Type III secretion and effectors shape the survival and growth pattern of Pseudomonas syringae on leaf surfaces.

The bacterium Pseudomonas syringae pv syringae B728a (PsyB728a) uses a type III secretion system (T3SS) to inject effector proteins into plant cells, a process that modulates the susceptibility of different plants to infection. Analysis of GREEN FLUORESCENT PROTEIN-expressing PsyB728a after spray inoculation without additives under moderate relative humidity conditions permitted (1) a detailed analysis of this strain's survival and growth pattern on host (Nicotiana benthamiana) and nonhost (tomato [Solanum lycopersicum]) leaf surfaces, (2) an assessment of the role of plant defenses in affecting PsyB728a leaf surface (epiphytic) growth, and (3) the contribution of the T3SS and specific effectors to PsyB728a epiphytic survival and growth. On host leaf surfaces, PsyB728a cells initially persist without growing, and show an increased population only after 48 h, unless plants are pretreated with the defense-inducing chemical benzothiazole. During the persistence period, some PsyB728a cells induce a T3SS reporter, whereas a T3SS-deficient mutant shows reduced survival. By 72 h, rare invasion by PsyB728a to the mesophyll region of host leaves occurs, but endophytic and epiphytic bacterial growths are not correlated. The effectors HopZ3 and HopAA1 delay the onset of epiphytic growth of PsyB728a on N. benthamiana, whereas they promote epiphytic survival/growth on tomato. These effectors localize to distinct sites in plant cells and likely have different mechanisms of action. HopZ3 may enzymatically modify host targets, as it requires residues important for the catalytic activity of other proteins in its family of proteases. Thus, the T3SS, HopAA1, HopZ3, and plant defenses strongly influence epiphytic survival and/or growth of PsyB728a.

A major virulence component of the extracellular bacterial plant pathogen Pseudomonas syringae involves the injection of large repertoires of different effector proteins (often .25) into plant cells using a type III secretion system (T3SS; Greenberg and Vinatzer, 2003). Once inside plant cells, effectors can suppress host defenses, promote disease, and/or possibly elicit the release of nutrients (Chen et al., 2000;Lee et al., 2001;Nomura et al., 2005;Abramovitch et al., 2006;Chisholm et al., 2006;Rosebrock et al., 2007). Some plants can recognize individual effectors and activate defenses that result in the restriction of bacterial growth. Effector recognition is usually conditioned by a cognate plant resistance (R) protein(s), which when activated often induces localized programmed cell death, the accumulation of the signal molecule salicylic acid (SA), and global transcriptional reprogramming (Greenberg and Yao, 2004;Caplan et al., 2008;Lee et al., 2008). SA is important for basal resistance to P. syringae as well as some R-gene-mediated resistance responses (Delaney et al., 1994;Nawrath and Métraux, 1999;Feys et al., 2001;Rate and Greenberg, 2001). Exogenous application of SA or the synthetic agonist benzo (1,2,3) thiadiazole-7carbothioic acid S-methyl ester (BTH) results in reduced pathogen growth (Uknes et al., 1992;Lawton et al., 1996).
Effectors that induce strong disease resistance were historically termed avirulence (Avr) effectors (Greenberg and Vinatzer, 2003). A number of characterized Avr effectors have large (qualitative) effects on pathogen fitness on specific hosts, causing large reductions in pathogen growth. However, recently it was found that P. syringae pv syringae strain B728a (PsyB728a) harbors several effectors that can quantitatively restrict pathogen growth on Nicotiana benthamiana even when disease ensues (Vinatzer et al., 2006). One possibility is that these effectors prevent the pathogen from causing too much host damage, and thus facilitate the cosurvival/coevolution of the pathogen and host.
Some P. syringae strains, such as PsyB728a, have a strong epiphytic phase (Wilson et al., 1999), in which the bacteria colonize leaf surfaces, called the phylloplane. Epiphytic bacteria must be able to withstand fluctuating wet and dry conditions, exposure to UV light, temperature fluctuations, and nutrient limitations. To protect themselves from these drastic conditions, one strategy is for bacteria to form aggregates. Surface-associated aggregates are often surrounded by exopolysaccharides that provide protection against desiccation, free radicals, and UV light and are important for epiphytic fitness (Yu et al., 1999;Wright et al., 2005;Chang et al., 2007). Bacteria also communicate with each other and other bacterial species on leaf surfaces through quorum sensing; this can also affect aggregate size (Dulla and Lindow, 2008). On snap bean (Phaseolis vulgaris) leaf surfaces, which PsyB728a colonizes very well, PsyB728a sometimes exists as some single cells, but a majority of the cells are found in aggregates of various sizes (Monier and Lindow, 2004). Many areas of leaf surfaces are not hospitable to bacteria, prompting the suggestion that access to nutrients is nonuniform in the phylloplane (Morris and Monier, 2003).
Another strategy for P. syringae to withstand fluctuating environmental conditions is to invade mesophyll regions and grow endophytically. Some strains of P. syringae produce coronatine, a metabolite mimic of Ile-conjugated jasmonic acid, which can facilitate bacterial entry into substomatal chambers (Melotto et al., 2006). PsyB728a does not produce coronatine, but instead makes syringolin A, which causes stomatal opening and counteracts stomatal innate immunity in snap bean and Arabidopsis (Arabidopsis thaliana; Schellenberg et al., 2010).
A majority of bacteria associated with leaves, including P. syringae, are on leaf surfaces (Wilson et al., 1999). Even before effectors were identified at the molecular level, epiphytic populations of P. syringae pv syringae on susceptible cultivars of snap bean were found to be larger than the populations on more resistant cultivars (Daub and Hagedorn, 1981). This is possibly due to the action of effectors. Field studies showed that T3SS-deficient PsyB728a does not survive well on snap bean and this reduced survival was hypothesized to be due to reduced epiphytic fitness . Interestingly, PsyB728a lacking the effectors HopAA1 or HopZ3 have increased populations of bacteria that can be washed off N. benthamiana leaves several days after spray inoculation, a phenotype that can be reversed by adding back each gene on a plasmid (Vinatzer et al., 2006). PsyB728a lacking a functional T3SS also show a greatly reduced epiphytic population using this assay (Vinatzer et al., 2006). In contrast, endophytic HopAA1-or HopZ3-deficient PsyB728a grow similarly to the wild-type strain in the mesophyll region of the leaves (Vinatzer et al., 2006). This suggests a role for specific effectors in the epiphytic niche on some plants. An important caveat to this assertion is that leaf disc washes might be contaminated with endophytic bacteria.
In this study, we used quantitative microscopy to perform a detailed analysis of the roles of the T3SS, HopAA1, and HopZ3 in epiphytic fitness and the timing of the expression of effectors during PsyB728a associations with leaf surfaces kept under moderate relative humidity conditions. We show not only the importance of the T3SS in early epiphytic survival when effectors are beginning to be expressed, but also that HopAA1 and HopZ3 have host-specific roles in epiphytic fitness and localize to different subcellular sites in plant cells. We also show that activating the SA pathway results in the reduced population size of epiphytic bacteria. This suggests that effector interactions with epidermal cells and the signaling status of a major plant defense pathway play important roles in the early infection process.

