• Striga hermonthica is a root hemiparasite of cereals that causes devastating loss of yield. Recently, a rice cultivar, Nipponbare, was discovered, which exhibits post-attachment resistance to this parasite and quantitative trait loci (QTL) associated with the resistance were identified.
• Changes in gene expression in susceptible (IAC 165) and resistant (Nipponbare) rice cultivars were profiled using rice whole-genome microarrays. In addition to a functional categorization of changes in gene expression, genes that were significantly up-regulated within resistance QTL were identified.
• The resistance reaction was characterized by up-regulation of defence genes, including pathogenesis-related proteins, pleiotropic drug resistance ABC transporters, genes involved in phenylpropanoid metabolism and WRKY transcription factors. These changes in gene expression resemble those associated with resistance to microbial pathogens. Three genes encoding proteins of unknown function, within a major resistance QTL on chromosome 12, were highly up-regulated and are excellent candidate resistance genes.
• The susceptible interaction was characterized by large-scale down-regulation of gene expression, particularly within the functional categories plant growth regulator signalling and metabolism, biogenesis of cellular components and cell division. Up-regulated genes included nutrient transporters, enzymes of amino acid metabolism and some abiotic stress genes.
Striga species are obligate root hemiparasites that infect the staple cereal crops (maize, sorghum, millet, and rice) of sub-Saharan Africa, causing severe losses in grain yield (Ejeta, 2007). The effects of the parasite are greatest on infertile soils and those most severely affected are the poorest subsistence farmers upon whom the weed exerts major impacts on poverty and health. Control of Striga has proved difficult for several reasons. First, the parasite and host lifecycles are intimately linked via an exchange of chemical signals; parasite seed germination and haustorial development occur only in response to host-derived chemical cues (Keyes et al., 2001; Yoder, 2001; Palmer et al., 2004; Humphrey & Beale, 2006). Second, the mechanisms underlying the negative impact of Striga on crop yield are complex. The parasites have a severe effect on host plant growth and development within days of attachment. Even a very small parasite biomass causes a shortening of internode length, leading to severe stunting of the plant (Gurney et al., 1999). The basis of this phenomenon is unknown but it has been suggested that Striga produces a toxic compound that moves into the host (Musselman & Press, 1995) or that the parasite perturbs plant growth regulator metabolism (Drennan & El Hiweris, 1979; Taylor et al., 1996; Frost et al., 1997). Later in the infection cycle, the parasites compete effectively for host carbon and nutrients, further reducing growth and yield (Frost et al., 1997; Gurney et al., 1999). As Striga has such a devastating impact on the host soon after attachment, the best control measures will be those that either prevent attachment or kill the parasite rapidly once attached.
It is clear that the use of Striga-resistant cultivars would represent a cost-effective control measure, as their cultivation does not require costly inputs from farmers. However, the use of resistant cultivars is limited by a lack of resistant germplasm and by a lack of understanding of the molecular genetic basis of both host susceptibility and resistance to Striga. Large numbers of sorghum and maize genotypes have been screened for resistance to Striga (Oswald & Ransom, 2004; Rich et al., 2004). Although a limited number of cultivated and wild relatives of sorghum show some post-attachment resistance to this parasite (Maiti et al., 1984; Mohamed et al., 2003; Rich et al., 2004), little is known about the molecular nature of the resistance. Recently, we have identified cultivars of rice that range from extremely susceptible (IAC 165) to highly resistant (Nipponbare) to an ecotype of S. hermonthica (Gurney et al., 2006). This resistance was characterized by the development of some necrosis around the site of parasite attachment and by the inability of the parasite to penetrate the endodermis (Gurney et al., 2006), although the endodermis did not appear to be more heavily lignified or thickened when compared with uninfected plants.
The discovery of resistance in rice to Striga is of significance as it is currently the best model cereal for molecular genetic studies. In order to begin the identification of the genetic basis of post-attachment resistance in Nipponbare, Gurney et al. (2006) carried out a quantitative trait loci (QTL) analysis using a mapping population of backcross inbred lines (BILs). QTL explaining a large proportion of resistance were discovered on seven chromosomes; six alleles conferring resistance were derived from Nipponbare (Gurney et al., 2006). Although QTL mapping allows regions of a chromosome associated with a particular phenotypic trait to be identified, the regions are often large and contain hundreds or even thousands of potential genes. A number of recent studies have combined QTL mapping with gene expression profiling using microarrays to identify potential candidate genes. This novel approach allowed the successful identification of candidate genes for ovariole number, a quantitative trait, in Drosophila melongaster (Wayne & McIntyre, 2002) and genes conferring resistance to Marek's disease, a herpes virus-induced T cell cancer in chicken (Liu et al., 2001).
In this study we aim to identify suites of genes and pathways underlying susceptibility in IAC 165 and resistance in Nipponbare to S. hermonthica by profiling changes in gene expression at different stages of infection using rice Affymetrix oligonucleotide arrays. In addition to global gene expression profiling, genes that are significantly up-regulated within the QTL for post-attachment resistance in Nipponbare are identified and their role as putative candidate resistance genes discussed.
