Endogenous silencing of Puccinia triticina pathogenicity genes through in planta-expressed sequences leads to the suppression of rust diseases on wheat


For Correspondence (e-mail Guus.Bakkeren@agr.gc.ca).


Rust fungi are destructive plant pathogens. The draft genomes of several wheat-infecting species have been released and potential pathogenicity genes identified through comparative analyses to fungal pathogens that are amenable to genetic manipulation. Functional gene analysis tools are needed to understand the infection process of these obligate parasites and to confirm whether predicted pathogenicity genes could become targets for disease control. We have modified an Agrobacterium tumefaciens-mediated in planta-induced transient gene silencing (PITGS) assay for use in Triticum spp. (wheat), and used this assay to target predicted wheat leaf rust fungus, Puccinia triticina (Pt) pathogenicity genes, a MAP kinase (PtMAPK1), a cyclophilin (PtCYC1) and calcineurin B (PtCNB), to analyze their roles in disease. Agroinfiltration effectively delivered hairpin silencing constructs in wheat, leading to the generation of fungal gene-specific siRNA molecules in infiltrated leaves, and resulting in up to 70% reduction in transcription of the endogenous target genes in superinfected Pt. In vivo silencing caused severe disease suppression, compromising fungal growth and sporulation, as viewed by confocal microscopy and measured by reductions in fungal biomass and emergence of uredinia. Interestingly, using the same gene constructs, suppression of infection by Puccinia graminis and Puccinia striiformis was also achieved. Our results show that A. tumefaciens-mediated PITGS can be used as a reverse-genetics tool to discover gene function in rust fungi. This proof-of-concept study indicates that the targeted fungal transcripts might be important in pathogenesis, and could potentially be used as promising targets for developing RNA interference-based resistance against rust fungi.


Cereal rust fungi are amongst the most devastating plant pathogens and their continuously increasing prevalence presents major constraints to Triticum spp. (wheat) production, posing a serious threat to global food security. Conventional breeding strategies to reduce yield losses relying on natural, race-specific host resistance genes have limited durability, presumably because of the rapid evolution of these pathogens. This demands a continuous search to identify effective sources of resistance. Three types of rust fungi are known to infect wheat: stem (black) rust (Puccinia graminis Pers. f. sp. tritici, Pgt), leaf (brown) rust (Puccinia triticina Eriks, Pt) and stripe (yellow) rust (Puccinia striiformis Westend f. sp. tritici, Pst). These obligate biotrophic fungi are heteroecious, with a sexual cycle restricted to an alternate host, but the major phase of the life cycle occurs on graminaceous hosts (Bushnell and Roelfs, 1984). The different rust species infect their host in a similar fashion. Urediniospores land on leaves and, depending on various chemical and physical cues from the host surface (Hoch et al., 1987; Uppalapati et al., 2012), initiate an intricate series of developmental stages and structures essential for establishing a successful parasitic relationship with the host. After penetration into the substomatal cavity, the fungus attaches to a mesophyll cell and breaches the cell wall, thereby invaginating the host plasma membrane to form a specialized feeding structure: the haustorium. This haustorium is surrounded by an extrahaustorial matrix (EHM) that represents the interface for signal exchange as well as nutrient uptake (Voegele and Mendgen, 2003). Within 8–10 days following infection, new urediniospores are produced in uredinia that can be wind-disseminated over long distances and might result in epidemics.

Rust fungus genomic resources are being generated at a fast pace (e.g. Duplessis et al., 2011; Xu et al., 2011), allowing the prediction of many genes. However, because of the lack of functional genomic tools for rust fungi the biological functions of these genes remain elusive. Although procedures to genetically transform rust fungi exist (Webb et al., 2006; Lawrence et al., 2010), they are not routine or easily adaptable for large-scale analysis of gene functions. RNA interference (RNAi) has become a powerful tool to study functions of target genes in a transient way in organisms refractory to genetic manipulation (Hellens et al., 2005). In dicot plants, transient expression of genes from constructs delivered by Agrobacterium tumefaciens (Agroinfiltration), is well established and commonly used for gene-silencing approaches to rapidly create gene knock-down phenotypes (Johansen and Carrington, 2001). In monocot plants, gene-silencing methods make use of viral systems (Holzberg et al., 2002), but Agrobacterium-mediated delivery of silencing constructs has not been very successful, even though monocots were shown to be susceptible to genetic transformation by this organism (e.g. Grimsley et al., 1987; Hensel et al., 2009).

RNAi mechanisms seem to be present in most fungi studied, although some ascomycetes such as Saccharomyces cerevisiae (Scannell et al., 2007) and some Candida species (Nakayashiki et al., 2006), as well as basidiomycete fungi such as the corn smut Ustilago maydis (Laurie et al., 2008), and some Cryptococcus species (D'Souza et al., 2011) do not possess such genome defense mechanisms. It is unknown whether Pt has a functional silencing system, although genes coding for the necessary components of the silencing machinery seem to be present in its genome (http://www.broadinstitute.org). It was reported that organisms that live within and/or develop intimate contact with their host, such as nematodes (Huang et al., 2006), insects (Baum et al., 2007; Mao et al., 2007) and parasitic plants (Tomilov et al., 2008), are sensitive to RNA silencing molecules generated in the host, but are targeted to endogenous genes in the respective parasites. Recently, using the Barley stripe mosaic virus (BSMV), the production of RNAi molecules in barley targeted at several genes in the powdery mildew fungus, Blumeria graminis, seemed to affect fungal development in the host (Nowara et al., 2010). Using the same system, effector genes were targeted in the wheat stripe rust fungus Pst, and although a reduction in steady-state levels of corresponding mRNAs was reported, indicating the existence of a functional RNA silencing mechanism, no effect on disease suppression was observed (Yin et al., 2011).

