Subcellular localization and functional analysis of the Arabidopsis GTPase RabE

Membrane trafficking plays a fundamental role in eukaryotic cell biology. Of the numerous known or predicted protein components of the plant cell trafficking system, only a relatively small subset has been characterized with respect to their biological role in plant growth, development, and response to stresses. In this study, we investigated the subcellular localization and function of an Arabidopsis thaliana small GTPase belonging to the RabE family. RabE proteins are phylogenetically related to well-characterized regulators of polarized vesicle transport from the Golgi apparatus to the plasma membrane in animal and yeast cells. The RabE family of GTPases has also been proposed to be a putative host target of AvrPto, an effector protein produced by the plant pathogen Pseudomonas syringae , based on yeast two-hybrid analysis. We generated transgenic Arabidopsis plants that constitutively expressed one of the five RabE proteins (RabE1d) fused to GFP. GFP-RabE1d and endogenous RabE proteins were found to be associated with the Golgi apparatus and the plasma membrane in Arabidopsis leaf cells. RabE down-regulation, due to cosuppression in transgenic plants, resulted in drastically altered leaf morphology and reduced plant size, providing experimental evidence for an important role of RabE GTPases in regulating plant growth. RabE down-regulation did not affect plant susceptibility to pathogenic P. syringae bacteria; conversely, expression of the constitutively active RabE1d-Q74L enhanced plant defenses, conferring resistance to P. syringae infection.


INTRODUCTION
Eukaryotic cells are compartmentalized by membranes that surround organelles having specific functions. Communication and transport between these membrane-bound compartments are vital to the cell and are accomplished through complex and tightly regulated pathways.
Trafficking pathways and their players have been extensively described in yeast and mammalian cells, but are still poorly characterized in plants (Lippincott-Schwartz et al., 2000;Jurgens, 2004). The main effectors and regulators of these pathways appear to be shared between all eukaryotes. Among these, Rabs are a group of small monomeric GTPases that act as molecular switches to mediate vesicle transport between membrane-bound cellular compartments (Segev, 2001). Rab GTPases participate in vesicle budding from a donor compartment, transport along the cytoskeleton toward a target compartment, and, eventually, tethering and fusion of the vesicles with the target membrane. Like other small GTPases, Rabs alternate between a GTPbound active form and a GDP-bound inactive form. Their functional specificity is determined, in part, by their unique subcellular distribution (Stenmark and Olkkonen, 2001;Zerial and McBride, 2001). The A. thaliana genome encodes 57 Rab proteins, divided into eight subfamilies (RabA to RabH) based on sequence similarities (Rutherford and Moore, 2002;Vernoud et al., 2003). Recent studies have indicated that members of various plant Rab subfamilies participate in secretion and recycling of cell wall components (Nielsen et al., 2008), apical growth of pollen tubes (de Graaf et al., 2005;Cheung and de Vries, 2008;Cheung and Wu, 2008) and root hair (Preuss et al., 2004;Preuss et al., 2006), formation of the phragmoplast in dividing cells (Chow et al., 2008), endocytosis (Ueda et al., 2004), and potentially osmotic stress tolerance (Mazel et al., 2004).
Fungal and bacterial infections in plants are often associated with activation (or suppression) of extracellular defense responses, including secretion of defense-related proteins and antimicrobial phytoalexins into the apoplast, and formation of callose-rich cell wall appositions, known as papillae (Snyder and Nicholson, 1990;Snyder et al., 1991;Brown et al., 7 RabE family of GTPases (Rutherford and Moore, 2002;Vernoud et al., 2003;Supplemental Fig. S1). As described in the Results section, we independently isolated a RabE GTPase in a Y2H screen for AvrPto-interacting Arabidopsis proteins.
The identification of Rab GTPases as AvrPto interactors in both tomato and Arabidopsis suggested that, as part of its virulence mechanism, this effector may perturb intracellular vesicle trafficking in the plant (Bogdanove and Martin, 2000). Interestingly, small GTPases regulating cytoskeleton dynamics and membrane trafficking are among the most common host targets of TTSS effectors produced by bacterial pathogens of animals (Harrison et al., 2004;Machner and Isberg, 2006;Murata et al., 2006;Rzomp et al., 2006;Smith et al., 2007). The biological role of tomato Api2 and Api3 and Arabidopsis RabE GTPases in plant development and defense is unknown. Localization and function of RabE proteins have only recently begun to be investigated. Arabidopsis yellow fluorescent protein (YFP)-RabE1d, transiently expressed in heterologous tobacco epidermal cells, was detected in the Golgi apparatus and in the cytoplasm, and functional analyses indicated that this protein participates in promoting post-Golgi secretory trafficking (Zheng et al., 2005). To gain further insights into the RabE GTPases subcellular localization in native Arabidopsis cells and their function in growth, development, and/or defense at the whole plant level, we generated transgenic plants constitutively expressing GFPfused wild-type RabE1d and RabE1d-Q74L (predicted to be preferentially GTP-bound, and therefore in an active state) and examined these plants for morphological and developmental phenotypes, and response to P. syringae infection. 8

