A method for high-throughput gene expression signature analysis

Abstract

Genome-wide transcriptional profiling has shown that different biologic states (for instance, disease and response to pharmacologic manipulation) can be recognized by the expression pattern of relatively small numbers of genes. However, the lack of a practical and cost-effective technology for detection of these gene expression 'signatures' in large numbers of samples has severely limited their exploitation in important medical and pharmaceutical discovery applications. Here, we describe a solution based on the combination of ligation-mediated amplification with an optically addressed microsphere and flow cytometric detection system.

Figure 1

Method overview. Transcripts are captured on immobilized poly-dT and reverse transcribed. Two oligonucleotide probes are designed against each transcript of interest. The upstream probes contain 20 nt complementary to a universal primer (T7) site, one of 100 different 24 nt barcode sequences, and a 20 nt sequence complementary to the 3'-end of the corresponding first-strand cDNA. The downstream probes are 5'-phosphorylated and contain a 20 nt sequence contiguous with the gene-specific fragment of the upstream probe and a 20 nt universal primer (T3) site. Probes are annealed to their targets, free probes removed, and juxtaposed probes joined by the action of ligase to yield synthetic 104 nt amplification templates. PCR is performed with T3 and 5'-biotinylated T7 primers. Biotinylated barcoded amplicons are hybridized against a pool of 100 sets of optically addressed microspheres each expressing capture probes complementary to one of the barcodes, and incubated with streptavidin-phycoerythrin to label biotin moieties fluorescently. Captured labeled amplicons are quantified and beads decoded by flow cytometry. nt nucleotides.

Figure 2

Reproducibility of the method. Mean expression levels for each transcript under each condition were computed and the deviation of each individual data point from its corresponding mean was recorded. A histogram of the fraction of data points in each of 12 bins of fold deviation values is shown. This plot represents 1,800 data points (two conditions × 90 transcripts × 10 replicates).

Figure 3

Performance in a representative gene space. Total RNA from HL60 cells treated with tretinoin or vehicle (DMSO) alone were analyzed by LMF in the space of 90 transcripts selected from microarray analysis of the same material. Plots depict log ratios of expression levels (tretinoin/DMSO) reported by both platforms for each transcript, in each of nine classes. Correlation coefficients of the log ratios between platforms within each class are shown. Yellow bars represent microarray findings and green bars LMF findings. Ratios were computed on the means of three parallel hybridizations of the pooled product from three amplification and labeling reactions (microarray) or 10 parallel amplification and hybridization procedures (LMF) for each condition. Basal expression categories are 20-60 (low), 60-125 (moderate), and >125 (high). Differential expression categories are 1.5-2.5× (low), 3-4.5× (moderate), and >5× (high). DMSO, dimethyl sulfoxide.