It has been reported that filament-forming surface proteins such as hydrophobins are important virulence determinants in fungi and are secreted during pathogenesis. Such proteins have not yet been identified in obligate biotrophic pathogens such as rust fungi. Rust transferred protein 1 (RTP1p), a rust protein that is transferred into the host cytoplasm, accumulates around the haustorial complex. To investigate RTP1p structure and function, we used immunocytological, biochemical and computational approaches. We found that RTP1p accumulates in protuberances of the extra-haustorial matrix, a compartment that surrounds the haustorium and is separated from the plant cytoplasm by a modified host plasma membrane. Our analyses show that RTP1p is capable of forming filamentous structures in vitro and in vivo. We present evidence that filament formation is due to β–aggregation similar to what has been observed for amyloid-like proteins. Our findings reveal that RTP1p is a member of a new class of structural effectors. We hypothesize that RTP1p is transferred into the host to stabilize the host cell and protect the haustorium from degradation in later stages of the interaction. Thus, we provide evidence for transfer of an amyloid-like protein into the host cell, which has potential for the development of new resistance mechanisms against rust fungi.
A characteristic feature of numerous plant pathogens is their biotrophic lifestyle. Biotrophic pathogens depend on a living host for successful colonization and completion of their lifecycle (Mendgen and Hahn, 2002). Sequencing several genomes of obligate biotrophic pathogens including oomycetes and fungi has led to a broader understanding of biotrophy. While metabolic pathways often show convergent gains and losses between biotrophic pathogens, the ‘effector’ complement has significantly diverged (Kemen et al., 2011a; McDowell, 2011). Effectors are defined as small secreted molecules/proteins that facilitate infection as virulence factors on some hosts but trigger defence responses as avirulence factors on others (Hogenhout et al., 2009). Biotrophic fungi and oomycetes form haustoria, specialized structures that provide intimate contact with host cells (Voegele and Mendgen, 2003). Haustoria are hyphal branches that penetrate the plant cell wall and invaginate the host plasma membrane so that pathogen and plant cytoplasm are separated by the haustorial membrane, the extra-haustorial matrix and a modified plant plasma membrane called the extra-haustorial membrane (Mendgen and Hahn, 2002). Haustoria are connected to intercellular hyphae by the haustorial neck. A neckband seals the extra-haustorial matrix and membrane from the apoplast and plant plasma membrane (Chong et al., 1985). The extra-haustorial matrix represents a unique structure for exchange of nutrients as well as the transfer of effector proteins into the host, as has been shown for rust transferred protein 1 (RTP1p) from Uromyces sp. (Kemen et al., 2005) and AvrM from Melampsora lini (Rafiqi et al., 2010). It has been postulated that the extra-haustorial matrix may represent an equivalent to the parasitophorous vacuole of apicomplexans (Birch et al., 2006). For Plasmodium falciparum, parasitophorous membrane-derived structures called Maurer's clefts (Tilley et al., 2008) and a pore-like structure (de Koning-Ward et al., 2009) have been described as being involved in effector delivery. In comparison with the plant plasma membrane, the extra-haustorial membrane lacks several proteins and therefore appears smooth after freeze fracturing in the electron microscope (Knauf et al., 1989). Tubular invaginations of the extra-haustorial membrane reaching far into the plant cytoplasm have been observed in several rust–host interactions (Mims et al., 2002), but their function remains elusive. Similar tubular structures connect the parasitophorous vacuolar membrane with Maurer's clefts (Henrich et al., 2009). Little is known about the composition of the extra-haustorial matrix or the parasitophorous vacuole, as isolating these membrane structures and their contents is difficult due to limited knowledge of their composition. It has been reported that the extra-haustorial matrix consists of a thin fungal wall and a matrix that consists of plant cell wall-related carbohydrates and enzymes (Mendgen and Hahn, 2002) but excludes fungal cell-wall proteins of inter-cellular aerial hyphae and spores such as hydrophobins or repellents (Teertstra et al., 2006). Hydrophobins are typical cell-wall proteins of hyphae that are capable of forming amyloid-like filaments and higher aggregates called ‘rodlets’ (Mackay et al., 2001). It has been reported that these filament-forming surface proteins are important virulence determinants that are secreted during pathogenesis (Whiteford and Spanu, 2002; Kim et al., 2005). This property of filament-forming amyloid-like surface proteins is not only known for fungi, but has also been reported for human pathogens such as P. falciparum (MSP2) (Anders et al., 2009). Amyloid-like fibrils may be formed by numerous proteins, showing high structural heterogeneity (Udgaonkar and Kumar, 2010). The mechanism that allows certain proteins to adopt rare conformational structures that allow amyloids to be formed was described recently (Radford et al., 2011).
In this study, we show that RTP1p is a representative of a new class of structural effectors that is able to form filaments inside the extra-haustorial matrix and the host cytoplasm. We further show that RTP1p accumulates in sub-compartments of the extra-haustorial matrix and is distributed throughout the host cytoplasm in late stages of infection.
