Cellular and molecular characterization of a stem rust resistance locus on wheat chromosome 7AL

Stem rust resistance phenotyping

Resistance against Pgt was assessed on seedlings, 11–15 days post-inoculation (DPI) with rust spores. Col was moderately susceptible (Infection type-IT 2⁺3), whereas Col-NS765 and Col-NS766 were resistant (IT ;1) (Fig. 1). CTH-K was fully susceptible (IT 3⁺4), whereas CTH-NS1 and CTH-NS2 were moderately resistant (IT 22⁺) (data not shown). Susceptible control Morocco, hardly having any known resistance genes for Pgt, was fully susceptible (IT 4) (data not shown). These results were in agreement with Pujol et al. [9].

Microscopic observations of the stem rust resistance response

Fluorescein isothiocyanate wheat germ agglutinin staining

Stem rust was visualized by staining with fluorescein isothiocyanate (FITC)-labelled wheat germ agglutinin (WGA), which binds to chitin present in fungal cell walls. Within 1 h of Pgt inoculation (0 days post inoculation—DPI), adhered urediniospores to the leaf surface were observed (Fig. 2-0). At 1 DPI (24 h after inoculation), most spores had germinated, but very few had developed an appressorium (Fig. 2-1). No difference between genotypes was observed at these time points. At 2 DPI, appressoria were formed over stomata as the pathogen started entering the leaf (Fig. 2-2). Some guard cells of the resistant lines showed blue autofluorescence under UV light, indicative of cell death (Fig. 2-2R2). This phenomenon was always associated with an infection site (i.e. the location where appressoria formed, typically at the stoma), and a penetration peg was often present. Although also observed in Col, autofluorescence of plant cells around stem rust infection sites was less frequent in the susceptible line than in the resistant ones (Fig. 2-2S2), and this applies to any time point from 2 DPI (Fig. 3). At 4 and 5 DPI, the rust colonies were well established inside the leaf of Col as many hyphae were observed, forming a small network (Fig. 2-5S). Haustoria were also visible. For the resistant lines, most fungi were still at the penetration stage and few hyphal networks and haustoria were observed. Epidermal cells adjacent to these infection sites were often autofluorescent under UV (Fig. 2-5R). Some mesophyll cells were also autofluorescent. At 7 DPI, hyphal networks spread further in Col (Fig. 2-7S), whereas their progression was limited in Col-NS765 and Col-NS766 (Fig. 2-7R). Finally, at 9 DPI, hyphal networks were large in Col, and urediniospores had begun to form (Fig. 2-9S). Many of the leaf cells adjacent to the large rust pustules showed blue autofluorescence. In the resistant genotypes, the relatively few fungal networks that progressed beyond the penetration stage were mainly smaller than in Col, and also showed autofluorescence of surrounding plant cells (Figs. 2-9R, 4). These results were in agreement with macroscopic observations at later time points, i.e. more and bigger uredinia in the susceptible genotypes (Fig. 1).

Observations in the susceptible CTH-K were comparable to those of Col, while the resistant lines, CTH-NS1 and CTH-NS2, were comparable to Col-NS765 and Col-NS766. At 9 DPI, more and highly extended hyphal networks were present in the susceptible lines than in the resistant ones (Fig. 4). Also, more rust infection sites were arrested at the penetration stage in lines with Col background than in lines with CTH background, reflecting the slightly higher susceptibility of the CTH background (Fig. 4).

Trypan blue staining

As well as dead plant cells, rust structures on the leaf surface (spores, elongation tubes and appressoria) were visible after trypan blue (TB) staining, when observed within 24 h of de-staining (Fig. 5). Observations were similar to those with WGA-FITC: urediniospores attached to the leaf surface at 0 DPI, germinated at 1 DPI and developed appressoria over stomata between 1 and 2 DPI. The first sign of plant cell death was observed at 2 DPI (Fig. 5-2R). Guard cells were sometimes TB stained in the resistant lines but almost never in Col. TB staining of leaf cells was associated with the presence of urediniospores or appressoria. At 5 DPI, epidermal cells adjacent to infection sites were often stained in the resistant genotypes (Fig. 5-5R). At 9 DPI, pustules with urediniospores, surrounded by TB stained epidermal cells, were observed in Col (Fig. 5-9S), but not in the resistant lines (Fig. 5-9R). At this stage, the staining in Col was probably due to the extensive damage caused by the rust development and the formation of uredinia breaking through the plant epidermis.

