Genomics Day 2005: Genomics Day 2005: Public Health Genomics at CDC
Group 2: Biomarker Discovery
Biomarker discovery for early detection of cervical neoplasia: gene expression profiling in exfoliated cervical cells
Steinau M1 , Lee DR1 , Rajeevan MS1 , Vernon SD1 , Ruffin MT2 ,Verma M2 , Srivastava S3, Unger ER1(1) Division of Viral and Rickettsial Diseases, NCID , DC ,
(2) University of Michigan,
(3) National Cancer Institute
Background: Biomarkers discovered in microdissected lesional tissues or cells often fail when tested in the mixture of cells collected by the noninvasive sampling methods useful in cancer screening. To avoid this problem we focused our discovery efforts on routinely collected exfoliated cells, hypothesizing that field effects associated with neoplastic expression could be identified. To further explore the possibility of using exfoliated cells for molecular screening, we compared the microarray gene expression of samples with cervical intraepithelial neoplasia grade 3 (CIN3) to those without or only low grade lesions (CIN1, CIN0)
Materials and Methods: Samples from 15 women with cervical intraepithelial neoplasia grade 3 (CIN 3) and 15 without (CIN 0/1) were selected from a population-based study. Total nucleic acid was extracted from exfoliated cells collected in methanol-based media. 500 ng RNA was labeled with biotin and hybridized to MWG A, B and C 10K Arrays (~30,000 genes). Signal was detected with Resonance Light Scattering particles and analyzed with ArrayVision. Expression data were normalized and statistically analyzed with BRB Array Tools, Significance Analysis of Microarrays (SAM) and FOCUS. We used the web based database DAVID for functional annotation and ontology, and the expression analysis systematic explorer (EASE) for identification of enriched biological themes.
Results: Using three different statistical approaches to identify differentially expressed genes resulted in lists of candidate genes with little overlap and relatively high estimates of false discovery rates. Six genes were found by all methods and all were upregulated in CIN3 samples two to four fold. Interestingly the within-class variation was greater for CIN 3 than for no disease or combined suggesting that cytologically identical CIN3 lesions may represent different molecular pathways to oncogenesis.
Conclusion: The found inconsistencies reflect the relatively small differences between disease classes compared to within-class variation. A more quantitative methodology such as real time RT-PCR would probably more suitable for the task, however the interim results support the feasibility of this approach for marker discovery.
Supported in part by NCI's Early Detection Research Network, Interagency Agreement Y1-CN-0101-01