Phytophthora infestans RXLR-WY Effector AVR3a Associates with Dynamin-Related Protein 2 Required for Endocytosis of the Plant Pattern Recognition Receptor FLS2

Pathogens utilize effectors to suppress basal plant defense known as PTI (Pathogen-associated molecular pattern-triggered immunity). However, our knowledge of PTI suppression by filamentous plant pathogens, i.e. fungi and oomycetes, remains fragmentary. Previous work revealed that the co-receptor BAK1/SERK3 contributes to basal immunity against the potato pathogen Phytophthora infestans. Moreover BAK1/SERK3 is required for the cell death induced by P. infestans elicitin INF1, a protein with characteristics of PAMPs. The P. infestans host-translocated RXLR-WY effector AVR3a is known to supress INF1-mediated cell death by binding the plant E3 ligase CMPG1. In contrast, AVR3aKI-Y147del, a deletion mutant of the C-terminal tyrosine of AVR3a, fails to bind CMPG1 and does not suppress INF1-mediated cell death. Here, we studied the extent to which AVR3a and its variants perturb additional BAK1/SERK3-dependent PTI responses in N. benthamiana using the elicitor/receptor pair flg22/FLS2 as a model. We found that all tested variants of AVR3a suppress defense responses triggered by flg22 and reduce internalization of activated FLS2. Moreover, we discovered that AVR3a associates with the Dynamin-Related Protein 2 (DRP2), a plant GTPase implicated in receptor-mediated endocytosis. Interestingly, silencing of DRP2 impaired ligand-induced FLS2 internalization but did not affect internalization of the growth receptor BRI1. Our results suggest that AVR3a associates with a key cellular trafficking and membrane-remodeling complex involved in immune receptor-mediated endocytosis. We conclude that AVR3a is a multifunctional effector that can suppress BAK1/SERK3-mediated immunity through at least two different pathways.

receptor-mediated endocytosis. We conclude that AVR3a is a multifunctional effector that can suppress BAK1/SERK3-mediated immunity through at least two different pathways. membrane and trans-Golgi network in a BAK1-independent manner to maintain steady-state levels at the cell surface [25,41]. In addition, FLS2 undergoes ligand-induced re-localization to endosomal vesicles in a BAK1-dependent manner [41][42][43][44]. Therefore, FLS2 traffics through two different endocytic pathways depending on its activation status. Furthermore, BRI1 and BAK1/ SERK3 also undergo constitutive recycling via endosomes and can localize to overlapping endosomal compartments [32]. Receptor-mediated endocytosis was initially thought as a mechanism for attenuation of signaling through depletion of activated receptor complexes. Further studies in animal systems revealed that receptor internalization contributes to additional signaling at endocytic compartments [45]. In animal cells, late endosomal compartments, which contain internalized receptors, regulate signaling events such as pro-inflammatory signaling, growth, and development [45][46][47][48][49]. In plants, localization of BRI1 and BAK1/SERK3 as well as a brassinosteroid analog at the same endosomal compartments pointed to a possible link between internalization and signaling in growth regulation [32,50,51]. Nonetheless, defense-related endocytic signaling has yet to be unambiguously demonstrated in plants.
Mechanisms of receptor-mediated endocytosis involve clathrin-mediated and clathrin-independent pathways, resulting in the recruitment of a plasma membrane cargo [52][53][54] that is later invaginated and pinched off into the cytoplasm often by the action of the large GTPase dynamin [53,55]. Dynamin is a~100 kDa protein that self-assembles into rings and helices to promote structural reorganization to mediate membrane fission [55]. The mechanistic details of clathrin-mediated endocytosis have been well established in animal cells and dynamin has been implicated in the internalization of the immunity-related Interleukin-2 Receptor (IL-2R) [56]. However, in plants, mechanisms of endocytosis have not been explicitly studied [57,58] and the extent to which dynamin or dynamin-like proteins play a role in receptor mediated endocytosis or plant immunity remains poorly understood.
