Rusts are among the most important diseases of legumes [1] and grass pea (Lathyrus sativus L.) is not an exception [2]-[4]. Rusts are caused by biotrophic fungi that keep infected host cells alive for their development. They form elaborate intracellular accommodation structures called haustoria, which maintain an intimate contact between fungal and plant cells over a prolonged period of time [5].

Rust in Lathyrus spp. is caused by Uromyces pisi (Pers.) Wint and U. viciae-fabae (Pers.) J. Schröt [6],[7], but and in addition to Lathyrus, U. pisi infects a broad range of other legumes too [7],[8]. Plants have developed multifaceted defence responses, many of which are induced only upon pathogen attack. These responses may include induction of pathogenesis related (PR) genes, the production of secondary metabolites (as e.g. phytoalexins), as well as the reinforcement of cell walls [9]. Associated with these responses may be the production of reactive oxygen species (ROS) and the induction of localized cell death (the hypersensitive response, HR) [10]. The induction of this basal plant defence machinery occurs upon the recognition of conserved molecules which are present in a variety of microbial species, but absent in the host. These pathogen associated molecular patterns (PAMPs) are molecular components highly conserved within a class of microbes, where they have essential functions for their fitness or survival [11]. These include, for example, fungal chitin, β-glucan and ergosterol. The specific virulence factors of the pathogen, known as fungal effectors, are recognized by corresponding resistance (R) genes of the host plant. Both rust-causing pathogens of Lathyrus are able to efficiently overcome R-gene based resistance [12]. To date, most fungal effectors identified are lineage-specific small secreted proteins (SSP) of unknown function [13],[14]. The U. viciae-fabae rust transferred protein 1 (Rtp1) was the first fungal effector visualized in the host cytoplasm and nucleus after in planta secretion by the rust fungus [15]. Rtp1 belongs to a family of cysteine protease inhibitors that are conserved in the rust species (order Pucciniales formerly known as Uredinales) [16].

Gene-for-gene resistance is associated with the activation of, for instance, the salicylic acid (SA)-dependent signalling pathway, leading to expression of defence-related genes like PR1, the production of ROS and finally to programmed cell death [17],[18]. Other phytohormones involved in plant/pathogen interaction are ethylene (ET) and jasmonates (JA). Plant defence responses appear specifically adapted to the attacking pathogen, with SA-dependent defences acting mainly against biotrophs, and JA- and ET-dependent responses acting mainly against necrotrophs [5],[19],[20].

Grass pea is a diploid species (2n = 14) with a genome size of approx. 8.2 Gbp [21]. Although grass pea is primarily self-pollinated, a 2 to 36% outcrossing rate was reported, depending on location and genotype [22]-[24]. Outcrossing is mainly driven by pollinators, and therefore can be minimized when grown in isolation [22]. There is a great potential for the expansion of grass pea in dry areas and zones that are becoming more drought-prone as a result of climate change [25]. Partial resistance to U. pisi has been reported in grass pea as a clear example of prehaustorial resistance, with no associated necrosis. This resistance is due to restriction of haustoria formation accompanied by frequent early abortion of the colonies, reduction in the number of haustoria per colony and decreased intercellular growth of infecting hyphae [26]. Though prehaustorial resistance is typical for non-hosts, it has also been implicated in host partial resistance [27],[28] and is common in resistance of major cool season grain legumes against rusts [1],[29]. Additionally, resistant Lathyrus genotypes may serve as a source of new and useful genetic traits in the breeding of related major legume crops such as peas, lentils and vetches. Cross-incompatibility has been reported between pea and L. sativus, but successful fusion of Pisum sativum and L. sativus protoplasts [30] creates new possibilities for gene transfer between these species. However, the slow progress in understanding the genetic control of important traits, such as disease resistance, in Lathyrus species hampered the development of modern cultivars or the introgression of their interesting traits into related species.

In economically important warm season legumes such as common bean and soybean, complete monogenically controlled resistances to rusts and associated rust resistance genes have been described together with closely linked markers for use in marker assisted backcrossing [29],[31]-[34]. By contrast, most rust resistances described so far in cool season food legumes are incomplete in nature and the genetic basis of resistance is largely unknown. Although QTL mapping studies confirmed the polygenic control of resistance as e.g. in pea [35], faba bean [36] and chickpea [37], no markers suitable for marker assisted selection (MAS) are available yet.

