Extracellular transport and integration of plant secretory proteins into pathogeninduced cell wall compartments Dorit Meyer, Simone

Pajonk, Cristina Micali, Richard OConnell, Paul Schulze-Lefert; 9 DEC 2008 [http://onlinelibrary.wiley.com/doi/10.1111/j.1365-313X.2008.03743.x/full] Keywords: secretion; exosomes; haustoria; plant defense; SNARE proteins; plant cell wall Summary Many fungal parasites enter plant cells by penetrating the host cell wall and, thereafter, differentiate specialized intracellular feeding structures, called haustoria, by invagination of the plants plasma membrane. Arabidopsis PEN gene products are known to act at the cell periphery and function in the execution of apoplastic immune responses to limit fungal entry. This response underneath fungal contact sites is tightly linked with the deposition of plant cell wall polymers, including PMR4/GSL5-dependent callose, in the paramural space, thereby producing localized wall thickenings called papillae. We show that powdery mildew fungi specifically induce the extracellular transport and entrapment of the fusion protein GFPPEN1 syntaxin and its interacting partner monomeric yellow fluorescent protein (mYFP)SNAP33 within the papillary matrix. Remarkably, PMR4/GSL5 callose, GFPPEN1, mYFPSNAP33, and the ABC transporter GFPPEN3 are selectively incorporated into extracellular encasements surrounding haustoria of the powdery mildew Golovinomyces orontii, suggesting that the same secretory defense responses become activated during the formation of papillae and haustorial encasements. This is consistent with a time-course analysis of the encasement process, indicating that these extracellular structures are generated through the extension of papillae. We show that PMR4/GSL5 callose accumulation in papillae and haustorial encasements occurs independently of PEN1 syntaxin. We propose a model in which exosome biogenesis/release serves as a common transport mechanism by which the proteins PEN1 and PEN3, otherwise resident in the plasma membrane, together with membrane lipids, become stably incorporated into both pathogeninduced cell wall compartments.

Introduction Secretion via vesicle-mediated exocytosis plays an important role in diverse biological processes in plants including development (e.g. creation of a cell plate during cell division, primary cell wall deposition, pollen and root hair tip growth) and immune responses (Kwon et al., 2008a; Lipka et al., 2007; Samaj et al., 2006; Somerville et al., 2004; Van Damme and Geelen, 2008). In Arabidopsis, SNARE (soluble N-ethylmalemide-sensitive factor attachment protein receptor) domain-containing PEN1 syntaxin, SNAP33 and VAMP721/722 assemble into a ternary SNARE complex which is thought to tether secretory vesicles containing unknown cargo to the plant plasma membrane (Kwon et al., 2008b). This mobilization of the secretory pathway was shown to be precisely localized at sites of attempted penetration by non-adapted (Blumeria graminis f. sp. hordei) as well as adapted (Golovinomyces orontii and Golovinomyces cichoracearum) powdery mildew pathogens (Assaad et al., 2004; Bhat et al., 2005; Kwon et al., 2008b). Loss of PEN1 specifically compromises penetration resistance against the non-adapted powdery mildew fungi B. g. hordei and Erysiphe pisi (Collins et al., 2003; Lipka et al., 2005). Furthermore, depletion of VAMP721/722 enhances susceptibility to the adapted fungus G. orontii, the oomycete