Identification of Arabidopsis RabE proteins as Y2H interactors of AvrPto
We conducted a Y2H screening of two separate Arabidopsis cDNA libraries, using AvrPto as bait. Several AvrPto-interacting Arabidopsis proteins were identified, including a member of the RabE family of small GTPases (At5g59840), a putative cytoplasmic kinase (At4g11890), an auxin signaling repressor, IAA7 (At3g23050), two hypothetical proteins (At3g26600, At5g16840), and several putatively chloroplast-or mitochondria-targeted proteins.
The interaction with the small GTPase RabE was particularly interesting because RabE was predicted to be membrane-localized, as is AvrPto (Shan et al., 2000;He et al., 2006), and, as mentioned above, because screening of a tomato cDNA library had also yielded two AvrPtointeracting proteins homologous to Rab8 (Bogdanove and Martin, 2000), the mammalian counterpart of Arabidopsis RabE.
To characterize the specificity of the AvrPto-RabE interaction, we investigated whether AvrPto interacts with other members of the RabE family and with other Rab proteins. Of the other Arabidopsis Rab protein families (i.e., RabA-D, F, G), we cloned and expressed in yeast RabA1a, B1b, C1, D2a, F2a, and G3a; none of these representatives interacted with AvrPto in the Y2H system (Fig. 1A). Of the five RabE genes, all but RabE1c were successfully cloned and expressed in yeast. All four RabE proteins tested (RabE1a, b, d, and e) interacted with AvrPto ( Fig. 1B). Thus, it appears that AvrPto interacts specifically with the Arabidopsis RabE family of GTPases.
Small GTPases normally cycle between the active GTP-bound, and the inactive GDPbound states. Mutation of highly conserved residues can be used to alter the nucleotide binding and hydrolysis activities of Rab proteins (Nielsen et al., 2008) (Supplemental Fig. S1). The mutant forms RabE1d-S29N (predicted to be GDP-bound) and RabE1d-Q74L (predicted to be GTP-bound) were engineered. In the Y2H system, AvrPto interacted only with wild-type RabE1d or with GTP-restricted RabE1d-Q74L, but not with GDP-restricted RabE1d-S29N (Fig.   1C), suggesting that AvrPto preferably binds to the active form.

RabE gene expression in Arabidopsis leaves
Analysis of the Arabidopsis genome showed that 44 of the 57 Rab genes reside in duplicated regions. Of the five RabE genes, RabE1d and E1e appear to be derived from a major duplication event between chromosomes III and V, and the same holds true for RabE1b and E1c (Rutherford and Moore, 2002) ( Fig. 2A). The high degree of sequence identity among the five RabE proteins, equal to or higher than 86%, suggests functional redundancy. However, members of a gene family could be preferentially expressed in different tissues, at specific developmental stages, or in response to stresses. We used the Genevestigator Gene Atlas Tool (Zimmermann et al., 2004) to investigate whether this is the case with the five RabE genes. The in silico analysis revealed that all five genes are expressed in all Arabidopsis tissues and developmental stages.
RabE1d and E1e (encoding 94% identical proteins) are the only two gene family members whose expression is much lower in pollen than in all other tissues. Based on publicly available microarray data (Genevestigator;Zimmermann et al., 2004), RabE1d appeared to be the most highly expressed RabE gene in Arabidopsis Col-0 rosette leaves, followed by RabE1c. The RabE1a, E1e and E1b genes had the lowest expression levels (Fig. 2B). We conducted reverse transcription (RT)-polymerase chain reaction (PCR) analysis on rosette leaves total RNA, and confirmed a high expression level for RabE1d (Fig. 2C).
To investigate potential up-or down-regulation of the RabE family members in response to pathogens or other stresses, we analyzed the expression patterns of the RabE genes with the AtGenExpress Visualization Tool (http://jsp.weigelworld.org/expviz/expviz.jsp) (Schmid et al., 2005), across several publicly available microarray datasets. None of the five genes appeared to be significantly (more than 2.5-fold) up-or down-regulated in response to pathogens or elicitors.
In the absence of any expression-based indication on whether the five RabE proteins have redundant or distinct functions, we chose the highly expressed RabE1d (At5g03520) for further study.
RabE1d is associated with not only the Golgi apparatus, but also the plasma membrane in