Here we show that A. tumefaciens wheat leaf infiltration can be used to deliver T-DNA from which hairpin RNA (hpRNA) structures are expressed, which subsequently trigger transient gene silencing. When targeted to Pt genes, small interfering RNA (siRNA) molecules generated in wheat can silence these genes inside superinfecting rust fungi. This novel strategy is easier to use than inoculations with a virus or the generation of stable transgenics. We demonstrate here in a ‘proof of concept’ the feasibility of this reverse-genetic approach to study the likely roles of selected Pt genes, an MAP kinase (PtMAPK1), a cyclophilin (PtCYC1) and calcineurin B (PtCNB), the functional orthologs of which are known to be involved in pathogenicity or virulence in other pathosystems (Hu et al., 2007a; Cervantes-Chávez et al., 2011). PITGS specifically reduced the transcript levels of the target genes in the fungus, resulting in decreased fungal development and sporulation. In addition, the same silencing constructs were sufficient to suppress disease development of Pgt and Pst also on wheat, indicating the potential for developing possibly durable resistance at the genetic level against these important plant pathogens.

Results and Discussion

Development of an Agrobacterium-mediated transient RNAi assay for wheat

The generation and assessment of transgenic wheat lines is a time-consuming process, limiting the large-scale study of pathogen genes. The development of high-throughput transient RNAi approaches based on T–DNA delivery by A. tumefaciens has provided an alternative rapid tool for functional genomic studies (Hellens et al., 2005), although the technology has not been extensively studied and applied in monocots, especially in cereals. To optimize the production of RNA silencing molecules in wheat by this approach, we first tested the phenotypic response caused by silencing of the wheat phytoene desaturase (TaPDS) gene. The PDS gene has been used as a visual marker in many studies on gene silencing, and suppression of this gene by RNAi results in easily recognizable photobleaching symptoms (Scofield and Nelson, 2009). A binary vector, pRNAi-TaPDS, transcribing the TaPDS gene sequence in an hpRNA conformation from the Zea mays (maize) ubiquitin promoter (Figure 1), was moved into three different Agrobacterium strains, COR308, LBA4404 and GV3101, to test for optimal T-DNA delivery and expression. Infiltration of these Agrobacterium strains resulted in efficient endogenous TaPDS silencing in the wheat cv. Thatcher. At 7–10 days post inoculation (dpi), bleached sectors were observed in infiltrated leaves (Figure 2a). The areas lacking pigmentation did not have sharp boundaries and non-responsive cells were apparent. Leaves infiltrated with Agrobacterium alone showed no TaPDS silencing phenotype, and were indistinguishable from leaf areas infiltrated with buffer. The Agrobacterium strains COR308 and LBA4404 produced more apparent TaPDS-silencing symptoms compared with strain GV3101. This could indicate that strain GV3101 has a low efficiency in the transfer of T-DNA to cereals (Chen et al., 2010). As the silencing symptoms induced by Agrobacterium strains COR308 and LBA4404 appeared similar, strain COR308 was employed for further analysis and optimization of the transient RNAi assay.

Figure 1.

Constructs used in the Agrobacterium-mediated in planta-induced transient gene silencing (PITGS) assays: 185-bp TaPDS and 520-bp PtMAPK1 gene fragments were recombined in the cereal-specific binary expression vector pIPKb007 (Himmelbach et al., 2007), allowing expression of hpRNAi molecules from the Zea mays ubiquitin promoter (Ubi1P) and terminator (Ter) sequences. R1 and R2, attB1 and attB2 recombination site of the GateWay system (Invitrogen), respectively. The 463-bp CNB and 501-bp PtCYC1 gene fragments were cloned into binary vector pMCG161 (ChromeDB), allowing their expression in hpRNAi form from the cauliflower mosaic virus 35S promoter (35SP) and octopine synthase terminator (Ter) sequences. Intr, intron. These constructs produce a constitutive supply of gene-specific silencing molecules in host cells that have acquired T-DNA.

Figure 2.

Agrobacterium-mediated in planta-induced transient gene silencing (PITGS) assay targeting the phytoene desaturase (TaPDS) gene in wheat cv. Thatcher. (a) Silencing phenotype caused by different Agrobacterium strains carrying a TaPDS gene segment. No photobleaching was observed for Agrobacterium alone (ck) or buffer controls (BC). (b) siRNA molecules specific to the TaPDS gene were detected in total RNA extracted from pooled leaf material from three different plants 3 days after agroinfiltration with pRNAi-TaPDS. An autoradiograph of RNA blot hybridized with a 32P-labelled gene-specific probe. No signal was detected in the control sample treated with Agrobacterium alone (ck). The bottom panel shows ethidium bromide-stained rRNA as the gel loading control; siRNA size indicated. (c) qRT-PCR measurements of TaPDS transcript abundance relative to an endogenous reference gene, Ta_EF1, in leaves infiltrated with pRNAi-TaPDS or the control treated with Agrobacterium alone (Agt-COR308, set at 100%). cDNA generated from total RNA isolated from pooled leaf material from three different plants. Bars represent mean values ± SDs of three independent sample collections. The reduction in the target TaPDS gene transcription level in silenced leaves, compared with non-silenced leaves treated with Agrobacterium alone, was significant at P < 0.05 in a Student's t-test.

The concentration of Agrobacterium cells is reported to have a significant effect on the outcome of infiltration (Bhaskar et al., 2009). We therefore standardized the infiltration assay by testing various cell concentrations. Bacterial suspensions with an OD600 of 0.5–0.7 resulted in a maximum level of TaPDS-related silencing symptoms. A low concentration of bacteria (OD600 < 0.4) resulted in a poorly developed or almost negligible photobleaching phenotype, whereas high concentrations (OD600 > 1.0) often caused leaf senescence, starting with the production of necrotic spots around the infiltration zone.

