RabGAP22 Is Required for Defense to the Vascular Pathogen Verticillium longisporum and Contributes to Stomata Immunity

Verticillium longisporum is a soil-borne pathogen with a preference for plants within the family Brassicaceae. Following invasion of the roots, the fungus proliferates in the plant vascular system leading to stunted plant growth, chlorosis and premature senescence. RabGTPases have been demonstrated to play a crucial role in regulating multiple responses in plants. Here, we report on the identification and characterization of the Rab GTPase-activating protein RabGAP22 gene from Arabidopsis, as an activator of multiple components in the immune responses to V. longisporum. RabGAP22Pro:GUS transgenic lines showed GUS expression predominantly in root meristems, vascular tissues and stomata, whereas the RabGAP22 protein localized in the nucleus. Reduced RabGAP22 transcript levels in mutants of the brassinolide (BL) signaling gene BRI1-ASSOCIATED RECEPTOR KINASE 1, together with a reduction of fungal proliferation following BL pretreatment, suggested RabGAP22 to be involved in BL-mediated responses. Pull-down assays revealed SERINE:GLYOXYLATE AMINOTRANSFERASE (AGT1) as an interacting partner during V. longisporum infection and bimolecular fluorescence complementation (BiFC) showed the RabGAP22-AGT1 protein complex to be localized in the peroxisomes. Further, fungal-induced RabGAP22 expression was found to be associated with elevated endogenous levels of the plant hormones jasmonic acid (JA) and abscisic acid (ABA). An inadequate ABA response in rabgap22-1 mutants, coupled with a stomata-localized expression of RabGAP22 and impairment of guard cell closure in response to V. longisporum and Pseudomonas syringae, suggest that RabGAP22 has multiple roles in innate immunity.


Introduction
Soil is a complex matrix that hosts a rich diversity of organisms. In the soil, plant roots must compete with other root systems of neighboring plant species for space, water, mineral nutrients, and with the excess of soil organisms feeding on an abundant source of organic material. In the microenvironment surrounding a plant root, the rhizosphere, some soil microorganisms take the advantage of root exudates to invade the root tissue. At the same time the plant immune system is activated upon perception of microbial-associated molecular patterns (MAMPs). It has been described how Arabidopsis roots respond to three selected MAMPs in a highly orchestrated and tissue-specific manner [1]. Still we have limited knowledge on MAMP signaling incited by saprophytes, deleterious soil-borne pathogens and beneficial soil organisms, required to suppress plant immunity pathways upon close interaction and invasion.
A specific host-microbe interaction involves recognition of the foreign organism by the host immune system, leading to amplification of immune responses and subsequent attenuation or elimination of the pathogen. The defense mechanisms include pattern recognition receptor proteins (PRRs), which recognize conserved pathogen-associated molecular patterns (PAMPs) or MAMPs [2]. Activation of PRRs leads to PAMP or MAMP triggered immunity (PTI/MTI), referred to in plants as basal or non-host resistance [3]. A second layer of plant defense consists of recognition of pathogen effector molecules by plant intracellular receptors, the effector-triggered immunity (ETI). In addition, the plant surveillance system is further shaped by a plethora of developmental, hormonal and specific signaling cascades, as well as epigenic components [4,5,6,7,8]. In all of these processes the cellular machinery plays important roles in receiving and executing proper responses, for example by the vesicle-associated and SNARE protein-mediated exocytosis pathways [9]. The opposite process, endocytosis, is also of importance, exemplified by ligand-triggered internalization of the flagellin-recognition receptor FLS2 [10] and the ethylene-inducing xylanase receptor LeEix2 [11].
The fungal genus Verticillium harbors many soil-borne plant pathogens, among them V. longisporum that incites disease on several plant species particularly within the family Brassicaceae [12,13]. Microsclerotia, the long-lived soil-infesting resting struc-tures of this hemibiotrophic pathogen, germinate when exposed to root exudates from a host plant. Hyphal growth around lateral roots is followed by direct penetration of the epidermal cells, and subsequent colonization of the xylem vessels [14,15]. Individual vessels can be filled with growing mycelia whereas others remain normal resulting in vague disease symptoms, particularly at an early disease stage. In the later stages of infection, symptoms including chlorosis, stunting and premature senescence emerge [16]. Formation of microsclerotia in the dying foliage then completes the disease cycle [14,17]. Knowledge on the plant immune responses taking place during fungal colonization and growth is fragmentary but is gradually expanding. A Verticilliuminduced transdifferentiation of bundle sheath cells to functional xylem elements suggests that a tissue-specific developmental reprogramming occurs during the infection process [18].
The aim of this study was to increase the knowledge on the components involved in the defense against V. longisporum. The starting point chosen was the phenotypic differences in response to V. longisporum earlier found in Brassica oleracea germplasm [19]. We reasoned that the findings from B. oleracea could be exploited in the Arabidopsis system to expand our basic understanding of this particular plant-pathogen interaction. Using a cDNA-AFLP approach, we identified a member in the RabGTPase activating gene family, RabGAP22, of importance for plant immunity. RabGTPases make up a large family in Arabidopsis, and the 57 members take part in mechanisms underlying intercellular membrane trafficking, hormone signaling, and stress responses [20,21,22,23]. The GTPases cycle between active GTP-bound, and inactive GDP-bound states. The inherent hydrolysis of GTP to GDP by individual GTPases is a slow process, but is increased by several orders of magnitude by GTPase activating proteins (GAPs). Subsequently, GDP-GTP exchange factors (GEFs) replace the hydrolyzed GDP to GTP, completing the cycle. A family of 24 RabGAP genes is present in Arabidopsis [24]. Many eukaryotic RabGAPs contain a Tre-2/Bub2/Cdc16 (TBC) domain architecture. In a comparative evolutionary scan of eukaryotic organism groups using TBC sequences, plant RabGAPs are classified into the TBC-B domain group [25], comprising the Arabidopsis class II subfamily in which RabGAP22 is clustered along with Rab-GAP11, 19 and 20 [24]. The closest orthologs of RabGAP22 are TBC1D25 in mouse and TBC1D14 in human, the latter being involved in vesicle transport leading to starvation-induced autophagy by interactions with the autophagy kinase ULK1 [26]. In the present study we found an involvement of RabGAP22 in jasmonic acid (JA) accumulation and JA and brassinolide (BL) signaling. We also uncovered a role for RabGAP22 in stomatal responses against both V. longisporum and Pseudomonas syringae DC3000. Together, our data revealed RabGAP22 to be a mediator of multiple responses in plant immunity.