Arabidopsis leaf cells
All Rab proteins are normally present in cells in two pools, one cytoplasmic, and the other membrane-associated (Novick and Brennwald, 1993). Nucleotide-binding state and interaction with accessory proteins determine whether a Rab is in the cytosol or the membrane, at  any given time. A hallmark feature of Rab proteins is that they localize to the specific membrane compartments in which they function. It was previously reported that Arabidopsis RabE1d, when transiently and heterologously expressed in tobacco epidermal cells as a fusion with YFP, was detected in the Golgi apparatus and in the cytoplasm (Zheng et al., 2005). To determine RabE localization in native Arabidopsis cells, we created stable transgenic plants that express RabE1d fused with enhanced green fluorescent protein (GFP), under the control of the CaMV 35S promoter. By convention, GFP was fused to the N-terminus of RabE1d to preserve the Cterminal CAAX geranylgeranylation site that is critical for membrane association and function (Calero et al., 2003) (Supplemental Fig. S1). Several independent transgenic lines were analyzed by confocal laser scanning microscopy (CLSM). GFP fluorescence was observed not only in intracellular punctate structures consistent with the Golgi apparatus, as detected in heterologous tobacco cells (Zheng et al., 2005), but also prominently at the cell periphery ( Fig. 3A-E).
Leaf epidermal cells typically contain a very large vacuole that accounts for most of the cell volume. Fluorescence detected at the cell periphery may represent the plasma membrane (PM), the vacuolar membrane (tonoplast), or the thin layer of cytoplasm that is between the PM and the tonoplast. To more precisely determine whether GFP-RabE1d was also localized at the PM, we stained live leaf tissue with the lipophylic dye FM4-64 (Fischer-Parton et al., 2000;Bolte et al., 2004). FM4-64 produces a bright red fluorescence in membranes but not in aqueous solutions and is rapidly incorporated into the PM. It is often used in microscopy as an endocytic tracker, because it is retained in the portions of PM that are internalized by endocytosis (Ueda et al., 2001). Within the first 15 to 30 minutes after incorporation (depending on the system used), FM4-64 selectively stains the PM. CLSM analysis revealed a precise overlap of GFP-RabE1d fluorescence with FM4-64 fluorescence immediately after staining, indicating that RabE1d is associated with the plasma membrane (Fig. 3D).
To investigate whether the punctate structures labeled by GFP-RabE1d correspond to the Golgi apparatus, we examined co-localization with rat sialyl transferase, a Golgi marker protein (Wee et al., 1998) fused to DsRed (ST-RFP). ST-RFP was transiently expressed in the GFP-RabE1d transgenic leaves via particle bombardment. Observation of cells co-expressing GFP-RabE1d and ST-RFP revealed overlapping fluorescence signals, confirming RabE1d association with the Golgi apparatus (Fig. 3E). Taken together, our analysis of GFP-RabE1d localization revealed that, in the native Arabidopsis leaf cells, membrane-associated RabE1d is found not only in the Golgi apparatus, as previously reported in heterologous tobacco cells, but also in the PM.