Transcriptome analysis

Differential expression analysis

Read counts for each gene in each sample were first normalised using the R DESeq package [13]. Hierarchical clustering of all samples revealed three main clusters, each representing a time point (Fig. 6). Genotypes were grouped into 3 sub-clusters at 2 and 5 DPI, but not at 0 DPI. In addition, Col-NS765 and Col-NS766 were closer to each other, at 2 DPI, than to Col. At 5 DPI, Col and Col-NS765 clusters remained distinct, whereas it was not as clear for Col-NS766. Differential expression analysis was performed using DESeq. Eighteen pairwise comparisons were explored, corresponding to two categories: (1) genotype comparisons at each time point and (2) time comparisons for each genotype.

As seen above, most of the Unigenes were represented by relatively few reads per sample. In addition, mRNA was isolated from whole leaves, whereas gene expression differences were likely to be concentrated in cells in contact with the rust pathogen, at least early after inoculation. To allow for these attenuating factors, the conditions for calling DE genes were not made too stringent, with a two-step selection, in order to maximize the number of candidate genes. Genes were at first considered differentially expressed (DE) with false discovery rate (FDR) ≤ 0.05 and absolute log₂ fold change (FC) ≥ log₂(2). On average, 8130 (12.1%) and 470 (0.7%) genes were DE between time points and between genotypes, respectively (Fig. 7). Once DE genes were determined for a particular comparison (e.g. 0 vs 2 DPI in Col), these same genes were re-evaluated in similar comparisons (e.g. 0 vs 2 DPI in Col-NS766), using less stringent thresholds [FDR ≤ 0.075 and |log₂(FC)| ≥ log₂(1.8)], which allowed detection of false negatives with respect to differential expression; for example, Ta.34224 was initially found to be DE between 0 and 2 DPI in Col-NS765 and Col-NS766 (FDR ≤ 0.05 and FC = 2.92 and 2.07, respectively), but not in Col (FDR ≤ 0.05 and FC = 1.97), but a closer analysis showed their expression patterns were very similar in all three genotypes.

Genes of interest for the stem rust resistance

Compared to 0 DPI, more than 15,000 and 11,000 genes were found to be DE in at least one genotype at 2 and 5 DPI, respectively (Fig. 7, Top). Around 40% of them were regulated in the same way in all genotypes, while several thousand genes were DE across time points in only one genotype. More genes were DE in Col across time points than in Col-NS765 and Col-NS766. A total of 1502 genes were actually DE between genotypes. At any time point, more genes were DE between Col and Col-NS765 than between Col and Col-NS766 (Fig. 7, Bottom).

Genes that were DE at any time point between Col and both backcrossed independent resistant lines, Col-NS765 and Col-NS766 (hereafter referred to collectively as Col-NS), were selected as genes of interest (GOIs); as these genes were potentially involved in the resistance response mediated by the 7AL locus. A total of 353 genes were selected, of which 198, 256 and 279 were DE at 0, 2 and 5 DPI, respectively (Fig. 7, Bottom). More than half of the GOIs were DE by at least eightfold (Fig. 8a). Fold changes were usually similar between Col and either Col-NS765 or Col-NS766. More than 80% of the genes that were DE between Col and Col-NS at 0 DPI were also DE at 2 and 5 DPI (161 genes, Fig. 8b). Also, 23 and 11% were never expressed in Col or Col-NS, respectively.

Unigenes with a minimum of 100 reads aligned (67,156) were assigned to a wheat chromosome by aligning the Unigene reference sequences to the IWGSC wheat chromosome survey sequencing (CSS) database [15]. More than 99% of the Unigenes were unambiguously associated with a chromosome. The 67,156 Unigenes were found to be uniformly distributed across the genome (Additional file 3: Figure S2). However, among the GOIs (353) that were DE at 0 DPI (198), a disproportionate number were found to be located on chromosomes 6AL, 6AS and 7AL (Additional file 3: Figure S2), which carried 41 (20.7%), 18 (9.1%) and 16 (8.1%) genes, respectively.

