A bone-seeking clone (MDA-231BO) of the human breast cancer cell line MDA-231 was constructed by sequential passages in mice and the metastatic cells collected from bone [
17]. The biological characteristics of MDA-231BO were identified with higher occurrences in metastasis to the bone compared with the MDA-231 parental cells (MDA-231 cell line) [
17]. In order to discover a sensitive and specific biomarker for detection of early bone metastases in breast cancer, we analyzed the differential gene expressions of MDA-231BO, a bone metastatic breast cancer cell line and compared it to MDA-231, a non-bone metastatic breast cancer cell line. The total RNA was extracted from MDA-231BO and MDA-231 cell lines, as previously described [
18]. The total RNA (100 ug) was used to produce labeled cDNA by anchored oligo (dT)-primer reverse transcription using SuperScript II reverse transcriptase (Life Technologies, Inc, Carlsbad, CA) in the presence of fluorescent dye, Cy5-dUTP or Cy3-dUTP (Amersham, Piscataway, NJ), respectively. The robot in the cDNA microarray facility at the Albert Einstein College of Medicine (AECOM) has the precision to spot 9568 PCR products onto a single glass slide. From Genome Systems, we have obtained 9568 unique human cDNAs from the I.M.A.G.E. consortium that represent 15–20% of all human genes [
19] (AECOM website:
http://microarray1k.aecom.yu.edu). The fluorescent cDNA probes were then hybridized to Silane glass slides with 9568 cDNA human gene spots were hybridized according to AECOM standard protocol in which each slide was probed with bone metastatic breast cancer and non-bone metastatic breast cancer. Slides were scanned in our microarray facility scanner. Data from the hybridization reactions are collected using a two-colored laser scanning confocal microscope that is custom designed and built at AECOM specifically for maximum sensitivity necessary to measure low abundance mRNAs. The images were exported to the GenePix Pro 3.0 software (Axon Instruments, Inc, Union City, CA) for signal intensity analysis. Poor-quality spots were flagged and excluded from analysis. Signal intensity information was exported to Excel. The data was normalized for statistical analysis. Locally weighted linear regression (LOWESS) analysis is a widely used normalization method to reduce systematic errors in the measured expression levels. Normalization algorithms can be applied either globally (to the entire data set) or locally (to some physical subset of the data) by GenePix software. The
output file of normalized data can be saved directly to your hard drive, and can be used as the input file for the "Filtering" module. The "Normalization" module also generates a graphical file containing data plots of Ratio vs. Geometrical Mean of R&G Intensity (before and after normalization). This module performs data filtering by eliminating bad/absent flag spots and weak signals (lower than "A" cutoff). The
input file was the result file from the "Normalization" module. The
output file can be saved directly to your hard drive, and can be used as an input file for the "Comparison" module. This module performs Ratio filtering. For one dataset (chip), this module performs Ratio filtering (Up & Down regulated genes). When comparing multiple datasets, the Ratio filtering test must be satisfied in all datasets for a given gene to be selected. Testing is satisfied by expression ratios in either the same or opposite direction. The
input file was the result file from the "Filtering" module. Data were sorted based on "fold change," and genes with at least 1.5-fold up-regulation or down-regulation were accepted as significant alteration and the genes have a significant differential expression in at least four of five experiments chosen for further consideration.