Endogenous RabE proteins co-fractionate with PM and Golgi markers
GFP-RabE1d expression in the transgenic plants was driven by the CaMV 35S constitutive promoter. To exclude the possibility that the observed RabE localization reflects patterns of only the transgenically expressed protein, we conducted additional analyses to determine the localization of endogenous RabE in transgenic as well as wild-type Arabidopsis plants. We performed subcellular membrane fractionation by centrifugation of clarified plant extracts on sucrose step gradients (Zeng and Keegstra, 2008). This method allowed separation of the PM from fractions containing lighter membranes (Golgi, tonoplast). Western blot analysis with an anti-RabE polyclonal antibody that reacts with all RabE family members revealed that the bulk of membrane-associated endogenous RabE proteins, as well as transgenically expressed GFP-RabE1d, were in the same fraction as the PM marker H + -ATPase (Morsomme et al., 1998).
A lower amount of endogenous RabE, as well as GFP-RabE1d, co-fractionated with xylosyl transferase (XT1), a Golgi apparatus resident protein (Faik et al., 2002;Cavalier and Keegstra, 2006;Zeng and Keegstra, 2008) (Fig. 3F and G). A tonoplast marker, gamma-tonoplast integral protein (γTIP), was found predominantly in the same fraction as the Golgi marker. Independent membrane fractionation experiments yielded similar results, although minor differences in protein levels and distribution among fractions were occasionally observed across western blots, as a result of inter-experiment variability. Overall, the membrane fractionation experiments complemented the live-cell imaging results; together, they indicate that endogenous and ectopically expressed RabE proteins are not only localized at the Golgi apparatus, but the majority of membrane-associated RabE is in the PM in Arabidopsis leaf cells.

GTP-restricted RabE1d-Q74L displays altered subcellular localization
We also produced transgenic plants expressing RabE1d-Q74L, which is predicted to be constitutively active, as a GFP fusion. Unlike wild-type GFP-RabE1d, the GFP-RabE1d-Q74L fusion was not detected in intracellular punctate structures (i.e., Golgi), but was primarily found at the cell periphery (Fig. 4A). PM staining with FM4-64 revealed that the bulk of GFP-RabE1d-12 Q74L fluorescence did not overlap with the PM, but was most likely localized in the tonoplast ( Fig. 4B). We obtained from ABRC (Arabidopsis Biological Resource Center, Ohio State University) Arabidopsis transgenic lines expressing various endomembrane markers fused to GFP (Cutler et al., 2000). Comparison of GFP-RabE1d-Q74L fluorescence to that of either a PM-localized channel protein (line Q8) or a tonoplast marker (line Q5) revealed a localization pattern similar to that of the tonoplast, both in intact and plasmolyzed leaf tissues (Supplemental

Challenge with P. syringae bacteria promotes focal accumulation of GFP-RabE1d
To investigate whether intracellular distribution of the RabE1d protein was perturbed in the presence of bacteria, we performed microscopic observation of transgenic GFP-RabE1d plants challenged with various strains of Pst DC3000. Leaves were syringe-inoculated with bacteria at 1x10 8 CFUs/ml, and analyzed by CLSM 5 to 6 hours post-inoculation. We observed that bacterial inoculation caused polarized GFP-RabE1d accumulation in mesophyll cells (Fig.   6A). Interestingly, whereas such a phenomenon was restricted to a few sparse cells in leaves challenged with wild-type virulent Pst DC3000 (Fig. 6D) or with the non-pathogenic TTSSdefective mutant hrpA - (Fig. 6C), focal accumulation of GFP-RabE1d was widespread in leaves inoculated with the avirulent strain Pst DC3000 (avrRpt2) (Fig. 6E). These results suggest that the GFP-RabE1d subcellular localization is dynamic and can respond to bacterial infection, especially in a gene-for-gene interaction.

13
plants invariably correlated with a distinct morphological phenotype ( Fig. 7B and C). Rosette leaves developed normally for the first 3 weeks, when Arabidopsis development is usually slower, and plants were indistinguishable from Col-0. In the following two weeks, when Arabidopsis size increases more rapidly, the leaves of RabE-cosuppressed plants did not fully elongate (Fig. 7D); midribs remained short, while the leaf lamina continued to expand, producing a characteristic wavy phenotype. Mature (6-to 7-week-old) RabE-cosuppressed plants were significantly smaller than wild-type and had short midribs and petioles. RabE-cosuppressed plants flowered at the same time as wild-type Arabidopsis, although their stems were much shorter than wild-type, and the plants produced fertile seeds. The progeny of selected silenced lines (B8, B11, and B13) maintained silencing and had the same phenotype as the parental plants.