Materials and Methods
Plant growth and infection with Striga hermonthica
Seedlings of the rice cultivars Nipponbare and IAC 165 (Oryza sativa L. subspecies japonica) were grown for 3 wk in a controlled-environment growth room under a 14 h photoperiod with an irradiance of 400 µmol m−2 s−1, day- and night-time temperatures of 27 and 23°C, respectively, and 60% relative humidity. Seeds were surface-sterilized with 10% bleach for 5 min. They were then placed between two sheets of moist glass fibre filter paper (GF/A Whatmann; BDH, Poole, UK) held between two blocks of moistened horticultural rockwool (Aquaculture, Sheffield, UK) for 6 d until they had germinated. Seedlings were then transplanted into root observation chambers (rhizotrons) as described previously (Gurney et al., 2006). Rhizotrons were supplied with 250 ml of 40% Long Ashton nutrient solution (Hewitt, 1966) containing 1 mol m−3 ammonium nitrate daily. Rhizotrons were randomized twice a week. Twenty-one-day-old seedlings were infected with 25 mg of preconditioned and pre-germinated Striga hermonthica L. seeds by aligning them along the roots using a paintbrush (Gurney et al., 2003). When investigating differences in post-attachment resistance and susceptiblity, pre-germination of Striga seed is essential to allow synchronous attachment of the parasites to the host roots (eliminating the possibility of differences in germination induction by each cultivar). Control plants were treated in a similar manner. Thirty-six replicate rhizotrons were established in the absence or presence of Striga for each rice cultivar.
Harvesting root material for microarray analysis
Root material was harvested 2, 4 and 11 d post-inoculation (dpi). Each rhizotron was opened, the roots were removed and rapidly, but gently, rinsed in distilled water to remove Striga seeds and, where present, Striga seedlings from the roots. Control roots were treated in an identical manner. Two independent replicate samples (each derived from the roots of six individual plants) were taken for infected or control plants for each cultivar at each time point. Samples were frozen in liquid nitrogen. The whole procedure took less that 2 min. In order to determine whether Striga RNA would cross-hybridize to the rice array, germinated Striga seeds were also collected and frozen in liquid nitrogen.
Extraction of RNA and hybridization to the rice genome oligonucleotide array
RNA was isolated using TRI reagent (Ambion, Huntingdon, UK) and then purified using an RNA purification column kit (Qiagen, Crawley, UK) according to manufacturer's instructions. RNA was quality-checked and quantified using a Bioanalyzer 2100 (Agilent Technologies, Cheadle, UK). RNA from two of the four independent replicate samples for each treatment were combined at this point, providing two replicates (each containing RNA from six different plants) for hybridization to the commercially available rice genome array (Affymetrix Inc, USA). Preparation of cDNA, cRNA, hybridization to the arrays and quality control checks were carried out at Syngenta Biotechnology Inc. (Research Triangle Park, NC, USA), as described in Zhu et al. (2003).
Analysis of microarray data
Data analyses were carried out using GeneSpring 7.2 (Silicon Genetics, Redwood City, CA, USA). Background-corrected expression data from each of the 24 CEL files was imported into GeneSpring. Normalization of the data was carried out independently for each cultivar. Two normalization procedures were carried out; a per-chip 50th percentile normalization followed by a per-gene median-centred normalization. Following normalization, probe sets whose flag call was ‘Absent’ in all treatments (i.e. the gene was not expressed) were removed from the analysis. A stringent statistical analysis consisting of a two-way ANOVA was performed between infected and control samples over time, and the Benjamini-Hochberg multiple testing correction was applied to the data (P ≤ 0.05). Following this analysis, only those genes with a hybridization signal of 50 or above and whose expression was at least twofold greater or lower in infected samples relative to uninfected tissue are reported here. There was very little hybridization of Striga RNA to the rice Affymetrix arrays. Of the 57 405 probes on the array, only 6304 (11%) showed any hybridization signal. Even when a hybridization signal was present, it was very low and did not interfere with the interpretation of the changes in gene expression in Striga-infected rice roots. Data files have been submitted to GEO database (accession number GSE10373).
Annotation of significantly regulated genes
Annotation of genes was updated regularly using the GeneSpider function in GeneSpring 7.2. Where appropriate, TIGR locus identifiers were used to retrieve the coding sequence for significantly regulated genes of unknown function, and these were used to perform translated BLAST searches of GenBank using the ‘blastx’ translated search to retrieve a best match identity for each element from the plant protein sequence database. Functional categorization of significantly regulated genes was carried out manually based on the MIPS (Munich Information Centre for Protein Sequences) functional catalogue (FunCat). Acquisition of a genome position for each probe set was carried out using two approaches. First, a physical position was obtained for each probe set with a unique locus match in the TIGR genome browser. Second, where no such match was possible, the probe set sequence was used to search the Oryza sativa genome in a standalone BLAST.