Results
RabGAP22 is Required for Immunity to V. Longisporum The white cabbage accession line BRA723 was previously identified as highly resistant to V. longisporum infection [19]. Hence, this material was selected for a cDNA-AFLP approach to further explore defense responses in this plant-pathogen system. A number of differentially expressed transcripts in roots during fungal interaction were detected. At two days post inoculation (dpi), 104 transcript-derived fragments were isolated and sequenced (Table S1). When comparing the sequences with the Arabidopsis database, 32 of the differentially expressed genes were identified as potential defense gene candidates. Out of these, the most strongly up-regulated transcript, RR86 (Accession number: KF258677), was further examined. BLASTP analysis identified four Arabidopsis genes homologous to RR86: At3g49350 (E = 3e 21 ), At5g24390 (E = 2e 205 ), At5g41940 (E = 2e 205 ) and At5g53570 (E = 5e 252 ), corresponding to the earlier classified RabGAP11, 19, 20 and 22, respectively [24]. Phylogenetic analysis on amino acid sequences from all 24 Arabidopsis RabGAPs and RR86 also placed RR86 in the same cluster as these four RabGAP genes ( Figure 1A).
To determine if any of the four RabGAP gene candidates were linked to plant immunity, their corresponding Arabidopsis T-DNA insertion mutants ( Figure S1A) were investigated. None of these showed any difference in developmental phenotype compared to wild-type Col-0. When screened for responses to V. longisporum, the homozygous rabgap22-1 plants showed a clear susceptible phenotype ( Figure 1B), whereas the rabgap11-1, rabgap19-1 and rabgap20-1 mutants displayed responses similar to that of wild-type ( Figure  S1b). To validate theses phenotypic responses quantitative realtime PCR (qRT-PCR) analyzes were performed. All T-DNA insertion mutant plants were found to be significantly downregulated in their target genes ( Figure S1C), and only the RabGAP22 gene was triggered by the fungus in wild-type Col-0 ( Figure 1C). These results, together with complementation analysis using a RabGAP22 Pro :RabGAP22 construct that restored the susceptible rabgap22-1 phenotype to the resistant wild-type phenotype ( Figure 1B), demonstrated the importance of RabGAP22 in defense to V. longisporum. Fungal growth in the plant roots was also monitored using GFP-tagged V. longisporum. At 14 dpi, the fungus was detected in roots of both Col-0 and mutant plants. In agreement with the earlier results the degree of colonization was found to be significantly higher in rabgap22-1 and rabgap22-1 plants ( Figures 1D, E).

RabGAP22 is Expressed in Vascular Tissues and Stomatal Guard Cells
In order to monitor the tissue and organ-specific expression of RabGAP22, transgenic Col-0 plants harboring a RabGAP22 Pro :GUS construct were made. GUS-staining revealed expression throughout the plant, most strongly in root meristems, vascular tissues, stomata, trichomes and flower tissues such as style and receptacle (Figures 2A, B and S2A-G). This gene expression pattern was only very subtly changed at 2 and 6 days post V. longisporum inoculation (data not shown). We further searched for the subcellular localization of RabGAP22. No information could be acquired by TargetP [27] but the search tools MultiLoc2 [28] and AtSubP [29] suggested a nuclear localization. To clarify this, Nicotiana benthamiana plants were infiltrated with a RabGAP22 Pro :RabGAP22-GFP construct, which demonstrated a nuclear localization of the RabGAP22-GFP protein ( Figure 2C). This observation was further supported by co-infiltration of N. benthamiana leaves with RabGAP22 Pro :RabGAP22-GFP and the nuclear marker VirD2NLS-YFP construct, showing a clear co-localization of the GFP and YFP signals in the nucleus ( Figure 2C).