Hyaloperonospora parasitica, as well as to non-adapted E. pisi (Kwon et al., 2008b), implicating the involvement of additional syntaxins in vesicle-mediated defense. A parallel antimicrobial delivery system relies on the activity of the plasma membrane ABC transporter PEN3. PEN3, like the PEN1/SNAP33/VAMP721/722 complex, is concentrated at the site of pathogen attack (Stein et al., 2006) and is thought to deliver PEN2-generated toxic compounds into the apoplastic space (Kwon et al., 2008a; Stein et al., 2006). PEN2 is an unconventional myrosinase, associated with the surface of peroxisomes, producing presumably toxic indole glucosinolate hydrolysis products (Bednarek et al., in press). Loss of PEN3 is associated with enhanced susceptibility to the non-adapted powdery mildew fungi B. g. hordei and E. pisi as well as the necrotrophic fungus Plectosphaerella cucumerina (Stein et al., 2006). PEN1 and VAMP721/722 are required for the timely formation of localized thickenings on the interior face of the plant cell wall at sites of attempted fungal entry (also called papillae) (Assaad et al., 2004; Kwon et al., 2008b), suggesting that at least part of the vesicle cargo may be cell wall material or other components found in papillae. Papilla formation is an early defense reaction in response to penetrating pathogens, but is also seen adjacent to bacterial colonies in the intercellular space (Bestwick et al., 1995). Papillae represent agglomerations of a variety of plant cell-derived components including cellulose, -1,3-glucan (callose), pectins, phenolics such as lignin, hydroxyproline-rich glycoproteins, and antimicrobial polypeptides such as thionins (Aist, 1976; Bolwell et al., 2002; Snyder and Nicholson, 1990; Thordal-Christensen et al., 1997). Collectively, these are thought to provide localized structural reinforcement and targeted chemical defense (von Rpenack et al., 1998). Papillary callose, synthesized by the PMR4/GSL5 glucan synthase-like complex, was long believed to serve as a structural barrier, and was also recently shown to negatively regulate salicylic acid-mediated defenses (Jacobs et al., 2003; Nishimura et al., 2003). Although papilla formation is typically seen at bacterial and fungal/oomycete contact sites (Kwon et al., 2008a), it could only present a structural barrier to invasion by pathogens that penetrate into the host cell lumen and form haustoria or intracellular hyphae e.g. powdery mildew and rust fungi (OConnell and Panstruga, 2006). Differentiation of these specialized feeding structures causes expansion and modification of the plants plasma membrane and this is thought to facilitate the exchange of nutrients (Hahn and Mendgen, 2001) and virulence factors between host and pathogen. During the formation of haustoria, additional callose and other plant cell wall material can be deposited to form a collar around the haustorial neck, which can subsequently expand to encase the entire haustorial body (Bushnell, 1972). Electron microscopic studies in an incompatible interaction between cowpea (Vigna unguiculata) and the rust fungus Uromyces phaseoli var. vignae revealed that the encasement is formed after full maturation of the haustorium and results in the termination of fungal growth (Heath and Heath, 1971), suggesting a role for the encasement in plant defense. As with papillae, encasements develop in both compatible and incompatible plantpathogen interactions (Chou, 1970; Donofrio and Delaney, 2001; Mellersh and Heath, 2003; Skalamera et al., 1997). In Arabidopsis, callose encasements have previously been reported in the compatible interaction with the powdery mildew fungus G. orontii and the oomycete H. parasitica (Donofrio and Delaney, 2001; Jacobs et al., 2003). However, the molecular basis of encasement and the mechanisms by which pathogen growth is restricted were never addressed. In particular, the deployment of specific defense pathways during encasement formation is unclear. Herein, we report a cell biological study of papillae and encasements produced by Arabidopsis in response to cell wall-penetrating pathogens. We demonstrate the specific integration of extracellular

defense components, PEN1, SNAP33, VAMP722, and PEN3 in both papillae and haustorial encasements. We present a cellular mechanism accounting for this unusual extracellular transport that is based on exosome biogenesis/release. In animals, this non-canonical secretory pathway involves the formation of multivesicular bodies (MVBs) and the release of membranebound vesicles (exosomes) into the extracellular space (Johnstone, 2006; Schorey and Bhatnagar, 2008). The phenomenon described in this study could serve as suitable model for elucidating non-canonical secretion and its likely contribution to other plant biological processes.

Results PEN1 specifically accumulates at powdery mildew contact sites The PEN1 syntaxin has been shown to accumulate at sites of attempted fungal entry in interactions between Arabidopsis and the non-adapted barley (Hordeum vulgare) powdery mildew B. g. hordei, as well as the adapted powdery mildew G. cichoracearum (Assaad et al., 2004; Bhat et al., 2005). To investigate whether this is a specific response to powdery mildews, Arabidopsis plants expressing a functional GFPPEN1 fusion protein were challenged with a range of different ascomycete pathogens that penetrate epidermal cells through the cuticle and cell wall. Cells underlying fungal appressoria were analyzed by confocal microscopy at 24 h postinoculation (hpi) of seedling cotyledons. Accumulation of GFPPEN1 was observed beneath the appressoria of two non-adapted powdery mildews, B. g. hordei (Figure 1a,b) and the pea powdery mildew E. pisi (Figure 1c,d), which typically fail to enter Arabidopsis epidermal cells, as indicated by the absence of haustoria. GFPPEN1 also accumulated beneath appressoria of the adapted powdery mildew G. orontii (Figure 1e,f), both at successful penetration sites as well as failed entry sites. Quantitative analysis of interaction sites revealed that the adapted and nonadapted powdery mildews induced accumulation of GFPPEN1 beneath differentiated appressoria with comparable frequency (Table S1). Thus, accumulation of PEN1 was not associated uniquely with failed mildew entry attempts and was triggered by both adapted and non-adapted mildew species. Appressoria of the non-adapted rice blast fungus, Magnaporthe grisea, were never observed to penetrate Arabidopsis epidermal cells, as shown by the absence of intracellular hyphae (Figure 1h). Callosic papillae were deposited by host cells beneath appressoria (Figure S1ad), but accumulation of GFPPEN1 was never seen at these sites (Figures 1g and S1ad). Similarly, the non-adapted Medicago anthracnose fungus, Colletotrichum destructivum, typically failed to enter Arabidopsis epidermal cells (Figure 1i,j), although 25% of sporelings entered epidermal cells, as indicated by the formation of intracellular primary hyphae beneath appressoria (Figure S1eh). Callosic papillae were present at both successful (Figure S1eh) and unsuccessful entry sites of C. destructivum (Figure 1i,j), but the accumulation of GFPPEN1 was never observed at these sites. Entry rates of the adapted crucifer anthracnose fungus, Colletotrichum higginsianum, were higher than C. destructivum: typically 6080% of appressoria formed primary hyphae. Following penetration by both Colletotrichum species, the host plasma membrane invaginated around the primary hyphae was labeled by GFPPEN1, but the fluorescence intensity was similar to that of the normal plasma membrane (Figures 1k,l and S1eh). Although the darkly melanized appressorial cell walls of M. grisea and Colletotrichum species may attenuate the apparent fluorescence intensity of GFP beneath appressoria, we obtained no evidence that any of these ascomycete fungi induce accumulation of GFPPEN1 at attempted penetration sites, even though they penetrate epidermal cells in a similar manner to powdery mildew fungi (Bourett and