The 353 GOIs were classified into categories (Table 1), depending on the time points at which they were DE between Col and Col-NS and their expression through time. 134 (38%) genes were categorized as constitutively DE, that is, always DE between Col and Col-NS, but never through time (Table 1 “Constitutive”). 85 (24%) and 78 (22%) genes were categorized as putatively involved in the early and late resistance response, respectively, with 25 genes being in common between both resistance responses (Table 1 “Early” and “Late”). Finally, 27 (8%) genes were putatively involved in the initiation of the resistance response, as they were located on chromosome group 7L and DE between Col and Col-NS at 0 DPI (Table 1 “Origin”). Genes on 7BL and 7DL were also considered, in case they corresponded to 7AL homoeologues that were missing in the Unigene or wheat genome databases.

Category	Definition	Criteria	Total		More expressed in
					Col		Col-NS
			#	%	#	%	#	%
Constitutive	Genes constitutively expressed	Always DE between Col and Col-NS, and never DE through time in any genotype	134	38%	59	44%	75	56%
Early	Genes putatively involved in the early resistance response	DE at 2 DPI between Col and Col-NS, but not at 0 DPI unless DE at 2 DPI, compared to 0 DPI	85	24%	36	42%	49	58%
Late	Genes putatively involved in the late resistance response	DE at 5 DPI between Col and Col-NS, but not at 0 DPI unless DE at 5 DPI, compared to 0 DPI	78	22%	32	41%	46	59%
Pathogen	Genes putatively from the pathogen	Not expressed at 0 DPI in any genotype, highly up-regulated in Col at 5 DPI	4	1%	4	100%	0	0%
Miscellaneous	Remaining genes		77	22%	9	12%	18	23%
Origin	Genes that map at the origin of the resistance (i.e. possible R gene)	DE between both Col-NS and Col at 0 DPI, and on chromosome 7L	27	8%	7	26%	20	74%

Criteria used for categorizing the genes found DE between Col and both Col-NS765 and Col-NS766, and the number and percent of the 353 genes of interest in each category

Genes were annotated using the rice, Arabidopsis and NCBI Refseq databases. Among the GOIs, 153 (43%) were annotated, including 40 genes (11%) annotated as repetitive or transposable elements, and 200 (57%) were poorly annotated or not at all. All details concerning the GOIs can be found in Additional file 2.

Microscopic observations

Stem rusts enter a leaf by forming an appressorium over a stoma, after which a penetration peg enters the leaf though the stomatal opening, eventually invading, usually a mesophyll cell, and forming a haustorium [16]. In this study, the different steps of stem rust development in planta were observed, from the landing of urediniospores on the leaf surface, to the formation of new spores 9 days later. Various rust structures were detected, such as appressoria and haustoria. Results indicate a two-phase resistance response, preceding and following plant invasion by the pathogen, which ultimately led to fewer uredinium sites with fewer spores in the resistant lines (Fig. 1).

The pre-invasion, or early resistance response, was visible from 2 DPI, after the fungus had developed an appressorium and started entering the leaf, but before haustorium formation (Figs. 2, 5). The response was localized to the stomatal guard cells and adjacent epidermal cells at the site of penetration, which showed greater autofluorescence or TB coloration in the resistant lines than in the susceptible line. Both autofluorescence and TB coloration are indicative of dead plant cells [17, 18], suggesting that the resistance involves host cell death, predominantly in the epidermal cell layer. This is possibly due to HR, which is typical of ETI [8]. HR is reported to usually involve the death of host mesophyll cells, after the development of haustoria [19]; however, others have also observed pre-haustorial resistance responses in incompatible, ETI-mediated interactions. For example, a pre-haustorial response to stem rust was recently reported by Wang et al. [20], who observed that early responses to avirulent stem rust races in wheat carrying resistance genes Sr5 and Sr36 involved callose deposition in stomatal guard cells. Another notable example of a pre-haustorial response to wheat stem rust is the case of barley durable stem rust resistance gene Rpg1, which encodes a receptor-like kinase (RLK) that is autophosphorylated within minutes of inoculation with the avirulent pathogen [21, 22]. Rpg1 is believed to interact, presumably in the epidermis, with two avirulence proteins present in urediniospores, resulting in the initiation of HR well before haustoria formation. Thus, the stem rust resistance mediated by the 7AL rust resistance locus shows some similarity with the mechanism of action reported for barley Rpg1.