RabGAP22 Interacts with the Photorespiratory Protein AGT1
To gain further insight into the function of RabGAP22 in plant immunity, we looked for potential interacting proteins under stress conditions. Transgenic Arabidopsis plants harboring 35S Pro :Rab-GAP22-His were made, and immunoprecipitation experiments with anti-His antibodies were performed. In leaves from mocktreated plants, only the ,60 kDa band corresponding to the RabGAP22-His protein was observed, whereas in leaf samples from V. longisporum soil-inoculated plants, a protein of ,45 kDa co- precipitated with the RabGAP22-His protein ( Figure S3A). The amino acid sequence of this new protein (Accession number: KF242188) was identified by MALDI-MS/MS as SERINE:-GLYOXYLATE AMINOTRANSFERASE 1 (AGT1, At2g13360) ( Figure S3B), an enzyme in the photorespiratory pathway [30]. An AGT1 homolog is known to be indispensable for appressorium function in the Magnaporthe oryzae fungus [31], thus we found it important to clarify if the identified AGT1 protein was of plant or fungal origin. When AGT1 protein sequences from V. longisporum, M. oryzae and Arabidopsis were aligned, the protein fragments from MALDI-MS/MS showed a perfect match to Arabidopsis AGT1 verifying a plant origin ( Figure S3C). qRT-PCR analysis also revealed an increase in AGT1 transcript levels in roots of V. longisporum inoculated plants at 2 dpi ( Figure S3D). To generate more data along this line, a bimolecular fluorescence complementation (BiFC) assay [32] was performed to show if interaction of the two proteins occurs in planta. Two plasmids, in which RabGAP22 and AGT1 were fused to the GFP C-and Ntermini respectively (pSITE-cEGFP-RabGAP22 and pSITE-nEGFP-AGT1) were constructed and infiltrated into N. benthamiana leaves. A clear fluorescence signal was detected in the peroxisomes, demonstrating interaction of the two proteins ( Figure 2D). The peroxisomal localization was further supported by co-infiltrations of the two BiFC constructs and a mCherry-tagged peroxisomal marker ( Figure 2D). As a control step, the pSITE-cEGFP-RabGAP22 and pSITE-nEGFP-AGT1 plasmids were infiltrated individually into tobacco leaves to check for potential interactions, but no fluorescence signals were detected from those samples. The observed peroxisomal localization of RabGAP22-AGT1 was not unlikely, as AGT1 is a peroxisome-localized protein [33]. Presence of RabGAP22 in both nuclear and peroxisomal cellular compartments is also in line with the role of RabGAP proteins in vesicle trafficking between membrane compartments [34]. When tested for the response to V. longisporum, agt1 T-DNA insertions mutants showed a phenotype indistinguishable to that of Col-0 ( Figure  S3E), indicating that the immune response is mainly mediated by RabGAP22. The lack of a peroxisomal localization of RabGAP22-GFP in unstressed plants ( Figure 2C) suggests that a specific relocalization of RabGAP22 to peroxisomes takes place under biotic stress conditions, here exemplified in response to V. longisporum. Changes in subcellular localization and activity of GTPases, GAPs and GEFs frequently occur in response to changes in the GTP/ GDP-bound state [35] and also in response to phosphorylation of individual residues in these molecules [36]. The rice Rab protein OsRab11 is characterized both for regulating vesicular trafficking between the trans-Golgi network and vacuole [37,38] and for interaction with the peroxisomal protein OsOPR8 [39].
Peroxisomal proteins are involved in a number of processes associated to pathogen defense responses, not least H 2 O 2production and jasmonic acid (JA) biosynthesis, reviewed by Sørhagen et al. [40]. Consequently we assayed these responses in rabgap22-1 plants, starting with measurement of H 2 O 2 levels by 3,39-diaminobenzidine (DAB) staining. The basal level of H 2 O 2 was lower in the mutant compared to Col-0, and in contrast to the wild-type, no increase in the levels were detected in response to V. longisporum ( Figures S3F, G). We interpreted the data as a very