Howard, 1990; OConnell et al., 2004; Perfect et al., 1999). Furthermore, GFPPEN1 did not accumulate at successful entry sites of the adapted oomycete pathogen H. parasitica (data not shown), which enters Arabidopsis epidermal cells from intercellular hyphae (Koch and Slusarenko, 1990). Taken together, these results suggest that accumulation of PEN1 may be a unique response of plant cells to attempted entry by powdery mildews. GFPPEN1 accumulates in papillae In previous studies on the adapted powdery mildew, G. cichoracearum, optical sectioning by confocal microscopy detected the presence of GFPPEN1 within the interior of host papillae (Assaad et al., 2004). However, in the ArabidopsisB. g. hordei interaction, plasmolysis experiments indicated that focal accumulation of GFPPEN1 was associated with a plasma membrane microdomain rather than the papilla itself (Bhat et al., 2005). To clarify the location of PEN1 accumulation at attempted entry sites, we used both optical cross-sectioning and plasmolysis experiments. In the ArabidopsisB. g. hordei interaction, optical sectioning at 1216 hpi demonstrated that GFPPEN1 fluorescence was uniformly distributed through the interior of papillae, but was excluded from fungal penetration pegs encased within papillae (Figure 2a,b). Next, we performed plasmolysis experiments on transgenic lines expressing either GFPPEN1 (Figure 2c) or monomeric yellow fluorescent protein (mYFP)SNAP33 fusion proteins (Movie S1). Cotyledons were plasmolyzed in 5 m sorbitol and imaged by confocal microscopy. In plasmolyzed epidermal cells, accumulation of GFPPEN1 and mYFPSNAP33 remained associated with papillae beneath appressoria. Weaker fluorescent labeling by both fusion proteins was uniformly distributed along the protoplast plasma membrane, without visible enrichment into membrane microdomains (Figure 2c and Movie S1). No GFP signal was detected in papillae formed in nontransgenic Arabidopsis inoculated with B. g. hordei, and the fluorescence emission spectrum associated with papillae formed in GFPPEN1 transgenic plants was characteristic of GFP (not shown). Overall, this extends previous observations that in addition to the syntaxin PEN1 resident in the plasma membrane, its membrane-associated interacting partner SNAP33 also accumulates beneath powdery mildew appressoria. Furthermore, their concentration is extracellular, resulting from incorporation into discrete cell wall thickenings. Remarkably, GFP PEN1 fluorescence could still be detected in papillae at 14 days post-inoculation (dpi; not shown), indicating that GFP remains stable outside the cell for extended periods of time.

GFPPEN1 co-localizes with callose inside haustorial encasements The haustoria of rust and powdery mildew fungi, as well as oomycetes, frequently become encased by a layer of callose-containing material (Aist, 1976; Donofrio and Delaney, 2001; Heath and Heath, 1971; Mangin-Peyrard and Ppin, 1996; Skalamera et al., 1997). In interactions between Arabidopsis and adapted powdery mildew species, the occurrence of haustorial encasements has only been reported with G. orontii (Jacobs et al., 2003). To investigate the localization of PEN1 around haustoria, Arabidopsis seedlings expressing GFP PEN1 were inoculated with four different haustorium-forming pathogens, and the infected cotyledon epidermal cells were examined by confocal microscopy at 24 hpi. For the simultaneous localization of callose in vivo, the cotyledons were infiltrated with an aqueous solution of Sirofluor. Haustorial encasements were found to be intensely labeled by GFPPEN1 in three different powdery mildew interactions involving the non-adapted B. g. hordei (Figure 3a,b) and E. pisi (Figure 3c,d) and the adapted G. orontii (Figure 3e,f). In each interaction, the GFPPEN1 labeling coincided precisely with the callose labeling, and optical sectioning showed that GFP