P. infestans is the causal agent of potato and tomato late blight and a major threat to food security [59]. This oomycete pathogen deploys a large set of effectors that target multiple host cellular sites. Cytoplasmic effectors include the RXLR class, whose members are modular proteins that translocate inside host cells [60]. The biochemical activity of RXLR effectors is carried out by their C-terminal domains, which often contain variations of the conserved WYdomain fold [61,62]. One example of RXLR-WY effector is AVR3a of P. infestans. In P. infestans populations, Avr3a has two major allelic variants encoding the proteins AVR3a KI and AVR3a EM , which differ in two amino acids in the mature protein and are differentially recognized by the potato immune receptor R3a [63][64][65][66][67][68]. Contrary to AVR3a EM , AVR3a KI induces R3a-mediated resistance and confers avirulence to homozygous or heterozygous strains of the pathogen [63]. In host plants that do not carry R3a, AVR3a KI effectively suppresses the cell death induced by P. infestans INF1 elicitin and is thought to contribute to pathogen virulence through this and other immune suppression activities [64,65,69]. Remarkably, AVR3a KI-Y147del , a mutant with a deleted C-terminal tyrosine residue, is not affected in activation of R3a but fails to suppress INF1-mediated cell death, demonstrating that distinct amino acids condition the two AVR3a activities [65]. Moreover, AVR3a KI-Y147del neither binds nor stabilizes the plant E3 ubiquitin ligase protein CMPG1, which is required for INF1-mediated cell death, further uncoupling the effector activities [65,67]. The current model is that AVR3a, but not AVR3a-KI-Y147del , binds and stabilizes CMPG1 to suppress BAK1/SERK3-regulated immunity triggered by INF1 during the biotrophic phase of P. infestans infection [67].
The current study was prompted by our discovery that natural variants of the P. infestans effector AVR3a (AVR3a KI and AVR3a EM ) and the AVR3a KI-Y147del mutant suppress FLS2-dependent early responses. This contrasts sharply with the differential activities of these three AVR3a variants in suppressing INF1-mediated cell death, another BAK1/SERK3-dependent pathway. The ability of the AVR3a KI-Y147del mutant to suppress FLS2-dependent responses revealed that AVR3a can suppress BAK1/SERK3-dependent responses in a CMPG1-independent manner. Furthermore, AVR3a reduced the internalization of the activated FLS2 receptor but did not interfere with its non-activated plasma membrane localization, indicating that this effector might target cellular trafficking initiated at the cell periphery. Consistent with this model, we found that AVR3a associates with a plant GTPase dynamin-related protein 2 (DRP2) involved in receptor-mediated endocytosis, implicating AVR3a in a cellular trafficking complex. Furthermore, we found that DRP2 is required for internalization of FLS2 and that overexpression of DRP2 suppressed PRR-dependent accumulation of reactive oxygen species (ROS). We conclude that AVR3a associates with a key cellular trafficking and membraneremodeling complex that may be required for PRR endocytic trafficking.

AVR3a suppresses PTI in a CMPG1-independent manner
To determine the degree to which AVR3a suppresses PAMP-elicited defense responses besides INF1-mediated cell death, we measured reactive oxygen species (ROS) production and defense gene induction in N. benthamiana triggered by bacterial and oomycete elicitors. Plants transiently expressing epitope tagged variants of AVR3a (FLAG-AVR3a KI , FLAG-AVR3a EM , FLA-G-AVR3a KI-Y147del ) or a vector control (pBinplus::ΔGFP) were treated with flg22 (100 nM) or INF1 (10 μg/ml) and the transient accumulation of reactive oxygen species (ROS) was followed over 45 minutes or 22 hours, respectively. We observed that all variants of AVR3a reproducibly suppressed flg22-induced ROS accumulation to the same extent whereas AVR3a KI suppressed INF1-induced ROS accumulation more effectively than the other two variants, which is consistent with previous reports [64,65] (Fig 1A and 1C). We obtained similar results after flg22 treatment in N. benthamiana and A. thaliana plants stably expressing AVR3a, validating the transient expression assays (Panel A and B in S1 Fig).