V. Longisporum Promotes Increased JA and JA-Ile Levels
Jasmonic acid (JA) and JA signaling play important roles in most plant defense systems, in particular in responses to necrotrophic and hemibiotrophic pathogens. To elucidate if the increased susceptibility of rabgap22-1 was linked to alterations in JA-mediated defense responses, endogenous levels of both JA and the bioactive JA-Isoleucine (JA-Ile) were measured at 2 dpi. As expected, both compounds increased in V. longisporum challenged Col-0 plants ( Figure 3A, B). However, the levels of JA and JA-Ile were significantly higher in the rabgap22-1 mutant and strikingly induced by the fungus, indicating an important role of RabGAP22 in the JA biosynthesis pathway. In order to understand more on JA-associated events we analyzed transcript levels of key genes in the JA signaling pathway. The basal transcript levels of the repressor of JA-signaling JASMONATE ZIM-DOMAIN 10 (JAZ10) were 4-fold higher in rabgap22-1 compared to wild-type, but did not change upon fungal challenge ( Figure 3C). Whereas levels of the JA signaling marker CORONATINE INSENSITIVE 1 (COI1) only increased slightly in Col-0 in response to V. longisporum, elevated levels were seen in both mock and inoculated rabgap22-1 plants ( Figure 3D). In agreement with previous reports [17], both the JA marker VEGETATIVE STORAGE PROTEIN 2 (VSP2) and the JA/ethylene marker PLANT DEFENSIN 1.2 (PDF1.2) were induced by V. longisporum in the wild-type ( Figures 3E, F). In contrast, the induced levels of VSP2 in rabgap22-1 were lower compared to Col-0, whereas PDF1.2 levels appeared constitutively high in the mutant. Overall, our results indicate that RabGAP22 is a negative regulator of JA responses and that over-production of JA and JA-Ile in the rabgap22-1 mutant is accentuated during V. longisporum infection, contributing to a distorted JA defense signaling and an enhanced susceptible phenotype.

RabGAP22 is Involved in Brassinosteroid-associated Responses
To further expand our understanding of RabGAP22 defense responses, we searched for co-expressed genes. A comparative analysis of all available microarray datasets deposited in Genevestigator [41] was performed. The similarity search tool limited to perturbations for RabGAP22 showed MITOGEN-ACTIVATED PROTEIN KINASE PHOSPHATASE (MPK1) and BRI1-ASSOCIATED RECEPTOR KINASE 1 (BAK1) to be the two most correlated genes with a score of 0.59 and 0.56, respectively ( Figure S4A). BAK1 is known to affect multiple processes, including brassinosteroid signaling and defense to plant pathogens [42,43,44]. Hence the bak1-4 Arabidopsis null mutant was challenged with V. longisporum to investigate its potential contribution in this pathosystem. The bak1-4 plants displayed a clear susceptible phenotype as well as a massive fungal colonization at 14 dpi compared to the wild-type ( Figures 4A, B). In addition, BAK1 Pro :GUS transgenic Arabidopsis plants displayed elevated levels of GUS expression in response to V. longisporum at 1 and 2 dpi ( Figure S5). Further, the finding of attenuated levels of RabGAP22 in the bak1-4 mutant ( Figure 4C) indicates that BAK1 may be required for RabGAP22 function.
When our Genevestigator search was limited to hormone responses, brassinolide (BL) treatment was found to induce the highest RabGAP22 expression ( Figure S4B). Brassinosteroids (BRs) are known to modulate plant defense responses and are involved in many other developmental processes including xylem differentiation [45,46]. We therefore analyzed the responses to 24epibrassinolide (BL) in our plant materials. No difference in the BL-induced root growth inhibition was found in rabgap22-1 compared to Col-0 plants grown in vitro, revealing that rabgap22-1 mutants were not compromised in the uptake of BL (data not shown). However when we pretreated plants with BL prior to V. longisporum inoculation, fungal colonization was reduced 4-fold in rabgap22-1 and 2-fold in wild-type ( Figure 4D). Together these results suggest that the increased fungal colonization in rabgap22-1 is in part due to impaired BL signaling, and that RabGAP22 is a downstream component of BL-mediated signaling.
RabGAP22 Affects Stomata Closure through Attenuated ABA Levels The observed RabGAP22 Pro :GUS expression in guard cells ( Figure 2B) resembles that of Arabidopsis ROP11 GTPase [47], a negative regulator of ABA-mediated stomatal closure. ABA is a central regulator of the stomatal apparatus [48], and Genevestigator expression data suggested a reduction in RabGAP22 expression in response to ABA treatment ( Figure S4B). These features prompted us to test the behavior of stomata in response to V. longisporum challenge. At 14 dpi we found stomatal apertures in wild-type Col-0 to be reduced by 75% compared to mock treated plants. A feature that could explain the increased drought tolerance of V. longisporum inoculated Col-0 seen at a similar time-point [18]. At 14 dpi stomata were in contrast partially open (reduced by 34%) in the rabgap22-1 plants upon pathogen attack ( Figure 5A). We wanted to know if this impairment in stomata closure could be linked to altered endogenous ABA levels in our plant materials. Congruent with previous findings [17] we detected a significant increase in ABA content in leaves of Col-0 in response to V. longisporum ( Figure 5B), whereas the induction of ABA in the rabgap22-1 mutant only reached half the levels compared to the wild-type. These results were followed by a closer monitoring of the stomatal apparatus in Col-0 and rabgap22-1 plants in response to ABA treatment. No anomalies in size and numbers were found but stomata in rabgap22-1 were impaired in ABA-induced closure, with stomatal apertures only being reduced by 14% in response to the hormone application ( Figure 5C). Based on the partial impaired stomatal closure in rabgap22-1 in response to V. longisporum, experiments were also run to clarify if a similar influence on the stomata apparatus could be seen in the wellstudied Pseudomonas syringae DC3000 (Pst DC3000) system. When inoculated with Pst DC3000, leaves of rabgap22-1 plants showed a strong increase in bacterial proliferation compared to wild-type at 3 dpi ( Figure 5D). PAMPs from Pst DC3000 are known to trigger stomatal closure in Arabidopsis within 1 to 2 h post inoculation through ABA biosynthesis and signaling [49,50,51], as well as via an ABA-independent oxylipin pathway [52]. The bacterial effector coronatine (COR) inhibits this effect by reopening the stomata within 3 to 4 h after infection. Treatment with flg22 (a PAMP of Pst DC3000) on our materials triggered stomatal closure in wildtype Col-0 plants as expected but this response was partially impaired in rabgap22-1 ( Figure 5E). Stomatal apertures were also measured in Pst DC3000 inoculated plants, revealing at 1 hpi a weak, but significant reduction in the stomatal closure response of the rabgap22-1 plants (Figure 5F). At 4 hpi, stomatal apertures in both Col-0 and rabgap22-1 were again open indicating that CORmediated re-opening of stomata was not effected in the mutant. That the stomatal closure responses to flg22 and Pst DC3000 were not as strong as those for V. longisporum, could be attributed to involvement of additional defense components in the case of this leaf pathogen.