PEN1 was distributed uniformly throughout the interior of the callose encasement. Furthermore, after plasmolysis of the infected cells, the GFPPEN1 fluorescence remained associated with the haustorial encasement rather than the retracted host plasma membrane (not shown). Encased haustoria of the oomycete H. parasitica were also labeled by GFP-PEN1 (Figure 3g,h), but in this case the co-localization with callose was not complete: around partially encased haustoria, some GFPPEN1 fluorescence was external to the callose labeling, possibly located in the host cytoplasm (Figure 3h, inset). Although the non-adapted B. g. hordei succeeded in producing very few haustoria on Arabidopsis, those that did form were partially, but never fully, encased (Figure 4c), possibly reflecting the onset of an epidermal cell death response once the fungus had succeeded in overcoming pre-invasion defense responses (Lipka et al., 2005). In contrast, in barley leaves callosic papillae were detected beneath appressoria, around the haustorial neck, but encasement of the haustorial body and haustorial lobes was never observed even at 72 hpi (Figure S2). This suggests that on its natural host, barley, B. g. hordei can effectively suppress callose deposition around its haustoria. Haustorial encasements contain membrane material Since PEN1 is a SNARE protein resident in the plasma membrane, the detection of GFPPEN1 within haustorial encasements raises the possibility that host plasma membrane material becomes incorporated into the encasement. To detect membrane material in encasements, we used the amphiphilic styryl dye FM4-64, which only fluoresces strongly when in a hydrophobic environment, such as the lipid bilayer of a membrane (Bolte et al., 2004). Arabidopsis seedlings expressing GFPPEN1 were inoculated with four different haustorium-forming pathogens and after 24 or 72 h, cotyledons were infiltrated with FM4-64 and examined by confocal microscopy. The labeling of haustorial encasements by GFPPEN1 spatially coincided with the FM4-64 labeling (Figure 4). This precise co-localization was observed with the non-adapted pathogens B. g. hordei (Figure 4ac) and E. pisi (Figure 4df), as well as the adapted pathogens G. orontii (Figure 4gi) and H. parasitica (Figure 4jl). Taken together, these results suggest that not only integral membrane proteins such as PEN1 but also membrane lipids become incorporated into haustorial encasements. Moreover, this phenomenon is not restricted to powdery mildew fungi, because membrane lipid was also detected in oomycete haustorial encasements. Encasement correlates with haustorial age in Golovinomyces orontii To determine when haustorial complexes of G. orontii become encased, a time-course study was performed, using GFPPEN1 accumulation as a marker for encasement. Arabidopsis plants expressing GFPPEN1 were analyzed at intervals between 16 and 25 hpi. At each time-point, fully differentiated primary haustoria (i.e. those formed by the primary appressorium) were scored for the presence or absence of encasements containing strong GFP fluorescence (Figure 5). Both partially and fully encased haustoria were included, but when GFPPEN1 labeling was restricted to accumulation around the haustorial neck (Figure 5b), the haustorium was scored as not encased. The first haustoria completely surrounded by a GFPPEN1-labeled encasement became visible at 23 hpi (Figure 5a). Thereafter, the proportion of encased haustoria increased over time. Thus, encasement only occurred around fully differentiated, mature haustoria. The agedependent nature of encasement is also illustrated by Figure S3, showing two haustoria at different stages of development: the older primary haustorium has a GFPPEN1-labeled encasement whereas the immature secondary haustorium is not encased. Interestingly, all partial encasements were cup-shaped and extended varying distances from the haustorial neck,

suggesting that these represent intermediate stages leading to fully encased haustoria. We never observed simultaneous deposition of encasement material all around the haustorium periphery or randomly scattered patches of encasement material. We infer from this that the engulfment process is initiated at the papilla/haustorial neck region and proceeds through continuous rim growth (Figure 6). Defense-related membrane proteins are preferentially incorporated into haustorial encasements To investigate whether SNARE proteins other than PEN1 accumulate in haustorial encasements, we tested transgenic Arabidopsis lines expressing t- or v-SNAREs tagged with fluorescent proteins, namely GFPPEN1, cyan fluorescent protein (CFP)SYP122, mYFPSNAP33 and GFP VAMP722. In addition, to determine whether plasma membrane-resident proteins unrelated to secretion also accumulate in haustorial encasements, we tested lines expressing five GFP-tagged plasma membrane marker proteins. These included the ABC transporter PEN3, which is known to accumulate at attempted powdery mildew entry sites (Stein et al., 2006), as well as two aquaporins (SIMIP and PIP2a), one low temperature-induced protein (LTI6a) (Cutler et al., 2000), and a truncated SNARE protein (Vamp3) (Koh et al., 2005), that were all shown to be excluded from the extrahaustorial membrane of G. cichoracearum (Koh et al., 2005). Seedlings of the transgenic lines were inoculated with G. orontii and at 24 hpi haustoria were scored for the presence of encasements labeled by the fluorescent fusion proteins (Figure 7b), and independently for the presence of callose encasements (Figure 7a). Among the SNARE proteins, haustorial encasements were more frequently labeled by GFPPEN1/mYFPPEN1 and mYFP SNAP33 than by GFPVAMP722 (Figures 7 and S4), but the closest homologue of PEN1, SYP122, rarely accumulated in encasements. The incidence of fluorophore marker accumulation was comparable in lines expressing GFPPEN1 from the strong 35S promoter or mYFPPEN1 from native 5 regulatory PEN1 sequences. Since the level of functional mYFPPEN1 in this transgenic line (in the pen1-1 null mutant background) is comparable to PEN1 abundance in wild-type plants (Pajonk et al., 2008), physiological amounts of the syntaxin appear to be sufficient for the visible accumulation of the fusion protein in haustorial encasements. Encasements were rarely labeled by the four plasma membrane marker proteins SIMIP and PIP2a, LTI6a, and Vamp3 (2% of interaction sites). In contrast, the ABC transporter PEN3, resident in the plasma membrane, labeled haustorial encasements most frequently of all the fusion proteins (35% of interaction sites). However, this transgenic line displayed an unusually high frequency of callose encasements around G. orontii haustoria (52%), which may indicate incomplete complementation of the pen3 mutation by the GFP fusion protein. For this reason, data for the GFPPEN3 line are not included in Figure 7. Overall, the results suggest that membrane proteins involved in defense-related processes are preferentially incorporated into haustorial encasements, and among the SNARE proteins, the plasma membrane-associated PEN1 and SNAP33 are incorporated more frequently than the vesicle-associated VAMP722. Deposition of PMR4/GSL5 callose into haustorial encasements occurs independently of PEN1 syntaxin We tested whether PMR4/GSL5-derived callose deposition at haustorial encasements is dependent on PEN1 accumulation. Golovinomyces orontii haustoria were isolated from wild-type, pmr4/gsl5, and pen1 mutant plants at 7 dpi, fixed and embedded in resin. Sections through the haustoria were then stained with aniline blue to label callose. Haustoria formed in pen1 plants (Figure S5gi) showed callosic encasements similar to those isolated from wild-type Arabidopsis (Figure S5ac), whereas fluorescence was undetectable in haustoria purified from pmr4/gsl5