The T3SS Is Necessary for Early Survival and Formation of Medium-Sized Bacterial Aggregates on Leaf Surfaces
Using leaf wash assays, T3SS-deficient (HrcC -) Psy-B728a show reduced epiphytic populations on sprayinoculated N. benthamiana (Vinatzer et al., 2006). In this assay, surface-associated bacteria harvested by gentle vortexing of leaf discs submerged in liquid are quantified by counting colony forming units (cfu). Endophytic bacteria might contaminate the epiphytic pool during this procedure. Therefore, we directly visualized GFP-labeled bacteria carrying Ptrp-gfp (gfp driven by a constitutive promoter) to assess the role of the T3SS in the epiphytic growth of PsyB728a on both N. benthamiana (a good host; Vinatzer et al., 2006) and tomato (Solanum lycopersicum 'Rio Grande-76R', a poor host; Lin and Martin, 2007). This approach permits the quantitative assessment of distributions of bacteria on leaf surfaces using epifluorescence microscopy (Monier and Lindow, 2003). Bacteria are quantified as fluorescence area per field imaged, analyzed using many random fields from multiple independent samples (see "Materials and Methods").
At 3, 24, and 48 h after spray inoculation of N. benthamiana, the average fluorescence area (5-7 mm 2 , about three to six bacteria) of epiphytic PsyB728a did not significantly change (Fig. 1A). (Three hours was the earliest time at which the leaves were dry enough to accurately perform the assay.) At these early times, PsyB728a were found in similar distributions as isolated individual cells and in small groups, the majority of which (.95%) were less than 25 mm 2 .
Between 48 and 72 h, there was a significant increase in the population size of PsyB728a (P = 0.031, Mann-Whitney test, n = 47; Fig. 1A).
The average 3 and 24 h population sizes of HrcCand PsyB728a were similar. However, at 48 and 72 h the average fluorescence area of HrcCwas greatly reduced relative to PsyB728a (Fig. 1A). This indicates a failure of the HrcCstrain to survive well on leaf surfaces after 24 h. Furthermore, the distributions of bacteria per image were significantly different for PsyB728a and HrcC - (Fig. 1B). At 72 h, the majority of the HrcCimages either contained no or few bacteria, ranging from one to eight (0.2-10 mm 2 ). In contrast, for PsyB728a there were more images that contained a few bacteria (,10 mm 2 ). Importantly, there were also PsyB728a images with small (10-100 mm 2 ) and medium aggregates (.100 mm 2 ) that were not present at all following inoculation with the HrcCstrain. In contrast, PsyB728a and HrcCgrew similarly in culture and formed colonies of similar sizes and at similar rates on agar medium (data not shown).
To test if GFP-labeled PsyB728a were dead but fluoresced due to persistent GFP, bacteria on leaf discs were counterstained with propidium iodide. This dye does not penetrate intact membranes and therefore only stains dead bacteria. Red, dead bacteria were found in some instances. However, they were very few and the dead bacteria did not colocalize with greenfluorescing bacteria (Fig. 1C).
When epiphytic bacteria were quantified using the leaf wash assay, there were significantly fewer HrcCbacteria than PsyB728a on N. benthamiana (Vinatzer et al., 2006) and tomato (Fig. 1E). After washing with vortexing, the epiphytic populations on leaf discs were greatly reduced as compared to before washing, as determined by epifluorescence microscopy indicating successful assessment of bacterial populations. A few bacteria (0.2-10 mm 2 ) remained on the leaf discs, but all aggregates greater than 10 mm 2 were removed. PsyB728a and T3SS-deficient HrcC insertion mutant (HrcC -) were sprayed onto 17-to 21-d-old N. benthamiana and tomato 76R at an OD 600 of 0.01. These experiments were repeated four or more times with similar results. All bars indicate SEs. A, B, and D, Fluorescence area (or percent of images with a given fluorescence area) as a measure of bacterial area of GFP-expressing strains carrying Ptrp-gfp on the leaf surfaces at various times after spray inoculation. Leaf disks were viewed using epifluorescence microscopy. A, Population sizes of PsyB728a and HrcCon N. benthamiana. The PsyB728a population size did not change between 3, 24, and 48 h (P . 0.08, Mann-Whitney test, n = 47). At 72 h, the PsyB728a population was significantly greater than at 48 h (P = 0.031, Mann-Whitney test, n = 47). At 3 and 24 h, there was no significant difference between PsyB728a and HrcC -(P . 0.12, Mann-Whitney test, n = 46-47). At 48 and 72 h, the HrcCepiphytic population was significantly reduced relative to PsyB728a at those times (**P = 0.0074, ***P , 0.0001, Mann-Whitney test, n = 47). B, The number of micrographs from 72 h populations that fell into different total fluorescence area categories was counted and the percentage from two independent experiments of the total images was determined. The categories were: zero (no bacteria present), 0.2 to 10 mm 2 (a few bacteria present), 10 to 100 mm 2 (a small aggregate), and .100 mm 2 (a medium aggregate). PsyB728a had significantly more micrographs that had larger amounts of bacteria and aggregates than HrcC -(P , 0.0001, x 2 test, n = 48). Using the Fisher's exact test with two categories: ,10 mm 2 or .10 mm 2 , also indicated that there were significantly more wild-type bacteria (P , 0.0001). C, Aggregates of PsyB728a/Ptrp-gfp that fluoresced green on N. benthamiana were primarily alive. A representative epifluorescence micrograph of PsyB728a/Ptrp-gfp on N. benthamiana 4 d after infection counterstained with propidium iodide, which preferentially stains dead cells with damaged membranes.
A total of 2523 magnification, scale bar is 20 mm. Green fluorescence marks all GFP-fluorescing bacteria. Red fluorescence marks dead, propidium-iodide-stained bacteria. Section shows green and red fluorescence merged images. D, Fluorescence area of GFP-expressing bacteria on the surface of tomato 76R leaves after 72 h. PsyB728a had significantly more micrographs that had larger amounts of bacteria and aggregates than HrcC -(P , 0.0001, x 2 test, n = 48). E, HrcChad a significantly lower bacterial population than PsyB728a as assayed using leaf washes after 72 h on tomato 76R (*P , 0.05, Mann-Whitney test, n = 12).
These results show (1) that conclusions drawn from the quantitative microscopy analysis and leaf wash assays are similar results, and (2) that possible endophytic bacterial contamination from the mesophyll region does not significantly affect conclusions about strain-to-strain differences in epiphytic populations using the wash assay. Furthermore, the microscopy analysis shows that the T3SS is important for survival prior to when bacteria populations increase on host leaf surfaces and probably also for survival on nonhost leaf surfaces when there is very limited bacteria growth or disease, as was the case on tomato.