Verification of changes in gene expression using qRT PCR
In order to validate the results of the microarray experiment, a range of genes that were up- or down-regulated in infected compared with control tissue and which exhibited different temporal profiles of expression were verified by quantitative RT-PCR. An independent set of rice plants from those used for the microarray experiments were grown and infected with S. hermonthica as described previously. Root material was harvested at 2, 4 and 11 dpi. Three replicate samples were taken for each treatment at each time point. Each replicate consisted of small segments of an individual root system. RNA was isolated, purified and quantified as described earlier in this paper. Before the synthesis of cDNA, any residual genomic DNA was removed by treating the samples with Turbo DNA-free DNase treatment (Ambion, Huntingdon, UK) according to the manufacturer's instructions. First-strand cDNA was synthesized from 2 µg RNA using BioScript (Bioline, London, UK) and subsequently purified using a PCR purification kit (Qiagen) according to the manufacturer's instructions.
Gene-specific oligonucleotide primers were designed against genes that were selected from the microarray analyses based on their functional identities and expression profiles. (details of the primers are given in Supplementary material, Table S1.) The specificity of the primers was tested using PCR, employing DNA and cDNA from Nipponbare and IAC 165. Polymerase chain reactions were carried out using a Techne PCR machine (Scientific Laboratory Supplies, Nottingham, UK) in a reaction containing five units Taq polymerase (Promega, Madison, WI, USA), 10 mm Tris-HCl pH 9.0, 50 mm KCl, 0.1% Triton X-100, 1.5 mm MgCl2, 0.2 mm deoxynucleotide triphosphate (dNTP), 4 µm forward primer and 4 µm reverse primer. Reaction conditions were 95°C for 5 min, followed by 35 cycles at 95, 60 and 72°C for 30, 60 and 60 s, respectively, and finally 72°C for 6 min. Following PCR, each sample was mixed with 5× gel-loading buffer (Sigma-Aldrich) and electrophoresed on polyacrylamide gels (12% acrylamide (Protogel, Gentaur, Brussels, Belgium), 0.07% (w/v) ammonium persulphate (APS), 0.02% (v/v) TEMED, 90 mm Tris-borate, 0.5 mm EDTA). Electrophoresis was performed using a running buffer of 90 mm Tris-borate, 0.5 mm EDTA, after which gels were stained for 10 min in 40 mm Tris-acetate, 1 mm EDTA (0.005% ethidium bromide). Gels were visualized using an Epi Chem II Darkroom gel-documentation system (UVP Laboratory Products, Cambridge, UK).
The expression of these genes was quantified by quantitative real-time RT PCR as follows. Template cDNA from control and infected material was amplified in a 96-well optical reaction plate (Applied Biosystems, Warrington, UK) in a 25 µl reaction containing SensiMix (Quantace Ltd, London, UK), SYBR Green I, 0.2 mm forward primer and 0.2 mm reverse primer. qPCR was carried out using an ABI Prism 7700 Sequence Detector (Applied Biosystems) under the following cycling conditions: 50°C (2 min), 95°C (10 min), and then 35 cycles of 95°C (15 s) and 60°C (1 min). SYBR Green fluorescence was detected in real time, and threshold cycle (CT) values were obtained using Sequence Detector version 1.7 software (Applied Biosystems). Initial template amount was calculated by relating CT values to a standard curve constructed by amplifying genomic DNA of known concentrations. Expression values were normalized for differences in cDNA input using parallel reactions employing primers designed against a reference gene (Os01g52460; regulator of ribonuclease activity or Os01g16930; presenilin; Table S1) whose expression was not altered by infection with S. hermonthica (in the microarray experiment or as measured by qRT-PCR).
Linking changes in gene expression with QTL for resistance
An aim of this study was to interrogate the microarray data in regions predicted to contain QTL for resistance to S. hermonthica in Nipponbare as reported in Gurney et al. (2006). First we used a conservative 2 LOD drop rule (Lynch & Walsh, 1998) to define the interval containing the mutations responsible for each QTL. This process defined start and end points of each confidence interval measured in cM. To convert these locations to Mbp on the genome sequence, we next positioned the RFLP markers used in the mapping study (Gurney et al., 2006) onto the rice genome using a standalone BLAST search (i.e. we inferred a Mbp position of each RFLP). Finally, we converted the cM positions of the start and end of the QTL confidence interval into Mbp by interpolation, following a linear regression of RFLP position (Mbp) on RFLP position (cM). Using GeneSpring software, all probes set sequences that were significantly up- or down-regulated by more than twofold within the predicted QTL regions were identified.