leaves (Figure S5df) although encasements were clearly present (Figure S5e). These findings indicate that the deposition of PMR4/GSL5 callose in haustorial encasements occurs independently of the PEN1 syntaxin. This also shows that the formation of encasements can take place in the absence of PMR4/GSL5 callose, and that encasements presumably contain additional plant cell wall polymers. We note that the shape of haustorial encasements isolated from pmr4/gsl5 plants was irregular in outline (Figure S5e), suggesting that callose contributes to the shape of the encasements.

Discussion Specificity of papillary PEN1 accumulation in interactions with powdery mildews PEN2 and PEN3 each restrict the growth of a broad spectrum of pathogens on Arabidopsis, including adapted and non-adapted powdery mildews, the non-adapted oomycete Phytophthora infestans and the necrotrophic fungus Plectosphaerella cucumerina (Lipka et al., 2005; Stein et al., 2006). In contrast, PEN1 has so far only been shown to restrict invasive growth of the nonadapted powdery mildews B. g. hordei and E. pisi (Collins et al., 2003; Lipka et al., 2005). This correlates with our finding that the accumulation of GFPPEN1 in papillae is triggered by the entry attempts of powdery mildews but not by three other ascomycete pathogens or the oomycete H. parasitica. Furthermore, accumulation of GFPPEN1 was not detectable at attempted entry sites of adapted and non-adapted Colletotrichum species (Shimada et al., 2006). It is possible that the apparent specificity of this response reflects differences in the entry mechanisms used by pathogens that penetrate the cell wall. Alternatively, only powdery mildews fail to effectively suppress the concentration of PEN1 into papillae. Notably, PEN1 only restricts the entry of non-adapted powdery mildew sporelings, whereas those of G. orontii enter Arabidopsis leaf epidermal cells in PEN1 and pen1 genotypes with a similar high frequency of 70% (Lipka et al., 2005). However, we have shown here that non-adapted powdery mildews and G. orontii trigger the papillary accumulation of GFPPEN1 with comparable frequency (Table S1). This implies that the adapted fungus has evolved means to detoxify the potentially harmful vesicle cargo secreted by hetero-oligomeric PEN1SNAP33VAMP721/VAMP722 SNARE complexes or inhibits the formation of productive PEN1 ternary SNARE complexes (Kwon et al., 2008b; Pajonk et al., 2008), rather than suppressing extracellular accumulation of PEN1. Although the apparent specificity of papillary accumulation of PEN1 in response to powdery mildew attack is remarkable, such a pathogen-specific engagement of a plant protein is not without precedent. For example, functionally conserved MLO proteins in barley, Arabidopsis and tomato (Solanum lycopersicum) are specifically required for successful host cell entry of a wide range of adapted powdery mildew species, and barley MLO co-accumulates with ROR2 syntaxin, the barley homolog of PEN1, beneath appressorial contact sites (Bai et al., 2008; Bhat et al., 2005; Bschges et al., 1997; Collins et al., 2003; Consonni et al., 2006). Since mlo resistance in Arabidopsis is partially compromised by loss-of-function mutations in PEN1 (mlo pen1 genotype; Consonni et al., 2006) and, similarly, barley mlo resistance is partially broken by a mutation in ROR2 syntaxin (mlo ror2 genotype; Freialdenhoven et al., 1996), MLO might directly or indirectly antagonize PEN1/ROR2-dependent secretion in response to powdery mildew attack. We note, however, that whilst PEN1 accumulation in papillae appears to be a unique plant response to powdery mildew attack, GFPPEN1 accumulates in encasements around the haustoria of both G. orontii and the oomycete H. peronospora (Figure 3eh), thus revealing differences in the