Detection of low molecular weight RNA (siRNAs) corresponding to the target TaPDS gene in leaf tissue agroinfiltrated with the pRNAi-TaPDS construct confirmed the degradation of TaPDS hpRNA by endogenous dicer-like nucleases (Figure 2b). Subsequent quantitative real-time reverse transcription PCR (qRT-PCR) analysis showed significant reductions (60%) in TaPDS gene transcript levels in leaves expressing the silencing construct, compared with controls (Figure 2c). The presence of siRNA molecules and the reduction in TaPDS mRNA levels suggest that the observed phenotypic effects are attributable to gene silencing achieved at the post-transcriptional level. RNAi-induced silencing has been reported to be non-cell autonomous and, once initiated, can spread systemically in a ‘source-to-sink’ direction (Dunoyer et al., 2010b; Brosnan and Voinnet, 2011). However, we did not observe TaPDS-related chlorophyll bleaching in distal and newly emerging leaves, and silencing was more or less confined to the infiltration zone. There are indications that silencing signals (siRNA molecules) can move across 10–15 cells through plasmodesmata without the need for amplification (Dunoyer et al., 2010b). However, systemic movement over longer distances requires an amplification of the original signal (Himber et al., 2003), which might be inefficient for the TaPDS target. Together, these results provided a benchmark for the application of this transient approach for the functional analysis of fungal genes.

Selection of P. triticina target genes

Of prime interest to the host–rust fungus interaction are genes encoding proteins with pathogenicity or virulence functions. The silencing of such genes is likely to result in a quantifiable phenotype, and if successful may offer an effective alternative to combat rust fungi. A previously generated Pt expressed sequence tag (EST) database (Xu et al., 2011) provided candidate gene sequences. PtCYC1 codes for a cyclophilin, a member of a large group of proteins generally possessing peptidylprolyl-cis/trans-isomerase (PPIase) activity, and having well-defined functions as folding catalysts and chaperones involved in cell signaling. Cyclophilins play roles in virulence in several pathogens (Wang et al., 2001; Viaud et al., 2002). As a second candidate we used the first identified and characterized rust fungus mitogen-activated protein kinase gene, PtMAPK1 (Hu et al., 2007a), a member of a family of serine/threonine protein kinases that are conserved in most eukaryotes, and are known to play roles in the phosphorylation of transcription factors required for growth and differentiation processes. We showed earlier that PtMAPK1 complemented deletion mutants of the homologs Ubc3/Kpp2 and Kpp6 in the corn smut fungus U. maydis, impaired in mating and pathogenicity (Hu et al., 2007a). In addition, in our database we identified an EST sequence with homology to CNB, coding for the calcineurin regulatory subunit. Comparison with the recent public Pt genome sequence (http://www.broadinstitute.org) indicated that this EST sequence seemed derived from a (possibly contaminating) ascomycete. It shared 65% nucleotide identity positions with the homolog PtCNB, and would allow us to test whether diverged sequences could also be used to silence endogenous homologs (Figure S1). CNB complexes with the catalytic subunit CNA to form calcineurin. In the fungal kingdom, the calcium–calcineurin pathway is involved in morphogenesis, circadian rhythm, cell cycle progression, stress response and virulence (Nguyen et al., 2008; Egan et al., 2009). Previously we showed that the calcineurin subunits are important virulence factors in the barley smut fungus Ustilago hordei (Cervantes-Chávez et al., 2011). These lines of evidence suggested that these three genes might have a similar role in Pt parasitism of wheat.

Transient expression of candidate fungal hpRNA is processed by the host silencing machinery

Selected 3′ regions of PtMAPK1, PtCYC1 and PtCNB-like sequences were amplified and cloned in hpRNA-generating silencing vectors (Figure 1). Small RNA molecules, specific to the PtMAPK1, PtCYC1 or CNB sequences, were detected in total RNA isolated from leaves agroinfiltrated with the respective silencing constructs 3 days earlier (Figure 3a). This showed that Agrobacterium infection could result in the induction of the transient expression of fungal double-stranded RNAs in wheat cells, which subsequently acted as substrates for plant nucleases catalyzing their cleavage into corresponding siRNA molecules. Various quantities of these molecules were detected in different plants, and similar variation was seen for the three constructs, indicating that both the 35S and maize ubiquitin promoters resulted in significant expression.

Figure 3.

Host-generated silencing RNAs and in planta-induced transient gene silencing (PITGS) effect on transcription of Pt target genes. (a) siRNA molecules specific to the PtMAPK1, PtCYC1 and PtCNB-like sequences were detected in total RNA extracted from pooled leaf material from three different plants, 3 days after agroinfiltration with the respective silencing vectors. Autoradiograph of RNA blots hybridized with 32P-labeled gene-specific probes. Bottom panels, rRNA gel loading controls. Lanes 1–3, different individual samples; Ck, control treated with Agrobacterium alone. siRNA size indicated. (b) Transcript abundance of the PtCYC1, PtMAPK1 and PtCNB genes, as indicated under each panel, quantified by qRT-PCR 5 days after superinfection with Pt on plants previously infiltrated with the respective RNAi constructs (MA, RNAi-PtMAPK1; PD, RNAi-TaPDS; CY, RNAi-PtCYC1; CN, RNAi-CNB), (BC, buffer control) or treated with Agrobacterium alone (CO, Agt-COR308). Transcript abundance was measured relative to a fungal endogenous reference gene (Experimantal procedures), and the CO control was set at 100%. Transcript reduction is specific to the targeted gene. cDNA was generated from total RNA isolated from pooled leaf material from three different plants. Values represent means ± SDs of three independent sample collections. The asterisks indicate that the reduction in the target gene transcription levels in silenced leaves, compared with in non-silenced leaves (CO), was significant at P < 0.05 in a Student's t-test.