Discussion
Soil-borne fungal pathogens are causal agents of diseases of increasing economic importance. In comparison to plant responses to foliar pathogens, relatively little is known about responses to root infecting pathogens primarily due to the difficulty in observing the early stages of the interaction. At present only a few important components in the plant immune response pathways to Verticillium fungi are known. Using a positional cloning strategy, the Ve1 and Ve2 genes conferring resistance to V. dahliae and V. albo-atrum were cloned in tomato [53]. Both Ve1 and Ve2 genes encode cell surface receptors with an extracellular leucine-rich repeat (LRR) domain but are lacking a cytoplasmic signaling domain, characteristic for receptor-like proteins [54]. The responses of the Ve genes have also been evaluated in transgenic Arabidopsis because no closely related Ve orthologs are present it its genome. Ve1 in the Arabidopsis genomic background conferred resistance to V. dahliae but not to V. longisporum [55] suggesting that another PRR is operating in the latter case. An Arabidopsis mutant in a cell wall-associated member of the receptor-like kinase (RLK) gene family mediating resistance to Fusarium oxysporum (rfo1) was earlier shown to promote susceptibility to V. longisporum infection [12], and could be indicative of important groups of RLKs to be further examined.
The fungal avirulence gene Ave1, a natriuretic peptide, is present in several fungal species including V. dahliae and is a proposed interactor to Ve1 [56]. We queried our V. longisporum genome sequences for the presence of Ave1 orthologs but no similar sequences could be found. In our efforts to find a candidate effector capable of binding to RabGAP22, we extended the searches to include a wider group of organisms. Effectors for example from the intracellular Shigella flexneri bacteria exhibit RabGAP activity [22]. Via GTP hydrolysis of the host RabGTPase Rab1, VirA disrupts ER-to-Golgi trafficking, allow-  ing the bacteria to escape the autophagy-mediated defense. In common with other RabGAPs, VirA contains a TBC domain. Hence, we used TBC sequences retrieved from various organisms in the search for orthologs in the V. dahliae genome [57], and in our genomic sequences from V. longisporum, without being able to identify an obvious effector candidate of this nature. Members of the somatic embryogenesis receptor-like protein kinase (SERK) gene family in Arabidopsis, SERK1 and SERK4 together with SERK3, also known as BAK1, are required for Vemediated defense against V. dahliae [55]. Here we revealed BAK1 to be an important defense component also to V. longisporum contributing to RabGAP22 expression. It could be speculated whether RabGAP22 functions together with BAK1 in a network similar to that in the Magnaporthae oryzae/chitinrice interaction. In that system the small GTPase OsRac1 is an essential component of a receptor complex in chitin-induced immunity, together with OsRacGEF1, chitin-binding OSCEBiP and the receptor-like kinase OsCERK1 [58]. As the BR receptors BAK1 and BRASSINOSTEROID INSENSITIVE 1 (BRI1) constantly recycle via endosomes [59], RabGAP22 could also take part in BAK1 recycling, thereby indirectly regulating appropriate BL or PAMP responses. In the present study, BL treatment led to a significant reduction in disease susceptibility towards V. longisporum in both wild-type Col-0 and the rabgap22-1 mutant. A similar induced resistance response was found on tomato inoculated with V. dahliae after an exogenous application of BR [60]. This type of BRenhanced resistance has been suggested to be a result of cross-talk with other plant hormones, and thus different from the BAK1mediated responses.
The formation of an AGT1-RabGAP22 complex characteristic for V. longisporum challenged plants suggests an activation of peroxisomal processes by RabGAP22, in line with the emerging role of peroxisome-produced H 2 O 2 in pathogen defense responses [61,62]. Proteomic analysis of the xylem sap from V. longisporum inoculated B. napus plants identified fungal genes involved in oxidative stress, whereof a catalase peroxidase was found to play a key role for the late infection phase [63]. We observed however only a mild attenuation of H 2 O 2 production in the susceptible rabgap22-1 plants, indicating that H 2 O 2 is not a main component of the RabGAP22-mediated defense.
Jasmonic acid is synthesized through the octadecanoid pathway that involves the translocation of lipid intermediates from the chloroplast membranes to the cytoplasm and later on into peroxisomes where further processing steps take place to complete the JA biosynthesis, reviewed by Kombrink [64]. JA-Ile, conjugated in the cytoplasm then triggers the JA-dependent signaling cascades. Arabidopsis OPDA REDUCTASE 3 (OPR3) catalyzes the reduction of 12-oxo-phytodienoic acid (OPDA) in the first step of JA biosynthesis taking place in the peroxisomes. In rice, Rab11 interacts with OPDA REDUCTASE 8 (OsOPR8) in the peroxisomes, and presumably this association activates the OPDA-reducing activity of OsOPR8, leading to induced JAmediated defense signaling [39]. In our case however, RabGAP22 has an inhibitory effect on the JA signaling, possibly via its interaction with peroxisomal AGT1. Modification of OPDA to JA includes several steps, including three rounds of b-oxidation and each of the various enzymes required for the modifications need to be imported into the peroxisome mediated by peroxins (PEXs). The small RabGTPase RabE1c in Arabidopsis has recently been shown to interact with PEX7 on the peroxisomal membrane [23], a process important for peroxisome biogenesis and subsequent boxidation. The photorespiratory gene AGT1 has multiple functions and currently we can only speculate that together with RabGAP22 it interferes with other unidentified factors required for JA biosynthesis. In either case, the accentuated JA signaling in the rabgap22-1 plants may also impact abscisic acid (ABA) signaling, as extensive crosstalk takes place between the JA and ABA pathways, not least in pathogen responses [4,65]. In the case of the rabgap22-1 mutant, the reduced ABA levels indicate that ABA may be negatively impacted by the substantial increases in JA and JA-Ile levels.
Stomata consist of two guard cells that function as turgoroperated valves regulating the gas exchange in plants. The guard cells are also capable of closing in response to PAMPs such as flg22 and lipopolysaccharides from P. syringae [49,50]. flg22-triggered stomatal closure, acting through the induction of SA and ABA biosynthesis and signaling is rapidly inhibited by the bacterial effector COR which re-opens the stomata 3-4 h after infection, through the actions of NAC transcription factors regulated by MYC2 [66,67]. The partial failure of rabgap22-1 stomatal closure in response to V. longisporum and ABA, but also to some extent to flg22 and Pst DC3000, points to a presumed role for RabGAP22 in mediating correct ABA-and stomatal immunity-associated signals. The Arabidopsis GTPase ROP11 negatively regulates stomatal closure, by direct interaction with abscisic acid insensitive 1 (ABI1), a signaling component downstream of the ABA receptor RCAR1/PYL9 [68]. Whether RabGAP22 functions individually or in concert with other GTPases such as ROP11 in regulating this process is currently unknown.

