Differential gene expression in nearly isogenic lines with QTL for partial resistance to Puccinia hordei in barley

Plants have evolved complex mechanisms to defend against pathogen attack. Two types of immunity have been described: Pathogen-Associated Molecular Pattern (PAMP)-Triggered Immunity (PTI) and Effector-Triggered Immunity (ETI). PTI is induced at an early stage when PAMPs are recognized by Pattern Recognition Receptors (PRRs), whereas ETI is induced by direct or indirect association of a Resistance (R) protein with a pathogen-derived effector [1–4]. The outcomes of the two immune systems appear to be partial or quantitative resistance and non-host resistance (PTI), and qualitative resistance (ETI). Recently, Niks and Marcel [5] proposed that the varying efficacy of PTI suppression by pathogen effectors may explain partial resistance. In cereal crops, the barley-Puccinia hordei Otth (barley leaf rust) pathosystem is a model for investigating partial and non-host resistance. Microscopic studies on resistance levels in relation to the pathogen developmental phases has indicated plant cell wall penetration and haustorium formation by P. hordei as critical phases determining the success or failure of the infection [6]. Pre-haustorial resistance reduces the chance of successful haustorium formation by the fungal pathogen in the host cells. Failed attempts are typically associated with cell wall appositions [6–10]. Such pre-haustorial basal host defence is a typical reaction to Ph-infection in most (if not all) barley lines exhibiting partial resistance [6]. Post-haustorial resistance is usually due to R gene-mediated hypersensitive response after haustorium formation [9].

These two types of resistance have strategic significance in plant breeding for resistance to diseases. Quantitative or partial resistance has become increasingly important because of its broader spectrum and higher durability compared to R-gene mediated race-specific resistance. Many of the genes underlying partial resistance have plant developmental stage-dependent effectiveness [11]. Currently, over 20 quantitative trait loci (QTL) for quantitative basal resistance to leaf rust from five different mapping populations have been mapped to barley chromosomes [11–16]. They are named Rphq genes [R esistance to P uccinia h ordei (q uantitative)]. Of these, 10 were effective during the seedling stage, and were detected by QTL analysis of the latency period exhibited by the rust fungus on seedling leaves [15]. Considerable effort has been expended in an attempt to identify the genes underlying these QTL. Notably, a set of NILs and reciprocal NILs have been developed that contain single (Rphq 2, 3, 4) or combined (Rphq 2+ 3) introgressed segment(s) carrying resistance and susceptibility QTL allele(s) that were identified in an L94 × Vada RIL population [11, 16, 17]. L94 is an Ethiopian landrace and highly susceptible to barley leaf rust. Vada is a Dutch cultivar expressing a high level of partial resistance. Following a positional cloning strategy, Marcel et al. [18] have fine-mapped Rphq 2, the QTL with largest effect, to an interval of 0.11 cM corresponding to less than 200 kb in physical length.

Microarray technology is being widely used to address various biological, biochemical and genetic questions. Microarray-based gene expression studies can be generally grouped into two major categories. The first aims to address specific biological questions by monitoring the differential expression of genes under contrasting conditions or over time. The most common studies in this field are the investigations on host-pathogen interactions. Profiling changes in genome-wide expression in response to pathogen challenge has identified a large spectrum of genes that are responsive to pathogen attack or are associated with plant resistance in various pathosystems (reviewed by Wise et al. [19]). The second category is based on the more recently emerged concept of 'genetical genomics' [20] or expression QTL (eQTL) mapping that combines highly parallel gene expression studies with the power of genetic segregation. eQTL studies have been performed on maize, eucalyptus and Arabidopsis[21]. eQTL analyses in barley have addressed the global genetic architecture of transcript abundance in

[22], the phenomenon of limited pleiotropy [23] and as an approach to identify the causal or candidate genes underlying partial resistance to fungal diseases [24, 25]. While both categories of microarray studies are based on variation in transcript abundance, eQTL analysis provides a genetic dimension that can differentiate cis- from trans- regulation and the genetic locations of a large number of genes through the co-location of high LOD eQTL (i.e. highly differentially expressed) and their structural genes [26]. This is particularly valuable for a crop with large and unsequenced genome like barley.

