Exploring plant–microbe interactions using DNA microarrays


  • Box  1 The DNA array technique

    The DNA array technique is in principle very simple. Thousands of DNA sequences (typically presynthesized oligonucleotides or inserts from cDNA libraries) are printed onto glass slides or nylon sheets using a robotic arrayer. To compare the abundance of these genes in a sample, RNA or DNA is extracted (the ‘target’), labelled and hybridized to the arrayed DNA (the ‘probe’). After washing, the probe is detected by fluorescence scanning or phosphor imaging. The primary data in microarray experiments consist of scans of the array (images). The spots on the images are quantified and the intensities are normalized. The final step is to identify genes that are significantly up- or down-regulated and to identify clusters of coregulated genes (regulons). The rationale behind the approach is that genes displaying similarity in expression pattern might be functionally related and governed by the same genetic control mechanism (Brown & Botstein, 1999).

Functional genomics of plant–microbe interactions – the 10th New Phytologist Symposium, Nancy, France, October 2002

Functional genomics, facilitated by DNA microarray technology, has vast potential for our understanding of plant–microbe systems. But how useful are the data when there is limited genomic information and the organisms cannot yet be genetically manipulated? During the 10th New Phytologist Symposium in Nancy, France, the potential and the problems associated with using DNA microarray technology for studying the molecular background of plant–microbe interactions were discussed.

‘There is a danger that microarray experiments will lead to a vast accumulation of data that cannot be meaningfully interpreted’

The power of DNA microarray technology

Since their introduction in the mid-1990s, DNA microarrays (Box 1) have become one of the major tools in functional genomics for exploring the genome-wide patterns of gene expression in an organism (Colebatch et al., 2002a). The main application of the DNA microarray technique has so far been in the analysis of gene expression in model organisms such as Saccharomyces cerevisiae, Arabidopsis thaliana, Drosophila melanogaster and Caenorhabiditis elegans, simply because their complete genome sequences are available. However, DNA microarray technology is now rapidly being applied to studies of other organisms, including plant pathogens and symbionts (Martin, 2001). Complete genome sequences are available for a number of important bacterial pathogens and symbionts and genome sequencing of several fungi is under way. In the absence of fully sequenced genomes, information from large sets of expressed sequence tags (ESTs) is well suited for constructing cDNA arrays. Genome sequences are also becoming available for several plant hosts including rice, legumes and poplar.

There is, clearly, great potential in applying DNA microarray technology to examining plant–microbe systems, as this will substantially increase our knowledge of the genetic background behind these interactions. However, concerns have been raised about the usefulness of the data obtained from microarray experiments in organisms for which there is limited genomic information and that cannot be genetically manipulated. To what extent can the complex and large data sets generated from microarray experiments in nonmodel organisms be meaningfully interpreted and validated? How can microarray data from different experiments and labs be compared?

Expression profiles

The first demonstrations of the applicability of the DNA array technique for monitoring the gene expression of plant–microbe interactions originate from studies on defence reactions in Arabidopsis. Schenk et al. (2000) analysed the expression of genes in Arabidopsis either infected by the incompatible fungal pathogen Alternaria brassicicola or treated with the defence-related signalling molecules salicylic acid (SA), methyl jasmonate (MJ), or ethylene. Analysis of the expression data obtained indicated the existence of a considerable network of regulatory interactions and coordination of signalling pathways, which had not been observed previously when analysing a few genes at a time. Maleck et al. (2000) monitored changes in gene expression induced during the systemic acquired resistance response (SAR) in Arabidopsis. Various levels of the SAR response were observed following chemical treatment and when analysing mutants with a constitutive or repressed SAR phenotype. Groups of genes with common regulatory patterns were identified. In addition, a common promotor element in one of the regulons was identified by searching for binding motifs in the upstream regions of the predicted translation start sites. During the meeting in Nancy, Nikolaus Schlaich (RWTH-BioIII Pflanzenphysiologie, Aachen, Germany) reported on a study in which cDNA arrays were being used to examine changes in the metabolism of Arabidopsis during infection with the bacterial plant pathogen Pseudomonas syringae pv. tomato. Based on the patterns of genes expressed during infection, it was possible to identify major shifts in the metabolism of the host (Scheideler et al., 2002). In addition, Laurent Zimmerli (Department of Plant Biology, Stanford University, CA, USA) presented data from recent experiments comparing the expression of genes in Arabidopsis leaves infected by compatible and incompatible powdery mildew species.

As a result of the ever-increasing rate at which genomes and ESTs are being sequenced (Tunlid & Talbot, 2002), the DNA array technique is now rapidly being applied to studies of a number of parasitic and symbiotic microorganisms and their corresponding host plants. For example, arrays have been constructed to examine the interactions between arbuscular mycorrhizal (AM) fungi and legumes (Martin Parniske, John Innes Centre, Norwich, UK; Philipp Franken, Max Planck Institute for Terrestrial Microbiology, Marburg, Germany; Maria Harrison, The Samuel Roberts Noble Foundation, Ardmore, OK, USA; Franken & Requena (2001)), nitrogen-fixing bacteria and legumes (Michael Udvardi, Max Planck Institute of Molecular Plant Physiology, Golm, Germany; Colebatch et al. (2002b); Sprent (2002)), ectomycorrhizal fungi and trees (Sébastien Duplessis, INRA, Nancy, France; Voiblet et al. (2002); Anders Tunlid, Department of Microbial Ecology, Lund, Sweden), parasitic nematodes and plants (Pierre Abad, INRA, Antibes Cedex, France), and between pathogenic fungi and host plants (Regine Kahmann, Max Planck Institute for Terrestrial Microbiology, Marburg, Germany). Notably, sequences (probes) have been obtained in several of these projects from both the microbe and the host, which will allow direct examination of the interaction between the pathogen/symbiont and the host at the transcript level.