specificity of PEN1 accumulation between pre- and post-invasion defense responses to haustorium-forming parasites. Mechanism(s) of papillary PEN1 accumulation In a previous study, confocal optical sectioning detected the presence of GFPPEN1 at 1724 hpi within the interior of papillae induced by B. g. hordei and E. cichoracearum sporelings (Assaad et al., 2004; Figure 2a,b). Here we have shown by a combination of plasmolysis and optical sectioning that papillary GFPPEN1 can be detected at earlier stages of papilla formation (at 12 hpi) in response to attack by B. g. hordei. Moreover, its interacting partner SNAP33, which normally resides on the cytoplasmic face of the plasma membrane, co-accumulates with the syntaxin in the newly synthesized papilla matrix (Figure 2c and Movie S1). However, in another study (Bhat et al., 2005), GFPPEN1 labeling completely retracted away from entry sites of B. g. hordei appressoria in the majority of epidermal cells inspected at 12 hpi, suggesting that focal accumulations of PEN1 are physically separable from papillae. The apparent disparity between the present results and those reported by Bhat and co-workers may relate to differences in plant age, the osmoticum, and the type of plant tissue used for plasmolysis experiments. However, the observation of circular areas of filipin-induced fluorescence beneath appressoria, indicating a localized redistribution of membrane sterols, was interpreted as evidence for the formation of plant plasma membrane micro-domains that concentrate PEN1 at powdery mildew entry sites (Bhat et al., 2005). One possibility is that these micro-domains are formed only transiently and serve as exit sites for the localized secretion of extracellular PEN1 and its SNAP33 partner. What is the underlying cellular mechanism for the unusual transport of a plasma membraneresident protein and its associated partner into the paramural space? We found that GFP fluorescence persists in papillae for at least 14 days, suggesting that the fluorophore is protected within papillae against degradation/turnover. Lack of GFPPEN1 turnover beneath powdery mildew appressoria is consistent with previous fluorescence recovery after photobleaching (FRAP) experiments (Bhat et al., 2005), showing that the process of GFPPEN1 accumulation is triggered only once following powdery mildew attack while, at a distance, replenishment of GFP PEN1 in the plasma membrane occurs rapidly (within minutes) and continuously. The detection of strong papillary GFP fluorescence is perplexing, since secreted GFP (secGFP) does not fluoresce, or fluoresces only weakly, in plants, probably because the low pH of the apoplast is not conducive to effective GFP fluorescence (Zheng et al., 2004). One transport mechanism that could account for the extracellular accumulation of GFP in papillae and the exceptional stability of this fluorophore within the apoplast is exosome formation (An et al., 2006a; van Niel et al., 2006; Schorey and Bhatnagar, 2008). In animal cells, exosomes are membrane-bound vesicles with a diameter of 30100 nm and a buoyant density equivalent to cytoplasm. The release of exosomes involves fusion of multivesicular bodies (MVBs) with the plasma membrane and discharge of the vesicles they contain into the extracellular space. Multivesicular bodies, in turn, originate from the plasma membrane via endocytosis (Figure 8). The ESCRT protein complex appears to drive invaginations of the MVB membrane, thereby generating multiple intraluminal vesicles (ILVs). Although MVB cargo is normally destined for terminal degradation following fusion with lytic vacuoles, retrograde transport and fusion of MVBs with the plasma membrane leads to the extracellular release of ILVs, which are then called exosomes (Figure 8; van Niel et al., 2006). Importantly, the biogenesis of putative papillary exosomes from the plant plasma membrane predicts a topology of membrane-resident GFPPEN1 and associated mYFPSNAP33 (Figure 8) that would protect the fluorophores against the low apoplastic pH and could thus explain the strong and persistent extracellular fluorescence detected in this and previous studies (Assaad et al., 2004; Bhat et al., 2005).