Here, using a previously reported Agilent 15 k custom array [25], we performed differential expression analysis of QTL-NILs and their recurrent parental lines at 18 hours post-inoculation (hpi) with Puccinia hordei. Our major objective was to identify candidate genes for Rphq 2 and Rphq 3. In addition, transcript profiles between Ph-infected parents and their respective mock-inoculated controls allowed the establishment of transcriptomic signatures for each line in response to Ph-infection. Our results indicate that transcriptional differentiation between QTL-NILs and their respective recurrent parents reveals components of a regulatory transcriptional relay induced in response to Ph-infection. The datasets generated offer a basis for further studies on defence signalling in relation to partial resistance to P. hordei in barley.

In this study, we performed differential expression analysis of two reciprocal QTL-NILs and compared them with their respective recurrent parents. As QTL-NILs differ genetically from their recurrent parent only in the selected QTL regions, we would anticipate that genetic polymorphism between these QTL regions would account for any differential expression observed. However, due to the complexity of gene regulation, differentially expressed genes may not necessarily be located in the introgressed QTL regions, which may themselves contain regulatory genes affecting the expression of other genes spread throughout the genome. We therefore established the map positions of differentially expressed genes by exploiting previously generated gene mapping datasets. Of the 55 genes highlighted in our comparisons between NILs and recurrent parents, 40 detected eQTL in at least one of the three previous eQTL studies and co-located at the QTL regions, most (83%) having high LOD/LRS scores [22, 25] (Table S5). eQTL with high LOD scores have been demonstrated previously to be almost always cis-eQTL [22, 25, 26] placing these genes within the Rphq 2 or Rphq 3 QTL regions. The observation that so many significantly differentially expressed genes appeared to be regulated in cis- is in agreement with previous studies [25, 29]. An exception was unigene 6636, encoding HvERF4. This gene mapped to 71 cM on chromosome 1H, consistent with the position of its rice homologue Os05g41780.1 predicted by conservation of synteny [28]. This observation raises the possibility that the introgressed regions at either Rphq 2 or Rphq 3 contain a polymorphic trans-acting regulator that differentially modulates expression of HvERF4. No eQTL for HvERF4 was detected at the regions corresponding to Rphq 2 or Rphq 3 in the St/Mx DH mapping population, consistent with the fact that it does not segregate for Rphq 2 or Rphq 3.

HvERF4 is a member of a family of plant transcription factors functionally involved in defence signalling pathways related to ethylene, jasmonic acid and abscisic acid. Over-expression of Arabidopsis AtERF4 represses the expression of pathogenesis-related (PR) genes such as basic chitinase and beta-1,3-glucanase genes and genes containing a GCC-box [30], the core sequence element of promoters required for responsiveness to ethylene [31]. In our previous experiment with Steptoe and Morex, cultivars with similar but intermediate levels of partial resistance to leaf rust, we also observed that HvERF4 was significantly up-regulated by Ph-infection but no differential expression (p > 0.2) was observed between the parents [25]. Here, HvERF4 was induced in Ph-infected L94 (susceptible) (FC = 4.42) and Vada (partially resistant) (FC = 2.42) as compared to mock-inoculated controls (Table S5), and the expression level of the Vada allele was only a third (FC = 0.34) of that of the L94 allele after induction. The association of resistance/susceptibility with lower/higher expression of HvERF4 appears to be in agreement with the negative regulatory role of HvERF4 on the expression of PR and other defence responsive genes. However, consistent association of higher expression of PR genes with resistance was not observed in Vada and L94. This may reflect the general complexity of natural resistance response coupled with allelic variation at PR genes between these two lines. While this train of inference highlights HvERF4 as potentially important in this specific defense interaction, none of the so far reported 20 QTL for partial resistance to leaf rust, nor any of the QTL for resistance to heterologous rusts is co-located with HvERF4 at 71 cM on chromosome 1H [11–16]. Thus, HvERF4 is not a positional candidate for any of the reported QTL. However, of direct relevance is a previously highlighted eQTL hotspot for