The phenotype of the resistant and susceptible interaction
Representative sections of infected roots from Nipponbare and IAC 165 were harvested 2, 4 and 11 dpi and photographed using a Leica MZFLIII stereomicroscope fitted with a RTKE Spot CCD (charge coupled device) camera (Diagnostic Instruments Inc., Sterling Heights, MI, USA). In order to examine the extent of parasite development within the host root cortex, sections of root plus attached parasite were dissected from host roots 4 and 11 dpi. Samples were fixed, embedded, sectioned, stained and photographed as described in Gurney et al. (2003, 2006).
Results and Discussion
Analysis of global changes in gene expression in resistant and susceptible cultivars following infection by Striga hermonthica
Striga hermonthica is a root hemiparasite that severely affects the growth and yield of maize, sorghum millet and upland rice, yet we know relatively little about changes in gene expression that occur during susceptible or resistant interactions. Although total post-attachment resistance to Striga is rare, Gurney et al. (2006) reported very robust resistance in the rice cultivar Nipponbare to S. hermonthica; out of the 100 or so attachments to the root system of individual plants, on average, only one to two parasites formed xylem connections with the host. Thus, to gain an insight into the molecular basis of susceptibility and resistance, a genome-wide expression analysis was carried out at three time points, 2, 4 and 11 dpi, following infection of IAC 165 and Nipponbare roots by S. hermonthica. The phenotype of the interactions at these time points is shown in Fig. 1. In the compatible interaction (IAC 165), the parasite radical had attached to the host root by 2 dpi (Fig. 1a), and by 4 dpi the haustorium was clearly visible (Fig. 1b). It had already penetrated the host root cortex and breached the endodermis to form parasite–host xylem connections (Fig. 1d). Parasites then developed rapidly so that by 11 dpi they had between two and four leaf pairs (Fig. 1c) with fully differentiated haustoria (Fig. 1e). In contrast to the susceptible cultivar, most parasites that attached to Nipponbare roots failed to develop. The initial stage of infection was similar to that on IAC 165; by 2 dpi parasites had attached to the host root and had begun to penetrate the host root cortex (Fig. 1f). However, by day 4 a ring of necrosis was visible around the parasite (Fig. 1g), although the endophyte had penetrated into the host root cortex (Fig. 1i). By 11 dpi the ring of necrotic host cells had intensified (Fig. 1h) and the parasites were clearly dying. Most parasites were unable to form connections with the host xylem and, consistent with the observations of Gurney et al. (2006), the parasitic endophyte often encircled the host vascular system (Fig. 1j).
The Affymetrix rice genome array contains approx. 57 405 probe sets and, of these, 33 572 (59%) were expressed in root tissue from Nipponbare and 32 754 (57%) in root tissue from IAC 165. In Nipponbare undergoing a resistance reaction, approx. 4.9% of the expressed genes (1653 genes) were significantly up- or down-regulated by more than twofold, compared with approx. 6.3% (2079 genes) in IAC 165 undergoing a susceptible interaction. Although there was some overlap in the identities of up- and down-regulated genes in both the resistant and susceptible interactions, approx. 200 genes were unique to each (Fig. 2), providing a novel insight into the molecular basis of resistance and susceptibility to S. hermonthica. A key difference between the susceptible and resistant interactions was the extent to which gene expression was repressed. In IAC 165, 1749 genes were down-regulated as Striga developed on the roots, compared with 1295 genes in Nipponbare (two-way ANOVA; Benjamini-Hochberg false discovery rate) (Fig. 2). This pattern of gene expression (i.e. greater down- than up-regulation in the susceptible cultivar) is in agreement with the results of a proteomic study of resistance and susceptibility in pea to O. crenata (Castillejo et al., 2004). In the susceptible cultivar Messire, 34 proteins decreased in abundance, one increased and three were novel when compared with uninfected tissue. By contrast, during the resistant interaction between O. crenata and cultivar Pc 624, 15 proteins increased and only three decreased in abundance.
The differentially regulated genes were annotated (using the rice TIGR database) and assigned to functional categories derived from FunCat (Munich Information Centre for Protein Sequences, MIPS). Fig. 3 shows the total number of genes up- and down-regulated in different functional categories in roots of Nipponbare and IAC 165 following infection by S. hermonthica. In Striga-infected Nipponbare roots, a greater number of genes were up-regulated within the functional categories gene expression/protein fate and plant defence responses when compared with infected roots of IAC 165 (Fig. 3a,b). Conversely, in IAC 165, more genes were up-regulated within the metabolism and cellular transport categories (Fig. 3a). In both Nipponbare and IAC 165, large numbers of genes were down-regulated in all categories.
In order to validate the microarray data, a number of genes were selected from different functional categories and their expression was measured in control and infected tissue using quantitative RT-PCR. RT-PCR profiles of these genes revealed that they exhibited the same temporal pattern and direction of change in gene expression in infected compared with control tissue (up- or down-regulated) as observed in the microarray experiments (Fig. 4). In total, over 95% of genes profiled by microarray and RT-PCR exhibited the same change in gene expression, thus giving excellent confidence in the microarray data.