RabE-cosuppressed plants are not compromised in resistance to Pst DC3000
We performed pathogenesis assays with Pst DC3000 to establish whether the partially RabE-cosuppressed plants are impaired in their defenses against P. syringae. Pst DC3000 consistently caused disease symptoms and multiplied on RabE-cosuppressed plants, reaching similar population levels as on wild-type Arabidopsis (Supplemental Fig. S2). In several instances, however, we observed that older (6-to 7-week-old) or environmentally stressed RabE- 14 cosuppressed plants displayed some degree of basal resistance, possibly due to stress caused by RabE down-regulation (Supplemental Fig. S2). We did not observe a consistent and reproducible difference between wild-type Col-0 plants and RabE-cosuppressed plants in the multiplication of Pst DC3000 (avrRpt2), or a TTSS-defective mutant of Pst DC3000, the hrpAmutant (Supplemental Fig. S2).

Transgenic expression of RabE1d-Q74L confers resistance against Pst DC3000
GFP-RabE-Q74L plants did not exhibit significant alterations in growth and development, other than appearance of minute sparse indentations in mature rosette leaves, about two weeks prior to bolting. The origin and significance of these indentations is unclear; however, we have been unable to show, using Trypan Blue staining, that they develop as a consequence of cell death.
Although transgenic expression of constitutively active GFP-RabE1d-Q74L did not globally affect plant growth or development (as did RabE cosuppression), it had a remarkable effect on plant responses to P. syringae infection. Upon challenge with Pst DC3000, the GFP-RabE1d-Q74L-expressing plants displayed a considerable degree of resistance, reflected by bacterial multiplication being consistently restricted 10-to 100-fold, compared to multiplication on wild-type Arabidopsis (Fig. 8A). This observation was consistent across several experiments on different transgenic lines. Visible disease symptoms, namely chlorosis and necrosis, were also markedly reduced (Fig. 8B). This enhanced resistance phenotype apparently requires the constitutively active form of RabE1d, as transgenic expression of GFP-RabE1d did not result in enhanced resistance (Supplemental Fig. S2). days later, total protein secretion in the apoplast and secretion of the extracellular defense protein PR1 (Pathogenesis-Related protein 1) were monitored. Intercellular wash fluid (IWF) collected from water-treated GFP-RabE1d-Q74L-expressing plants contained PR1 and several unknown proteins that were absent from the water-treated Arabidopsis wild-type IWF, indicating a constitutive activation of secretory and defense pathways. BTH application resulted in similar levels of secreted PR1 and several other proteins in the apoplast of both wild-type and transgenic plants (Fig. 8C). Interestingly, some protein bands were exclusively detected in the IWF of water-and BTH-treated RabE1d-Q74L-expressing plants, but not in the BTH-treated wild-type plant IWF (indicated by arrowheads in Fig. 8C). These unique extracellular proteins associated with expression of RabE1d-Q74L suggest that other secretory pathways, in addition to the SAR pathway, are activated in these plants.