genes that was associated with OPA-SNP 11_20157 [25] at 70 cM (98 cM on the consensus map [28]) on chromosome 1H spanning the region containing HvERF4. This hotspot comprised 127 eQTL in less than a 10 cM interval and contained genes primarily involved in defence response [25]. Given its known role in PR-protein regulation, we speculate that HvERF4 represents a key regulatory relay component of the signalling pathway that controls expression of at least a portion of the genes with eQTL located at the hotspot on chromosome 1H. Considering these observations together we hypothesise that the causal genetic polymorphism at either Rphq 2 or Rphq 3 differentially regulates HvERF4 in trans (possibly through direct or indirect modulation of ethylene, jasmonic acid or abscisic acid levels, known in Arabidopsis to alter levels of AtERF4 expression [30]), the consequence of which is differential regulation of down-stream defence responses. In this scenario, the candidate genes for Rphq 2 or Rphq 3 would be those acting up-stream rather than down-stream of HvERF4 and possibly involved directly or indirectly in plant hormone signalling pathways. While we did not find such a candidate from the annotated functions of the QTL-specific and differentially expressed candidates for Rphq 2 and Rphq 3, the gene controlling expression of HvERF4 may, however, not be differentially expressed between L94 and Vada, may not be on our expression platform (which probably contains less than half of the barley genes) or may not be at the orthologous position in rice. An alternative to identifying the causal gene for Rphq 2 or Rphq 3 could be through map-based cloning of the trans-eQTL for HvERF4.

Marcel et al. [18] narrowed down the genetic interval for Rphq 2 to 0.11 cM corresponding to a physical length of 183 kb in barley (Marcel and Niks unpublished data) and a 69.7 kb syntenic region on rice chromosome 4. Inspection of all predicted genes in the Rphq 2 syntenic interval in rice identified a cluster of six peroxidase genes and a MAP3K gene [18] as potential candidates because of their functional involvement in defence responses. In this study, we identified four barley genes at Rphq 2 that were differentially expressed and had homologues located in the syntenic region in rice (Table S5, unigene1852 (no.7), unigene2111 (no.9), unigene13865 (no. 8) and unigene8521 (no. 46)). Unigene1852 and 2111 both encode peroxidases and are within the 0.11 cM interval containing Rphq 2. The other two, according to the fine mapping data of Marcel et al. [18] fell just outside the candidate interval. However, given the frequent breakdown in conservation of synteny, positional candidate gene identification using this approach alone remains problematic. Differential expression in the QTL-NILs identified an additional five candidate genes (Figure 5 and Table S5: no. 2, 3, 19, 21and 51) that were not apparently present in the syntenic region of rice. Two of these encode proteins that are functionally involved in signal transduction (Additional file 5, Table S5, no.19 and 21 encoding a kinase and a receptor-like kinase respectively), one PR protein, and one with homology to human NIPA1, implicated in Prader-Willi/Angelman syndrome 1 [32]. One gene showed no homology to known genes. Thus, these five genes, together with the two peroxidase genes, are potential positional candidates for Rphq 2. Further refinement of the candidate gene list will require knowledge of the role of these genes in defence response and correlation of transcript levels with resistance/susceptibility. Many more genes (i.e. 31) were identified as being differentially expressed and located at the Rphq 3 region. This is expected given the larger interval of the QTL (28 cM for Rphq 3 vs. 4 cM for Rphq 2) and Rphq 3 may, therefore, account for differential expression of most of the genes with unassigned map positions. Functionally, none of the differentially expressed genes at Rphq 2 or Rphq 3 appear to be obvious candidates for a regulator of HvERF4.

Many defence genes encoding PR proteins and components of the phenylpropanoid pathway such as phenylalanine ammonia lyase (PAL) were, as would be expected, Ph-responsive. PR genes encoding beta-1,3-glucanases, chitinases and thaumatin-like proteins exhibit strong in vitro anti-fungal activity [33] and numerous studies have shown that transgenic plants expressing PR-proteins have significant improvement of disease resistance [34–37]. PAL, the first committed enzyme in the phenylpropanoid pathway, is involved in synthesis of both phytoalexins and lignin. Phytoalexins are antimicrobial while lignin synthesis contributes to formation of papillae, which are physical barriers against cell wall penetration by the pathogen [38]. As part of the general response to pathogen infection, few of the genes fell into these categories co-located at the two QTL for partial resistance. One exception, unigene25195, encoding a chitinase (PR3), co-located at Rphq 2, was Ph-responsive and differentially expressed between Rphq 2 and Rphq 2. However, it was not prioritized as a candidate for Rphq 2 since the higher level of gene expression was associated with the susceptibility allele Rphq 2. Whereas a number of defence genes were activated in response to Ph-infection, none was found to be a promising candidate for Rphq 2 or Rphq 3. Our results support the notion that components of the general defence response have incremental, rather than deterministic, roles in the outcome of an interaction between a plant and a pathogen [39]. Many attempts to identify genes for disease resistance have highlighted those involved in signal transduction [40, 41] or physiological and cellular functions [42, 43] rather than defence per se[44, 45].