The molecular basis of the resistance response in Nipponbare
Plants exhibit a range of pre-existing and induced physical and chemical defence responses against microbial pathogens and insect pests. Induced defences include strengthening of cell walls, the production of phytoalexins and the synthesis of pathogenesis-related proteins (PRs) (van Loon et al., 2006). Rapid activation of programmed cell death (the hypersensitive response (HR)) is a characteristic feature of many plant defence mechanisms, particularly those governed by gene-for-gene resistance (Greenberg & Yao, 2004). Complex signalling pathways, often involving salicylic acid (SA), jasmonic acid (JA) and/or ethylene, lead to the activation of suites of defence genes, the exact response being dependent on the host–parasite interaction (Feys & Parker, 2000; Glazebrook, 2001). Although significant progress has been made in understanding the molecular nature of defence against microbial pathogens, much less is known about plant–plant defence responses. This study has revealed that many of the genes and pathways that are up-regulated in roots of Nipponbare undergoing a resistance response to S. hermonthica resemble those observed in defence against fungal and bacterial pathogens (Table 1).
Table 1. Representative defence-related genes up-regulated only in the roots of the rice cultivar Nipponbare undergoing a resistant interaction with Striga hermonthica
Days post-inoculation (dpi)
Expression values represent fold change in expression (infected relative to control tissue). Genes were significantly up-regulated according to a two-way ANOVA (using Benjamini-Hochberg multiple testing correction P < 0.05). (a–i) indicate that these genes were profiled by RT-PCR and are shown in Fig. 4.*These genes were up-regulated in both Striga-infected roots of Nipponbare and IAC 165. Data presented are for Nipponbare.
A characteristic phenotype associated with the resistance response to S. hermonthica was host cell necrosis (Fig. 1). Although it is not clear whether this phenotype represents a classic HR reaction, a gene encoding a hypersensitive response protein homologue PrMC3, (a 2-hydroxyisoflavanone dehydratase) (Walden et al., 1999) and another putative hypersensitive-induced response protein were up-regulated (Table 1). The accumulation of reactive oxygen species (ROS) and the expression of NADPH oxidases and peroxidases are often associated with the development of an HR (Torres et al., 2005), and there was evidence of necrosis involving ROS in sunflower resistant to Orobanche cumana (Letousey et al., 2007). In this study, there was little evidence for enhanced expression of these genes; indeed, many genes encoding peroxidases were repressed in infected roots (data not shown). An HR-like phenotype has also been reported in some sorghum cultivars which exhibit resistance to S. asiatica (Mohamed et al., 2003) and in some cowpea cultivars resistant to specific races of S. gesnerioides (Botanga & Timko, 2005).
Recognition of an invading microbial pathogen is often followed by the de novo synthesis and accumulation of a range of PR proteins and secondary metabolites (van Loon et al., 2006). Although expression of genes encoding some PR proteins was enhanced in both the susceptible and resistant interaction to S. hermonthica (Fig. 3b), many were only up-regulated in the resistant interaction and included genes encoding endochitinases (PR-3), glucanases (PR-2) and thaumatin-like proteins (PR-5) (Table 1). Up-regulation of these suites of PR genes is characteristic of defence against fungal pathogens, as chitinases and β 1-3 glucanases hydrolyse chitin and β 1-3 glucan in fungal cell walls, respectively (Kim et al., 2004; Zhu et al., 2006). The mode of action of such PR proteins against an invading parasitic plant is unclear. Recent studies have revealed that chitinases are structurally related to a class of xylanase inhibitors (Durand et al., 2005), which interfere with degradation of cell wall material. A gene encoding a xylanase inhibitor was also up-regulated in Nipponbare roots (Table 1), and it is possible that the large number of endochitinases up-regulated during the resistance response could interfere with parasite growth in the root by altering cell wall extensibility. A probenazole-induced gene, PBZ1, encodes a protein with homology to intracellular PR proteins. This gene was induced in rice in response to both incompatible and compatible isolates of M. grisea, but the gene was induced more quickly and to a greater extent in the incompatible interaction (Midoh & Iwata, 1996). PBZ1 was also induced in roots of Nipponbare and IAC 165 in response to Striga infection, but again the gene was induced more quickly and to a higher degree during the resistance response (Table 1 and Fig. 4e). The mode of action of the PBZ1 protein is at present unknown.