The immune responses triggered by another bacterial PAMP, EF-Tu, overlap and share signaling components with those triggered by flagellin [28,70,71]. Therefore, we tested the effect of AVR3a on ROS accumulation triggered by the EF-Tu derived peptide elf18. We found that elf18 ROS production was significantly impaired to a similar extent by all variants of AVR3a (Panel A in S2 Fig). In contrast, the BAK1/SERK3-independent ROS production in response to chitin was not affected by any of the AVR3a variants (Panel B in S2 Fig). These results may indicate that AVR3a suppresses BAK1/SERK3-dependent pathways but not BAK1-independent immune responses.
One of the outcomes of PAMP elicitation is transcriptional reprogramming, a defense response occurring downstream of ROS production [9]. Therefore, we monitored the effect of AVR3a on gene expression of the previously characterized PTI marker genes NbCYP71D20 and NbACRE31 [26,72]. NbCYP71D20 expression was induced~12-fold by flg22 and~80-fold by INF1 treatment in control plants (Fig 1B and 1D). All AVR3a variants reduced the induction of NbCYP71D20 by flg22 by approximately 80% (Fig 1B) whereas reduction of INF1-elicited gene induction was about 40% for AVR3a KI and AVR3a EM with no reduction observed for the AVR3a KI-Y147del variant (Fig 1D). Similarly, induction of NbACRE31 by flg22 treatment decreased by 80% in the presence of all variants of AVR3a (Panel C in S1 Fig). In contrast, AVR3a did not suppress NbACRE31 expression after INF1 elicitation, although INF1 induction of this gene was very low (Panel D in S1 Fig).
Overall, our results indicate that in addition to suppressing INF1-mediated cell death [64,65,67], AVR3a has the capacity of supressing PTI responses mediated by the BAK1/ SERK3-dependent cell surface receptors FLS2 and EFR. Remarkably, all the variants of AVR3a, including AVR3a KI-Y147del , which neither suppresses INF1-mediated cell death nor interacts with the E3 ligase CMPG1, suppressed flg22 responses to the same extent. These findings indicate that the newly identified suppression activity is CMPG1-independent and that the AVR3a effector may suppress PTI through multiple mechanisms.
AVR3a does not alter receptor levels at the cell surface or receptor complex formation The bacterial effector AvrPtoB targets FLS2 for degradation to suppress plant immunity [73]. This prompted us to determine whether the suppression of flg22-triggered responses by AVR3a involved perturbation of FLS2 or BAK1 protein accumulation or complex formation [74,75]. To address this question, we transiently co-expressed AVR3a variants with FLS2-GFP or BAK1/SERK3-YFP in N. benthamiana and assessed the fusion protein levels. We found that AVR3a did not alter FLS2 or BAK1/SERK3 protein accumulation in planta (Fig 2A). Next, we tested whether AVR3a KI perturbed heterodimerization of FLS2 with BAK1/SERK3 after flg22 treatment [25,26,28]. We used leaves of transgenic N. benthamiana plants stably expressing AVR3a KI or a vector control and transiently expressed FLS2-GFP and BAK1/SERK3-HA. In the presence of AVR3a KI a double band signal appeared (WB:HA) for BAK1/SERK3 total protein extracts ( Fig 2B) that was not seen after immunoprecipitation. However, this observation was not consistent between experiments and probably is the result of protein degradation during protein extraction. After co-immunoprecipitation, AVR3a did not prevent the flg22-triggered association between FLS2 and BAK1/SERK3 (Fig 2B). In summary, these results suggest that AVR3a does not interfere with receptor protein accumulation or complex formation and that the effector suppression of flg22 responses most likely occurs downstream of FLS2/BAK1 heterodimerization.

