” (developed) and “tested” (validated) on primary tumor samples, and as such, provided that their reproducibility can be confirmed, provide important new ways to design and carry out clinical trials. To provide a current summary, we have gathered together in Table 1 the large majority of such trials correlating expression profiles of tumor cells before therapy with the tumor response or survival after such therapy. In all of these studies, tumor cells are obtained and undergo expression profiling, the patients are treated with similar regimens, and the tumor response, survival, and time to progression (“response phenotype”) are determined. From this, various biostatistical approaches often comparing the extremes of complete responders to tumor progressors are used to develop expression profiles that would identify these two clinically important groups. The ultimate clinical use would be in standard practice to profile tumors and enable the use of only those therapies to which patients are predicted to respond. Of course, the sensitivity and accuracy of these predictions is of vital importance, as are whether different treatment regimens exist that provide a selection so that the majority of patients have the possibility of an excellent response to at least one regimen. Thus, tumors could respond well to a chemotherapy regimen, but would also respond well to all of the standard chemotherapy regimens—a phenotype of general chemosensitivity. Although this would be of use to know, clearly it would be much better to have the situation where one tumor would respond to regimen “A” but not to “B,” whereas another tumor would have the converse phenotype. Buried within these questions is whether there exists “cancer stem-cell” populations, and how the drug response phenotype of this subpopulation compares with the tumor population as a whole, and whether the expression phenotype predicting cancer stem cells’ response can be determined from the population as a whole or only from a cancer stem-cell subpopulation. If the latter were true, then the task would become technically more difficult. Although this editorial focuses on genome-wide mRNA profiles as predictive biomarkers, all of these same features apply to consideration of using genome-wide DNA copy number changes (eg, with high-density single nucleotide polymorphism arrays), genome-wide DNA methylation changes, proteomics, and expression profiling of large numbers of proteins (such as on reversephase protein array), mRNA profiling, or profiling for expression of the proteome or specific cytokine and angiogenic factors in a patient’s blood. Finally, this editorial discusses “tumor autonomous” changes in mRNA expression profiles. Clearly, the tumor interacts with its microenvironment through a variety of autocrine and paracrine mechanisms, and there undoubtedly will be biomarkers of tumor response that focus on the microenvironment that will need to be developed because that microenvironment may properly be the therapeutic target (eg, the tumor vasculature, which is targeted with the anti–vascular endothelial growth factor monoclonal antibody bevacizumab). One disadvantage of these types of studies involving clinical specimens is the limitation in the number of chemotherapeutic drugs that can be tested, as well as their dependence on wellestablished and approved drug therapies. To circumvent this, many groups have been using preclinical models that make use of human tumor cell lines and/or xenografts to investigate gene expression profiles associated with in vitro sensitivity (drug response phenotypes) to hundreds or even thousands of drugs. This JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 25 NUMBER 28 OCTOBER 1 2007

Extracted Key Phrases

3 Figures and Tables


Citations per Year

682 Citations

Semantic Scholar estimates that this publication has 682 citations based on the available data.

See our FAQ for additional information.

Cite this paper

@article{Minna2007TumorME, title={Tumor mRNA expression profiles predict responses to chemotherapy.}, author={John D. Minna and Luc Girard and Yang Xie}, journal={Journal of clinical oncology : official journal of the American Society of Clinical Oncology}, year={2007}, volume={25 28}, pages={4329-36} }