As detailed above, Col was moderately susceptible (IT 2⁺3) to stem rust race 313, while the two Col-NS lines were resistant (IT;1). Death of epidermal cells adjacent to rust invasion sites was also observed in Col, although these responses were less frequent than in Col-NS. Therefore, the resistance mechanism in Col-NS seemed to be an enhancement of a weak resistance present in Col. Interestingly, a resistance response similar to the one described here was microscopically observed in all three genotypes following inoculation with an avirulent race (data not shown). This suggests that the same resistance mechanism is effective in all genotypes, but it was not activated in Col against race 313, whereas it was in Col-NS through the specific recognition of the pathogen, indicative of an ETI response. Further studies will be needed in order to establish whether the enhanced resistance in Col-NS is derived from a basal PTI response or a weak and early ETI response present in Col.

The early resistance response correlated with fewer rust infection sites progressing beyond penetration in Col-NS compared to Col (Fig. 4). For those that successfully invaded the plant, hyphal networks and uredia were usually smaller than in Col (Figs. 1, 4). This might be due to the early resistance response, which delayed rust propagation, or to the establishment of a late resistance response. This resistance response might also involve HR, as suggested by the autofluorescence and TB staining of mesophyll cells and the leaf necrosis/chlorosis at a later stage. However, compared to the early resistance response, cell death was not common, and the propagation of fungi that evade the early resistance could be hindered by another resistance mechanism.

Transcriptome analysis

RNA-Seq is a powerful technique for quickly obtaining an overview of the molecular changes during plant-pathogen interaction. Wheat transcriptome studies during rust infection often rely on other techniques (e.g. microarray) and deal with leaf and stripe rusts [23–30]. To the best of our knowledge, the current study is the first wheat whole-transcriptome analysis during Pgt infection.

mRNA from the susceptible Col and the two resistant lines Col-NS765 and Col-NS766 was isolated at three time points, based on the microscopic observations: (1) 0 DPI to determine gene expression differences between the genotypes before the initiation of the response to rust inoculation, (2) 2 DPI to investigate gene expression differences in the early resistance response during leaf penetration, and (3) 5 DPI to investigate gene expression differences in the late resistance response during fungal invasion.

Hierarchical clustering analysis of the similarity between samples revealed a grouping into three main clusters, each representing one time point after inoculation (Fig. 6). Thus, the transcriptome profiles of all genotypes, at a given time point, were more similar to each other than those of a single genotype through time. In other words, gene expression changes through time after rust inoculation were broadly similar for all three genotypes. However, the genotypes grouped into three sub-clusters at 2 and 5 DPI, but not at 0 DPI, with Col-NS765 and Col-NS766 being clearly closer to each other than to Col at 2 DPI. This indicates that responses to stem rust diverged between the sensitive and resistant genotypes by 2 DPI. At all three time points, there were higher numbers of DE genes between Col and Col-NS765 than between Col and Col-NS766 (Fig. 7), indicating that the transcriptome profile of Col was more similar to that of Col-NS766. A possible explanation for these results is that more of the CTH-NS genetic background remains in the backcrossed line Col-NS765 than in Col-NS766.

Plant material

All Canthatch (CTH-K, CTH-NS1 and CTH-NS2) and Columbus (Col, Col-NS765 and Col-NS766) lines/mutants were provided by Kerber [11], Agriculture Canada Research Station in Winnipeg, Canada. CTH-K is a single seed derived and purified selection of CTH descent from which the NS EMS mutants originated. CTH-NS1 and CTH-NS2 are two independent EMS mutants, in which the 7D-Sup has been inactivated [11]. Col-NS765 and Col-NS766 are near-isogenic lines, derived from independent backcrosses (BC₅F₄) of each of the two CTH-NS mutants with the recurrent parent Col. Both CTH and Col are closely related to Thatcher.