DISCUSSION
Plant RabE GTPases are predicted to be involved in mediating secretory vesicle traffic from the Golgi to the PM. They could therefore play fundamental roles in secretion of extracellular matrix molecules, thereby influencing plant growth and development, as well as in secretion of extracellular defense molecules in response to pathogen infections. RabE proteins from both tomato (Bogdanove and Martin, 2000) and Arabidopsis (this study) interact with a pathogen effector protein, AvrPto, in Y2H assays. Because a number of studies have implicated vesicle trafficking in pathogen defenses (Collins et al., 2003;Assaad et al., 2004;Wang et al., 2005;Nomura et al., 2006;Kalde et al., 2007) Arabidopsis size increases more rapidly, in the 4 th and 5 th weeks, the leaves of RabEcosuppressed plants were not able to fully elongate; leaf midribs and petioles remained short, yielding characteristically wavy leaves. Inflorescences emerged at the same time as in wild-type, but the overall stature of the RabE-cosuppressed plants remained much shorter. These observations lead us to suggest that RabE-mediated vesicle traffic from the Golgi to the PM is required for rapid elongation of certain tissues (e.g., leaf midribs and stems) associated with rapid plant growth. The dwarf phenotype of RabE-cosuppressed plants was likely caused by simultaneous down-regulation of more than one RabE gene, as suggested by reduced expression of RabE1d and E1e and, to a lesser extent, RabE1a, E1b, and E1c (Fig. 6E) defect in growth and development (data not shown). Notably, cosuppression did not affect the expression of RabD genes, the closest homologues of RabE genes and, therefore, the observed phenotypes can be ascribed to partial silencing of multiple RabE gene family members.
Besides our interest in determining the subcellular localization of RabE GTPases in native Arabidopsis cells and a possible role of RabE GTPases in growth and development at the whole plant level, another major motivation for this work was to investigate a potential role of RabE GTPases in plant defense. Bogdanove and Martin (2000) first identified RabE GTPases as AvrPto interactors in tomato and discussed the possibility that these GTPases could be virulence targets of AvrPto. In an independent Y2H screen for AvrPto-interacting Arabidopsis proteins, we also isolated a RabE GTPase. We found that AvrPto interacted with all tested members of the RabE family, but not with members of other Rab families of Arabidopsis (Fig. 1). Moreover, AvrPto interacted with a RabE mutant (RabE1d-Q74L) predicted to be GTP-bound, but not with a RabE mutant (RabE1d-S29N) predicted to be GDP-bound. Finally, we found that the majority of membrane-associated RabE1d is localized at the PM, where AvrPto is known to be located in tomato and Arabidopsis cells (Shan et al., 2000;He et al., 2006). Despite the intriguing specificity in Y2H assays and the apparent co-localization of AvrPto and RabE1d to the same host membrane, we have been unable to detect AvrPto-RabE1d interaction using in vivo co- 18 v-SNAREs) at the site of pathogen penetration has also been observed in fungal infections (Assaad et al., 2004;Huckelhoven, 2007;Kwon et al., 2008).
RabE cosuppression in Arabidopsis, under the conditions reported in this study, did not result in increased susceptibility to P. syringae bacterial strains. RabE proteins, therefore, may not be required for establishing defenses against this pathogen. Alternatively, partial downregulation of RabE proteins may be insufficient to confer a discernable defense phenotype. A complete knockout of all five RabE genes would be necessary to resolve this question. However, the dwarf phenotype of partially RabE-cosuppressed plants suggests it is unlikely that completely

RabE-deficient plants would be viable and/or suitable for bacterial infection assays.
Transgenic expression of the RabE1d-Q74L variant, on the other hand, conferred in Arabidopsis a significant degree of resistance to Pst DC3000. It remains to be determined whether this resistance is caused by a direct effect of RabE1d-Q74L, due to enhancement of defense-related vesicle traffic, or rather is triggered by an indirect effect, due to overall perturbation of cellular vesicle traffic. We found that the IWF collected from water-treated GFP-RabE1d-Q74L plants contained PR1 and several unknown proteins that were absent from the IWF of water-treated wild-type Arabidopsis, indicating constitutive activation of secretory and defense pathways in these plants. BTH application resulted in similar levels of secreted PR1 and other proteins in the apoplast in both wild-type and RabE1d-Q74L transgenic plants (Fig. 8C).
However, some of these extracellular proteins are detected only in the IWF of water-and BTHtreated RabE1d-Q74L-expressing plants, but not in the IWF of BTH-treated wild-type plants.
These unique extracellular proteins associated with expression of RabE1d-Q74L suggest activation of other secretory pathways in these plants in addition to the SAR pathway.
Interestingly, activation of SAR and resistance to Pst DC3000 in the GFP-RabE1d-Q74L transgenic plants did not correlate with a dwarf phenotype, a common phenotype of Arabidopsis mutants that are constitutively resistant to pathogens (Lorrain et al., 2003). Further analysis of the RabE-Q74L transgenic plants may provide novel insights in plant defense, and indicate possible avenues for engineering resistance in crops. We identified Arabidopsis proteins that interacted with AvrPto of Pst DC3000 by using the Matchmaker LexA-based yeast two-hybrid system (Clontech Laboratories Inc., Palo Alto, CA). Two Arabidopsis cDNA libraries, constructed from infected and uninfected Landsberg erecta plants (kindly provided by J. Jones), were screened. The avrPto coding sequence was amplified from Pst DC3000 genomic DNA by PCR (sense primer: 5'-GCGAATTCCGAACCATGGGAAATATATGTGTC-3'; antisense primer: 5'-GCCTCGAGATTGCCAGTTACGGTA-3') and cloned into pNLexA, to serve as bait in the yeast two-hybrid screen.