Epiphytic Growth of PsyB728a on N. benthamiana Is Not Correlated with Endophytic Growth after Spray Inoculation After spray inoculation, increased numbers of epiphytic PsyB728a may in part result from invasion to the intercellular mesophyll region followed by release of endophytic bacteria to leaf surfaces. To assess the degree to which bacteria originating from endophytic microcolonies can contribute to epiphytic populations, Figure 2. PsyB728a endophytic microcolonies are rarely detected after spray inoculation, and endophytic populations are not correlated with epiphytic populations. A, Single bacteria (white arrow) detected by confocal microscopy were detectable 3 h after low-dose infiltration (OD 600 of 0.00001) into the intercellular mesophyll region. B and C, Microcolonies of GFP-labeled PsyB728a carrying Ptrp-gfp or PavrPto-gfp 24 h after infiltration directly into the mesophyll region were common, whereas 72 h after spray inoculation (as in Fig. 1) microcolonies were rare (see text). A to C, 4003 magnification. Scale bar is 20 mm. D. Epiphtyic bacteria were quantified in leaf washes and endophytic bacteria were quantified after macerating washed and surface-sterilized leaf discs 72 h after spray inoculation. Bacteria were quantified from 12 leaf discs from independent plants. Each diamond represents epiphytic and endophytic bacteria enumerated from the same leaf disc. Some endophytic samples contained no bacteria, but are plotted as having a value of 1 due to the log scales of the axes. Arrow indicates the bacterial population at the start of the experiment. These experiments were repeated two or more times with similar results. Figure 3. The epiphytic population of PsyB728a is reduced on BTHtreated N. benthamiana. Twenty-four hours after spraying 17-to 21-dold N. benthamiana with 100 mM BTH, treated and control (untreated) plants were sprayed with PsyB728a at an OD 600 of 0.01. These experiments were repeated three times with similar results. All bars indicate SEs. A, Epiphytic bacterial area was significantly reduced on BTH-treated plants as measured by epifluorescence microscopy of PsyB728a/Ptrp-gfp on d 3 (**P , 0.005, Mann-Whitney test, n $ 55 or *P , 0.03, unpaired t test using square-root-transformed data to account for differences in the data distribution). B, The number of micrographs from A that fell into the different indicated size categories were counted and the percentage of the total images was determined. There was a significant difference between BTH-treated and untreated plants using a x 2 test (P , 0.0001, n $ 55; Fisher's exact test, two categories: ,10 mm 2 or .10 mm 2 , P , 0.0004). C, Bacterial population of BTH-treated plants was significantly reduced as compared to untreated plants when measured using leaf washes (Mann-Whitney test, **P , 0.005, n = 10). D, PsyB728a disease (small pin-prick spots) ratings were reduced on d 3 in BTH-treated plants as compared to untreated plants (P , 0.0001, Mann-Whitney test, n = 25-31). Ratings, expressed as percent diseased leaves/plant, were: 1, no disease; 2, #50%; 3, .50; and #75; 4, 100%.
Lee et al.
we characterized: (1) the propensity for bacteria to be released from newly formed endophytic microcolonies to leaf surfaces, (2) the frequency of endophytic microcolony formation after spray inoculation, and (3) whether the number of endophytic bacteria is correlated with epiphytic population size.
We used confocal microscopy to completely scan leaf discs for endophytic GFP-labeled PsyB728a after inoculation (approximately 1,791 fields/leaf disc). In two experiments, 3 h after very-low-dose infiltration (OD 600 of 0.00001) into the intercellular mesophyll region, individual endophytic GFP-labeled PsyB728a/ Ptrp-gfp bacterium were detected ( Fig. 2A). By 24 h, we observed up to 130 fields among eight leaf discs with 1,635 6 129 mm 2 (mean 6 SE) endophytic microcolony fluorescence area/field with bacteria ( Fig. 2B) in substomatal chamber (31.8%) and mesophyll cell regions (68.2%). Leaves were surface sterilized at the time of infiltration to allow monitoring of bacterial release to leaf surfaces. After 24 h, at sites where we found epiphytic bacteria, their population size averaged 44 bacteria on the leaf surface; most sites were associated with a microcolony under the epidermal layer (Supplemental Fig. S1). Most such microcolonies (83.3%) were associated with substomatal chambers, but some were mesophyll associated (9%). Thus, within 24 h a single bacterium in the mesophyll region can become a microcolony, which in some cases can release a small number of bacteria to the leaf surface (Fig. 2B). Later (after 48 h), lesions were visible on the leaves, at which time endophytic microcolonies were enlarged to 9,077 6 331 mm 2 (n = 313) fluorescence area.
In three experiments, we also analyzed endophytic PsyB728a colonization after spray inoculation using confocal microscopy in which the entirety of eight discs per experiment was examined. Three and 24 h after spray inoculation, there were no GFP-labeled PsyB728a/Ptrp-gfp endophytic bacteria detected in any trials. By 48 h, among the three experiments, we observed a total of 0, 1, or 2 fields per eight leaf discs, with GFP-labeled endophytic PsyB728a clusters that showed 127.8 6 16.8 mm 2 fluorescence area/field containing bacteria. By 72 h, endophytic microcolonies were present but rare: Each leaf disc had one or no fields containing GFP-labeled PsyB728a microcolonies at 1,791 6 649 mm 2 fluorescence area/field containing PsyB728a (Fig. 2C). The timing of appearance and the size of most endophytic microcolonies indicates that most were only newly formed by 72 h after spray inoculation, similar to 24 h microcolonies after infiltration. If these microcolonies behave similar to those observed 24 h after infiltration, each leaf disc would have ,50 bacteria that were released from the endophytic pool to the leaf surfaces, or a small fraction of the total epiphytic pool, at 72 h after spray inoculation.
At 72 h, we also enumerated epiphytic bacteria after leaf washes and subsequently quantitated endophytic bacteria in the same leaf discs. Some leaf discs had similar epiphytic populations (7 3 10 3 -2 3 10 4 cfu), but had widely differing endophytic populations (from 0-5 3 10 5 cfu; Fig. 2D). In this experiment, some endophytic bacteria may be lost during the leaf washes, but this should occur similarly between leaf discs and would not account for the wide differences White triangles indicate GFP-labeled bacteria. 4003 magnification, scale bar is 20 mm. B, Data from A was quantified. White bars: fluorescence area of bacteria expressing GFP from PavrPto-gfp. Black bars: fluorescence area of bacteria expressing constitutive GFP from Ptrp-gfp. At least 48 pictures were analyzed for each strain. Bars indicate SEs. There was no significant change in the fluorescence area of PsyB728a/Ptrp-gfp at the time points shown. The fluorescence area of PsyB728a/PavrPto-gfp was initially undetectable, but increased over time. These experiments were repeated three or more times with similar results. C and D, Cell types with which bacteria expressing GFP from Ptrp-gfp (C) and PavrPto-gfp (D) were associated on leaf surfaces. Bars indicate the percentage of GFP-expressing bacteria observed that were associated with epidermal cell junctions, epidermal cell surfaces, trichomes, and stomata of N. benthamiana at 24 h following inoculation of PsyB728a carrying PavrPto-gfp and Ptrp-gfp.
in endophytic populations that we observed. Additionally, epiphytic PsyB728a attained a population of nearly 10 5 cfu on one leaf disc that yielded only 200 cfu from the intercellular mesophyll region (Fig. 2D). Thus, after spray inoculation, there was no correlation between epiphytic and endophytic populations; endophytic bacteria were too few in most samples to account for the significant increases in epiphytic populations at 72 h.