In vivo silencing of fungal target genes

Host cell-produced fungal siRNAs are postulated to navigate across haustorial membranes and matrix to induce degradation of cognate mRNAs in fungi (Nowara et al., 2010; Yin et al., 2011). qRT-PCR analysis of total RNA isolated from silenced leaves, 5 days after superinfection with Pt urediniospores, revealed specific reductions in transcript levels of the fungal target genes. The endogenous transcript levels of the PtCYC1 and PtMAPK1 genes showed an approximate 63 and 70% downregulation, respectively, in plants agroinfiltrated with the corresponding RNAi constructs, compared with controls (Figure 3b). Interestingly, the endogenous PtCNB gene transcript levels were also reduced by 59%, although this was caused by a CNB sequence sharing 65% nucleotide identity with the PtCNB homolog. There are, however, several stretches of identical nucleotides along the target sequence, the longest match being 14 bases and part of a 32-bp homology, with only two mismatches, and it has been shown that as few as between eight and sixteen nucleotides of contiguous homology between an siRNA and target mRNA can attenuate gene activity (Jackson et al., 2006; Haley et al., 2010). The silencing induced by each respective RNAi construct was specific to the corresponding fungal gene targeted for PITGS, and was not found to influence the expression of other, unrelated genes such as the other two fungal genes tested (Figure 3b). The silencing induced against TaPDS was also not observed to influence the expression of candidate fungal genes (Figure 3b). The reduction in the relative abundance of fungal gene transcripts strongly suggested that the translocation of silencing molecules from host cell to fungal cytoplasm occurred. The induction of in vivo silencing of the Pt target genes, in conjunction with the detection of gene-specific siRNA molecules in the wheat host, indicate RNAi as the underlying mechanism.

PITGS of candidate Pt pathogenicity genes compromises fungal development

To study whether the PtMAPK1, PtCYC1 and PtCNB genes have any roles in pathogenicity or virulence, we investigated the possible effect of gene silencing on fungal development in wheat host leaves superinfected with Pt urediniospores after agroinfiltration with the corresponding RNAi constructs. Disease incidence was first assessed by a physiological parameter. A quantitative measurement was made of the number of sporulating uredinia developing on a defined surface area of the infiltrated leaves at 10 days after fungal inoculations (Experimental procedures). The plants infiltrated with either of the three silencing constructs exhibited much reduced Pt disease symptoms, as indicated by significantly lower (51–68%) pustule densities compared with those infiltrated with Agrobacterium alone, which in turn were as susceptible to Pt as buffer-inoculated control plants; silencing per se, induced by the pRNAi-TaPDS construct, also did not affect sporulation (Figure 4a,b). This indicated a clear PITGS effect on urediniospore production. The suppression was, however, local (mainly confined to the infiltration zone), and no systemic silencing effect on disease development was observed in upper, non-agroinfiltrated leaves.

Figure 4.

Effect of the presence of gene-specific siRNA molecules in wheat cv. Thatcher on Pt infection. (a) Agroinfiltration with the PtMAPK1, PtCYC1 or CNB RNAi constructs resulted in the suppression of Pt disease symptoms on leaves, whereas no effect on disease susceptibility was observed on leaves infiltrated with Agrobacterium alone (Agt-COR308), buffer (BC) or pRNAi-TaPDS silenced control plants. Plants photographed 10 days after Pt inoculation. (b) Quantification of pustule density on leaves caused by superinfection with Pt after PITGS of the target fungal genes by the respective RNAi construct, compared with controls. (c) Fungal biomass measurements by qRT-PCR analysis of total genomic DNA, extracted 5 and 10 days after Pt inoculation of infiltrated leaves. Ratio of fungal nuclear to wheat nuclear genomes using single-copy genes PtRTP1 (rust transferred protein 1) and TaEF1, respectively (Song et al., 2011). Relative ratio of fungal biomass in leaves silenced with RNAi constructs targeting Pt genes compared with control leaves infiltrated with Agrobacterium alone. DNA was extracted from pooled leaf material from three different plants. Mean values ± SDs from three independent sample collections. Differences were significant at P < 0.05 (Student's t-test).

To test whether the observed reduction in sporulation was correlated with mycelial growth in host tissue, fungal biomass measurements were made in the infected leaves. Total genomic DNA was extracted from agroinfiltrated leaves superinfected with Pt, and the relative levels of single-copy Pt and wheat genes were quantified by qRT-PCR. At 5 and 10 days after Pt inoculation on leaves previously agroinfiltrated with silencing constructs, fungal biomass was significantly reduced by 60–65 and 59–69%, respectively, compared with controls infiltrated with Agrobacterium alone or buffer, or silenced control plants (Figure 4c). This result indicated that fungal development was impaired, probably as a result of the virulence penalty imposed by the silencing of the targeted fungal genes.

The formation of mature haustoria and the development of secondary hyphae are prerequisites for a successful interaction between hosts and biotrophic fungi. Therefore, we further investigated the possible role of the PtMAPK1, PtCYC1 and PtCNB genes in host invasion by microscopic assessment of Pt development in foliar tissue. Wheat leaves infiltrated with Agrobacterium without silencing constructs showed widespread colonization of mesophyll cells by Pt 5 days after inoculation, typified by an extensive hyphal network and numerous established haustoria (Figure 5). In contrast, the presence of the various fungal siRNA molecules in the host cells clearly suppressed fungal development in silenced leaves. Primary haustoria development was limited and hyphal growth was severely reduced, remaining mainly confined to the first few host mesophyll cells (Figure 5). A differential effect on the developmental pattern of the secondary hyphae caused by the presence of the various siRNA molecules was also observed. This was most pronounced when caused by the PtMAPK1 silencing construct, which seemed to keep the secondary hyphae shorter compared with the growth inhibition imposed by the targeting of the PtCYC1 and PtCNB genes. We reported before that PtMAPK1 is differentially expressed, with high levels found during early infection and sporulation (Hu et al., 2007a). A similar expression profile was also observed for a Pst homolog (Guo et al., 2011). The cytological observations corroborated these expression studies and further suggested that the MAPK signaling cascade plays an important role during Pt development in wheat. The differential effects seen by microscopy suggest that the targeted genes may be regulating different development stages during Pt infection, warranting further investigation, but indicating the feasibility of using this approach for the functional analysis of rust fungus genes. The observed suppression of fungal growth inside silenced host tissue correlated well with the measured reduction in fungal biomass and uredinia formation. Overall, these results strengthen the view that the targeted fungal genes might have important roles in pathogenesis.