RTP1p specifically localizes to matrix protuberances at the interface between pathogen and host
Uromyces fabae and Uromyces striatus RTP1p, two haustoria-specific proteins, are localized inside the host cytoplasm (Kemen et al., 2005). It was further found that transfer into the host is dependent on the developmental stage of the haustorium. Immunoelectron microscopy revealed that RTP1p accumulates in the extra-haustorial matrix before being transferred into the host cytoplasm (Kemen et al., 2005). In this study, we performed detailed analyses to identify sites of RTP1p localization and translocation. We used a specific antibody that had been tested by Kemen et al. (2005). Looking at the distribution of RTP1p within the matrix using the ApoTome technique (Carl Zeiss, Jena, Germany), we were able to reconstruct the 3D distribution of the immuno signal. This method revealed that RTP1p predominantly accumulates in mature parts of the extra-haustorial matrix and not at the growing haustorial tip (Figure 1a). No signal was detectable beyond the neckband (Figure 1a, arrow, and Figure S1a). Immunoelectron microscopy revealed that RTP1p is predominantly localized at the most outer layer of the extra-haustorial matrix (Kemen et al., 2005). We used isolated haustoria that have been stripped of the extra-haustorial membrane (Hahn and Mendgen, 1992) to analyse the outermost layer of the extra-haustorial matrix. Using electron microscopy combined with anti-RTP1p antibodies and immunogold labelling on whole-mount samples of isolated haustoria, we detected a signal on the haustorial surface, including labelling of protuberances arising from the extra-haustorial matrix (Figure 1b).
We established a high resolution cryo-scanning electron microscopy technique following high-pressure freezing to further investigate the nature of the identified protuberances and to exclude artefacts due to chemical fixation and embedding. Applying this technique to infected plant material, we were able to visualize protuberances of the extra-haustorial matrix reaching far into the plant cytoplasm (Figure 1c). These protuberances were in close proximity to the plant endomembrane system such as the Golgi (Figure 1c, inset), as well as to the nuclear outer membrane (Figure 1c, left arrow) and resembled those described for other rust fungi and oomycetes (Harder and Chong, 1984; Mendgen et al., 1991; Mims et al., 2004). Consistent with previous findings (Mims et al., 2002), these protuberances are restricted to mature parts of the haustorium, and were not observed in young haustoria or the growing tip of older haustoria.
To link structures seen in whole-mount samples with the protuberances that RTP1p localizes to, we used high pressure-frozen, freeze-substituted samples for immunolocalization. We detected RTP1p signals within protuberances and close to the membrane of these structures (Figure 1d and Figure S1b). To address whether our observations are species-specific, we used the pathosystem U. striatus on Medicago sativa to detect RTP1p localization (UsRTP1p) and obtained comparable results (Figure S2a,b).
In summary, we conclude that protuberances of the extra-haustorial matrix are sub-compartments that reach into the host cytoplasm and contain RTP1p.
RTP1p accumulates in the entire cytoplasm of infected cells
Most interaction studies between haustoria-forming pathogens and their host are restricted to early stages of infection (Heath, 1997; Catanzariti et al., 2007), although reproduction and therefore completion of the lifecycle of obligate biotrophs relies on a much longer phase of interaction. To close this gap in knowledge, we performed live-cell imaging at various stages of haustorial differentiation (Figure 2 and Table S1). We observed the plant nucleus moving towards the haustorium upon haustorial cell penetration, but moving away after the initial penetration event. At this stage, we did not detect any disturbance in cyclosis (Figure 2, tA, and Video S1). Once the rust haustorium starts to show secondary growth, cyclosis slows down and chloroplasts start to accumulate around the haustorium (Figure 2, tB). These results are in accordance with previous observations (Heath et al., 1997; Kemen et al., 2005). To ensure RTP1p is not diffusing into the host cell due to a ruptured extra-haustorial membrane, we tested the semi-permeability of this membrane. We showed that the extra-haustorial membrane osmotically expands in the presence of 0.9% NaCl and 2% sucrose (Figure S3). In later stages of infection, when sporogenous tissue starts to develop and mature haustoria with extensive secondary growth and branching are visible within the cells, cyclosis of the nucleus and chloroplasts ceases (Figure 2, tC). We performed immunolocalization of cells at stages where cyclosis of chloroplast and nucleus ceased, and found that the whole cytoplasm shows high concentrations of RTP1p (Figure 3a–c).
Based on these findings indicating a correlation between the localization of RTP1p and cessation of cyclosis, we hypothesize that RTP1p may be involved in causing the cessation in nuclear and chloroplast movement. Our live-cell imaging from early to late stages (Videos S1–S3) reveals a strong cytoplasmic current that is still visible in the form of fast-moving microbodies despite the fact that movement of chloroplasts and nucleus ceases in late stages of infection.