Genomic resources for grass pea are still scarce (e.g. in April 2014 the NCBI database contained only 178 EST sequences from L. sativus [38]), and the two linkage maps existing for grass pea do not contain sufficiently informative markers to bridge between them [4].

The advent of next-generation sequencing (NGS) technologies was an important breakthrough enabling the sensitive and quantitative high-throughput transcriptome analysis referred to as RNA-seq [39],[40]. RNA-seq discriminated between microbial and host transcriptomes, during plant-microbe interactions, using original or phylogenetically related genomes as a reference for transcript annotation [41]-[44]. RNA-seq gene expression patterns provided also information on complex regulatory networks and on variations in expressed genes, such as SNPs and SSRs, in an increasing number of non-model plants [45] and thus may be well suited to overcome the bottleneck of lacking genomic resources in Lathyrus.

Here we employed RNA-seq to study the response of L. sativus to U. pisi infection. We used MapMan and metabolic pathway analyses to interpret the results and assessed allelic diversity in transcripts as a source for genic markers for future (comparative) mapping studies. In addition, the expression of a set of selected genes was measured by qRT-PCR to validate the RNA-seq results.

To our knowledge, this is the first study on the global expression profiling of genes in grass pea/pathogen interaction using NGS. Our results will assist the elucidation of pathways and genes associated with resistance to rust in grass pea and related species. This approach may represent one of the initial steps towards the development of effective strategies for resistance breeding against such a quickly evolving pathogen.

Lathyrus spp. is a potential source of resistance to several pathogens [4],[25] and especially L. sativus provides resistance to several fungal and bacterial diseases [26],[38],[54],[55]. However, the lack of genetic and/or genomic information was a barrier to further identify resistance-related genes and to use them in breeding.

In the present study we therefore attempted to improve this unfavourable situation by identifying ESTs and SNPs, potentially involved in resistance, that may be used in future smart breeding approaches. We describe for the first time a high-throughput transcriptome assembly of grass pea/pathogen interaction, using genotypes contrasting in response to rust infection, to unravel the involved partial resistance mechanism and associated resistant genes.

Our study has identified a large number of differentially expressed genes corresponding to biological categories that are thought to be most relevant in grass pea response to rust. A limitation of our study is the fact that only a single pooled sample was investigated for each genotype and condition. Although the biological variance could not be assessed in the bulked approach, the large number of individual samples in the pool is likely to level out many of possible outliers. Nevertheless, the validation of twelve genes by RT-qPCR, using three biological replicates, provided a good correlation with RNA-seq results.

Another motive that could also be influencing our results is that we used different cDNA synthesis primers, oligo(dT) for the RNA-seq and poly(A) for RT-qPCR, what might yield different quantities of poly-adenylated and non-adenylated transcripts.

Our study was severely hampered by the low number of annotated sequences, which is due to the lack of a reference genomic sequence for Lathyrus. Nevertheless, we could annotate between 34% and 46% of differentially expressed contigs to hits in plant databases, depending on the genotype and the infection status of the plants. We further developed new gene-based molecular tools as e.g. expressed sequence tags, gene-based simple sequence repeats (EST-SSR) and SNP-based markers. Moreover, we identified a number of U. pisi effectors in the infected tissues though the overall low number of observed fungal transcripts probably reflects the low quantity of fungal structures in early-infected leaves [56]. Thus, our present study will help to overcome the problems we encountered in previous work, where the transfer of molecular markers from close related species had a very low rate of success (18% for pea EST-SSRs and 6% for pea genomic SSRs, [57]) Therefore, the present RNA-seq libraries will boost the availability of specific EST-SSRs and SNP-based markers that will be equally important for future development of more effective grass pea resistance breeding approaches.

The high amplification rate of the developed EST-SSRs validates the quality of the RNA-seq data. The few primers that failed to produce amplification products or produced amplicons with an unexpected pattern may be caused by the location of the respective primers across splice regions or the presence of a large intron, since genomic regions are absent from cDNA. In addition also primers could be derived from chimeric cDNA clones [58].