The application of microarray technology to these plants and microorganisms will provide a considerable amount of new data for the scientific community. Concerns were expressed that the characterization of gene expression patterns in these organisms will be much less valuable than those of model organisms such as Arabidopsis and S. cerevisiae. The interpretation of the data obtained from DNA array studies in these organisms is, to a large extent, dependent on our knowledge of the many genes that have been characterized by the classical methods of genetics, molecular biology and biochemistry. Corresponding data are lacking for most parasitic and symbiotic microorganisms and their host plants. However, the rate of evolution of many genes with respect to both sequence and function has been so slow that characterization in one organism can suffice for many or all. For example, even when sequences from such evolutionarily distant organisms as S. cerevisiae and C. elegans are compared, the function of 20% of the genes encoded by the nematode could be indicated by knowing the function of the yeast orthologue (Chervitz et al., 1998). This suggests that clusters of coregulated genes identified in DNA array experiments of plant–microbe interactions will, in many cases, contain at least some genes that encode proteins with orthologues that have been functionally characterized in other organisms including one or more of the models (Brown & Botstein, 1999). In many cases this information can serve as a starting point for generating new hypotheses for the mechanisms of pathogenesis and symbiosis. There are several other ways in which investigators applying the DNA microarray technique to plant–microbe systems can benefit from the efforts of those working with model organisms.

Design of microarray experiments

DNA microarray experiments are costly in terms of equipment, consumables and time. Therefore, it is important that the experiments be carefully planned and executed. Well-designed array experiments improve the quality and reliability of the data (Yang & Speed, 2002). The community of scientists working on functional genomics in Arabidopsis have devoted substantial efforts to determining the best practice for DNA array experiments. Experiments have shown that careful probe selection, physical design of the array, and experimental design, including the number of biological replicates, can have a considerable impact on the quality of the microarray data obtained. However, the best methods of scanning, extraction, normalization and data analysis have not yet been determined. Until this is done, it is recommended that each microarray experiment be run through a series of quality tests (Finkelstein et al., 2002). Further information can be found on the homepages of the Arabidopsis Information Resource (TAIR, http://www.arabidopsis.org) and the Genomic Arabidopsis Resource Network (GARNET, http://www2.york.ac.uk/res/garnet/garnet.htm).

Validation of results

Validation of DNA array results using other techniques is critical for establishing the biological significance of the data. Expression levels of key genes identified in array analysis should be confirmed using RT-PCR or Northern blots. In many cases the function of these genes cannot be inferred by sequence similarities to well-characterized genes, and their function must be examined by genetic and molecular methods. Although there are many plant symbionts and pathogens that cannot be genetically manipulated, there are a few that can be transformed to generate knockout and conditional mutants. Analysis of such mutants can provide important information on gene function. In addition, DNA array analysis in organisms exhibiting a loss of function mutation or over-expression of a transcription factor or genes involved in signalling pathways may result in the identification of downstream genes. Such analyses are now under way in the corn smut fungus Ustilago maydis (Regine Kahmann, Max Planck Institute for Terrestrial Microbiology, Marburg, Germany).

Furthermore, it should be remembered that expression profiling using DNA arrays only assays functionality in an indirect way. mRNA molecules are just transmitters of the instructions for synthesizing proteins, while it is proteins and metabolites that are the functional entities in the cell. Extensive efforts have been devoted to developing methods for analysing the proteome and metabolome in model organisms, but there are still major technical and conceptual difficulties in analysing these ‘omes’ (Oliver, 2002). Michael Udvardi (Max Planck Institute of Molecular Plant Physiology, Golm, Germany) presented data showing how metabolome analysis using GC-MS has been used to profile the metabolites in nitrogen-fixing nodules in legumes and to follow leads on potentially new aspects of metabolism uncovered by DNA array analysis. In addition, this group is using proteome analysis to identify nutrient transporters in the symbiosome membrane which separates the rhizobia from the plant cell cytoplasm.

Sharing microarray data

There will soon be a need to share DNA microarray data between research groups working on plant–microbe interactions. There are several reasons for this. One is that most array experiments will identify dozens (if not hundreds) of genes that are differentially regulated and only a few of them can be studied in detail by one laboratory. Another reason is that a common database on transcript profiles from a number of different experiments and plant–microbe systems can function as a ‘compendium’ for comparative studies on gene expression in different species, tissues, treatments and growth conditions. For such analysis it is essential that researchers have access to each other's raw transcriptome data, so that they can apply the normalization procedure that is most appropriate for the biological problem that they are studying (Oliver, 2002). Therefore, current work towards defining international standards for depositing microarray results in public databases is very welcome (Brazma et al., 2001) (http://www.mged.org).


The DNA microarray technique was developed for gene expression profiling in model organisms with complete genome sequences, but is now being widely applied in studies of other organisms including important plant pathogens, symbionts and their hosts. Exciting new information will be gained from these studies, including broadened knowledge on the molecular background of plant–microbe interactions, and the gap in information on gene function between these organisms and the currently favoured model organisms will narrow. However, there is a danger that microarray experiments may lead to a vast accumulation of data that cannot be meaningfully interpreted. For this reason, DNA microarray experiments should be carefully designed, array data must be validated using other techniques, and the expression data should be made available in databases with public access.


The 10th New Phytologist Symposium, ‘Functional genomics of plant–microbe interactions’ in Nancy, 23–25 October 2002 was sponsored by the New Phytologist Trust, INRA and The Noble Foundation. The workshop participants are grateful to the organizers Francis Martin, Maria Harrison, Nicholas Talbot and Jonathan Ingram.