Incompatible interactions, tested using the pathosystem U. striatus on M. truncatula ecotype GRC.098, revealed an even stronger secretion of RTP1p once defence reactions are induced (Figure 3d–f and Figure S4). We therefore hypothesize that RTP1p is of importance for the pathogen, especially once defence cannot be efficiently suppressed anymore.
RTP1p forms filament-like structures in the extra-haustorial matrix and within the plant cytoplasm
Using cross-linking experiments on infected plant material and detection by anti-RTP1p antibodies, we revealed that most RTP1p present in the native system exists either bound to high-molecular-weight interaction partners or in the form of multimers (Figure 4a). In addition to the high-molecular-weight band, we detected monomers (approximately 25 kDa) and dimers (approximately 50 kDa). To obtain more information about localization and association as higher-order structures of RTP1p within the extra-haustorial matrix and within the cytoplasm, we used two methods: (i) a negative-stain electron microscopy whole-mount technique on isolated haustoria to investigate RTP1p within the extra-haustorial matrix, and (ii) a deep-etch immunolocalization method to obtain insights from the host cytoplasm. As RTP1p first accumulates within the extra-haustorial matrix, we used ConA affinity purification to isolate haustoria that have been stripped of their extra-haustorial membrane (Hahn and Mendgen, 1992). Using this approach, the extra-haustorial matrix became accessible to immunocytochemistry (Figure 4b,c). Immunogold signal was detected across the matrix (Figure 4b) and associated with protuberances extending from isolated haustoria (Figure 1b). Close-ups of the matrix revealed signals associated with filamentous structures (Figure 4c).
Despite our deep-etch method being destructive to membranes, it enhances antigenicity of proteins and therefore enables observation of proteins and protein aggregates within the cytoplasm, particularly if structures extend into three dimensions (Figure 4d). In the cytoplasm surrounding the haustorium, we identified microfilaments in close proximity to the extra-haustorial matrix as previously described (Heath and Skalamera, 1997; Takemoto et al., 2003). Immunogold grains labelling RTP1p showed a pearl necklace-like localization within the host cytoplasm, comparable to what was identified within the extra-haustorial matrix (Figure 4d,e). Stronger contrasted areas revealed that the immuno signal followed filament-like structures of approximately 2 nm diameter (Figure 4f). These results indicate that either RTP1p is attached to filamentous structures within the extra-haustorial matrix and the host cytoplasm, or is able to form filamentous structures itself. As the signal in the cytoplasm is lower than expected from immuno light microscopy, we compared conventionally fixed samples used for immuno light microscopy with high pressure-frozen samples used for our deep-etch method by probing with anti-RTP1p and anti-phosphoenolpyruvate carboxylase (PEPC), an antibody recognizing the cytoplasmic enzyme phosphoenol pyruvate carboxylase (Figure S5). There was no difference in recognition of phosphoenol pyruvate carboxylase between samples but a significant difference in recognition of RTP1p, indicating that cytoplasmic RTP1p is present in a folded or multimeric configuration, and is only weakly recognized by our antibody under such conditions.
Heterologously expressed RTP1p forms aggregates and filaments
As RTP1p has been shown to be glycosylated (Kemen et al., 2005) and over-expression in Escherichia coli results in formation of inclusion bodies, we used the expression system Pichia pastoris for structural studies. Performing cross-linking experiments as described for the native protein but using purified heterologously expressed RTP1p, we obtained similar results: RTP1p was detected as a monomer and as multimers (Figure 5a). Even if no cross-linker was added, we observed RTP1p precipitation after several hours. These findings are independent of glycosylation, as de-glycosylated protein still showed aggregates in Western blotting (Figure S6), suggesting that de-glycosylated homomers are sufficient for the formation of high-molecular-weight multimers. Expression of full-length UfRTP1p fused to GFP in the cytoplasm of P. pastoris revealed that, in this heterologous system, RTP1p shows punctate localization and is not distributed over the cytoplasm (Figure S7a). The cytoplasmic spots are not associated with degradation (Figure S7b), and probably correspond to the cytoplasmic spots that have been implicated in amyloid formation in Saccharomyces cerevisiae (Alberti et al., 2009). Applying UfRTP1p to Vicia faba protoplasts revealed a comparable localization, with RTP1p being localized in speckles rather than freely diffused within the cytoplasm (Figure S7c).
We identified amorphous aggregates by analysing the precipitate of purified heterologously expressed RTP1p by electron microscopy after negative staining (Figure 5b). We then used the method described by Lee and Eisenberg (2003) that allows conversion of amorphous aggregates of prion proteins into filaments. This method (known as ‘seeded conversion’) resulted in RTP1p forming filament-like structures (Figure 5c). The difference compared with the method described by Lee and Eisenberg (2003) was that RTP1p was able to form filaments de novo without adding filaments as a starter. Using electron microscopy, we found that the smaller twisted filaments formed larger filamentous structures (Figure 5d). To confirm that our antibodies were capable of detecting both forms, amorphous aggregates and filaments were mixed prior to immunodetection (Figure 5e). High antigenicity was observed for amorphous aggregates, while filamentous structures showed varying antigenicity based on their structural integrity. Folded filaments were barely detected, but some signal was observed for partially unfolded filaments. Completely unwound filamentous structures showed the best antigenicity (Figure 5e,f).