AVR3a interferes with FLS2 internalization
We hypothesized that AVR3a alters the subcellular distribution of FLS2 and/or BAK1/SERK3 to perturb their activities. To determine the effect of AVR3a on subcellular distribution of the receptors, we transiently co-expressed FLS2-GFP or BAK1/SERK3-YFP with AVR3a variants in N. benthamiana and assessed the localization of the non-activated receptors by confocal microscopy. In both cases, AVR3a did not alter the previously reported plasma membrane localization of FLS2 and BAK1/SERK3 [31,42] (Fig 3A). We then examined whether AVR3a has an effect in the non-activated subcellular localization of other cell surface receptors. We coexpressed EFR and CERK1 with all AVR3a variants in N. benthamiana and found that AVR3a did not alter the membrane localization of these immune receptors (S3 Fig). FLS2 activation after flg22 perception leads to endocytosis and accumulation of the receptor in mobile endosomal compartments [41,42,44]. To test whether AVR3a has an effect on the subcellular distribution of an activated receptor, we transiently expressed FLS2-GFP in N. benthamiana plants expressing a vector control. We detected FLS2-GFP vesicles at 80 minutes to 150 minutes after flg22 elicitation ( Fig 3B and S4 Fig), whereas we rarely detected vesicle-like structures in cells treated with water ( Fig 3B). Remarkably, the flg22-induced FLS2 endosomal localization was partially inhibited in plants expressing AVR3a (AVR3a KI or AVR3a EM or AVR3a KI-Y147del ) ( Fig  3B and S4 Fig). To evaluate the robustness of this phenomenon, we quantified the number of flg22-induced FLS2-GFP endosomes, seen as distinct fluorescent signal in punctate structures. We found that in the presence of AVR3a (AVR3a KI or AVR3a EM or AVR3a KI-Y147del ), the number of FLS2-GFP labelled endosomes was reduced down to almost half compared to the control ( Fig  3B and 3C). Thus, we conclude that AVR3a partially inhibits FLS2 internalization.
Next, we assessed whether AVR3a specifically inhibits the internalization of FLS2 or if it generally interferes with endocytic processes. For this we used BRI1, a receptor involved in development that also requires BAK1/SERK3 for its activity and shows constitutive internalization [32]. Using the same experimental procedure described above, we observed that BRI1-GFP fluorescent signal from vesicle-like structures was unaltered in the presence of AVR3a KI (Panel A and B in S5 Fig). These results suggest that AVR3a interferes with FLS2 internalization without affecting general receptor endocytosis, possibly by targeting a plant protein specifically required for FLS2 internalization.

AVR3a associates with DRP2, a plant protein involved in cellular trafficking
AVR3a was previously shown to bind the E3 ligase CMPG1 [67]. To determine which additional host proteins associate with AVR3a, we used immunoprecipitation of FLAG-AVR3a KI benthamiana of FLS2-GFP or BAK1-YFP with FLAG-AVR3a KI or FLAG-AVR3a EM or FLAG-AVR3a KI-Y147del or Vector Control (ΔGFP) as indicated. Total proteins were isolated at 2.5 days post infiltration (dpi) and total extracts were subjected to immunoprecipitation with 8 μl of anti-GFP agarose beads (Chromotek) to enrich for the GFP-tagged proteins, and detected using anti-GFP antibody. Equal amounts of protein were analyzed (Ponceau lane) in all cases. (B) Transgenic N. benthamiana expressing FLAG-AVR3a KI or vector control (ΔGFP) were transiently infiltrated with a mix (1:1) of expressed in N. benthamiana followed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis using previously described methods [76]. We identified four proteins that associated with AVR3a KI that had unique peptide counts compared to the controls in two biological replicates (S1 Table). We selected a homolog of the GTPase dynamin-related protein (DRP) (NCBE_074039.1) for further investigation based on the established role of dynamins in cellular trafficking and clathrin-mediated endocytosis in animal systems [77]. In mammals, dynamin is a five-domain protein consisting of a GTPase domain, an unstructured middle domain, a Pleckstrin-homology domain, a GTPase effector (GED) domain and a proline-rich domain, and its functions range from vesicle scission during endocytosis to the formation of the tubular-vesicular network during cytokinesis [77]. In Arabidopsis, DRP2A and DRP2B (previously known as ADL6 and ADL3, respectively) are the only dynamin-related proteins (DRP) with a domain architecture similar to other canonical dynamins [77][78][79][80]. Database searches using the tblastn algorithm and NCBE_074039.1 as the protein query sequence revealed that DRP2A and DRP2B are the most similar proteins for NCBE_074039.1. Therefore, we designed primers based on the Arabidopsis DRP2A and DRP2B sequences and the partial N. benthamiana sequences available at the time of the experiment. However, PCR reactions using these primers and N. benthamiana leaf cDNA as a matrix did not produce amplicons. As an alternative strategy, we cloned two putative DRP2 proteins from Nicotiana tabacum, termed NtDRP2-1 and NtDRP2-2, which share 99% amino acid sequence identity. Sequence analysis revealed that NtDRP2-1 and NtDRP2-2 have the classical five-domain structure of canonical dynamin proteins [77] and have 77% amino acid sequence similarity to the Arabidopsis DRP2 proteins (S6 Fig). To determine the evolutionary relationship of NtDRP2-1/2 to other plant DRPs, we first searched the genomes of solanaceous species, Arabidopsis and other dicot plants for proteins containing the dynamin signature [81]. Next, we constructed a maximum likelihood tree based on the amino acid sequence alignment of the conserved GTPase-domain of 45 identified DRPs (S7 Fig and S2 Table). The tree confirmed that NtDRP2-1/2 are most similar to AtDRP2A and AtDRP2B among the 11 Arabidopsis DRPs (S7 Fig).