Activation of the SA Defense Pathway Restricts Epiphytic Growth
To test whether known plant defenses affect epiphytic bacterial populations, the SA agonist BTH was sprayed onto N. benthamiana 24 h prior to spray inoculation with PsyB728a. Three days after infection, the fluorescence area of epiphytic PsyB728a/Ptrp-gfp on BTH-treated leaves was much smaller than the area found on untreated leaves (Fig. 3A). The distributions of aggregates between BTH-treated and -untreated leaves were different, with images from BTH-treated leaves lacking medium aggregates of .100 mm 2 (Fig.  3B). In addition, the bacterial population recovered using the wash assay was also significantly reduced in BTH-treated leaves as compared to untreated leaves (Fig. 3C). Spray inoculation resulted in disease symptoms of small, discolored patches on N. benthamiana leaves. As expected, BTH pretreatment suppressed these disease symptoms. On d 3, the percentage of inoculated leaves with symptoms was significantly reduced in BTH-treated plants (Fig. 3D). BTH is thought not to possess antimicrobial activity, but rather acts indirectly as a defense-inducing SA agonist . Therefore, these results suggest that SA-induced defenses can limit epiphytic bacterial growth and reduce the incidence of disease.

P. syringae B728a Expresses Type III Effectors on the Leaf Surface
The early reduced fitness of HrcCbacteria on leaf surfaces indicated the possibility that P. syringae uses the T3SS on the leaf surface. Since a direct in vivo secretion assay for surface-associated bacteria is not yet available, we used expression of a T3SS reporter as a proxy for possible secretion. Specifically, we intro- These experiments were repeated three or more times with similar results. Bars indicate SEs. Green fluorescence area expressed from Ptrpgfp was quantified as a measure of the amount of bacteria associated with the surface. A, Early population sizes (3, 24 h) of PsyB728a, HopAA1 -, and HopZ3were similar (P . 0.13, Mann-Whitney test, n = 47). At 48 h, HopAA1and HopZ3populations were increased compared to PsyB728a (at each time point, each letter group differs from other letter groups at a level of P , 0.004 or less, Mann-Whitney test, n = 47). Fisher's exact test using two categories: ,10 mm 2 or .10 mm 2 , also showed significant differences: P , 0.0039 (PsyB728a and HopAA1 -), P , 0.0001 (PsyB728a and HopZ3 -). This experiment was done together with HrcCbacteria (see Fig. 1A). Thus, the PsyB728a data are the same as that shown in Figure 1A. B and C, At 48 h (B) and 72 h (C) HopZ3and HopAA1had a greater amount of larger aggregates than PsyB728a. x 2 tests of the distributions at the two time points were each significantly different between the deletion mutants and PsyB728a (P , 0.0001, n = 47). D to I, Five days after spray inoculations, HopZ3and HopAA1had larger and thicker aggregates than PsyB728a on N. benthamiana. Confocal projections of aggregates near disease lesions of GFP-expressing (from Ptrp-gfp) PsyB728a (D), HopAA1 -(E), and HopZ3 -(F) from d 5 with a magnification of 4003. Scale bar is 20 mm. Cross-section views of the thicknesses of bacterial aggregates on leaf surfaces of PsyB728a (G), HopAA1 -(H), and HopZ3 -(I). Sagittal images were created from a collection of 26 to 91 images at a 0.5-mm interval. Autofluorescence of chlorophyll is depicted in red. Scale bar is 10 mm.  (Greenberg and Yao, 2004). However, epidermal cells in contact with fluorescent PsyB728a/PavrPto-gfp bacteria showed no change in their UV fluorescence and maintained their turgid appearance in differential interference contrast images (Supplemental Fig. S3; Fig. 4A).
Generally, the fluorescence area of the mesophyllassociated PsyB728a/PavrPto-gfp bacteria was smaller than that of PsyB728a/Ptrp-gfp (Fig. 2, B and C). Thus, effectors are expressed in a subpopulation of epiphytic bacteria before colonization of the mesophyll region, and only a subset of the total population may synthesize and secrete effectors during the bacterial associations with plant cells.
To determine whether effector expression occurred preferentially at particular sites of the leaf surface, we first quantified the relative distribution of the bacteria on N. benthamiana leaf surfaces 24 h after inoculation. Most PsyB728a/Ptrp-gfp bacteria were associated with the epidermal cell surface and cell junctions, although a small percentage of bacteria were associated with trichomes and stomata (Fig. 4C). Effector gene-expressing bacteria were observed in a similar pattern mainly at the epidermal cell surface and epidermal cell junctions (

Effectors Influence the Survival and Size of the Epiphytic Populations on Leaves
Previous experiments using leaf washes after spray infection suggest that in PsyB728a, HopZ3 and HopAA1 may have roles in the epiphytic niche (Vinatzer et al., 2006). We sought to rule out that the leaf wash results with HopZ3and HopAA1 -PsyB728a were biased due to contaminating endophytic bacteria and to compare the roles of HopZ3 and HopAA1 on host and nonhost plants. Therefore, we further characterized the epiphytic behaviors of the HopZ3and HopAA1strains carrying Ptrp-gfp after spray inoculation using microscopy.
The HopAA1or HopZ3strains each showed a similar level of survival to PsyB728a that did not change between 3 and 24 h on N. benthamiana, as measured using epifluorescence microscopy ( Fig. 5A). At 48 and 72 h, HopZ3and HopAA1showed significantly different population distributions that favored increased aggregate sizes, as compared to PsyB728a (Fig. 5, B and C). At 72 h, PsyB728a had no large aggregates (.1,000 mm 2 ), while these were present, albeit rarely, for HopZ3and HopAA1 - (Fig. 5C). Scanning confocal laser microscopy revealed that compared to PsyB728a aggregates on d 5, aggregates of HopZ3and HopAA1had more developed (Fig. 5, D-F; Table I) and/or thicker appearances (Fig. 5, G-I; Table I). Similar to the microscopy results, leaf washes confirmed increased growth of HopZ3and HopAA1relative to PsyB728a, which was not accompanied by increased endophytic populations of HopZ3or HopAA1at 48 h (Supplemental Fig. S5).
In contrast to the results with N. benthamiana inoculations, epiphytic populations of HopZ3and Ho-pAA1carrying Ptrp-gfp on tomato 76R were reduced, similar to HrcC -, compared with PsyB728a after 72 h. The majority of the HopZ3and HopAA1images Mann-Whitney test was performed for statistical analysis. c Height from the epidermal cell surface to the top of bacterial aggregates. Values are means 6 SEs of the means (n = 7-10). Cross-section images (z stack) were created from a collection of 26 to 91 plane images (x-y) at 0.5-mm intervals. Vertical cross-section (x-z) images were generated with a laser scanning microscope image browser.
either contained no or few bacteria (,10 mm 2 ). In contrast, there were more images of PsyB728a that contained a few bacteria (,10 mm 2 ) or small bacterial clusters (10-100 mm 2 ; Fig. 6A). There were also significantly fewer surface-associated HopAA1and HopZ3bacteria than PsyB728a on tomato, as assayed using leaf washes (Fig. 6B). No significant differences were found in the levels of endophytic HopAA1and HopZ3versus PsyB728a (Fig. 6C; Vinatzer et al., 2006). Thus, whether HopAA1 and HopZ3 suppress or promote epiphytic bacterial growth and/or survival depend on the host genotype/species on which the bacteria reside.

HopZ3, HopAA1, or a T3SS Does Not Affect Bacterial Attachment to Surfaces
It seemed possible that epiphytic population size might be influenced if the T3SS or individual effectors affect bacterial attachment to surfaces. First, we checked the early population size at 3 and 24 h after bacteria had been sprayed onto leaves (Figs. 1A and 5A). There was no difference in the levels of HrcC -, HopZ3 -, or HopAA1relative to PsyB728a, indicating that the same number of bacteria initially adhered to N. benthamiana. The relative abilities of HopZ3 -, HopAA1 -, HrcC -, and PsyB728a to attach to plastic and glass surface were also similar after culturing cells under conditions that promoted effector expression and secretion (He et al., 1993; Fig. 7, A and B). Thus, differences in attachment did not account for the strain-to-strain alterations in epiphytic population distributions that we observed.

HopZ3 or HopAA1 Do Not Affect the Release of Endophytic Bacteria onto Leaf Surfaces
Larger epiphytic communities might occur if effectors promote the release of mesophyll-associated endophytic bacteria onto leaf surfaces. Therefore, we examined the ability of endophytic PsyB728a/Ptrp-gfp and its derived strains to be released from the mesophyll region onto leaf surfaces. Bacteria were syringe inoculated into leaves that were subsequently surface sterilized and examined to ensure that there were no GFP-expressing epiphytic bacteria at the beginning of the experiment. After 48 h, more PsyB728a cells were released to the leaf surface from the intercellular region than HrcC -. This may reflect the reduced fitness of endophytic HrcC -PsyB728a on N. benthamiana (Vinatzer et al., 2006). In contrast, there was no difference in the numbers of released PsyB728a, HopZ3 -, and HopAA1or in their endophytic populations (Fig. 7, C-E; Supplemental Fig. S6; Vinatzer et al., 2006). Thus, increased epiphytic populations of HopZ3and HopAA1strains after spray inoculation is not due to enhanced endophytic growth or release to leaf surfaces.