There is increasing evidence that PR proteins are also induced in host roots in response to other parasitic plants. For example, Perez-de-Luque et al. (2006) demonstrated, by in situ hybridization, that a gene encoding a β-glucanase was up-regulated in pea roots exhibiting resistance to O. crenata, and Borsics & Lados (2002), using differential display, demonstrated the induction of several PR proteins in shoots of alfalfa infected with Cuscuta trifolii. A proteomic analysis (by two-dimensional electrophoresis) of roots of a resistant (Ps 624) and a susceptible (Messire) pea cultivar infected with O. crenata revealed that at least 22 different protein spots differentiated control, noninfected, Messire and Ps 624 accessions. Three proteins that were induced in the resistant interaction were a cysteine proteinase, a β-1,3-glucanase, and an endochitinase (Castillejo et al., 2004). Several genes encoding enzymes involved in defence-related secondary metabolism, including naringenin 3-dioxygenases (involved in the diterpene phytoalexin biosynthetic pathway), chalcone synthase and phenylalanine ammonia lyase (PAL), were highly up-regulated in Nipponbare roots infected with S. hermonthica (Table 1) and such pathways have also been implicated in resistance of vetch, sunflower and pea to Orobanche parasites (Goldwasser et al., 1999; Serghini et al., 2001; Griffitts et al., 2004; Perez-de-Luque et al., 2005a).
Plants utilize a wide array of cytochrome P450 monoxygenases (P450s) in biosynthetic and detoxification pathways and several genes encoding P450s were highly up-regulated during the resistance response to S. hermonthica (Table 1). In addition, several genes encoding proteins of the pleiotropic drug resistance (PDR) subfamily of ABC transporters were up-regulated from 2 dpi. The ABC binding cassette family of transporters use ATP hydrolysis to actively transport a diverse array of compounds across biological membranes, including toxins, drugs, glutathione conjugates, peptides and secondary metabolites (Theodoulou, 2000; Yazaki, 2006). It has been suggested that some of the detrimental effects of Striga on the growth of its host may result from the production of a toxic metabolite which moves from the parasite to the host, but such a toxin has yet to be identified. It is possible that one or more of these proteins may be involved in the transport or detoxification of a Striga-derived metabolite or may be involved with the transport of secondary metabolic defence compounds. There is certainly increasing evidence for the involvement of ABC transporters in plant defence. For example, Jasinski et al. (2001) identified a PDR protein in tobacco leaves (NpABC1) that was involved in the efflux of an antimicrobial diterpenoid compound, sclareol, on to the leaf surface, and Campbell et al. (2003) identified a gene from Arabidopsis encoding a PDR-like ABC transporter (AtPDR12) that was pathogen-responsive. More recently, Kobae et al. (2006) demonstrated that loss of an Arabidopsis plasma membrane-localized PDR ABC transporter (AtPDR8) leads to hypersensitive cell death upon pathogen infection.
Mechanical barriers to infection are common in resistance to microbial pathogens. In sunflower challenged with the parasitic plant O. cumana, callose depositions and an ‘encapsulation layer’ preventing entry of the parasite were reported (Labrousse et al., 2001; Letousey et al., 2007). In addition, endodermal and pericyclic thickening and/or lignification have also been seen in several host species displaying resistance to Orobanche species (Goldwasser et al., 1999; Perez-de-Luque et al., 2005b) and in one sorghum cultivar (N13) displaying resistance to an ecotype of S. asiatica (Maiti et al., 1984). However, no differences in cell wall architecture or the degree of lignification of cell walls in the cortex or endodermis were detected in Nipponbare roots exhibiting resistance to S. hermonthica (Fig. 1), and there was little evidence for the enhanced expression of genes involved in callose or lignin biosynthesis. In fact, genes encoding proteins involved in cell wall modification, for example expansins, cellulose synthases, polygalacturonases and arabinogalacturonases, were down-regulated (data not shown).
Parasite recognition and signal transduction in Nipponbare roots infected with S. hermonthica
Members of the WRKY family of transcription factors are commonly involved in regulating pathogen defence, stress and senescence (Eulgem et al., 2000; Ulker & Somssich, 2004; Eulgem & Somssich, 2007; Ross et al., 2007) and rice is predicted to contain over 100 WRKY genes (Ross et al., 2007). In keeping with the considerable impact of S. hermonthica infection on defence gene expression in Nipponbare roots, many genes encoding WRKY transcription factors (including OsWRKY19, OsWRKY45, OsWRKY62, OsWRKY76 and OsWRKY77) were up-regulated (Table 1, Fig. 3b). Interestingly, the expression of genes encoding OsWRKY45, OsWRKY62 and OsWRKY76 was also enhanced in rice leaves infected with an avirulent strain of M. grisea (Ryu et al., 2006) and two of these genes OsWRKY45 and OsWRKY62 were up-regulated in rice tissue treated with SA (Ryu et al., 2006). Although it is not clear whether the resistance response to S. hermonthica involves SA signalling, several other genes that were enhanced in Striga-infected Nipponbare roots encode proteins that have been shown to be SA-inducible in Arabidopsis and wheat, including chalcone synthase, PAL, germin-like proteins and thaumatin-like proteins (Table 1) (Schenk et al., 2000; Jayaraj et al., 2004). However, there was no evidence for the up-regulation of rice homologs of the A. thaliana NPR1 gene, a key regulator of the SA pathway in Arabidopsis, or other key markers of SA (De Vos et al., 2005).