Previous ultrastructural studies on papillae induced by powdery mildew and other plant pathogens lend support to the hypothesis that exosome-mediated secretion contributes to papilla formation. Multivesicular bodies are often abundant in the plant cytoplasm around papillae and label with probes specific for -1,3-glucans and H2O2, both of which accumulate in papillae (An et al., 2006a,b; Xu and Mendgen, 1994). Moreover, examination of both resin sections and freeze-fracture replicas has revealed apparent fusion events between large MVBlike structures and the plant plasma membrane around papillae, as well as the presence of smaller vesicle-like structures embedded within the papilla matrix (Aist and Williams, 1971; Aist, 1976; An et al., 2006a,b; Zeyen and Bushnell, 1979). In the present study, our finding that papillae are strongly labeled by the lipophilic dye FM4-64 is also consistent with the predicted incorporation of exosome-derived membrane material into these cell wall thickenings (Figure S6). Do common plant defense mechanisms act at papillae and haustorial encasements? The deposition of callose-containing encasements around haustoria appears to proceed by extension of the papilla into a collar around the haustorial neck, which expands over the haustorial body and may eventually enclose the entire haustorium (Bushnell, 1972). The underlying mechanisms of deposition and function of the encasement are poorly understood. Encasements might have a role in plant defense by serving as a barrier to block effective uptake of host nutrients through haustoria. Alternatively, these structures might be induced by the parasite to protect haustoria against toxic plant defense molecules and/or limit the exposure of fungal cell surface pathogen-associated molecular patterns (PAMPs) that could elicit PAMPtriggered immune responses. Here we have shown that encasements of B. g. hordei haustoria were never detected in epidermal cells of its natural host barley at 3 dpi, when extensive fungal growth occurs on the leaf surface (Figure S2). In contrast, essentially all B. g. hordei haustoria on the non-host Arabidopsis were partially encased at 24 hpi (Figure 4c). This is difficult to reconcile with the idea that the fungus induces the formation of encasements for self-protection, and rather suggests that B. g. hordei fails to suppress their formation in the non-host interaction. Importantly, we detected callosic encasements in approximately 20% of interaction sites with the adapted G. orontii fungus (Figure 7a), which were not seen in a previous study with the adapted G. cichoracearum mildew (Koh et al., 2005). This is consistent with the hypothesis of encasement suppression by adapted powdery mildew fungi and suggests that G. orontii, unlike G. cichoracearum, can only partially inhibit the formation of these structures in Arabidopsis. We found a preferential integration of fluorophore-tagged SNARE proteins PEN1 syntaxin, SNAP33, and, with lower frequency, VAMP722 into encasements of G. orontii haustoria (Figure 7a,b). In addition, the ABC transporter fusion protein GFPPEN3, which like PEN1 also accumulates at powdery mildew entry sites (Stein et al., 2006), labeled haustorial encasements most frequently of all the tested fusion proteins. The low frequency (2% of interaction sites) of detectable fluorophore in haustorial encasements formed in transgenic lines expressing four other plasma membrane marker proteins strongly argues against the possibility that non-specific cleavage products, rather than the full-length fusion proteins, become integrated inside encasements (Figure 7a,b). Furthermore, this finding suggests that a defense-related subset of plasma membrane proteins is selectively recruited and incorporated into encasements. It is possible that PEN1-dependent exocytosis and PEN3-dependent secretion of small molecules, which is normally seen before fungal entry, continues during the presumed rim growth extension of papillae into haustorial encasements (Figure 6). Thus, we hypothesize that shared plant defense mechanisms are activated during the formation of papillae and haustorial encasements. Moreover, because the labeling of haustorial encasements by GFPPEN1 spatially

coincides with FM4-64 labeling (Figure 4), indicating incorporation of membrane lipids as well as proteins, we speculate that exosome biogenesis/release is a common mechanism by which the otherwise plasma membrane-resident PEN1 and PEN3 become incorporated into extracellular papillae and haustorial casings (Figures 6 and 8). These inferred secretory defense responses at haustorial encasements could account for the recently reported hyper-susceptible G. orontii infection phenotypes of VAMP721+/VAMP722/ and VAMP721/VAMP722+/ plants, although this post-invasive VAMP721/722 defense function is likely to involve additional syntaxins other than PEN1 (e.g. SYP122; Kwon et al., 2008b). The detection of PMR4/GSL5-dependent callose in papillae and powdery mildew encasements demonstrates the engagement of a shared glucan synthase-like enzyme directing the synthesis of a structural component of both extracellular compartments (Jacobs et al., 2003; Nishimura et al., 2003). The formation of cytologically detectable papillae and haustorial encasements in pmr4/gsl5 mutants (Nishimura et al., 2003; Figure S5) demonstrates that other de novosynthesized cell wall polymers are sufficient to serve as scaffold(s) for these extracellular compartments. Whereas the irregular outline of haustorial encasements purified from pmr4/gsl5 plants points to an architectural role for PMR4/GSL5 callose (Figure S5), we cannot rule out that this anomaly is an indirect consequence of heightened salicylic acid-dependent defense responses in pmr4/gsl5 mutants (Nishimura et al., 2003). The formation of PMR4/GSL5 callosic encasements in the pen1 null mutant background demonstrates that the enzyme and its substrates must be delivered to appropriate cellular destination(s) independently of PEN1 syntaxin function. Although this makes it unlikely that rate-limiting amounts of the presumed plasma membrane-resident PMR4/GSL5 and its substrates form part of the vesicle cargo discharged by PEN1/SNAP33/VAMP721/722 ternary complexes, it is possible that PMR4/GSL5 targeting involves other plasma membrane-resident syntaxins such as SYP122. It remains to be determined whether the polytopic PMR4/GSL5 membrane protein is itself transported inside papillae and haustorial encasements by exosomes and, if so, whether it contributes to the extracellular synthesis of callose within these compartments. Whilst our study has provided evidence for the existence of common molecular processes leading to the formation of papillae and haustorial encasements, it remains to be shown whether the inferred exosomes in both extracellular structures directly contribute to the growth restriction of invading powdery mildew sporelings or are merely the by-product of focal secretion. Indirect evidence for the former interpretation comes from findings that pen1 mutants show delayed papilla formation upon attack by B. g. hordei (Assaad et al., 2004) and that in 35S::RNAiVAMP722 gene silencing lines papillary PMR4/GSL5 callose deposition is reduced (Kwon et al., 2008b). Thus, neither PEN1 nor VAMP721/VAMP722 are absolutely required for the formation of papillae, indicating the existence of redundant secretory pathways. We propose that termination of fungal pathogenesis requires multiple pathways, including exosomal and SNARE complex-mediated secretion, which act together at the plasma membrane to enable the timely and localized release/storage of cell wall components and antimicrobial cargo. Experimental procedures Barley (Hordeum vulgare, cultivar Ingrid-10) and Arabidopsis plants were grown as described previously (Lipka et al., 2005; Shen et al., 2007). The following Arabidopsis thaliana mutant and transgenic lines were used in this study (all Col-0 background): pen1-1 and p35S::GFPPEN1 in pen1-1 (Collins et al., 2003), pmr4-1 (Nishimura et al., 2003), p35S::GFPSIMIP, p35S::GFP Vamp3, p35S::GFPPIP2a and p35S::GFPLTI6b (Cutler et al., 2000), pn::mYFPPEN1 in pen1-1 (Pajonk et al., 2008), pn::GFPVAMP722 in vamp722 and p35S::mYFPSNAP33 in snap33 (the