Figure 5.

Confocal microscopic visualization of the in planta-induced transient gene silencing (PITGS) effect targeting the PtMAPK1, PtCYC1 and PtCNB genes on colonization of wheat leaf tissue by Pt. Plates represent a projection of a 3D Z-series of scans taken from inside wheat cv. Thatcher leaves 5 days after Pt inoculation; side panels provide the respective 3D views through the Z-series. Fungal structures stained with Uvitex 2B and acridine orange-stained plant cell walls are visible. Plants agroinfiltrated to deliver hpRNAi constructs, as indicated, show poorly developed fungal mycelium with very few developed haustoria (arrows). This contrasted starkly with the control treated with Agrobacterium alone (Agt-COR308), where an extensive mycelial network with many formed haustoria inside mesophyll cells could be observed. AP, appresorium; HA, haustorium; HMC, haustorium mother cell; IH, infection hypha; SSV, substomatal vesicle; ST, stoma. Scale bar:20 μm.

Sequence comparisons had indicated that effective RNAi targets for these candidate genes should be present in both Pgt and Pst (Figure S1). Therefore, wheat leaves agroinfiltrated with or without silencing constructs were also challenged with urediniospores from Pgt and Pst isolates known to infect wheat cv. Thatcher. A reduction in uredinia development of 43–61 and 48–69% in leaves infiltrated with target RNAi constructs, compared with controls treated with Agrobacterium alone, was also seen for Pgt and Pst, respectively, suggesting that these conserved genes might have similar roles in the virulence of the different Puccinia species (Figure 6).

Figure 6.

Response caused by the various silencing constructs on Puccinia graminis f. sp. tritici (Pgt, stem rust) and Puccinia striiformis f. sp tritici (Pst, stripe rust) infection of wheat cv. Thatcher. (a) The generation of PtMAPK1, PtCYC1 or CNB gene-specific siRNA molecules in the wheat host, caused by agroinfiltration with the respective hpRNAi constructs, resulted in the suppression of disease symptoms on leaves superinfected with Pgt (a) or Pst (c), whereas no effect on disease susceptibility was observed on leaves infiltrated with Agrobacterium alone (Agt-COR308) or buffer control (BC). Plants were photographed 10 days after Pgt inoculation. (b) Quantification of pustule density on leaves caused by superinfection with Pgt or Pst (d) after in planta-induced transient gene silencing (PITGS) of the indicated genes, compared with controls. Pustules were counted 10 days after fungal inoculation per area of leaf surface, spanning 3 cm around the infiltration site. The mean value is given, representing data from 30 randomly selected leaves from three experiments. Error bars depict standard errors. Statistical analysis was performed with a Student's t-test (P < 0.05 versus control).

Although RNAi is considered to be highly sequence specific, siRNAs can induce ‘off-target’ gene silencing effects (Jackson et al., 2006; Haley et al., 2010; Senthil-Kumar and Mysore, 2011). No obvious phenotypic effects were observed in wheat plants infiltrated with the various silencing constructs. Indeed, only low overall sequence similarity between the 3′ fungal target sequences and a large, currently available collection of wheat sequences could be revealed (Figure S1). However, we cannot completely rule out potential silencing effects by fungal-specific siRNA molecules on transcripts in the wheat genome. Overall though, as all three silencing constructs had a major effect on fungal development in the wheat host, and since the CNB sequence had virtually no sequence similarity to its closest homolog in wheat (being extremely fungal-specific), we think that a contribution from a host gene product(s) to the suppression effect is unlikely.

The off-target effect could be more pronounced in the fungus, and this might have contributed to the disease suppression we observed. With more extensive genome sequences now available, we identified possible fungal homologs using the blastn algorithm. Depending on the overall nucleotide similarity to the 3′ fungal target regions used, and judged significant and possibly able to trigger silencing effects, six potential Puccinia MAPK-related homologs and possibly one cyclophilin-related homolog were identified: the CNB sequence seemed to represent a single gene. Therefore, at this point we cannot ascertain that the disease suppression observed was solely the result of the silencing of the respective candidate fungal gene, possibly with the exception of the CNB gene. Therefore, the use of this tool to assess (virulence and pathogenicity) gene function probably depends on whether related homologs in the genome are expressed at the same time. To achieve crop protection, however, a ‘broad-spectrum’, multiple fungal gene-specific silencing strategy is preferable. The CNB-like sequence, sharing a higher level of identity with ascomycete homologs, efficiently silenced the endogenous PtCNB gene. This result illustrates the flexibility of this system, but we have not tested the interesting possibility that infection by other (ascomycete) fungal pathogens of wheat, such as the tan spot fungus Pyrenophora tritici-repentis, could be compromised as well.