Ph-infection triggers a broad range of biological responses with defence response genes being significantly over-represented. Of note is a set of genes encoding receptor-like kinase (RLK), receptor-like proteins (RLP), WRKY, MAPK and PR proteins (Additional File 1, Table S1), which form a complete and well-explored defence signalling cascade starting with the perception of PAMPs, activation of WRKY transcription factors and the subsequent induction of PR proteins [3, 46]. Our results also suggest that, in the absence of cognate R genes to P. hordei, plants still mount reactions similar to R-gene mediated responses as indicated by the significant up-regulation of genes coding for R gene (-like) proteins and marker genes for oxidative burst such glutathione S-transferase and peroxidase. Although no obvious R genes were identified as candidates for the QTL in this study, R gene-like mediated responses may contribute to basal resistance as a complementary mechanism to PAMP-triggered defence responses. Support for this is provided by observations that resistance QTL are often coincident with the location of R-gene homologues [47–51] and that mutated R genes can induce a resistance phenotype similar to quantitative resistance controlled by multiple genes [52–54].

One striking characteristic of the responses to Ph-infection was the activation of signalling pathways related to a broad range of plant hormones including ethylene, gibberellins, auxin, and brassinosteroid as indicated by the up-regulation of genes encoding ethylene-responsive transcription factors, ACC oxidase, auxin-responsive proteins, brassinosteroid insensitive 1-associated receptor kinase 1 (BAK1), gibberellin receptors and a DELLA protein (Additional File 1, Table S1). All of these hormones have been reported to be involved in plant defence responses [55–57] and various defence pathways are interconnected through hormone-mediated signalling pathways forming complex regulatory networks [55, 56, 58–60]. Here, the identification of the ethylene-responsive factor HvERF4 as a putative link between pathogen perception and response is consistent with a role for differential hormone signalling in partial resistance. Understanding the role of Rphq 2 of Rphq 3 in initiating and coordinating the response requires further work.

Substantial overlap of Ph-responsive genes was identified in super-susceptible (L94) and partially-resistant (Vada) lines. Over 70% of Ph-responsive genes were detected in both L94 and Vada and had the same expression patterns (up- or down-regulation) in both lines. An even higher percentage of overlapping Ph-responsive genes (79%) was discovered in both Steptoe and Morex, two cultivars with similar and intermediate level of partial resistance. Given that these lines are genetically diverse, we conclude that barley lines without known cognate R genes to P. hordei exhibit similar responses at the transcriptional level, and that observed differences are largely quantitative. Similar findings have been observed in the comparison between compatible and incompatible interactions [61–63]. A small proportion (7%) of Ph-responsive genes in this study did appear to be resistant/susceptible line-specific and it may be that they determine part of the observed phenotypic differences between lines. However, in Ph-infected leaves we found no evidence for their differential expression in the comparisons between the two QTL-NILs and their respective recurrent parents. Therefore, if the variation in resistance, accounted for by Rphq 2 or Rphq 3, is regulated at the transcriptional level, these are not strong candidate genes.

We generated a robust expression data set in reciprocal Rphq 2/Rphq 3 QTL-NILs at 18 hpi, which is the timepoint previously described as being the most critical during P. hordei invasion in barley [25]. However, we realise that transcriptional re-programming in response to pathogen infection is a dynamic and complex process and that defence-associated genes respond to input stimuli with different timing and amplitude. A limitation of our experiment is, therefore, that defence response scenarios constructed on the transcriptional profiles of the 802 Ph-responsive genes identified here is simply a snapshot of a dynamic process, at the point when infection hyphae have just attempted penetration of the host cells forming haustoria [25]. To extend our understanding of the complex regulatory mechanisms occurring during defence against P. hordei, a more comprehensive investigation would involve sampling at multiple timepoints covering the whole infection period.

Differential expression with QTL-NILs identifies genes predominantly located at the target region(s) providing both transcriptional and positional candidate genes underlying the QTL. Positional analysis of the differentially expressed genes relative to the QTL has the potential to discover regulatory relays initiated from genes within the QTL.