Microarrays in ecological research: a case study of a cDNA microarray for plant-herbivore interactions

Microarray technology allows researchers to simultaneously monitor changes in the expression ratios (ERs) of hundreds of genes and has thereby revolutionized most of biology. Although this technique has the potential of elucidating early stages in an organism's phenotypic response to complex ecological interactions, to date, it has not been fully incorporated into ecological research. This is partially due to a lack of simple procedures of handling and analyzing the expression ratio (ER) data produced from microarrays.

We describe an analysis of the sources of variation in ERs from 73 hybridized cDNA microarrays, each with 234 herbivory-elicited genes from the model ecological expression system, Nicotiana attenuata, using procedures that are commonly used in ecologic research. Each gene is represented by two independently labeled PCR products and each product was arrayed in quadruplicate. We present a robust method of normalizing and analyzing ERs based on arbitrary thresholds and statistical criteria, and characterize a "norm of reaction" of ERs for 6 genes (4 of known function, 2 of unknown) with different ERs as determined across all analyzed arrays to provide a biologically-informed alternative to the use of arbitrary expression ratios in determining significance of expression. These gene-specific ERs and their variance (gene CV) were used to calculate array-based variances (array CV), which, in turn, were used to study the effects of array age, probe cDNA quantity and quality, and quality of spotted PCR products as estimates of technical variation. Cluster analysis and a Principal Component Analysis (PCA) were used to reveal associations among the transcriptional "imprints" of arrays hybridized with cDNA probes derived from mRNA from N. attenuata plants variously elicited and attacked by different herbivore species and from three congeners: N. quadrivalis, N. longiflora and N. clevelandii. Additionally, the PCA revealed the contribution of individual gene ERs to the associations among arrays.

While the costs of 'boutique' array fabrication are rapidly declining, familiar methods for the analysis of the data they create are still missing. The case history illustrated here demonstrates the ease with which this powerful technology can be adapted to ecological research.