Although it is clear that resistance in cowpea to races of S. gesnerioides is controlled by one or, in some cases, two dominant resistance genes (Timko et al., 2007), little is known about the genetic/race structure of S. hermonthica and its relationship to host resistance. The QTL analysis of post-attachment resistance in Nipponbare to S. hermonthica (Gurney et al., 2006) suggested that resistance was polygenic, although it may be controlled by few genes of major effect rather than many genes of minor effect. In this study, a number of genes encoding resistance-like proteins or homologs of resistance genes were up-regulated in S. hermonthica-infected Nipponbare roots (Table 1, Fig. 3b). The products of such genes are often receptor-like kinases, and resistance proteins typically contain NBS LRR domains, which are highly diverse and are thought to be involved in pathogen recognition and signal transduction (Jiang et al., 2007). It is possible that such receptor-like R gene homologues could be functionally important in the resistance of Nipponbare to S. hermonthica but further work is required to clarify their role.
Integrating QTL and transcriptomics analyses to identify putative candidate resistance genes
The integration of information obtained from QTL analysis with microarray expression experiments is a potentially useful marriage of information that can shortcut conventional breeding or marker-assisted selection in identifying candidate genes (Wayne & McIntyre, 2002; Price, 2006). We have previously uncovered a number of QTL in Nipponbare (on chromosomes 1, 5, 6, 7, 8 and 12) that account for a large proportion of the resistance that this cultivar exhibits against S. hermonthica (Gurney et al., 2006). In order to identify those genes that were up-regulated within the QTL intervals, their start and end positions were converted from cM to Mbp positions on the rice genome (Table S2).
Twenty-four genes were significantly up-regulated within QTL only during the during the resistance response (Table 2). A number of the up-regulated genes encoded proteins characteristic of a resistance response and included endochitinase precursors (chromosomes 1 and 6), a receptor-like kinase, (chromosome 1) and a NAC domain-containing transcriptional regulator (chromosome 6 QTL) (Table 2). The QTL on chromosome 12 had the largest effect on post-attachment resistance (Gurney et al., 2006). Within this QTL interval, up-regulated genes included a resistance-response protein, R gene homologues and, immediately at the peak position of the QTL, a cluster of genes of unknown function. The latter (Os12g11620, Os12g11660 and Os12g11710) were highly up-regulated upon infection both on the microarray (Table 2) and by quantitative RT-PCR (Fig. 5a–c). These genes are particularly interesting as there was no hybridization signal for them when cRNA from IAC 165 (control and Striga-infected tissue) was hybridized to the microarray chips (the raw fluorescence values ranged from 0 to 12) and there was no amplification of gene products in quantitative RT-PCR reactions. In order to examine whether these genes were present in the genome of IAC 165 and another rice cultivar, Kasalath, a genomic PCR was carried out on DNA extracted from IAC 165, Kasalath and Nipponbare. Fig. 5(d) shows that PCR products were amplified successfully from Nipponbare but not from IAC 165 or Kasalath. Cloning and functional analysis of these genes is under way to determine whether they play a pivotal role in resistance to S. hermonthica. These genes would not have been identified as potential candidate genes by functional analysis of gene expression data alone, thus illustrating the value of integrating QTL and transcriptomic data.
Table 2. Up-regulated genes encoding proteins within regions of the Nipponbare rice genome predicted to contain quantitative trait loci (QTL) for resistance to Striga hermonthica (these genes were not differentially regulated during a susceptible interaction)
Values represent fold change in expression (infected relative to control tissue). Genes were significantly up-regulated according to a two-way ANOVA (using Benjamini-Hochberg multiple testing correction P < 0.05). (a–c) indicate that these genes were profiled by RT-PCR and are shown in Fig. 5.
Acidic endochitinase precursor
Aspartic proteinase nepenthesin-1 precursor
Receptor-like protein kinase 5 precursor
Dynein light chain LC6, flagellar outer arm
Protein disulphide isomerase
Elicitor-induced DNA-binding protein
Protein Z, putative
DNA-binding protein RAV1
Cell division protein AAA ATPase family
Basic endochitinase 1 precursor
NAC domain-containing protein 68
10-deacetylbaccatin III 10-O-acetyltransferase
Xylanase inhibitor protein 2 precursor
Disease resistance response protein 206
Caffeic acid 3-O-methyltransferase
Verticillium wilt resistance protein precursor (leucine-rich repeat)
NBS-LRR disease resistance protein
Changes in gene expression in the susceptible interaction
Cereals infected with Striga species exhibit characteristic changes in growth and allometry when compared with uninfected plants. These include severe stunting of the host and lower leaf, stem and root biomass although root : shoot ratio is usually greater than in uninfected plants (Frost et al., 1997; Gurney et al., 1999; Oswald & Ransom, 2004). It has been hypothesized that alterations in plant growth regulator metabolism may account for some of the changes in host allometry following infection by S. hermonthica. Many genes involved in auxin and gibberellin signalling were down-regulated during infection (Table 3). Gibberellins are involved in the regulation of various growth processes, including stem elongation (Sakamoto, 2006), and their absence can result in dwarfism in some plants. It is possible that the reduced internode extension in Striga-infected cereals is the result of perturbations in gibberellin signalling, an area that requires further investigation.