Conclusions
Evidence of involvement of RabGTPases and RabGAPs in diseases has its basis from human and animal model systems [69,70], and roles of these molecules in plant pathogen interactions are emerging [21,22,23]. Here, we characterize Arabidopsis RabGAP22 as a component required for resistance to V. longisporum, and in this early stage we can only speculate on its function in plant immunity ( Figure S6). Our results showed a clear interaction of RabGAP22 and AGT1 in the peroxisomes of V. longisporum-challenged plants. Together with the increased JA content in rabgap22-1, this warrants the hypothesis that Rab-GAP22-AGT1 may interfere with peroxisomal-localized steps in the JA biosynthesis. The strong reduction in V. longisporum colonization in rabgap22-1 upon BL treatment clearly indicates that BL signaling is important for RabGAP22 function. However the exact involvement of the BL signaling component BAK1 remains elusive at this stage. We also observed a stomatal closure response in V. longisporum-challenged wild-type plants. This response was impaired in rabgap22-1, and stomata in the mutant further failed to close in response to ABA, suggesting a role for RabGAP22 in ABA-mediated stomatal closure. The multiple effects of RabGAP22 found here in the Arabidopsis -V. longisporum apertures in four-week-old soil-grown plants, in response to ABA treatment. Data represent means 6 SE (.60 randomly selected stomata). (D) Quantification of bacterial growth in soil-grown Arabidopsis plants 0 and 3 days post spray inoculation with Pst DC3000. Average log colony forming units (CFU) per cm 2 leaf area are shown. Data represent means 6 SD (n = 5 pools of 3 leaves). (E, F) Stomatal apertures in four-week-old soil-grown plants, in response to (E) flg22 treatment and (F) Pseudomonas syringae DC3000 spray inoculation. Data represent means 6 SE (.60 randomly selected stomata). Asterisks indicate significant difference to the respective mock treated control. (Student's t-test; *p#0.05, **p#0.01, ***p#0.001). Different letters indicate significant difference (one-way ANOVA followed by a Tukey HSD test, 95% confidence interval). doi:10.1371/journal.pone.0088187.g005 interaction are intriguing. At present we only have a fragmentary knowledge on the role of RabGAP22 in the response to V. longisporum, and further efforts remain to gain a full understanding of the multiple processes it is involved in.