Besides the novel markers, the deep insights into pathogenesis-related mechanisms provided by this study are of particular interest. The most interesting pathogenesis-related protein that we identified, the “MLO-like protein” is involved in signalling in response to biotic stress. MLO was described for the first time in barley, where it conferred partial resistance to powdery mildew by inducing the thickening of the cell wall at fungal penetration sites [46]. Two “MLO-like” contigs were up-regulated exclusively in the resistant genotype (group E), and perhaps related to this, also in group E and more strongly up-regulated in the resistant genotype (group C), we identified cellulose biosynthesis genes. The exclusively resistance-up-regulated group E contained one “IRREGULAR XYLEM 3” (IRX3) gene and three “IRX1”. Additionally, one “cellulose synthase 1” (CESA1) was stronger up-regulated in the resistant genotype than in the susceptible one (group C). Consistent with the assumed importance of MLO signalling for rust resistance, some genes important for MLO function as e.g. “calmodulin”, involved in calcium signalling as a prerequisite for MLO function [51], were down-regulated in the susceptible genotype (group G). Another gene involved in MLO resistance and down-regulated in the susceptible genotype, is the glycosyl hydrolase “PENETRATION 2” gene [59]. Since rust resistance in grass pea is of prehaustorial type we consider MLO as a candidate R-gene. In order to confirm this assumption, callose deposition, as a potentially durable resistance mechanism against rusts, should be further investigated in rust-resistant and susceptible grass pea genotypes. The down-regulation of several “MLO-like” contigs in response to infection in both, susceptible and resistant genotypes, does not necessarily contradict our assumption, since several MLO orthologs were demonstrated to function as susceptibility genes [47]-[49].

Plant responses to biotic stressors are, i.a., controlled by phytohormones as e.g. salicylic acid (SA), abscisic acid (ABA), jasmonates (JA) and ethylene (ET). Differences in expression of hormone-related genes of the susceptible and resistant genotype, in response to the pathogen, also occurred in our gene expression patterns. For example, plant resistance to biotrophic pathogens is mainly controlled by the SA pathway [18] and the importance of SA in the induction of systemic acquired resistance in legumes against rust fungi has been reported [60],[61]. In our study an inducer of the SA pathway, the “ethylene response factor 5” (ERF5) gene, which at the same time inhibits the JA and ET biosynthesis pathways [62], was exclusively up-regulated in the resistance genotype (group E), whereas two Apetala2/Ethylene Responsive Factor (AP2/ERF) transcription factor genes, important for the regulation of defence responses [63], were down- regulated in the susceptible genotype (group G).

ABA regulates defence responses through its effects on callose deposition and production of ROS intermediates [18], activating also stomata closure as a barrier against pathogen infection [64]. In the resistant genotype, the transcript for “9-cis-epoxycarotenoid dioxygenase 2”, a key regulator of ABA biosynthesis in response to drought [65], and involved in the crosstalk between ABA and SA signalling in plant-pathogen interactions [66], was up-regulated (group E), whereas several transcripts engaged in ABA, auxin and JA signalling, were down-regulated in the susceptible genotype (group G). This is consistent with a susceptible response to a biotroph attack [18].

The image emerging from the transcription profiles, of the resistant and susceptible genotype, further highlights that pathogenesis related (PR) proteins are key players in Lathyrus-rust interactions since several PR genes were mainly up-regulated in the resistant genotype, after inoculation. Among these were two chitinases (PR-3 and PR-9) involved in the degradation of the fungal cell wall [9] and a thaumatin (PR-5) gene, which causes an increase of the permeability of fungal membranes by pore-forming mechanisms [67]. In group E, we found a “pathogenesis related protein 1” (PR-1) and a “protein kinase-coding resistance protein”, a receptor kinase with a thaumatin domain (PR5K), presumably involved in thaumatin signaling and described previously as delaying infection [68]. Another important PR-gene, an “acidic endochitinase” (PR3), was down regulated exclusively in the susceptible genotype. Genes involved in secondary metabolism were also detected. Legumes utilize flavonoids, notably isoflavones and isoflavanones, for defence against pathogens and as signalling molecules, with a number of phenylpropanoids having antimicrobial activity and restricting pathogen growth and disease symptoms [69]. In group G, we identified a “reticuline oxidase-like protein”, up regulated in non- race-specific resistance to stripe rust in wheat [70], an “isoflavone 2′-hydroxylase” from the isoflavonoid pathway [71] and a “dihydroflavonol-4-reductase” catalysing the first enzymatic step in anthocyanin biosynthesis, in the flavonoid pathway [72].