We conclude that RTP1p is able to form filamentous structures without seeding or the help of other proteins.
A β–aggregation domain computationally identified in RTP1p is able to form filaments
To unravel the filament-forming mechanism in RTP1p, we performed computational analyses using Tango and Waltz algorithms (Fernandez-Escamilla et al., 2004; Maurer-Stroh et al., 2010). We identified two potential aggregation domains in positions 139–151 (domain I) and 204–209 (domain II) (Figure S8). Based on Tango analysis and further secondary structure predictions, domain I resembled a β–aggregation domain consisting of two antiparallel β–strands and one loop. To analyse properties of this domain, we used a synthetic peptide covering positions 135–155 (RTP1p135-155). We added four amino acids to both sides of the predicted aggregation domain to increase solubility and stabilize the predicted structure. Using CD spectroscopy, we confirmed our secondary structure predictions and revealed that the peptide has two antiparallel β–strands with 40–50% of the amino acids contributing to this structure, 15–25% being involved in β–turns, and 20–30% being involved in random coils (Figure 6a,b). The numbers showed variability depending on the solvent.
We observed macroscopic precipitates of RTP1p135-155 after 24 h. Using negative staining for electron microscopy, we observed a precipitated peptide showing long filamentous structures consistent with the predicted β–aggregation properties (Figure 6c,d). As a control to determine whether our method favours filament formation of peptides non-specifically, we tested several RTP1p135-155 peptides in which amino acids have been exchanged (Figure S9). Replacing the phenylalanine in position 17 of the peptide by alanine (RTP1p135-155 F151A) did not block filament formation, but replacing the phenylalanine at position 7 by alanine (RTP1p135-155 F141A) or replacing both phenylalanines (RTP1p135-155 F141A F151A) lead to a complete block of filament formation. These results reveal sequence-dependent aggregation and highlight a possible role for phenylalanine at position 7 of RTP1p135-155 in filament formation. We also used thioflavin T to obtain biochemical evidence for β–aggregation of RTP1p135-155 and mutated peptides. Thioflavin T shows fluorescence at 510 nm when intercalated into filaments formed by β–aggregation (Voropai et al., 2003). We observed a significant increase in fluorescence using RTP1p135-155, thus validating our hypothesis (Figure S10). While the mutated form RTP1p135-155 F151A showed only a minor reduction in endpoint fluorescence (Figure S11a) and no significant difference in aggregation rates (Figure S11b), fluorescence levels and aggregation rates were low for RTP1p135-155 F141A, complementing our electron microscopy studies that revealed filament formation for RTP1p135-155 and RTP1p135-155 F151A but not RTP1p135-155 F141A.
To link these findings to in vivo aggregation, we used a non-denaturating tissue printing method (modified from Smallwood et al., 1994) on microscope slides, in combination with a thioflavin T-based fluorescent detection method for amyloid-like proteins (Westermark et al., 1999). The RTP1p immunofluorescence signal and the thioflavin T signal co-localized in the matrix surrounding the haustorium (Figure S12). Parts of the haustorium that did not show RTP1p signal also did not show thioflavin T staining. This finding suggests that, in addition to high concentrations of monomeric RTP1p, amyloid-like RTP1p structures exist within the extra-haustorial matrix.
We conclude that RTP1p forms aggregates and filamentous structures within the extra-haustorial matrix and the host cytoplasm based on β–aggregation.
RTP1p as a tool to study effector transfer into the host
Unlike bacterial effectors that are directly secreted into the host cytoplasm via type III secretion systems (Galan and Collmer, 1999), eukaryotic effectors have to cross at least one membrane to enter the host cell after being secreted by the pathogen (Catanzariti et al., 2007; Leborgne-Castel et al., 2010). In obligate biotrophic interactions, the most likely place of transfer is the extra-haustorial membrane that separates the haustorium from the host cytoplasm (Leborgne-Castel et al., 2010). We showed that, comparable to effector localization within the biotrophic interfacial complex of Magnaporthe grisea (Khang et al., 2010), RTP1p accumulates in distinct parts of the extra-haustorial matrix, particularly the older parts towards the neck of the haustorium.