HopZ3 and HopAA1 Localize to Different Sites within Plant Cells
To gain insight into where and how HopZ3 and HopAA1 might act within plant cells, C-terminal fusions of HopZ3 and HopAA1 to GFP were transiently expressed and localized in N. benthamiana. HopAA1-GFP fluorescence was visible at the cell periphery (Fig. 8A). Mature plant cells have a large vacuole and thin cytoplasm, making it difficult to distinguish cytoplasmic from plasma membrane or tonoplast membrane localization. However, HopAA1-GFP is likely at the plasma membrane, since the protein was only detected in the membrane fraction in N. benthamiana (Fig. 8B) and GFP was visible closer to the cell edges than chloroplasts, which are located in Figure 6. HopAA1 and HopZ3 are important for epiphytic fitness of PsyB728a on tomato 76R. PsyB728a and mutants carrying Ptrp-gfp at an OD 600 of 0.01 were sprayed onto 17-to 21-d-old tomato 76R. All bars indicate SEs. A, After 72 h, the populations of HopZ3and HopAA1were reduced compared to PsyB728a. A x 2 test indicated that the distributions were significantly different between the deletion mutants and PsyB728a (HopAA1 -, P , 0.0001, n $ 48; HopZ3 -, P , 0.0001, n $ 48). This experiment was done together with HrcCbacteria (see Fig.  1D). Thus, the PsyB728a data are the same as that shown in Figure 1D. B, Leaf disks were washed to remove and enumerate bacteria in the attached biomass. Reductions in the mutant bacterial populations compared with PsyB728a after 72 h were statistically significant (*P , 0.05, Mann-Whitney test, n = 12). C, Endophytic bacteria population of HopAA1and HopZ3from Figure 6B was not significantly different compared with PsyB728a after 72 h (HopAA1 -, P = 0.90, HopZ3 -, P = 0.3556, n $ 12 for both). These experiments were repeated twice with similar results. the region between the tonoplast and plasma membrane (Supplemental Fig. S7). Bioinformatic analysis using two different structure prediction programs indicated that HopAA1 may contain between one and three transmembrane regions, possibly exposing a portion of the protein to the extracellular milieu.
HopZ3-GFP was localized to the nucleus and the cell periphery (Fig. 8, A and C). In this case, the cell periphery is probably the cytoplasm, since HopZ3 was previously only found to be in the soluble fraction when produced in N. benthamiana (Lewis et al., 2008). The GFP fluorescence was from the intact 75-kD fusion proteins, since there was no evidence of the GFP monomer in extracts, as determined by western-blot analysis (Fig. 8D). These results indicate that HopAA1 likely exerts its function at the membrane, whereas HopZ3 may act in the nucleus and/or cytoplasm.

Conserved Potential Catalytic Residues in HopZ3 Are Involved in Its Function
Since HopZ3 has similarity to the C55 family of Cys proteases (Lewis et al., 2008), we tested whether the potential catalytic residues C300, H243, and E263 were important for HopZ3's function. Several mutations did not affect the stability or localization of HopZ3-GFP in N. benthamiana or HopZ3-HA accumulation in PsyB728a (Figs. 8D and 9A; Supplemental Fig. S8). Therefore, we tested the HopZ3 variants in three assays in which HopZ3 showed a phenotype.
Fortuitously, we found that HopZ3 elicited mild cell death on transgenic N. benthamiana that ectopically Figure 7. Effect of the loss of HopAA1, HopZ3, and HrcC on bacterial attachment to plastic and glass substrates and endophytic growth and bacterial release onto leaf surfaces. A, The ability of PsyB728a, HopAA1 -, HopZ3 -, and HrcCto attach to a plastic microtiter plate was not different (P . 0.19, t test, n = 8). PsyB728a, HopAA1 -, HopZ3 -, and HrcCwere inoculated at an OD 600 of 0.1 in effector-inducing minimal media in a microtiter plate. After 24 h, the supernatant was removed and the attached biomass stained with 1% crystal violet was measured after dissolving the dye in ethanol. Bars indicate SEs. B, Attachment of GFP-expressing PsyB728a, HopAA1 -, HopZ3 -, and HrcCcarrying Ptrp-gfp onto glass coverslips over 3 h was not different (P . 0.055, t test, n = 27-36). The bacterial area per image was determined for at least 25 randomly taken images. C and D, Release of mesophyll-associated bacteria onto leaf surfaces after 48 h was similar between PsyB728a, HopAA1 -, and HopZ3 -, while HrcCwas reduced. Four-week-old N. benthamiana were injected with strains carrying Ptrp-gfp and surface sterilized to remove GFP-expressing bacteria. C, The fluorescence area of each strain released to the surface was quantified using epifluorescence microscopy. HopAA1and HopZ3populations were not significantly different from PsyB728a (P . 0.66, Mann-Whitney test, n $ 48), while HrcChad a significantly reduced epiphytic population (***P , 0.0001, Mann-Whitney test, n $ 48). D, Bacterial populations of HopAA1and HopZ3were not significantly different from PsyB728a (P . 0.16, n = 12), while HrcChad a significantly reduced epiphytic population, as assayed using leaf washes (***P , 0.0001, Mann-Whitney test, n = 12). E, Bacterial population of HopAA1and HopZ3when inoculated to the intercellular region were not significantly different from PsyB728a (P . 0.38), while HrcChad a significantly reduced endophytic bacteria population on days 1 and 2 (***P , 0.005, n = 8). These experiments were repeated three or more times with similar results. Bars indicate SEs. expressed the tomato R gene Pto (Supplemental Fig.  S9) or when HopZ3 and Pto were transiently coexpressed (Table II; Supplemental Table S1). This phe-notype was unexpected, but not unique. AvrB3, an unrelated PsyB728a effector, but not several other effectors, also elicited Pto-dependent cell death (Supplemental Fig. S10). Elevated defense signaling due to Pto overexpression (as was documented in tomato; Tang et al., 1999) might explain why HopZ3 and AvrB3 elicited cell death in N. benthamiana/Pto. Alternatively, Pto might have a low-affinity interaction with HopZ3 and/or AvrB3 in planta that triggers cell death due to protein overexpression. However, we found no evidence for an interaction of Pto with HopZ3 or AvrB3 (Supplemental Fig. S11). All the HA-epitope-tagged HopZ3 point mutants were expressed (Fig. 9D) and all except H243A and H243L lost the ability to cause cell death on N. benthamiana/Pto (Fig. 9E). Data from the different assays suggests that the potential catalytic residues C300, E263, and H243 of HopZ3 are important for one or more phenotypes conferred by HopZ3.