fusion protein rescues the snap33 lethality; Kwon et al., 2008b), p35S::CFPSYP122 in pen1-1 (Assaad et al., 2004), and pn::PEN3GFP in pen3-1 (Stein et al., 2006). Growth conditions, inoculation methods and infection assays for the pathogens used in this study were described in Miklis et al. (2007) for B. g. hordei (isolate K1) and E. pisi (Birmingham isolate), in Gllner et al. (2008) for G. orontii (strain MPIZ), in Tanaka et al. (2007) for M. grisea (isolate Haku 1-007), in OConnell et al. (2004) for C. destructivum (isolate LARS 202) and C. higginsianum (isolate IMI34906I) and in Aarts et al. (1998) for H. parasitica (isolate Noco2). Light microscopy The frequency of GFPPEN1 accumulations beneath differentiated appressoria was scored on primary leaves of 2-week old plants expressing 35S::GFPPEN1 in pen1-1. Three to seven plants (one leaf per plant, 100 interaction sites per leaf) were inspected at 24 hpi and analyzed by bright field and epifluorescence microscopy to visualize differentiated appressoria and GFPPEN1 focal accumulations. To quantify callosic encasements, inoculated cotyledons were harvested at 3 dpi, cleared in ethanol/acetic acid (3:1) for 24 h and stained with aniline blue (in 150 mm KH2PO4, pH 9.5) and inspected by UV-light epifluorescence microscopy. In four independent experiments at least three plants (one leaf per plant, 100 interaction sites per leaf) were scored. Laser scanning confocal microscopy Two-week old Arabidopsis seedlings were inoculated as described above. To stain for membrane lipids and callose, aqueous solutions of FM4-64 (17 m; Invitrogen, http://www.invitrogen.com/) or Sirofluor (0.1 mg ml1; Biosupplies, http://www.biosupplies.com.au/products.htm) were vacuum-infiltrated into excised cotyledons 3 5 min. To analyze callosic papillae in barley, inoculated primary leaves were cut into pieces of length 1 cm and vacuum-infiltrated with Sirofluor (3 10 min). Marker proteins and fluorochromes were viewed using a Leica TCS SP2 confocal microscope (Leica, http://www.leica-microsystems.com/) equipped with an argon/heliumneon laser. The GFP was excited at 488 nm, mYFP was excited at 514 nm, FM4-64 was excited at 561 nm and a 405 nm diode laser was used to excite cCFP, Sirofluor and aniline blue. Images were analyzed and processed with Leica software (LCS Lite, Leica Microsystems, http://www.leica-microsystems.com) and Adobe Photoshop 8.0. To monitor plasmolysis, cotyledons from 10-day-old Arabidopsis seedlings expressing fluorochrome-tagged SNARE proteins were mounted on microscope slides in 5 m sorbitol and imaged immediately by confocal microscopy. Haustorial complex isolation and resin embedding Haustoria of G. orontii were isolated from 4- to 5-week-old Arabidopsis plants heavily infected with the fungus at 7 dpi, using a protocol developed for the isolation of Colletotrichum lindemuthianum primary hyphae by isopycnic centrifugation (Pain et al., 1994). Isolated cells were then processed for resin embedding as follows. Samples were fixed in 4% (w/v) formaldehyde, 0.5% (v/v) glutaraldehyde in 0.05 m sodium phosphate buffer (pH 7.2), washed and dehydrated through a graded ethanol series. They were then incrementally infiltrated with LR white acrylic resin (London Resin Company, http://www.lrgold.com/), and polymerized at 50C. Sections (700 nm) were cut and mounted onto poly-l-lysine-coated slides, stained with aniline blue and analyzed by confocal microscopy as described above.