Some biochemical evidence of how eukaryotic effectors may be internalized into host cells has been published (Kale et al., 2010; Yaeno et al., 2011; Wawra et al., 2012), but the mechanisms remain controversial (Ellis and Dodds, 2011; Yaeno and Shirasu, 2013). It is therefore crucial to use markers to understand where the transfer occurs and hence obtain an indication of the conditions under which effector proteins are transferred into the host cell. Studies using electron microscopy have been performed that showed potential effector proteins accumulating within the extra-haustorial matrix before being transferred into the host cytoplasm (Kemen et al., 2005; Rafiqi et al., 2010). Immunocytochemistry is one of the major tools to study obligate biotroph plant pathogens as functional tests are limited due to the lack of stable transformation systems. Unlike previous studies, we improved the resolution of membrane structures within the host cytoplasm significantly, and used RTP1p as a marker for pathogen-to-host protein transfer. We observed RTP1p in protuberances of the extra-haustorial matrix. Protuberances of the extra-haustorial matrix are a common feature of haustoria. These structures have been observed in rust fungi (Mendgen et al., 1991; Mims et al., 2002) and oomycetes (Mims et al., 2004; Baka, 2008). They morphologically resemble Maurer's clefts, protein trafficking compartments that are induced by P. falciparum inside red blood cells (Tilley et al., 2008). In our study, we showed that occurrence of protuberances correlates with sites of RTP1p immuno signals in the haustorium. In order to cross the membrane, it is likely that proteins need to bind to lipids or receptors within the extra-haustorial membrane in order to be internalized (Grouffaud et al., 2010; Kale et al., 2010). This is consistent with our findings showing that RTP1p localizes to the membrane within protuberances. We hypothesize that sub-compartmentalization within the extra-haustorial matrix is relevant for transfer of RTP1p into the host. It is therefore crucial to understand how theses protuberances emerge inside infected cells and how they are protected from fusion with the host endomembrane system or lytic compartments.
Rust infection has an inhibitory effect on plant chloroplast cyclosis
Using live imaging of infected cells during various stages of haustorial development, we focused on chloroplast and nuclear movement. For necrotrophic interactions, movement of the nucleus and chloroplasts to the site of infection has been reported (Oliver et al., 2009). In biotrophic interactions, chloroplasts and the nucleus do not stay at the side of pathogen penetration, but the nucleus moves towards the haustorium once haustoria show secondary growth (Heath et al., 1997). This is consistent with our live-cell imaging results. Further, we observed accumulation of chloroplasts surrounding the haustorial complex in late stages of infection while the haustorium is still growing inside the host cell. The mechano-triggered accumulation caused during penetration shows directed movement of chloroplasts towards the signal source (Sato et al., 2003; Wada et al., 2003). Unlike mechano-triggered accumulation, we observed normal cyclosis of chloroplasts in U. fabae-infected cells. Only during haustorium maturation did we observe a slow down of chloroplast cyclosis in close proximity to the haustorial complex, resulting in complete cessation in late stages of infection. Our immunolocalization results revealed that, during this stage, RTP1p becomes distributed over the complete cytoplasm and nucleoplasm of the host cell. Using live-cell imaging of young stages during which RTP1p has been observed to accumulate to restricted cytoplasmic zones around the haustorium (Kemen et al., 2005) revealed that cytoplasmic streaming is strong and therefore accumulation may only be explained by RTP1p being attached to structures that are not undergoing cyclosis or by RTP1p forming higher aggregates that are not affected by cytoplasmic streaming.
Analysing resistant plants that show cell death prior to secondary growth of haustoria revealed a strong accumulation of RTP1p inside the cytoplasm. From these findings, we conclude that RTP1p secretion is not triggered by the haustorial developmental status but by the status of its host cell. It has been hypothesized that biotrophic pathogens can keep their host cell alive as long as they are able to suppress host-induced cell death (Heath and Skalamera, 1997). As cyclosis and the resulting cytoplasmic streaming enables distribution of molecules and vesicles (Verchot-Lubicz and Goldstein, 2010), blocking chloroplast and nuclear cyclosis probably reduces the signal exchange between organelles, autophagy and the collapse of lytic vacuoles, and may therefore be a mechanism to slow down host cell death.