We then investigated the subcellular localization of NtDRP2 to determine whether it is consistent with the presumed function of DRPs in cellular trafficking. Transient expression of GFP-NtDRP2-1 and GFP-NtDRP2-2 in N. benthamiana followed by confocal microscopy revealed that GFP-NtDRP2-1 and GFP-NtDRP2-2 mainly localized to the plasma membrane, as confirmed by the formation of thin cytoplasmic strands at points of adhesion between the plasma membrane and the cell wall (Hechtian strands) after plasmolysis treatment (Panel B and C in S8 Fig). In addition, GFP-NtDRP2-1 and GFP-NtDRP2-2 localized to the cytosol in a punctate pattern (Panel C in S8 Fig). This subcellular distribution is consistent with the reported localization of canonical dynamins in animal, yeast and Arabidopsis [77,78,82,83].
Next we validated the association of AVR3a with the cloned NtDRP2-1 in planta by coimmunoprecipitation analysis. All three AVR3a variants co-immunoprecipitated with GFP-NtDRP2-1 when expressed in N. benthamiana as FLAG epitope tagged proteins (Fig 4). The RXLR effector AVRblb2 was used as a negative control in these experiments as it failed to co-immunoprecipitate with NtDRP2-1 (Fig 4). We also used the Phytophthora capsici effector PcAVR3a-4 [61] as an additional negative control because this AVR3a homolog does not FLS2-GFP/BAK1-HA and treated with flg22 (100 nM) or water for 15 minutes. Treated leaf-tissues were collected at 2.5 dpi and subjected to co-immunoprecipitation with 25 μl GFP agarose beads (Chromotek). Purified complex formation after flg22 treatment of FLS2-GFP and BAK1-HA was analyzed by immunoblotting with the specified antibodies. 35S:GFP/BAK1-HA was used as a control. Note that FLS2-GFP was almost never detected in the total extracts (Input) but became visible by western blot analysis after immunoprecipitation. Overall, these results confirm that all AVR3a variants associate with NtDRP2-1 and suggest this association may be linked to the ability of AVR3a to suppress responses elicited by flg22.

DRP2 dynamin is required for FLS2 internalization
In a recent independent study, Smith and colleagues [84] showed that DRP2B functions in flg22-signaling and bacterial immunity in Arabidopsis [84]. To further investigate the link between AVR3a activities and NtDRP2, and notably the degree to which DRP2 is required for FLS2 internalization, we used RNAi silencing experiments. We designed silencing constructs that target N. benthamiana DRP2 Nb05397 (NbDRP2-1) and Nb31648 (NbDRP2-2), which are the most similar proteins to NtDRP2-1 and NtDRP2-2 in N. benthamiana (S7 Fig). Reports in Arabidopsis showed that members of the DRP2 family are implicated in cell cytokinesis and post-Golgi vesicular trafficking, and are essential for development since drp2 double mutants exhibit pleiotropic developmental defects [78][79][80]85]. Indeed, using virus-induced gene silencing (VIGS) in N. benthamiana, we observed that plants silenced for NbDRP2-1 and NbDRP2-2 display severe developmental defects and ultimately become necrotic and died confirming the essential role of dynamins in plant development (S10 Fig). To circumvent VIGS-induced lethality, we transiently expressed a DRP2-targeted hairpin-silencing construct in fully developed N. benthamiana leaves to silence Nb05397 and Nb31648 (Panel A in S11 Fig). Using quantitative RT-PCR, we confirmed that the hairpin construct reduced the mRNA levels of Nb05397 and Nb31648 but not of the two other related N. benthamiana DRP2 genes Nb11538 and Nb09838, indicating that the RNAi silencing is specific to the targeted N. benthamiana DRP2 genes (Panel B in S11 Fig). We also confirmed that the silenced epidermal cells remained viable by staining with propidium iodide (Panel C in S11 Fig).
