Connecting genes, coexpression modules, and molecular signatures to environmental stress phenotypes in plants

David J Weston 2 Lee E Gunter 2 Alistair Rogers 0 1 Stan D Wullschleger 2 0 Department of Crop Sciences, University of Illinois at Urbana Champaign , Urbana, IL 61801 , USA 1 Environmental Sciences Department, Brookhaven National Laboratory , Upton, NY 11973-5000 , USA 2 Environmental Sciences Division, Oak Ridge National Laboratory , Oak Ridge, Tennessee 37831-6422 , USA Background: One of the eminent opportunities afforded by modern genomic technologies is the potential to provide a mechanistic understanding of the processes by which genetic change translates to phenotypic variation and the resultant appearance of distinct physiological traits. Indeed much progress has been made in this area, particularly in biomedicine where functional genomic information can be used to determine the physiological state (e.g., diagnosis) and predict phenotypic outcome (e.g., patient survival). Ecology currently lacks an analogous approach where genomic information can be used to diagnose the presence of a given physiological state (e.g., stress response) and then predict likely phenotypic outcomes (e.g., stress duration and tolerance, fitness). Results: Here, we demonstrate that a compendium of genomic signatures can be used to classify the plant abiotic stress phenotype in Arabidopsis according to the architecture of the transcriptome, and then be linked with gene coexpression network analysis to determine the underlying genes governing the phenotypic response. Using this approach, we confirm the existence of known stress responsive pathways and marker genes, report a common abiotic stress responsive transcriptome and relate phenotypic classification to stress duration. Conclusion: Linking genomic signatures to gene coexpression analysis provides a unique method of relating an observed plant phenotype to changes in gene expression that underlie that phenotype. Such information is critical to current and future investigations in plant biology and, in particular, to evolutionary ecology, where a mechanistic understanding of adaptive physiological responses to abiotic stress can provide researchers with a tool of great predictive value in understanding species and population level adaptation to climate change. - Background The advent of high-throughput genome sequencing coupled with breakthroughs in the field of functional genomics has provided an unprecedented opportunity to study the molecular mechanisms that govern the dynamic behavior of cells, organs, and organisms [1]. Indeed, there are excellent examples documenting interdisciplinary use of these emerging technologies, from human genome SNP scans diagnostic of human disease susceptibility [2,3] to discovery of the genetic mechanisms underlying beak morphology of Darwin's finches [4]. Applications are also apparent in plant biology, where the use of genomic technologies have uncovered stress-dependent behaviors in mechanistic detail (see [5] for a review). Such studies have led to the elucidation of highly complex and interacting networks of the abiotic stress response. For example, salinity, drought, and cold elicit a dehydration response that shares many common elements and interacting pathways [6,7]. These findings have spurred additional investigations searching for shared signaling cascades or molecules associated with pathway integration, or crosstalk, and have led to numerous candidates including reactive oxygen species (ROS) and calcium signaling [8,9], hormones [10,11] and others [12-14]. However, despite the advances made possible by "omics"-based technologies, we still struggle to accurately associate the genes, transcriptional cascades, and signaling networks with physiological performance and ecological fitness. One obstacle to this lack of association is perhaps the result of two opposing paradigms often used in comparative physiology [15]. The first approach, termed gene-tophenotype, is typified by that of many "omics"-based studies where the effects of specific genes on phenotypic performance and fitness are evaluated (e.g., a reverse genetics approach, [16]). This is in contrast to the phenotype-to-gene approach where the biologist attempts to determine the evolutionary potential of a given trait within a population without identifying the underlying genes (e.g., ecological genetics [17]). Thus, the latter approach is interested in the potential for a trait to evolve, while the former focuses on the underlying genetic mechanism of a particular trait. The integration of both approaches will be an important component of the emerging field of evolutionary and ecological genomics, which aims to study adaptation of natural populations to their environment [18]. To fully understand the genetic mechanisms underlying physiological adaptation to abiotic stress, we must first begin to understand the complex biological processes of how the resultant phenotype is generated from the genotype and then seamlessly coalesce our newfound understanding with population and evolutionary genetics. To initiate this task, we have adapted and integrated two recent analytical advances from the biomedical community. The first approach uses a novel weighted gene coexpression network to determine signaling networks and core genes underlying disease states and evolutionary diversification [19-21]. The second approach explores the genomic signature concept as recently defined by Lamb et al. [22], and is currently used to connect the disease state of an organism with the underlying genes and possible drug treatments [23]. Our purpose is to determine if these techniques can be used to associate the abiotic plant stress transcriptome with common and specific pathways underlying phenotypic response in a manner that is conducive to current and future genetic studies. We address this by combining gene coexpression networks with the genomic signature concept to investigate transcript profiles for plants exposed to drought, osmotic, salt, cold, heat, and UV-B stress. Our intent is not to describe in exhaustive detail the genes unique to or common among these stresses, although we do this to some extent, but rather to illustrate the power of this approach and provide sufficient information so that we and others can evaluate the full potential of this technique for plant biologists and evolutionary ecologists. Results Arabidopsis stress gene coexpression network It is known that the plant stress response is characteristic of highly complex and often integrated signaling pathways [6-12]. To help elucidate the transcriptional networks associated with exposure to abiotic environmental stress, a weighted gene coexpression network was constructed as described in Zhang and Horvath [20] and in Materials and Methods from a subset of the AtGenExpress abiotic stress dataset [24]. The data subsets were determined by first analyzing all abiotic stress datasets separately for differential gene expression between control and treatment conditions using the (...truncated)