RabE cloning and mutagenesis
We amplified the RabE1d (At5g03520) coding sequence from Arabidopsis Col-0 cDNA using the rabE-5' and rabE-3' primers (Table S1), containing the EcoRI and BamHI restriction sites, respectively. The PCR product was ligated into a TOPO vector (pCR2.1, Invitrogen) and sequenced. Single nucleotide changes were introduced in the RabE1d sequence by two-step overlapping PCR, to generate the RabE1d-S29N and RabE1d-Q74L mutant derivatives. RabE1d-S29N was obtained through a G→A substitution; in the first PCR step, two overlapping RabE1d fragments were amplified using the primer combinations rabE-5'/S29N-rev and rabE-3'/S29N-for (Table S1). The products were purified from agarose gel, mixed, and used as template for a second PCR amplification step, with the rabE-5' and rabE-3' primers. The presence of an overlapping region allowed annealing of the two gene fragments and amplification of the fulllength coding sequence. A similar procedure was used for introducing the Q74L mutation, through an A→T substitution. In this case, the following primer combinations were used in the first PCR step: rabE-5'/Q74L-rev and rabE-3'/Q74L-for (Table S1). RabE1d-S29N and RabE1d-Q74L amplification products were introduced in a TOPO vector by TA-cloning and sequenced. 20 dip method (Clough and Bent, 1998). Transgenic plants were selected based on resistance to the herbicide Basta (glufosinate). A solution containing 0.012% glufosinate (Finale concentrate, AgrEvo Environmental Health) and 0.025% Silwet L-77 was sprayed on 2-week-old seedlings growing in soil. Surviving T1 plants were screened for GFP fluorescence with a Zeiss Axiophot microscope, and expression of the correct size GFP-RabE fusion proteins was verified by western blot with an anti-RabE polyclonal antibody (described below).

Protein extraction and immunoblotting
Total proteins were extracted as follows: approximately 20 mg (fresh weight) of fresh or frozen leaf tissue were ground with a pestle in a microfuge tube in the presence of 100 μl of 1× SDS-PAGE loading buffer [90 mM Tris-HCl, pH 8.0, 100 mM DTT, 3% SDS, 22.5% sucrose, 10 μl/ml Protease Inhibitor Cocktail for plant cell extracts (Sigma), bromophenol blue (to saturation)]. Extracts were immediately heated at 80°C for 10 min and then frozen at -20°C.
Before loading, extracts were thawed at room temperature and centrifuged at 20,000×g for 2 min, to pellet debris. An equal volume of each sample was used for SDS-PAGE electrophoresis.

Cell membrane fractionation
Leaves were harvested and weighed immediately prior to extraction. Leaf tissue (2.5 g) was ground with a cold mortar and pestle in the presence 5 ml of ice-cold extraction buffer [50 mM HEPES, pH 7.5, 100 mM KCl, 10 mM EDTA, 1 mM DTT, and 10 μl/ml Protease Inhibitor Cocktail for plant cell extracts (Sigma)] containing 34% sucrose (w/v). The extract was homogenized with a Polytron immersion blender (3 pulses of 10 s each), filtered through a single layer of Miracloth, and centrifuged for 10 min at 10,000×g, to remove most unbroken www.plantphysiol.org on August 19, 2017 -Published by Downloaded from Copyright © 2009 American Society of Plant Biologists. All rights reserved. 21 chloroplasts and nuclei. The supernatant was adjusted to 40% sucrose in about 10 ml final volume (concentration was determined with a refractometer), and layered on a 5 ml cushion of 50% sucrose, in clear ultracentrifugation tubes. The homogenate was subsequently layered with 10 ml of 34% sucrose, 8 ml of 25% sucrose and 8 mL of 18% sucrose (w/v). All sucrose solutions were prepared in the same buffer used for extraction. Gradients were centrifuged at 100,000×g for 3 h, at 4°C, in an SW28 rotor (Beckman). After centrifugation, the membranecontaining interphases were collected and diluted with sucrose-free extraction buffer, and membranes were collected by ultracentrifugation (1 h at 100,000 ×g). Membrane pellets were resuspended in equal volumes of SDS-PAGE loading buffer and heated at 80°C for 10 min.
Equal volumes were loaded on SDS-PAGE gels. Protein electrophoresis and western blot were performed as described above.

