Temporal gene expression profiling applied to the hypoxic response

R. Seigneuric, F.J.M. Starmans, M.G. Magagnin, N.A.W. Riel, van, B.G. Wouters, Ph. Lambin

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review


Although DNA micro arrays are becoming more and more popular in the clinic, the temporal nature of the biological process being measured is seldom studied due to the large costs of micro arrays and the lack of appropriate statistical tools1. Our approach intended at profiling a temporal signature of hypoxia which is a common feature of solid tumors leading to resistance to therapy and increased malignancy. We used an experimental design to decrease the number of micro arrays needed2. RNA from a prostate cancer cell line (DU145) was collected at 8 different time points (0h, 1h, 2h, 4h, 8h, 12h, 16h, 24h) after initiation of hypoxia. For each time point, three independent biological replicates of the time series were obtained. They were pooled according to quality and quantity criteria, and hybridized to Affymetrix arrays (HG U133 Plus2.0). Then a supervised approach was applied to investigate the hypoxic response as a function of time. User-defined curves of interest were defined as templates and applied across all time points with a sliding window. After normalization, each of the 54,675 time series measured in parallel was rescaled on [0;1]. Probe sets with trends highly correlating (>0.9) to pre-defined templates were selected for the signature. This aggregated temporal signature was then tested on 2 independent clinical data sets. Each data set contains gene expression from breast tumors' biopsies of about 300 patients, together with their clinical annotation. A median cut was used to partition patients in 2 groups. The group with a high expression level of the signature was expected to exhibit poor survival (and vice versa). Interestingly, although our signature was extracted from a prostate cancer cell line, it was shown to be statistically significant on these 2 breast cancer data sets with a p-value of 0.00022 and 0.02 (on the van’t Veer and the Wang data set respectively). Thus, by focusing on the dynamics of a general feature of solid tumors, namely hypoxia, we provide a way to stratify patients so that their treatment can be individualized, and prognosis improved.ACKNOWLEDGEMENT: Maastro Lab, Genomic Center and BigCat in Maastricht, F. Buffa, A. Harris in Oxford, A. Begg, D. Nuyten, L. Wessels: NKI, Amsterdam.REFERENCES:1. Storey JD et al., Significance analysis of time course microarray experiments, Proc Natl Acad Sci USA, Sep 6;102(36):12837-42. (2005).2. Koritzinsky M, Seigneuric R, et al., The hypoxic proteome is influenced by gene-specific changes in mRNA translation, Radiother Oncol Aug; 76(2):177-86 (2005).
Original languageEnglish
Title of host publicationSeventh International Conference on Systems Biology
Place of PublicationJapan, Yokohama
Publication statusPublished - 2006


Dive into the research topics of 'Temporal gene expression profiling applied to the hypoxic response'. Together they form a unique fingerprint.

Cite this