RTP1p forms amyloid-like fibrillar structures
Our observation that native and heterologously expressed RTP1p form filamentous polymers, alongside identification of a β–aggregation domain, are crucial in understanding RTP1p function. Cross-β–aggregation as predicted by the TANGO algorithm may either lead to amorphous aggregation or amyloid-like filamentous structures (Rousseau et al., 2006). In the native system, we observed filament-like structures only, but protein heterologously expressed in P. pastoris forms amorphous aggregates when purified under non-denaturating conditions. After denaturation and controlled renaturation, amyloid-like filaments were formed. For amyloid-like proteins, it has been shown that seeding with infectious multimers initiates filament growth in vitro (Taylor et al., 1999; Lee and Eisenberg, 2003; Nonaka et al., 2010). In cases where the monomer adopts an amyloid-like conformation, filaments may grow without seeding, after a lag phase that is required to rearrange the nucleation complex (Wu and Shea, 2011). Most fungal amyloid-like proteins are cytoplasmic (Wickner et al., 2007), and seeding is therefore possible by existing filaments or chaperone assistance (Kryndushkin et al., 2011). However, RTP1p is targeted to the extracellular space, where self-assembly and therefore self-seeding are crucial to induce filament formation. Comparable to RTP1p, yeast cell adhesion molecules are secreted into the extracellular space and show β–aggregation in combination with amyloid-like filament formation (Ramsook et al., 2010). These proteins mediate attachment, colony and biofilm formation, and are therefore important virulence factors determining host specificity (Nobbs et al., 2010; Martin et al., 2011). RTP1p may have evolved from an adhesin-like haustorial cell-wall protein and gained further functions by being delivered into the host cell. Heat shock proteins (HSPs) have been shown to regulate the status of cytoplasmic amyloid-like proteins between monomeric and filamentous (Chernoff, 2007). Recent findings revealed HSPs not only within the haustorial cytoplasm but also within the extra-haustorial matrix of oomycete pathogens [E. Kemen, A. Kemen, M. E. Jørgensen, J. D. G. Jones (The Sainsbury Laboratory, Norwich, UK)]. In yeasts, the status of cytoplasmic amyloid-like proteins depends on the concentration of HSPs: low and normal concentrations favour filaments, whereas high concentrations favour monomers (Chernoff, 2007). Under stress conditions, human cells release HSPs into the extracellular space (Lancaster and Febbraio, 2005). Based on our results and previous findings, we hypothesize RTP1p is secreted conventionally into the extra-haustorial matrix in an amyloid-like stage, which leads to aggregation and accumulation within the extra-haustorial matrix. Secondary haustorial growth triggers defence reactions, as seen by the movement of the nucleus towards the haustorium (Heath et al., 1997). These defence reactions may cause stress to the haustorium and trigger the release of HSPs. As for cytoplasmic prion proteins (Chernoff, 2007), the increase in HSPs may release monomers that are transferred into the host cytoplasm, where, in the absence of fungal HSPs, formation of higher aggregates occurs upon reaching a critical concentration. This dualism of RTP1p as monomer or multimer without being degraded by plant proteases in the apoplast even during strong plant defence reactions is possible, as the monomer, and likely to an even greater extent the aggregated form of RTP1p, show protease resistance or even protease inhibitor function (Pretsch et al., 2013). The amorphous aggregates that are formed by expressing RTP1p in P. pastoris support the existence of a postulated rust-specific HSP protein that may be substituted by a controlled conversion process.
A new class of structural effectors
In this study, we have identified RTP1p as a filament-forming protein that accumulates in the host cell in late stages during biotrophic interaction. RTP1p may therefore be involved in the inhibition of cyclosis observed during late stages of infection (Figure S13). As chloroplasts are important in plant defence (Padmanabhan and Dinesh-Kumar, 2010), accumulating chloroplasts and therefore simulating high-light conditions may lower photosynthetic activity by shading. This in turn may be beneficial in reducing production of reactive oxygen species. Once cell death suppression is not effective anymore and hydrolases are released from collapsed lytic vacuoles (van Doorn et al., 2011), protease-resistant RTP1p filaments may be relevant to protect the haustorial complex from degradation.
We hypothesize that RTP1p is an effector protein that is relevant for extending the biotrophic phase and protecting the haustorium from defence by delivering host-stabilizing multimers during the process of infection. RTP1p is therefore a representative of a new class of structural effectors whose function is to stabilize the zone of interaction between pathogen and host. Our findings suggest a new class of targets for plant protection that have not been exploited although extensive resources for amyloid-like proteins are available.
Cultivation of plants and micro-organisms
Cultivation of Vicia faba cv. con Amore and inoculation with U. fabae uredospores was performed as described previously (Deising et al., 1991; Hahn and Mendgen, 1992). Cultivation of Medicago sativa L. ‘Europe’, M. truncatula Gaertn. ‘Jemalong’ A17, and M. truncatula Gaertn. GRC.098 and U. striatus uredospore infection were performed as described by Kemen et al. (2005).
Isolation of haustoria
Haustoria were isolated using ConA affinity purification as described by Hahn and Mendgen (1992). Pichia pastoris strain KM71 was used for heterologous protein over-expression according to the manufacturer's protocol for P. pastoris overexpression systems (Life Technologies GmbH, Darmstadt, Germany).
Plasmid construction and heterologous RTP1p expression
For expression of recombinant RTP1p in P. pastoris, UfRTP1p was amplified using the primers 5′-CGTAGAATTCCATTATGTCAAACCTTCGCTTAC-3′ and 5′-GCCGCCCTAGGTCAGTGGTGGTGGTGGTGG-3′, introducing unique EcoRI and AvrII sites (underlined) as well as a C–terminal His tag. After digestion with the respective enzymes, RTP1 was introduced into EcoRI/AvrII-digested pPIC3.5 (Invitrogen). Constructs expressing GFP or N–terminal GFP-tagged RTP1p (-signal peptide) were introduced into pPIC3.5 using BamHI/NotI digests from GFP–RTP1 fusion constructs used for transient expression in tobacco protoplasts (Kemen et al., 2005). Pichia pastoris strain KM71 was used for heterologous protein over-expression according to the manufacturer's instructions (Life Technologies GmbH). For deglycosylation studies, 30 μl protein samples were treated with 3000 units endoglycosidase Hf (New England Biolabs, Frankfurt, Germany, in 5 mm sodium citrate, pH 5.5, at 37°C for 3 h.