Confocal microscope analysis and imaging
Pieces of leaves were sampled randomly and mounted in water. Imaging was performed using an LSM510 META inverted confocal laser scanning microscope (Zeiss), and either a 20×

22
leaves were harvested and arranged on MS agar plates (4.3 g/l MS salts, 0.8% agar, pH 5.7). The DNA-coated particles were delivered into the lower leaf epidermis with a particle gun (Dupont), using 1100 psi rupture discs under a vacuum of 25 in Hg. After bombardment, leaves were incubated in the sealed plates at room temperature and fluorescence was observed 24 h post transformation. For co-imaging GFP-RabE1d and ST-RFP, the argon ion laser excitation lines of 488 nm (for GFP) and 543 nm (for DsRed) were used. GFP fluorescence was collected with a 505 to 530 nm band-pass filter, and DsRed fluorescence was collected with a 615-nm long-pass filter.

Plant growth and bacterial multiplication assay
Arabidopsis plants were grown in soil, in growth chambers, under a 12 h dark/12 h light cycle. The light intensity averaged 100 μE m -2 sec -1 , and the temperature was kept constant at 20°C. Pst DC3000 bacteria were cultured in low-salt LB medium (10g/l Tryptone, 5g/l Yeast Extract, 5g/l NaCl), supplemented with 100 μg/ml rifampicin. For multiplication assays in plants, bacterial liquid cultures were incubated at 30°C to the mid-to late-logarithmic phase.
Bacteria were collected by centrifugation and resuspended in sterile water with the addition of 0.004% Silwet L-77 (OSI Specialties, Friendship, WV). Titer of the bacterial inoculum was 1x10 5 colony-forming units (CFUs)/ml, unless otherwise indicated. Arabidopsis leaves were inoculated by syringe-infiltration, and bacteria enumeration in leaves was conducted as previously described (Katagiri et al., 2002).

23
amplification of the RabE and RabD transcripts, a single RT reaction was carried out in a total volume of 50 μl and incubated in a thermal cycler for 30 min at 45°C, followed by 5 min at 99°C and 5 min at 5°C. Five microliters of reverse-transcribed cDNA were used as template in each of ten PCR reactions with gene-specific primer pairs designed to amplify the five RabE gene family members, the four RabD genes, and the Actin8 gene as a control. Primer sequences are listed in Table S2. Each PCR reaction contained 2.5 mM MgCl 2 , 1×LA PCR Buffer II, 0.2 μM forward primer and 0.2 μM reverse primer, sterilized distilled water, and 5 μl of the RT reaction    Table S2). Equal volumes of each PCR product were loaded on 1% agarose gel.       A, Bacterial multiplication in GFP-RabE1d-Q74L-expressing plants (grey bars), compared to that in wild-type Arabidopsis (white bars). Pst DC3000 was syringe-infiltrated at a density of 1x10 5 CFUs/ml. B, Disease symptoms 3 d after vacuum-inoculation with Pst DC3000 at a y p density of 1x10 6 CFUs/ml. Left, Arabidopsis Col-0 gl1 (wild-type); right, Arabidopsis expressing GFP-RabE-Q74L C Accumulation of extracellular proteins in plants expressing expressing GFP RabE Q74L. C, Accumulation of extracellular proteins in plants expressing GFP-RabE1d-Q74L. Proteins in the intercellular wash fluid (IWF) from wild type (Col) and RabE1d Q74L expressing plants (Q74L) were separated by SDS PAGE In the top panel RabE1d-Q74L-expressing plants (Q74L) were separated by SDS-PAGE. In the top panel, Coomassie Blue-stained gel, representing total proteins; the arrowheads indicate bands that seem to be exclusive to the Q74L plants. In the bottom panel, western blot with the anti-PR-1 antibody (gift of Dr. X. Dong, Duke Univ.). y (g g )