Stem rust inoculation

Resistance to stem rust was evaluated with wheat stem rust Pgt culture no. 313 (pathotype 34-1,2,3,5,6,7 according to Australian nomenclature) from the Plant Breeding Institute, University of Sydney, Australia. This race is avirulent on Sr9e and Sr21, but virulent on Sr5, Sr6, Sr7b, Sr8a, Sr9b, Sr9 g, Sr11, Sr15 and Sr17. Seedlings were grown in a microclimate room (22/20 °C day/night) until the two-leaf stage (approximately 2 weeks). Inoculation was undertaken by spraying the plants with urediniospores suspended in light mineral oil. Plants were dried for 30 min and incubated for 2 days in incubation cabinets (22/20 °C day/night, high humidity). Finally, plants were transferred to a microclimate room, at a lower temperature (18/17 °C day/night), until sampling/scoring. Seedlings were scored for stem rust resistance between 11 and 15 DPI, using the stem rust scoring scale [43, 44].

Microscopic observations

For each genotype, the first leaf of two seedlings from different pots were pooled at several DPI and stained for microscopic observations, using the Zeiss Axio Imager. M1 microscope. The development of stem rust in planta was observed using WGA-FITC, which binds specifically to N-acetyl-glucosamine (i.e. chitin) and allows the observation of fungal structures as green fluorescence. WGA-FITC staining was done following a modified protocol described by Ayliffe et al. [45]. Briefly, leaf tissue was immersed in 1 M KOH and incubated at 37 °C for 2 days; KOH solution was replaced during incubation. Tissue was then neutralized in 50 mM Tris, pH 7.0, and stained with WGA-FITC at 20 µg/ml (Sigma-Aldrich, St. Louis). Fungal structures and autofluorescence were observed under blue (excitation 470–440 nm, emission 525–550 nm) and UV (excitation 365 nm, emission ≥ 420 nm) light, respectively. This experiment was performed twice: at 0, 1, 2, 5, 7 and 9 DPI on genotypes with Columbus background for the first one, and at 4, 7 and 9 DPI on all genotypes for the second one. For each genotype, hyphal network size at 9 DPI was quantified by measuring the area covered by hyphae, using the AxioVs40 software (v4.8.20). In the first experiment, all infection sites were analysed, i.e. 105, 79 and 61 infection sites for Col, Col-NS765 and Col-NS766, respectively, whereas in the second experiment, the first 30 infection sites observed were analysed. Trypan blue staining was done once on genotypes with the Columbus background, following the method described by Desmond et al. [46].

Transcriptome analysis

For each sample, the first leaf of two seedlings from different pots were pooled and quickly frozen in liquid nitrogen. Total RNA was extracted using the RNeasy Plant Mini Extraction Kit (Qiagen) and RNA integrity was checked on 1.4% agarose gel and on the Agilent 2100 Bioanalyzer. Following poly(A) isolation step performed by the sequencing company (Australian Genome Research Facility), mRNA samples were sequenced using Illumina HiSeq 2000 for 100 bp single-end and an expected yield of 30 million reads. Reads were aligned against the wheat (Triticum aestivum) Unigene database build #62 (http://www.ncbi.nlm.nih.gov/unigene), comprising 158,028 transcripts, using the BioKanga software (http://code.google.com/p/biokanga). Alignment allowed a mismatch rate of up to 10% of the length of the read. Only reads with a unique best hit were selected for differential expression analysis. Unigenes were filtered to those with a sum of counts greater than, or equal to, 100 across samples. Read counts were normalized and tested for differential expression, using the R DESeq package v1.12.0 with default parameters [13]. Hierarchical clustering was done using the Euclidian distances on the normalized reads count following DESeq variance stabilizing transformation. Eighteen pairwise comparisons were performed (between genotypes at each time point and between time points for each genotype). Genes were considered DE for FDR ≤ 0.05 and |log₂(FC)| ≥ log₂(2), and for FDR ≤ 0.075 and |log₂(FC)| ≥ log₂(1.8) if already found DE in similar comparisons. Genes were considered expressed if more than 0.15 read per kilo base per million (RPKM) was detected. Chromosomal location of each Unigene was determined using BLASTN (best hit, evalue ≤ 10⁵) against the flow-sorted IWGSC chromosome survey sequencing contigs of cultivar Chinese Spring [15]. Unigenes putatively involved in hormonal pathways (PR genes) were recovered using tBLASTN from wheat protein sequences. RT-qPCR checks were performed using the Applied Biosystems 7900HT Fast Real-Time PCR System and the relative expression level was calculated by comparing the target gene with the reference Unigene Ta.2291 (ADP-ribosylation factor) using the 2^−ΔCq formula.