Movement of silencing signals from host to fungus

The generation of self-complementary RNA, whether through antisense RNA transcription or hairpin formation, triggers a sequence-specific mRNA degradation, leading to gene silencing. RNAi signals are diffusible and can traverse across source-to-sink gradients (Brosnan and Voinnet, 2011). In plants, the transgene silencing signals mainly move short distances from cell to cell, but can also spread over long distances through the vasculature, the latter process being mostly evident through grafting experiments (Palauqui et al., 1997; Dunoyer et al., 2010a). Like other eukaryotes, most fungi are also sensitive to RNAi and recently it was shown that silencing signals can travel from host cell to fungus, presumably through the haustorial interface (Nowara et al., 2010; Yin et al., 2011). Uptake of genetic material, in the form of RNA, has also been demonstrated between host plants and parasites such as insects, nematodes or other parasitic plants feeding on them (Huang et al., 2006; Baum et al., 2007; Mao et al., 2007; Tomilov et al., 2008). However, the mechanism behind such cross-species transfer of silencing molecules remains to be elucidated. Our data show that in the Puccinia-wheat pathosystems, the silencing signals are likely to be the siRNA molecules, as these were generated in the host cells (Figures 2 and 3). Preliminary studies suggest that translocation of genetic material from host to infecting fungal cells may occur via the exosomal biogenesis pathway (Casadevall et al., 2009; Meyer et al., 2009; Lu et al., 2012). At sites of fungal penetration, multivesicular compartments, including multivesicular bodies (MVBs) and cell wall-associated paramural bodies, aggregate in host cytoplasm around haustorial complexes, facilitating an anterograde, and probably retrograde, polarized vesicle trafficking across the plant–pathogen cellular interface. These MVBs contain multiple intraluminal vesicles (ILVs), which upon fusion with the plasma membrane are released extracellularly as exosomes into the paramural space (Meyer et al., 2009; Lu et al., 2012). Interestingly, exosomes have been shown to contain both mRNA and non-coding small RNAs that can be delivered, and can be functional, in recipient cells, thereby facilitating genetic exchange between cells (Valadi et al., 2007). Transport of RNA by exosomes might require appropriate receptors at the cell surface for attachment, or follow other specialized transfer pathways. MVB-like compartments have been reported in trafficking mechanisms at intercellular channels called gap junctions, nanotubes or even internalization of sections of plasma membrane by neighbouring cells (Gibbings and Voinnet, 2010). It is possible that in our system the siRNA species generated in the host ‘silencing donor’ are transferred into the ‘fungal recipient’ through this exocytic/endocytic exchange mechanism at the haustorium interface (Figure 7). During plant infection, biotrophic and hemibiotrophic fungi secrete effector proteins, some of which translocate to the plant apoplast or cytosol where they can alter host responses to condition susceptibility. Recently, it was reported that some effectors secreted by oomycete pathogens, which have infection strategies similar to rust fungi, may enter host cells via receptor-mediated endocytosis (Kale and Tyler, 2011). However, similar effector delivery has not yet been discovered for biotrophic fungi. It is currently unknown whether a bidirectional trafficking across the host–parasite cellular interface can take place, and whether this could include small RNA species. Nutrient uptake has been reported to occur in bean rust fungal haustoria through specific transporters (Voegele and Mendgen, 2003), and it is possible that siRNAs use these or similar transporters. Alternatively, their sizes may be sufficiently small as to allow other (passive) ways for crossing various membranes and the EHM.

Figure 7.

Schematic presentation of possible in planta-induced transient gene silencing (PITGS) mechanisms by which siRNA molecules generated in host cells could be delivered into fungal cells. Fungal dsRNA, produced inside host cells, is cleaved by the plant silencing machinery using endonuclease-type DICER enzymes into small silencing molecules (siRNAs). Plasma membrane proteins (PMPs) and other specific materials destined for extracellular secretion are internalized by endocytosis, trapping small RNAs along the way to form sorting endosomes that subsequently get compartmentalized as intraluminal vesicles (ILVs) within multivascular bodies (MVBs). These MVBs can either fuse with lysosomes to permit degradation of the ILVs or follow a retrograde pathway, fusing with the plasma membrane (PM) and releasing the ILVs as exsosomes into the paramural space. This space acts as a nexus of communication and transport between host or pathogen cells and their respective exterior environments. Passage through the haustorial cell wall (HCW), either active or passive, occurs and once inside the fungal haustorium the silencing molecules trigger RNAi of their mRNA targets, and may act as primers in the fungal silencing pathway, leading to the generation of systemic silencing signals. HCW, haustorial cell wall; PCW, plant cell wall; HMC, haustorium mother cell; SSV, substomatal vesicle; UD, urediniospore.

Genomic resources for the agronomically important cereal rust fungi are being generated at a fast pace, yet functional assay systems are lacking to test for genes predicted to be involved in pathogenicity or virulence in these obligate biotrophic pathogens. We made use of an RNAi approach by developing a functional Atumefaciens-mediated PITGS assay for wheat that should allow the rapid testing of many genes. Using this approach we demonstrated the silencing of three endogenous Pt genes, suggesting the feasibility of this system to study loss-of-function phenotypes in rust fungi. Gene silencing by transient expression of hpRNA does not require stable genetic transformation, and is consequently a promising technique for functional genomic studies in organisms refractory to genetic manipulation. Our results demonstrate that wheat-expressed fungal double-stranded RNA (dsRNA) can trigger RNAi of corresponding homologous genes in several Puccinia species. Furthermore, silencing of PtMAPK1, PtCYC1 and PtCNB results in significant disease suppression of three major wheat rust pathogens, showing that these genes and the signaling cascades they control might be important in the disease process, and suggesting that they could be excellent targets to generate durable and broad-spectrum genetic resistance against these destructive fungi. Pyramiding multiple gene targets for silencing could provide further disease suppression, and could potentially be used for rust disease control in wheat.