Incubation of V. faba protoplasts with heterologous UfRTP1p
Vicia faba protoplasts were prepared as described by Okuno and Furusawa (1977), with modifications as described by Obi et al. (1989). His-tag-purified protein was re-dialysed into protoplast buffer (500 mm d–sorbitol, 1 mm CaCl2, 5 mm MES, pH 5.5), and diluted 1:16 v/v. A protoplast suspension (200 μl) was obtained by centrifugation at 80 g in a swing-out rotor, and re-suspended in 500 μl RTP1p-containing protoplast buffer. Protoplasts were incubated for 2 h at room temperature after two washing steps, prior to fixation using 3.7% v/v formaldehyde in protoplast buffer. Protoplasts were permeabilized for immunolocalization using TSW (10 mm Tris/HCl pH 7.5, 154 mm NaCl, 0.25% w/v gelatine, 0.02% w/v SDS, 0.1% w/v Triton X-100) buffer as described by Frigerio et al. (2000). Protein was detected using S844p as the primary antiserum for immunolabelling.
His-tagged RTP1p was purified from the filtered supernatant of P. pastoris cultures by immobilized metal ion affinity chromatography using a two-step gradient (113 and 181 mm imidazole). For cross-linking of heterologous RTP1p, a glutaraldehyde concentration of 0.1% v/v or 1.5 mm Ethylene glycol bis (sulfosuccinimidylsuccinate) were used. Native protein was extracted from U. fabae-infected leaves at 8 days post-infection. Leaves were finely cut in buffer (2.5 mm Tris/HCl pH 7.2, 0.1% v/v Triton X–100 and 200 μm phenylmethanesulfonyl fluoride) using a razor blade. The supernatant was cross-linked using a final glutaraldehyde concentration of 0.05% v/v. Samples were incubated for 1–60 min. Reactions were quenched using 10 mm glycine for 5 min.
Proteins were separated by 12% SDS–PAGE (Laemmli, 1970), and RTP1p was detected by immunoblotting using purified serum S746 (Kemen et al., 2005) at a 1:10 000 dilution. Anti-GFP antibody was kindly provided by R. Kissmehl (Department of Biology, University of Konstanz, Germany). Visualization was performed using peroxidise-conjugated anti-guinea pig IgG (Sigma-Aldrich, Hamburg, Germany) as the secondary antibody and ECL Western blot detection reagent (GE Healthcare Europe, Freiburg, Germany).
For immuno light and electron microscopy, plants were fixed and embedded as described previously (Kemen et al., 2005, 2011b).
In brief, for light microscopy, samples were fixed using acetic acid/ethanol (1:3) and embedded in acrylic resin. Prior to immunostaining, samples were deep etched using acetone, and developed using purified primary antibody S844p (UfRTP1p) or S849p (UsRTP1p) or anti-PEPC antibody (Rockland Immunochemicals, Gilbertsville, PA, USA) as a control. All primary antibodies were detected using Cy3-labelled goat anti-rabbit secondary antibody. For electron microscopy, samples were high pressure-frozen and freeze-substituted prior to embedding in epoxy resin. Samples were sectioned and developed using purified S844 as primary antibody, and 10 nm gold-labelled goat anti-rabbit secondary antibody.
For deep etching in electron microscopy, samples were high pressure-frozen and freeze-substituted followed by acrylic resin embedding. Sections were mounted on carbon-coated mesh grids. Prior to immunostaining, resin was removed using acetone, and samples were treated as described for conventional immunostaining.
For immuno whole-mount samples, leaves were cleared with acetic acid/ethanol (1:5) for 1 1/2 h, washed in ethanol and re-hydrated (80, 60, 40, 20 and 10% v/v ethanol/water) and transferred into 2.5 mm Tris/HCl pH 7.2. Samples were developed as described for light microscopy except that the primary antibody incubation step was extended to overnight at 4°C and secondary antibody incubation was extended to to 3 h. For optical slices, an Axioplan2 imaging system equipped with the ApoTome technique was used (Carl Zeiss).
For live-cell imaging, infected Vicia faba leaves were syringe-infiltrated with BG11 (Rippka et al., 1979) supplemented with 2 mm Tris/HCl and 1% sucrose, pH 6.2. Prior to mounting, the epidermis was removed. For observation, a Zeiss Plan Apochromat 100 × 1.4 oil-objective was used, and images were taken at 30 sec intervals.
High-resolution cryo-scanning microscopy
For high-resolution cryo-scanning electron microscopy, samples were high pressure-frozen and mounted on a Gatan cryo-stage sample holder with lockable clamp (Alto 2500; Gatan, Munich, Germany) under liquid nitrogen. We covered one clamp with indium foil to avoid cracks within the sample when the clamp was locked. Samples were fractured on the cryo-stage using a cold rotary fractioning device prior to etching and platinum coating. Samples were scanned using a Hitachi S–4700 cold-field emission scanning electron microscope (Hitachi High-Technologies Europe, Krefeld, Germany) with attached cryo-stage at 1–2 kV.