DISCUSSION
PsyB728a associated with plant leaves exhibit dynamic behaviors. The T3SS and specific effectors are important in shaping these behaviors early in the interaction when moderate relative humidity (60%-80%) is used. Spray-inoculated PsyB728a reside on leaf surfaces for 48 h prior to increasing in population size. From 24 to 48 h, a subset of bacteria expresses effectors. During this time period, a functional T3SS is required for a significant portion of the epiphytic population to survive, and HopAA1 and HopZ3 prevent an increase in the epiphytic population on N. benthamiana. In contrast, HopAA1 and HopZ3 are required for epiphytic survival on the nonhost tomato. Up to 72 h, endophytic colonization of N. benthamiana is rare and the population sizes of epiphytic and endophytic bacteria are not correlated. Furthermore, from 48 to 72 h, epiphytic PsyB728a populations can increase independently of whether bacteria significantly colonize the mesophyll region. Our data imply that there is an early and intimate interaction between bacteria and plant epidermal cells, in which the T3SS and effectors play an active role in affecting survival and/or growth in the epiphytic niche, prior to significant endophytic colonization.
Growth conditions play a critical role in the early population dynamics of PsyB728a after spray inoculation of plants. In contrast to our conditions, inoculation under high relative humidity conditions (.80%-100%) promotes immediate epiphytic growth of Psy-B728a on snap bean leaf surfaces without a lag period (Beattie and Lindow, 1994). When snap bean plants are kept at high humidity until 48 h, PsyB728a not only grows rapidly to large populations on leaf surfaces within 48 h, but bacteria also invade the mesophyll region within 24 h after spray inoculation (Quiñ ones et al., 2004). In high humidity, bacteria might easily enter into the host plant's mesophyll region through stomata, rapidly multiply, and become released onto leaf surfaces. Epiphytic communities play an important role in the transmission of pathogens, especially P. syringae, which can be carried in water droplets and may be part of the water cycle (Morris et al., 2008). More than 25 years ago, epiphytic bacterial populations of P. syringae were shown to be influenced by plant genotype in field studies (Daub and Hagedorn, 1981;Stadt and Saettler, 1981). Our study agrees with these results: PsyB728a shows larger epiphytic aggregates on N. benthamiana than are found on tomato. Furthermore, stimulating plant defenses using an SA agonist also restricts epiphytic fitness. Interestingly, epiphytic populations of Xanthamonas axonopodis on pepper (Capsicum annuum) are reduced in the field when specific effectors are deleted (Wichmann and Bergelson, 2004). However, because there is exchange between the epiphytic and endophytic populations as infections proceed, it was not clear from previous studies how and when the T3SS and specific effectors might influence bacterial populations in these two niches. Our work implicates a functional T3SS as important for early adaptation to and survival on leaf surfaces. We found no evidence that HopZ3 or HopAA1 affect bacterial release from the endophytic to the epiphytic pool. Since PsyB728a has numerous other effectors (Feil et al., 2005), one or more of these may have this role, similar to the effect of the AvrRpt2 effector, which promotes endophytic growth and release of P. syringae to susceptible Arabidopsis leaf surfaces (Guttman and Greenberg, 2001).
Direct imaging of GFP-labeled bacteria revealed the importance of the T3SS and specific effectors for the development of aggregates on leaf surfaces. On N. benthamiana, the very large aggregates of HopAA1and HopZ3strains and to some extent wild-type PsyB728a Figure 9. Conserved potential catalytic residues of HopZ3 are important for the function of HopZ3. A, Cys 300 and Glu 263 were important for HopZ3's effect on epiphytic bacteria growth. HopZ3complemented with HopZ3-HA and HA-epitope-tagged point mutants driven by the constitutive npt2 promoter were sprayed onto N. benthamiana at an OD 600 of 0.01. After 3 d, bacteria were quantified using leaf wash assays. An asterisk indicates significant differences from HopZ3 -+ vector at a given level (*P , 0.05, **P , 0.005, ***P , 0.0001, Mann-Whitney test, n = 12). HopZ3-HA and its variants were expressed in P. syringae shown by immunoblot analysis. Top blot, bacteria were grown in minimal media to mimic the plant environment. Bottom blot, bacteria were grown in KB media, similar to the conditions before spray inoculation. B, Immunoblot of extracts of N. benthamiana expressing HopZ3 and its variants or vector control cotransformed with AvrPto1-GFP. C, The ability of HopZ3 to interfere with AvrPto1elicited cell death on N. benthamiana requires C, H, and E catalytic sites. AvrPto1-GFP and HopZ3-GFP or HopZ3-GFP variants were coexpressed after A. tumefaciens-mediated transient transformation. Representative pictures of the resulting cell death (indicated by +) are shown. D, Immunoblot analysis of extracts of plants expressing HopZ3 and its variants. Note that in the bottom section, lane H243L was exposed about 10-times longer than the other samples. For sections B and D, equal amounts of protein were resolved on 12% SDS-PAGE gels. E, The ability of HopZ3-HA to elicit cell death on Pto-expressing N. benthamiana depends on E263 and C300. Representative pictures showing the effect of HopZ3-HA and its variants on N. benthamiana with and without the tomato R-gene Pto. Cell death is indicated by +. These experiments were repeated at least three times with similar results.
resemble microbial biofilms observed in aquatic environments (Morris et al., 1998). Biofilms are communities of microorganisms with an exopolysaccharide matrix that protects the constituent bacteria from environmental stress, antibiotics, and host defenses (Parsek and Fuqua, 2004;Branda et al., 2005). More work is necessary to determine whether the very-large aggregates documented here are potentially developing into biofilms. However, on N. benthamiana, PsyB728a has more single cells (not in aggregates) than HopAA1and HopZ3strains. SA and components of innate immunity from animals can prevent bacterial biofilm development of Pseudomonas aeruginosa on glass coverslips and Arabidopsis roots (Singh et al., 2002;Prithiviraj et al., 2005). Similarly, epiphytic PsyB728a aggregates, possibly biofilms, might be affected by plant defenses that are induced by HopZ3 and/or HopAA1.
The growth dynamics and effector expression seen with PsyB728a may only be germane/relevant to some strains of Pseudomonas spp., mainly those with a strong epiphytic growth phase. Like PsyB728a, a subset of epiphytic P. syringae pv phaseolicola strain NP3121/ PavrPto-gfp cells express GFP at 24 and 48 h, prior to endophytic colonization of N. benthamiana (J. Lee and J.T. Greenberg, unpublished data). Epiphytic expression of GFP in PphNP3121/PavrPto-gfp was not previously observed when expression was examined 6 h after inoculation (Xiao et al., 2004). P. syringae pv tomato DC3000, a strain without a strong epiphytic phase (Roine et al., 1998;Boureau et al., 2002), shows a different effector expression pattern and colonization strategy. Epiphytic GFP expression from PtoDC3000 carrying hrpA-gfp (hrpA is a type-III-secreted protein that forms the secretion pilus; Lee et al., 2005) only occurs late in the infection, coincident with endophytic growth and highly developed disease lesions (Boureau et al., 2002). Because PtoDC3000 uses coronatine to gain rapid access to subepidermal regions (Melotto et al., 2006), this strain may not have the need to establish large epiphytic populations.
How might HopZ3 and HopAA1 function in plant cells to affect bacterial survival and growth? These effectors promote survival and possibly growth on tomato, a nonhost plant. This may be due to the presence of virulence targets of these effectors in tomato. However, any virulence activity of HopZ3 or HopAA1 is not enough to allow PsyB728a to cause significant disease on this host. HopZ3, which has sequence similarity to the YopJ Cys protease effector family, localizes to the nucleus and cell periphery, probably in the cytoplasm. However, dual subcellular localization might result from the high expression level used in this study. At least two other members of this effector family, PopP2 from Ralstonia solanacaerum and XopD from Xanthomonas campestris pv vesicatoria also localize to plant nuclei (Deslandes et al., 2003;Hotson et al., 2003). XopD possesses Small Ubiquitinlike Modifier (SUMO) isopeptidase activity (Hotson et al., 2003). PopP2 has acetyltransferase activity (Tasset et al., 2010). We were unable to detect significant Cys protease or SUMO protease activity of HopZ3, but we did detect acetyltransferase activity in vitro (J. Lee and J.T. Greenberg, unpublished data). Mutation of two residues that align with the conserved active site, E263 and C300, results in the loss of three phenotypes conferred by HopZ3. Thus, HopZ3 may have enzymatic activity on an as-yet-unknown substrate(s) or may require plant cofactors. Additionally, a third conserved residue, H243, is only required for suppressing cell death elicited by AvrPto1. YopJ has both SUMO protease and acetyltransferase activities (Mukherjee et al., 2006). It is possible that HopZ3 has more than one biochemical activity and that H243 is only essential for one of these activities.
HopAA1 lacks sequence similarity to proteins with any reported biochemical activity. Although HopAA1 from PtoDC3000 localizes to mitochondria when expressed in yeast (Saccharomyces cerevisiae; Munkvold et al., 2008), there was no evidence for such localization in plants. In contrast, like some effectors that modulate host signaling (Maurer-Stroh and Eisenhaber, 2004), HopAA1 localizes to the plant cell periphery, probably the plasma membrane. HopAA1 and its orthologs are predicted to have one to three conserved membrane-spanning regions and at least one extracellular region. Interestingly, the Escherichia coli TIR effector localizes to the host plasma membrane and forms a receptor with a predicted extracellular loop region for E. coli to dock (Kenny et al., 1997). Whether HopAA1 has a similar specialized function due to its possible architectural similarity to TIR1 awaits further investigation.
In summary, the T3SS and individual effectors are important for PsyB728a bacteria to survive and grow on leaf surfaces. This raises the question as to how the T3SS functions across the plant cuticle. Several different bacterial epiphytes can alter leaf surface properties and affect cuticle permeability (Schreiber et al., 2005). We hypothesize that Psy728a can also modify cuticle properties to allow access of the T3SS secretion needle to the plant epidermal cell membrane. Alteration of the cuticle is consistent with the induction of the effector reporter expressed in epiphytic PsyB728a, since infection of Arabidopsis with reduced cutin, a cuticle component, results in strong induction of this reporter in another P. syringae strain (Xiao et al., 2004). Further characterization of how PsyB728a adapts to and alters leaf surfaces will be informative to understanding early events in the infection process.