Also, exclusively down regulated in the susceptible genotype, we found some genes important for defence response within the miscellaneous category, like “endo-beta-1 3-glucanase”, “glutathione S-transferase” (GST) and “peroxidase”. In plants, beta-glycosidases serve a number of diverse and important functions, including bioactivation of defence compounds, cell wall degradation in endosperm during seed germination, activation of phytohormones, and lignifications [73]. GSTs are detoxification-related proteins, protecting cells from secondary metabolites produced in response to pathogen attack, including peroxidases [74]. Finally, peroxidases function as resistance factors against invading fungi, inhibiting hyphal elongation, and when H₂O₂ is present, causing oxidative burst [75].

Effectors are expected to be excellent targets for the control of pathogens, but, unlike effectors from some other plant pathogens, relatively little is known about rust effectors [76]. In this study, several unigenes were identified in fungal databases. The most known rust effector identified was “rust transferred protein 1” (Rtp1). This effector aggregates into amyloid-like filaments in vitro [77]. Immunoelectron microscopy localized this effector to the extrahaustorial matrix protuberances extending into the host cytoplasm, although the exact role for this protein remains to be discovered [78]. Other Uromyces effectors identified in this study were “succinate dehydrogenase”, “invertase” and “permease”. From the total potential rust transcripts identified, a selection of effector proteins could be used as probes to identify the target host proteins as a first step in the development of effector-driven legume breeding, maximizing the durability of resistance against the quickly evolving rust pathogens [79].

From the DE contigs obtained in the present study, 2,634 presented SNPs between the resistant and the susceptible lines. The MapMan software aided in the functional categorisation of SNPs, revealing that the categories “RNA regulation of transcription” (9.5%) and “protein.degradation” (8.9%) contained by far the most SNPs. Within these categories, ubiquitins were most polymorphic (5.6%). Ubiquitins tag proteins for proteasome degradation and play a central role in signalling pathways [80]. Especially ubiquitin “E3 RING” and “SCF F-BOX” contigs contained a large number of SNPs (1.7% and 1.8% respectively). E3 RING and SCF F-BOX proteins are involved in several aspects of plant immunity ranging from pathogen recognition to both PTI to effector-triggered immunity (ETI). From the differentially expressed contigs identified as containing SNPs, we found one E3 RING, “PREDICTED: RING-H2 finger protein ATL2-like”, down regulated in the susceptible genotype. Four other functional categories, “RNA.regulation of transcription” (9.5%), “protein.postranslational modification” (4.3%), “protein.synthesis” (3.2%) and “signalling.receptor kinases” (2.5%), also contained significant numbers of polymorphisms. Especially the “signalling.receptor kinases” category may be of particular interest for further studies since receptor kinases recognize pathogen effectors and their rapid evolution, reflected by large numbers of polymorphisms, may represent plants adaptation to a rapidly changing spectrum of pathogens in the arms race between them and their hosts [81].

The large number of SNPs that we identified will be instrumental for the development of linkage and high-throughput association mapping approaches and for the expansion of our previous diversity studies in Lathyrus [57],[82].

Our results provide an overview of gene expression profiles of contrasting L. sativus genotypes inoculated with rust, offering a valuable set of sequence data for candidate rust resistant gene discovery.

Our transcriptome analysis provided comprehensive insight into the molecular mechanisms underlying prehaustorial rust resistance in L. sativus.

The differences in resistance between the two L. sativus genotypes investigated appear to be mainly due to the activation of the SA pathway and several pathogenesis related genes, including the ones regulated by MLO. The fastest-evolving pathways differentiating between the two genotypes are the general RNA’s regulation of transcription, followed by the Ubiquitin-26S proteasome system and having also as most mutated receptor-based signalling genes and biotic and abiotic stress related genes. The detected polymorphic SNPs will allow the development of new gene-based molecular tools. Altogether, 51 genes were identified as potential resistance genes, prioritizing them as specific targets for future functional studies on grass pea/rust interactions. Besides a plethora of pathogenesis-related host genes, 4,558 transcripts, including putative effectors, were also identified for the rust fungus U. pisi. As a consequence of the newly developed wider array of genetic and genomic resources, future work will focus on high throughput mapping of the genetic basis of disease resistance in L. sativus and eventual comparative mapping with other legume species, contributing all to an improved exploitation of this under used highly potential legume species.