Pathogens, Plant Inoculation and Pathogen Quantification
The Verticillium longisporum isolate CBS110220 (here renamed VL1) [74,75] was used throughout the study, except for studies of root colonization, where GFP-tagged V. longisporum VL 43 [15] was used. For soil infections, roots of two-week-old soil-grown Arabidopsis plants were dipped for 10 min in a 10 7 conidia ml 21 suspension. Inoculated plants were re-potted in sterile soil, humidity raised to 100% for 2 days, and disease progress monitored for 3-4 weeks post inoculation. Plants were scored for disease susceptibility based on the appearance and severity of chlorosis, stunting and premature senescence. Typically, three separate stages of disease were distinguished between: 1 = discoloration of leaf vascular tissues; 2 = stunting and leaf chlorosis being to appear; 3 = severe stunting and chlorosis compared to control. For in vitro liquid cultures, two-week-old (Arabidopsis) or 6-daysold (B. oleracea) plants were washed with sterile H 2 O and transferred to fresh MS medium without sucrose supplemented with 10 4 VL1 conidia ml 21 . Quantification of V. longisporum DNA in plants grown in hydroponic culture [55] was carried out 14 d post inoculation. Following a five-minute wash in 70% ethanol, biological replicates (roots from at least five plants) were collected. This method ensured the quantification of fungus inhabiting the vascular tissues. Quantitative real-time PCR (qRT-PCR) pathogen quantification was performed as described below with primers provided in Table S2. Pseudomonas syringae pv. tomato DC3000 was maintained at 28uC on King's B medium supplemented with rifampicin (50 mg ml 21 ). Four-week-old Arabidopsis plants were spray inoculated [76] with a 10 8 CFU ml 21 bacterial suspension and kept under high humidity for disease development. At 3 dpi, biological replicates (at least three leaves) were collected and surface sterilized with ethanol (70% v/v). Samples were ground in sterile water in a TissueLyser (Qiagen, http://www.qiagen.com/), serially diluted, and grown on King's B medium containing rifampicin (50 mg ml 21 ). Colonies were counted after 40 h.

Brassinolide Treatment
Quantification of V. longisporum upon brassinolide treatment was carried out as described above for hydroponic culture, with the following modifications. Two-week old in vitro-grown plants were transferred to MS medium supplemented with 10 nM 24epibrassinolide (Sigma-Aldrich, http://www.sigmaaldrich.com/). Seven days later, plants roots were dipped in V. longisporum suspension for 30 min and thereafter grown in hydroponic culture supplemented with 10 nM 24-epibrassinolide for the rest of the experiment.
In Situ Detection of H 2 O 2 H 2 O 2 was detected in two-week-old in vitro grown plants at 2 dpi, by vacuum infiltration of DAB solution for 30 min, followed by a 2 h incubation and subsequent destaining in absolute ethanol [77]. Quantification of DAB was done by measuring the intensity of stained tissues in Adobe Photoshop (Adobe Systems Inc., http://www.adobe.com/). At least three images were taken at 20x magnification on a Zeiss Axioplan microscope, on each of at least five individual plants per genotype and treatment. After conversion of images to black and white, the average intensity of DAB (0-250) was measured on three separate squares (50650 pixels) for each root image using the histogram function implemented in Photoshop.

RNA Isolation, cDNA-AFLP and Quantitative Real-time PCR
Total RNA was isolated from B. oleracea roots 2 days post VL1 inoculation using Spectrum TM Plant Total RNA Kit (Sigma-Aldrich, http://www.sigmaaldrich.com/). cDNA synthesis, restriction, adapters, primers, amplification steps, transcript profiling and sequencing of selected transcripts were performed as previously described [78]. qRT-PCR on Arabidopsis was performed on materials collected at 2 dpi. RNA isolations and cDNA synthesis were carried out on biological triplicates (roots from at least 25 plants). Each biological replicate was amplified in at least two separate reactions (two technical replicates) in qRT-PCR. Primers were designed using Primer3 [79] and are provided in Table S2. qRT-PCR data were analyzed by the comparative C T method [80] with qRT-PCR efficiency correction determined by the slope of standard curves. Fold-differences in transcript levels and mean standard error were calculated as outlined by Schmittgen and Livak [81].

Phylogenetic Analyses
Protein sequences from the coding region of all Arabidopsis RabGAP genes (http://www.arabidopsis.org) and B. oleracea RR86 were aligned using ClustalW (http://www.ebi.ac.uk/Tools/msa/ clustalw2/), and manually inspected. One hundred and sixteen conserved amino acids were used in the final analysis. The MEGA 5.1 software [82] suggested the JTT+4G model [83] for sequence evolution, and was used in Maximum likelihood analyses. Five hundred bootstrap replicates were performed.