Bacterial Strains and Growth Conditions
The bacterial strain PsyB728a and its previously characterized derivatives deleted for HopAA1 or HopZ3 or harboring an insertion in T3SS component HrcC Vinatzer et al., 2006) were grown on solid agar and liquid King's broth (KB;King et al., 1954) with 34 mg/mL rifampicin at 30°C. Pseudomonas syringae strains were tagged with GFP using the Ptrp-gfp plasmid (to achieve constitutive GFP; Miller et al., 2000) or the PavrPto-gfp plasmid (to monitor effector gene expression; Xiao et al., 2004) by electroporation using 2.5 kV, 500 , and 10 mF. GFP-tagged strains were grown using 50 mg/mL kanamycin.

Plant Genotypes and Growth Conditions
Nicotiana benthamiana and tomato (Solanum lycopersicom '76R') were grown in a walk-in growth chamber kept at 24°C, 60% to 80% relative humidity, and with 16-h day light. Plants were grown for 3 weeks for P. syringae infection assays and for 4 weeks for expression of proteins with A. tumefaciens-mediated transient transformation.
Transgenic N. benthamiana containing the tomato R gene Pto was developed using a similar method to that described previously (Thilmony et al., 1995).

Epiphytic Bacteria Population Assays
Overnight cultures of GFP-tagged PsyB728a and its derivatives were diluted 1:10 and grown for 2 h at 30°C. To make the inocula, cells were collected by centrifugation and resuspended in 10 mM MgSO 4 to an optical density at 600 nm (OD 600 ) of 0.01. Inocula were sprayed with spray bottles onto the tops of the first through fourth leaves of N. benthamiana plants (17-21 d old). Plants were sprayed to run-off to completely soak the leaves (about 1 mL per plant), covered with a dome with holes, and maintained at a relative humidity of 60% to 80%.
Epiphytic bacteria were quantified from leaf discs from independent plants using leaf wash (8-mm diameter discs, n = 10-12) and/or microscopy assays (12-mm diameter discs, n = 6-8 with at least three random images per disc analyzed). For the wash assay, each leaf disc was placed into 1 mL 10 mM MgSO 4 and vortexed for 5 s at maximum speed to remove the attached biomass. Samples were serially diluted and enumerated by viable counts after plating on KB agar. Sonication was also evaluated as a method for harvesting epiphytic bacteria, but this approach was less effective than vortexing for harvesting aggregates .10 mm 2 . For the microscopy assay, each leaf disc placed onto a microscope slide adaxial side up was covered with 1% agar and a coverslip was placed on the top (Monier and Lindow, 2003). Samples were then viewed with a Zeiss Axioskop epifluorescence microscope (Carl Zeiss) or Zeiss 510 laser-scanning confocal microscope using 4003 or 6303 magnification, respectively. For the epifluorescence microscopy, GFP was visualized using a 450 to 490 nm excitation, 515 nm emission filter (fluorescein isothiocyanate filter set). Images were captured with an AxioCAM HRC video camera (Zeiss) and data were processed with Axio vision Rel.4.6 software. For the confocal microscopy, GFP fluorescence was visualized by excitation using a single line 488-nm laser, and emission was collected with filter set to a 505 to 550 nm bandpass. The chlorophyll autofluorescence was excited with a single line 543 laser, and emission were collected at wavelengths .615 nm. Horizontal (x-y) images were taken at 0.5-mm intervals. Vertical cross-section (x-z) images were generated with a laser scanning microscope image browser.
Dead epiphytic bacteria were stained by placing leaf discs adaxial side down in 200 mL of 330 mg/mL propidium iodide (Invitrogen) in phosphatebuffered saline (PBS) for 15 min in the dark and washing once in 200 mL PBS prior to embedding in 1% agar. Stained bacteria were visualized by epifluor-escence microscopy using excitation at 515 to 560 nm and emission at 590 nm (Rhodamine filter set).
To perform quantitative analysis of GFP-tagged bacteria, random micrographs along transect of the leaves were taken as representative images of the attached bacterial populations. The area of green fluorescence was used to quantify the amount bacteria present on leaf surfaces. While there was some green autofluorescence from the leaves, this was diffuse and not as bright as the bacteria. The autofluorescence dropped out of the green-fluorescing area during image analysis because it was below the threshold value used.

Analysis of Micrographs
Bacterial area was quantified from epifluorescence micrographs using ImageJ available online from the National Institutes of Health. Images were thresholded using a maximum entropy threshold function developed by Jerek Sacha that is available on the ImageJ Web site. The area of the pixels that fluoresced green above the threshold amount were quantified using the analyze particles function. An example of an image used in the analysis is shown in Supplemental Figure S12.

BTH Treatment of Plants
N. benthamiana that was 17 to 21 d old was sprayed with 100 mM BTH (a kind gift from Robert Dietrich, Syngenta, Research Triangle Park, NC) dissolved in water. After 24 h, plants were sprayed with PsyB728a and sampled as described above.

Release of Endophytic Bacteria onto Leaf Surfaces
Leaves of 4-week-old N. benthamiana were sterilized by spraying with 70% ethanol and then infiltrated with PsyB728a and its derivatives at an OD 600 of 0.00001 using a needleless syringe. The leaves were wiped with 70% ethanol again to remove any surface bacteria and examined to ensure that GFPmarked bacteria were absent from the surface. Release of bacteria to leaf surfaces was monitored by leaf washes and epifluorescence microscopy (see above).