Experimental procedures

Growth conditions and infection procedures

Fungal isolates, Pt race 1 (BBBD), Pgt isolate 1373 (TPMK, Manitoba, Canada, 2011) and Pst isolate str9 (Manitoba, Canada) were maintained and increased on susceptible Triticum aestivum (L.) cv. Thatcher (RL 6101) plants. Inoculations were performed as described by Song et al. (2011). For PITGS assays, the inoculum strength was quantified, and 20 mg of urediniospores suspended in 2 ml of mineral oil was used to inoculate a batch of 18 10 × 10-cm pots containing four or five wheat plants each. This concentration of urediniospores was sufficient to cause a uniform infection in wheat plants, and was used in all experiments. For the inoculation of plants in each replicate within an experiment, the same procedure and concentration of inoculum was used.

Vector construction

A 520-bp fragment, representing the 3′ end of PtMAPK1 (Hu et al., 2007a; GenBank accession 68303937) was amplified by PCR using primers PtMAPK-F and PtMAPK-R (Table 1). The PCR product was directionally recombined into the vector pENTR/D-topo (Invitrogen, http://www.invitrogen.com) using the CACC string added during the PCR amplification and LR recombinase (Gateway technology; Invitrogen). The insert of the resulting entry clone was recombined with the binary destination vector pIPK007 (Himmelbach et al., 2007) using the LR recombination reaction to create the Agrobacterium binary construct pRNAi-PtMAPK1. Similarly, to generate the PDS silencing construct, a 185-bp 3′ end of the TaPDS gene was amplified from wheat (T. aestivum cv. Thatcher) cDNA using primers TaPDS-F and TaPDS-R (Table S1), recombined directionally in pENTR/D-topo, and its insert subsequently recombined into binary vector pIPKb007 to yield pRNAi-TaPDS.

Table 1. Primers used in this study
GenePrimer sequence
Pt succinate dehydrogenase-QPCR-FGGTTCCAGCGATAGATCGAG
Pt succinate dehydrogenase-QPCR-RCAACTACGACCAGCCACTCA

To generate pRNAi-CNB, an EST (PT0315.I05.C21.ptt, GenBank accession GR495416), isolated from a Pt teliospore cDNA library was used. This sequence was annotated as a calcineurin B regulatory subunit, CNB, but later, when the stem, leaf and stripe rust fungal genomes became available (Broad Institute, http://www.broadinstitute.org; Cantu et al., 2011), proved to be more closely related to ascomycete CNB gene sequences. For example, this EST sequence revealed 83% nucleotide and 96% amino acid identities with a homolog in Aspergillus kawachii (Figure S1). A 463-bp 3′-end fragment of this EST clone was amplified by PCR using the forward primer CNB-F, containing SpeI and AscI restriction sites, and reverse primer CNB-R containing SgfI and AvrII restriction sites (Table 1). The amplified fragment was then sequentially cloned into the SpeI/AscI and SgfI/AvrII sites of plasmid pUBleX1-RNAi, containing a fungal intron (Laurie et al., 2008). The 1250-bp BglII CNB-intron hairpin-containing fragment was made blunt with the Klenow PolI fragment, then digested with Spe1 and subsequently cloned into the EcoRV- and Spe1-digested binary vector pMCG161 (ChromeDB). This placed the 3′-CNB gene sequences behind the 35S promoter-Adh1 intron sequences for expression as a self-complimentary dsRNA form in cereals.

pRNAi-PtCYC1 was generated as follows: a Pt EST assembly, PtContig6674, consisting of ESTs originating from cDNA libraries constructed from several developmental stages (TR24, GenBank accession number BU672663; PT0012a.A01.B7.ptu.chim2, EC395818; PT0055c.B07.BR.ptu, EC399927; PT0121b.F04.BR.pth, EC414784; and PT0191.G22.CPTR.ptm, EC418145 (Hu et al., 2007b; Xu et al., 2011), encoded a protein with homology to a cyclophilin. In the Pt draft genome (Broad Institute), PtContig6674 matched 15 sequences with various homology using the Blastx algorithm (with probability of e = 0, PTTG_00425, to e < 10−4, PTTG_06503). In the Pgt draft genome, homology to at least 13 genes, with significance ranging from e = 0 (e.g. PGTG_14910 and PGTG_14936) to as high as e = 10−5 (PGTG_10482) was found. PtContig6674 seemed to represent a full-length cDNA, matching the best-predicted gene PTTG_00425 of 985 bp with six introns, coding for a protein of 138 amino acids, and was named PtCYC1. PtContig6674 contained two ESTs from the wheat infection stage, strengthening the idea that this cyclophilin might have a function during pathogenicity. Similarly as for CNB, a 501-bp 3′ part of PtCYC1 was amplified using primers PtCYC-F and PtCYC-R (Table S1) and cloned sequentially into the SpeI/AscI and SgfI/AvrII sites of pUBleX1-RNAi. The 1350-bp BglII PtCYC1-intron hairpin-containing fragment was cloned into pMCG161 to generate binary vector pRNAi-PtCYC1 similar to the CNB fragment. All constructs described were verified by sequencing and subsequently transferred into the respective Agrobacterium strains: COR308 (Hamilton et al., 1996), LBA4404 (Hoekema et al., 1983) or GV3101 (Koncz and Schell, 1986) by electroporation.

Agroinfiltration assay

A single colony of Agrobacterium was cultured in 10 ml of YEB medium (per liter, 1 g yeast extract, 5 g beef extract, 5 g peptone, 5 g glucose and 0.493 g MgSO4.7H2O, pH 7.2), supplemented with antibiotics as required, and grown overnight at 28°C with continuous shaking at 200 rpm on an orbital shaker. Of this bacterial culture, 250 μl was transferred to 50 ml fresh YEB with antibiotics as required, 10 mm 2-(N-morpholino)ethanesulfonic acid (MES, pH 5.6) and 20 μm acetosyringone, and grown at 28°C to an OD600 of 1 for induction. The cells were harvested twice by centrifugation at 1700 g for 5 min at 4°C and the pellet washed with 5 ml of ice-cold sterile distilled water. Cells were then resuspended in infiltration buffer (10 mm MES, pH 5.6, 10 mm MgCl2, 200 μm acetosyringone) to a final desired concentration. The bacterial suspension was then incubated for 3 h at 20–25°C before infiltrating leaf tissue. As controls, Agrobacterium strains without constructs, or infiltration buffer solution, was used in all experiments. The inoculum was infiltrated through the abaxial surface of wheat seedlings (8–9 days old, two-leaf stage) using a 1-ml sterile syringe (without a needle) by applying gentle pressure. A small incision was made at the site of infiltration using a sterile needle to enhance the efficiency of infiltration. Plants were maintained in growth chambers under a 16-h photoperiod with a temperature of about 22°C.