Table 3. Genes encoding proteins involved in plant growth regulator (PGR) biosynthesis and signalling, up- or down-regulated in the susceptible rice cultivar IAC 165 following infection with Striga hermonthica
Days post-inoculation (dpi)
Values represent fold change in expression (infected relative to control tissue). Genes were significantly up- or down-regulated according to a two-way ANOVA (using Benjamini-Hochberg multiple testing correction P < 0.05). (j) gene was profiled by RT-PCR and is shown in Fig. 4.
ABA response element binding factor
Cytokinin dehydrogenase 5 precursor
OsSAUR5 – auxin-responsive SAUR gene family member
OsSAUR33 – auxin-responsive SAUR gene family member
Auxin-induced protein PCNT115
IAA-amino acid hydrolase ILR1 precursor
OsIAA7 – auxin-responsive Aux/IAA gene family member
Gibberellin regulated GAST1 protein precursor
Gibberellin receptor GID1L2
Gibberellin receptor GID1L2
Gibberellin receptor GID1L2
Gibberellin receptor GID1L2
Gibberellin-regulated protein 1 precursor
Gibberellin-regulated protein 1 precursor
Abscisic acid (ABA) has also been suggested to play a role in the response of plants to Striga infection. It has been known for a long time that, following attachment of Striga to host roots, stomata close, rapidly leading to low host stomatal conductance and low rates of photosynthesis (Press et al., 1991; Frost et al., 1997). These symptoms together with lower stem height and the increase in the root : shoot ratio have been attributed to increased concentrations of ABA in xylem sap and leaves, although this response is very variable (Drennan & El Hiweris, 1979; Frost et al., 1997; Taylor et al., 1996). Furthermore, it has been suggested that the elevated concentrations of ABA might result from a wounding effect of the parasite as it penetrates through the root cortex, or to drought stress owing to the diversion of water from host to parasite (Taylor & Seel, 1998). In roots of Striga-infected IAC 165, there was evidence for the up-regulation of ABA-responsive genes, although this was not extensive (Table 3). In addition, genes encoding one wound-induced protein, two drought-response proteins and one salt stress-responsive protein were up-regulated immediately after infection (2 dpi), but thereafter gene expression declined and, by day 11, was similar to uninfected roots (Table 4). These results suggest that infected plants may experience a transient wound or drought stress at the time of parasite penetration into the root, but this was not maintained, at least at the transcriptional level, suggesting that these abiotic stresses are unlikely to be solely responsible for the alterations in ABA concentration in Striga-infected plants.
Table 4. Genes encoding proteins characteristically involved in abiotic stress, nutrient transport and amino acid metabolism up-regulated in the susceptible rice cultivar IAC 165 following infection with Striga hermonthica
Days post-inoculation (dpi)
Gene expression values represent fold change in expression (infected relative to control tissue). Genes were significantly up-regulated according to a two-way ANOVA (using Benjamini-Hochberg multiple testing correction P < 0.05).
Wound-induced protein Win2 precursor
Salt stress-induced protein
Desiccation-related protein PCC13-62 precursor
Early response to drought 3
Amino acid permease-like protein
Amino-acid permease C1039.01
Amino acid transporter
Ammonium transporter 1, member 1 precursor
Component of high affinity nitrate transporter
Sulphate transporter 1.2
Amino acid biosynthesis synthesis
Tryptophan synthase beta chain 1
Tryptophan biosynthesis protein trpCF
Striga hermonthica relies on its host for all of its nitrogen nutrition. Once the parasite has invaded the root cortex and established xylem continuity with the host, it can compete effectively for host nutrients. Pageau et al. (2003) demonstrated using 15N nitrate that, following uptake of nitrate by host roots, nitrogen in the form of nitrate and amino acids (particularly glutamine and asparagines) was rapidly transported to the parasite. The amounts of these amino acids also increased in Striga-infected roots. Consistent with these physiological measurements, genes encoding an ammonium transporter, a high-affinity nitrate transporter, amino acid permeases, tryptophan synthases, asparagine synthetase and aromatic L-amino acid decarboxylase were all up-regulated in roots of IAC 165 infected with Striga (Table 4). Up-regulation of these genes was not observed in Striga-infected Nipponbare roots, consistent with the lack of host–parasite xylem connectivity and the early death of attached parasites.
We gratefully acknowledge funding from Syngenta Foundation for Sustainable Agriculture and Gatsby Charitable Trust. We thank Dr Tong Zhu (Syngenta Biotechnology, Research Triangle Park, North Carolina 27709) for performing the microarray hybridization experiments. We also thank Dr Kay Tittcomb for assistance with qRT PCR.