Attachment Assays
The ability of PsyB728a and its derivatives to attach to surfaces was assayed by examining attachment to glass and plastic substrates. For the plastic surface, 400 mL of an 0.1 OD 600 culture in minimal media (Mudgett and Staskawicz, 1999) was added to wells of a 48-well tissue culture plate (Corning Life Sciences). Plates were incubated for 24 h at 20°C with shaking. The supernatant was removed and the attached biomass was washed three times with PBS. Attached biomass was then stained with 1% crystal violet for 20 min at room temperature and then washed another three times with PBS. The crystal violet was resuspended by dissolving it in 95% ethanol for 2 min at room temperature with shaking and then quantified by measuring the A 600 .
For the glass surface assay, P. syringae strains were grown overnight in minimal media (Mudgett and Staskawicz, 1999) at 20°C. The coverslips were immersed in 1 mL of 0.1 OD 600 culture cell and allowed to attach to the coverslips for 3 h at 20°C. Then the coverslips were washed three times in minimal media to remove unattached bacteria. The coverslips were then transferred to microscope slides and viewed with epifluorescence microscopy.

Construction of Point Mutants in HopZ3
Single amino acid changes in the catalytic domains of HopZ3 were made using the Stratagene quick change kit. Point mutants were created from HopZ3 that was cloned into the pDONR207 Gateway donor vector (Invitrogen; Vinatzer et al., 2006).Cys 300 was changed to Trp (C300W) using the forward primer CAATACTCGAAAACTGACTGGACCATGTTTGCGC and reverse primer GCGCAAACATGGTCCAGTCAGTTTTCGAGTATTG, both 5# to 3#. The catalytic His at 243AA was changed to Leu (H243L) using the forward primer CAAAAAGATGCCAGACCTCCATATAGCCTTGGATATC and reverse primer GATATCCAAGGCTATATGGAGGTCTGGCATCTTTTTG, both 5# to 3#. Glu 263 was changed to Ala (E263A) using the forward primer GATTGTCGGCTTT GCGTCTGCCCCTGGGAAC and reverse primer GTTCCCAGGGGCAGACGC AAAGCCGACAATC, both 5# to 3#. Point mutants were constructed using the specification in the manufacturer's manual and using 16 cycles of amplification.
In addition, point mutations were also made in HopZ3 in pDONR207 using the GeneTailor kit (Invitrogen). Cys 300 was changed to Ala (C300A) using the forward primer TCAATACTCGAAAACTGACGCCACCATGTTTGC and the reverse primer GTCAGTTTTCGAGTATTGAAGAAAATTTCCTACC, both 5# to 3#. His 243 was changed to Ala (H243A) using the forward primer TTCAAAAAGATGC-CAGACGCCCATATAGCCTTG and the reverse primer GTCTGGCATCTTTTT-GAATGGGGGGTAAAC. HopZ3 point mutants were sequence validated and transferred into Gateway vectors (pBAV154, pBAV150, and pBAV226) for expression in A. tumefaciens and P. syringae following the method described in the Gateway manual (Invitrogen). HopZ3 point mutants were transferred into two vectors for expression in A. tumefaciens, pBAV154 and pBAV150. pBAV154 is a dexamethasone-inducible vector with a C-terminal HA tag, while pBAV150 is a dexamethasone-inducible vector with a C-terminal GFP tag (Vinatzer et al., 2006). Electroporation of plasmids into A. tumefaciens C58C1 was achieved using 2.5 kV, 400 , and 25 mF. For expression in P. syringae, HopZ3 point mutants were transferred into pBAV226, a low-copy-number expression vector that has an nptII promoter (Vinatzer et al., 2006) and a C-terminal HA tag. Plasmids were electroporated into PsyB728a as described above.

Transient Expression Assays, Western Analysis, and Membrane Fractionation
A. tumefaciens C58C1 containing effectors in pBAV154 or pBAV150 was grown in liquid culture in Luria-Bertani broth with kanamycin overnight at 30°C. One milliliter of overnight culture was added to 3 mL induction medium (Vinatzer et al., 2006) containing 50 mg/mL acetosyringone and placed at 30°C to induce for 5 h. An infiltration inoculum was made by collecting induced cells by centrifugation and resuspending them in infiltration medium (Vinatzer et al., 2006) at an OD 600 of 0.4. Inocula were infiltrated into 4-week-old N. benthamiana with a needleless syringe. For HopAA1 expression, samples for immunoblot analysis were taken at 24 h and cell death was scored visually after 48 h. For all other experiments involving cell death, plants were treated with 30 mM dexamethasone and 0.1% Tween 20 in water (dex) 2 d after A. tumefaciens infiltration. Cell death was scored by visual inspection on d 4 after dex treatments. For immunoblots, samples were taken 12 h after dex treatments.
For immunoblot studies, two 8-mm leaf disks were ground to a uniform consistency in 50 mL of resuspension buffer (Vinatzer et al., 2006). Cellular debris was removed by centrifugation, and supernatant was mixed with SDS sample buffer (Lewis et al., 2008). Equal amounts of extract for each sample were separated by eletrophoresis and were transferred to a polyvinylidene fluoride membrane (Millipore Corporation). Membranes were probed with a primary, anti-HA monoclonal antibody (Covance) at a dilution of 1:5,000, followed by a secondary, anti-mouse HRP (Pierce) at a 1:5,000 dilution.
Membrane fractionation was performed as described (Lewis et al., 2008). GFP monoclonal antibody (Covance) was used at a 1:5,000 dilution and secondary anti-rabbit antibody (Pierce) at a 1:5,000 dilution. Visualization of immunoreaction was achieved by ECL (Pierce).
For microscopy analysis, nuclei were counterstained with 10 mg/mL Hoechst 33342 solution in water (Invitrogen). Fluorescence was analyzed with UV (365nm excitation, 420-nm emission) or fluorescein isothiocyanate filter sets.

Yeast Interaction Tests
Possible interactions between Pto and HopZ3 or AvrB3 were tested using yeast (Saccharomyces cerevisiae) two-hybrid assays using the LEXA system as described .

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Examples of GFP-labeled microcolonies associated with substomatal chambers or mesophyll cells and bacteria being released from the endophytic pool to leaf surfaces.
Supplemental Figure S2. Comparison of bacterial populations of PsyB728a carrying Ptrp-gfp and PavrPto-gfp inoculated on N. benthamiana leaf surface.
Supplemental Figure S3. Lack of evidence for epidermal cell death at 48 h after spray inoculation with PsyB728a.
Supplemental Figure S4. Examples of the association of GFP-expressing bacteria with different epidermal cell types.
Supplemental Figure S5. Enumeration of epiphytic and endophytic bacterial populations after spray inoculation show that epiphytic but not endophytic bacterial populations of HopAA1 -, HopZ3 -, and HrcCare different from PsyB728a at 48 h.
Supplemental Figure S6. Fluorescence micrographs showing the effects of HopAA1 -, HopZ3 -, and HrcCmutations on the release of endophytic bacteria onto N. benthamiana leaf surfaces.
Supplemental Figure S8. Localization of variants of HopZ3-GFP expressed in N. benthamiana.
Supplemental Figure S9. Validation that Pto is expressed in transgenic N. benthamiana harboring Pto.
Supplemental Figure S11. Lack of interaction of Pto with AvrB3 or HopZ3 in yeast.
Supplemental Figure S12. Example of image processing to visualize GFPtagged bacteria.
Supplemental Table S1. Cell death phenotype elicited by HopZ3 (or HopZ3 variants) and Pto after coexpression on N. benthamiana.