Plasmid Construction and Plant Transformation
Promoter sequence for RabGAP22 was determined using AtcisDB [84]. cDNA or genomic DNA target sequences were PCR amplified using Phusion DNA polymerase (Thermo Scientific, http://www.thermoscientificbio.com/), cloned into suitable donor vectors (Table S3), and introduced to the Gateway system (Invitrogen, http://www.invitrogen.com/) for sequencing. Final 35S Pro :RabGAP22-His, RabGAP22 Pro :GUS, RabGAP22 Pro :RabGAP22 and RabGAP22 Pro :RabGAP22-GFP constructs were introduced to Agrobacterium tumefaciens strain C58, and Arabidopsis Col-0 or rabgap22-1 plants were transformed using floral-dip [85]. Confirmed T2 lines were used for GUS and GFP analyses and T3 homozygous complementation lines for inoculation assays. At least three independent transgenic lines per construct were used for all analyses.

GUS Expression, BiFC Analysis and Confocal Microscopy
Transgenic Col-0 plants harboring RabGAP22 Pro :GUS and BAK1 Pro :GUS [71] were used for GUS staining [86] and images were taken using a Leica Z16 APO microscope. Transient GFP expression of RabGAP22 Pro :RabGAP22-GFP was monitored in Agrobacterium-infiltrated Nicotiana benthamiana leaves. For BiFC analysis [87], AGT1 and RabGAP22 cDNAs were amplified by PCR using primers listed in Table S3. PCR products were ligated into the pCRTM8/GX/TOPO entry vector and sequenced, and then transferred into Gateway compatible BiFC vectors (pSITE-nEGFP-C1 and pSITE-cEGFP-C1) by LR reactions. The pSITE-cEGFP-AGT1 and pSITE-nEGFP-RabGAP22 plasmids were subsequently transformed into the Agrobacterium strain GV3101. For determination of organelle localization, the nuclear marker VirD2NLS-YFP [32] or a peroxisomal-targeted mCherry marker [88] were used in tobacco co-infiltration settings. For coinfiltrations, bacteria harboring the different plasmids were mixed in 1:1 or 1:1: 1 (v/v) ratios as appropriate and infiltrated into leaves of two-week-old N. benthamiana plants [89]. Fluorescence and brightfield images were taken using a Zeiss 780 confocal scanning microscope at 4 d post infiltration. GFP was excited at 488 nm, and detected at 493-530 nm, YFP excited at 514 nm and detected at 518-560 nm, chlorophyll excited at 633 nm and detected at 647-721 nm, and mCherry excited at 561 nm and detected at 600-650 nm. Images were analyzed using the built in Zen2011 software.

ABA and JA Analysis
In vitro grown Arabidopsis Col-0 and rabgap22-1 plants inoculated with VL1 or water were harvested at 2 dpi, in biological triplicates consisting of at least 50 whole plants. Endogenous levels of ABA and JA were extracted and quantified using GC-MS [90,91].

Stomata Assays
Leaf peels were collected from the abaxial side of mature leaves from four-week-old plants and placed in 300 ml H 2 O or buffer (25 mM MES, 10 mM KCl, pH 6.15) for 30 minutes. For ABA and flg22 treatments, the buffer was supplemented with either 10 mM ABA (Sigma-Aldrich), or 10 mM flg22 (EZBiolab, http:// www.ezbiolab.com/). For bacterial assays, a cell suspension of 1?10 8 Pst DC3000 ml 21 was sprayed on the leaves, and samples collected at 1 and 4 h post inoculation. Leaves were stained with 20 mM propidium iodide and stomatal measurements carried out according to Chitrakar and Melotto [92]. Leaf peels were observed at a Zeiss Axioplan microscope. The width of the stomatal pore opening was measured at the widest size on at least 60 randomly selected stomata for each genotype and treatment. Stomatal measurements on V. longisporum inoculated plants was performed in the same way as for Pst DC3000, except samples were taken 14 d post inoculation.   Figure S6 Hypothetical model for RabGAP22 in early defense to Verticillium longisporum. The RabGTPase activating genes are known to serve as molecular switches in a wide range of pathways. In congruency, we see multiple functions of RabGAP22 in the response to V. longisporum. In a pathogen recognition complex (dashed line), with a so far unidentified PAMP molecule and receptor, RabGAP22 could act together with BAK1, most likely interfering with the phosphorylation steps required for BAK1 activation. In response to V. longisporum RabGAP22 would then re-localize from the nucleus to the peroxisome, where it interacts with the photorespiratory protein AGT1. In peroxisomal JA biosynthesis, OPDA is stepwise converted to JA, which after release is subsequently conjugated to the bioactive JA-Ile in the cytosol. By a so far unidentified mechanism, RabGAP22/AGT may interfere with the multiple steps leading to formation of JA, thereby also indirectly impacts JA-Ile signaling. As extensive cross-talk takes place between the phytohormones JA and ABA, altered JA levels may also contribute to the impairment in stomatal closure responses seen in the rabgap22-1 mutants. (AGT1 = SERINE GLYOXYLATE AMI-NOTRANSFERASE 1; BAK1 = BRI1-ASSOCIATED RECEP-TOR KINASE 1; JA = jasmonic acid; JA-Ile = JA-Isoleucine; OPDA = 12-oxo-phytodienoic acid; PAMP = Pathogen-Associated Molecular Pattern). (TIF)