For conversion of amorphic protein aggregates into filaments, the method described by Lee and Eisenberg (2003) was modified. Heterologously expressed, purified RTP1p aggregates from P. pastoris supernatant were concentrated 3.6-fold using a Vivaspin 15R ultrafiltration unit (Vivascience, Littleton, MA, USA), and dialysed into monomerization buffer (50 mm Tris/HCl pH 7.6, 2.5 m guanidine/HCl, 3 m NaCl, 1 m dithiothreitol), followed by incubation for 14 h at room temperature. After a 24 h dialysis step into oxidation buffer (50 mm sodium acetate, 1 m guanidine/HCl, pH 3.8), dialysis into aggregation buffer (50 mm sodium acetate, pH 3.8) was performed. The filaments were incubated at 4°C for 24 h, and transferred onto Pioloform (Agar Scientific, Stansted, UK)-coated grids (200 mesh) for transmission electron microscopy. After 10 min sedimentation, the supernatant was replaced by H2Odistilled, nd 0.05% uranyl acetate was used for negative staining.
Negative staining of peptide aggregates
For detection of filaments in electron microscopy, the following peptides were dissolved in 50 mm potassium phosphate buffer, pH 6.0, and incubated for 4 h at room temperature: RTP1p135-155 (NH2-SPGDYVFVSYGTCATVFQNPQ-OH), RTP1p135-155 F151A (NH2-SPGDYVFVSYGTCATVAQNPQ-OH), RTP1p135-155 F141A (NH2-SPGDYVAVSYGTCATVFQNPQ-OH) and RTP1p135-155 F141A F151A (NH2-SPGDYVAVSYGTCATVAQNPQ-OH) (GenScript, Piscataway, NJ, USA), all with a purity of >95%. Peptide solutions (50 μl) were transferred to Pioloform-coated grits (200 mesh). After 10 min sedimentation, the supernatant was replaced by H2Odest, and 0.05% uranyl acetate was used for negative staining.
For CD and fluorescence spectroscopy, synthesized RTP1p135-155 peptide (NH2-SPGDYVFVSYGTCATVFQNPQ-OH) (Bio-Synthesis Inc., Lewisville, TX, USA) with a purity of >85% was used. The peptide was dissolved in four different media (see Figure 6b), and measured at 20°C using a JASCO-600 CD spectrophotometer (JASCO, Easton, MD, USA). Each measurement was performed using a 1 mm cuvette and a 0.1 mm cuvette. Each measurement was repeated three times. Secondary structures were calculated from peptides in solution using a range for measurement between 260 and 185 nm, and CDNN 2.1 analysis software (Böhm et al., 1992).
For fluorescence spectroscopy, the RTP1p135-155, TP1p135-155 F151A and RTP1p135-155 F141A peptides were dissolved in 50 mm potassium phosphate buffer, pH 6.0, and mixed 1:1 v/v with a thioflavin T stock solution (0.05 μm thioflavin, 50 mm potassium phosphate buffer, pH 6.0) (see above). Fluorescence was measured after excitation at 450 nm, or at 400 nm as a control.
Tissue printing and thioflavin T detection
The tissue printing method was modified from that described by Smallwood et al. (1994). In brief, infected V. faba leaves were cut on slides coated with Biobond (BBI International, Cardiff, UK), and immediately pressed at a 90° angle against the slide. Slides were air-dried for 1 min and immunostained as previously described (Kemen et al., 2005, 2011b). A 0.05 μm thioflavin T solution in Tris-buffered saline was used, and samples were incubated for 20 min prior to two 10 min washing steps in Tris-buffered saline. To separate fluorescence signals, we used Cy3-labelled secondary antibodies in combination with a Cy3 filter set (F41–007, AHF Analysentechnik, Tübingen, Germany), and a fluorescein isothiocyanate filter set (F41–012, AHF Analysentechnik) for detection of thioflavin T. Chlorophyll and background fluorescence were detected using filter set 05 (488005-0000, Carl Zeiss). Bisbenzimide was detected using a 4,6–diamidino-2–phenylindole filter set (F31–000, AHF Analysentechnik). Pictures were merged and analysed using ImageJ (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, http://imagej.nih.gov/ij/, 1997–2012).
We would like to thank Marie-Cécile Caillaud (The Sainsbury Laboratory, Norwich, UK) and Sebastian Schornack (The Sainsbury Laboratory, University of Cambridge, Cambridge, UK) for critically reading the manuscript. We thank Ewald Daltrozzo (Department of Chemistry, University of Konstanz, Konstanz, Germany) for his guidance in fluorescence spectroscopy, Ulla Neumann (Central Micrsocopy, Max Planck Institute for Plant Breeding Research, Cologne, Germany) for her technical support, and Rudolf Heitefuss (Plant Pathology, University of Göttingen, Germany) for his impact in rust research.