DNA isolation

Fungal genomic DNA was isolated as described by Hu et al. (2007b). Plant genomic DNA was isolated using the CTAB method (Allen et al., 2006). For biomass quantification, epiphytic fungal mycelium was wiped off wheat leaf surfaces using wet cotton pads prior to genomic DNA extraction.

RNA extraction and cDNA synthesis

Total RNA was isolated from leaf tissue around infiltrated areas using Trizol reagent (Invitrogen), following the manufacturer's instructions. For qRT-PCR analysis, RNA samples were treated with TURBO DNase I (Ambion, now Invitrogen, http://www.invitrogen.com). The absence of genomic DNA contamination was subsequently confirmed by the lack of PCR amplification of the RNA samples using Pt-specific primers for the PtRTP1 gene (rust transferred protein homolog, PTTG_03497; Song et al., 2011) and wheat-specific primers for the TaEF1 gene (GenBank accession M90077.1; Table 1). First-strand cDNA was synthesized from 1 μg of total RNA in a final volume of 20 μl using SuperScriptIII enzyme (Invitrogen) and oligo (dT)18 primers, according to the manufacturer's instructions.

Quantitative real-time PCR

Quantitative real-time PCR was performed on a CFX96TM Real-Time PCR machine (Bio-Rad, http://www.bio-rad.com). Specific primers for each gene were designed using Primer 3.0, and are listed in Table S1. qRT-PCR was conducted in a 10-μl volume using SsoFast EvaGreen Supermix (Bio-Rad). Thermal cycling parameters were: 98°C for 2 min, followed by 39 cycles of 95°C for 10 ses and 60°C for 30 sec. Three technical replicates were performed on each sample. All products were subjected to melting curve analysis between 65°C and 95°C, to determine the specificity of the PCR reaction. Experiments included a non-template control. Relative levels of gene transcripts were computed by qpcr 2.2 (Bio-Rad) using the comparative quantification method, with the Pt succinate dehydrogenase gene as the endogenous reference gene for normalization, previously selected from several candidates (Song et al., 2011). To distinguish the PtCNB gene transcripts from the CNB sequences derived from the EST, PtCNB-specific qRT-PCR primers were designed (Table 1).

To measure changes in fungal biomass, the relative quantification of the single-copy target genes PtRTP1 and TaEF1 was assessed (Song et al., 2011). Total genomic DNA of wheat cv. Thatcher or Pt was used to prepare standard curves derived from at least six serial dilutions for each. The correlation coefficients for the analysis of the dilution curves were above 0.99. The relative quantities of PCR product of PtRTP1 and TaEF1 in mixed/infected samples were calculated using the gene-specific standard curves to quantify the Pt and wheat gDNA, respectively.

siRNA detection

To detect small RNA species, total RNA was extracted from wheat leaves 3 days after agroinfiltration. For each sample, the same quantity of RNA was separated on a 15% polyacrylamide gel and transferred to neutral Hybond NX membrane (Amersham) by electroblotting. The transferred molecules were chemically cross-linked to the membrane as described by Pall and Hamilton (2008). As signal markers, gene-specific oligonucleotides were also loaded on the same gels. For hybridization, the ULTRAhyb Hybridization buffer (Ambion, Burlington, ON, Canada) was used, and to generate probes, PCR-amplified target gene fragments were labeled with [α32P]dCTP using a random primer DNA labelling kit (Amersham, now GE Healthcare Life Sciences, http://www.gelifesciences.com). Hybridization was carried out at 38°C for 18 h, after which membranes were washed twice in double-strength SSC (30 mM sodium citrate, 300 mm NaCl) + 0.1% SDS buffer for 15 min, each at 42°C, before exposure to HyperfilmTM MP (Amersham).

Uredinia count

The visual assessment of disease was performed by counting the number of uredinia that appeared 10 days after rust inoculation per cm2 of infiltrated leaf surface using ImageJ (http://rsb.info.nih.gov/ij). Uredinia were counted within a 3-cm2 area around the infiltration point (this was the area where solution could easily be infiltrated in leaves), and the average number per leaf was calculated. Only lesions that were sporulating were counted. To avoid bias among leaf samples, leaves from thirty treated plants from three experiments were randomly selected.

Confocal microscopy

Leaf samples were collected 5 days after fungal inoculation. Fungal structures were stained with Uvitex-2B as described by Moldenhauer et al. (2006). Acridine orange (Sigma-Aldrich, http://www.sigmaaldrich.com) was used to stain plant cell walls. Fluorescent microscopy was carried out using a SP2-AOBS laser scanning confocal microscope (Leica,http://www.leica.com). The fluorescence of Uvitex 2B and Acridine Orange was detected by excitation at 405 and 514 nm, respectively, and scanning with filter settings at 411–485 and 550–560 nm, respectively.


The authors wish to thank T. Fetch for Pgt isolates, M. Weis for help with confocal microscopy and D.L. Joly and H. Sanfaçon for their comments. This research was partially funded through the interagency NSF/USDA-CSREES Microbial Genome Sequencing Program (award 2008-35600-04693 to C. Cuomo, J. Fellers, L. Szabo and G. Bakkeren).