Development of a common oligonucleotide reference standard for microarray data normalization and comparison across different microbial communities

Abstract

High-density functional gene arrays have become a powerful tool for environmental microbial detection and characterization. However, microarray data normalization and comparison for this type of microarray remain a challenge in environmental microbiology studies because some commonly used normalization methods (e.g., genomic DNA) for the study of pure cultures are not applicable. In this study, we developed a common oligonucleotide reference standard (CORS) method to address this problem. A unique 50-mer reference oligonucleotide probe was selected to co-spot with gene probes for each array feature. The complementary sequence was synthesized and labeled for use as the reference target, which was then spiked and cohybridized with each sample. The signal intensity of this reference target was used for microarray data normalization and comparison. The optimal amount or concentration were determined to be ca. 0.5 to 2.5% of a gene probe for the reference probe and ca. 0.25 to 1.25 fmol/microl for the reference target based on our evaluation with a pilot array. The CORS method was then compared to dye swap and genomic DNA normalization methods using the Desulfovibrio vulgaris whole-genome microarray, and significant linear correlations were observed. This method was then applied to a functional gene array to analyze soil microbial communities, and the results demonstrated that the variation of signal intensities among replicates based on the CORS method was significantly lower than the total intensity normalization method. The developed CORS provides a useful approach for microarray data normalization and comparison for studies of complex microbial communities.

FIG. 1.

Signal intensities of the reference target (ca. 0.0125 to 12.5 fmol/μl) cohybridized with 500 ng of gDNA target of an equal mixture of D. vulgaris, R. palustris, and S. oneidensis. The amounts of the reference probe ranged from ca. 0.05 to 10% of a gene probe. Error bars showing the standard deviations are presented.

FIG. 2.

Box plots of variations of hybridization signal intensities among inter- and intrareplicates. A reduction of variations was observed when the CORS method was used for normalization compared to the mean signal intensity (mean) as evidenced by the paired Student t test with P < 0.001 for 1,500, 300, and 75 ng of gDNAs and P < 0.05 for 15 ng of gDNAs of equal mixtures of D. vulgaris, R. palustris, and S. oneidensis.

FIG. 3.

Quantitative analysis of relationships for expression ratios of all genes (Δfur mutant RNA/wild-type RNA of D. vulgaris) between different normalization methods. (A) Dye swap versus CORS; (B) gDNA versus CORS.

FIG. 4.

Ratios (Δfur mutant RNA/wild-type RNA of D. vulgaris) of gene expression obtained from three normalization methods, dye swap, gDNA, and CORS.

FIG. 5.

Box plots of variations of hybridization signals among replicates of high-, low-, and no-contamination soil microbial communities. A reduction in variation was observed when GeoChip data were normalized with the CORS method compared to the conventional mean signal intensity (mean) method as evidenced by the paired Student t test with P < 0.001.