Next, we used the hairpin RNAi system to determine the effect of silencing NbDRP2 in N. benthamiana leaves on flg22-induced internalization of FLS2-GFP. We found that FLS2-GFP containing vesicles were reduced in the NbDRP2 silenced leaves upon flg22 treatment, but were clearly visible in control-silenced leaves (Fig 5A and 5C). Moreover, silencing NbDRP2 did not alter accumulation of FLS2 at the plasma membrane in water treated samples (Fig 5A), indicating that DRP2 only affects FLS2 after activation. To determine whether the effect of NbDRP2silencing on receptor internalization is specific to FLS2, we performed the same NbDRP2 RNAi experiments with BRI1. Remarkably, BRI1-GFP constitutive endocytosis was not affected in We conclude that NbDRP2 is required for ligand-induced endocytosis of FLS2 and that this requirement might be specific for immune-related receptors. Importantly, these findings are consistent with our earlier finding that AVR3a interferes with the internalization of the activated FLS2 receptor but not BRI1 (Fig 3B and 3C, S4 and panel A and B in S5 Figs).

Overexpression of NtDRP2 dynamin suppresses flg22-induced ROS burst
To assess the role of Solanaceous DRP2 in early defense responses, we first determined the degree to which silencing of NbDRP2 affects flg22-induced ROS production. Although we observed a reduction in flg22-induced ROS accumulation in NbDRP2-silenced N. benthamiana leaves, the effect was inconsistent and was only noted in 40% of the experiments (n = 10) (S12 Fig). We then determined the effect of overexpressing NtDRP2 in N. benthamiana on ROS production upon flg22 treatment. Consistent with the finding that AtDRP2B is a negative regulator of flg22-triggered ROS [84], we found that NtDRP2-1/2 decreased flg22-induced ROS burst by more than 50% compared to plants expressing a vector control (Fig 6A). In addition, NtDRP2-1 significantly reduced the ROS production in response to INF1 and chitin treatments (Fig 6B and 6C). These results might indicate that the effect of DRP2 on ROS responses

Discussion
Even though perception of pathogen-associated molecular patterns significantly contributes to basal defense of plants against P. infestans, our knowledge of how P. infestans effectors subvert PTI is patchy. One of the most studied effectors of P. infestans is the RXLR-WY type effector AVR3a [63][64][65][66][67]69,86]. In this study, we further characterized the virulence activities of AVR3a and discovered that AVR3a suppresses early defense responses mediated by the cell surface  Phytophthora AVR3a Effector Associates with Dynamin immune co-receptor BAK1/SERK3, which contributes to basal immunity against P. infestans [36]. More specifically, we provide evidence that AVR3a reduces internalization of the activated pattern-recognition receptor FLS2 but neither interferes with the plasma membrane localization of non-activated FLS2 nor perturbs the steady state levels of this immune receptor or other PRRs. Furthermore, we found that AVR3a associates with DRP2, a Solanaceae member of the plant GTPase dynamin family whose members mediate endocytosis and membrane remodeling [87]. Interestingly, DRP2 is required for FLS2 internalization but does not affect internalization of the growth receptor BRI1.
Several bacterial type III secretion system effectors target cell surface immune receptor complexes to modulate their activities and suppress plant immunity [13,73,75]. AVR3a is an example of a filamentous plant pathogen effector that has evolved to deregulate plant immune signaling, including suppression of the cell death triggered by several pathogen molecules, among them the elicitin molecular pattern of oomycetes [64,67,69]. Our results further demonstrate that AVR3a contributes to suppression of basal defense responses similar to bacterial effectors. However, the exact molecular mechanisms by which eukaryotic effectors such as AVR3a subvert plant immunity remains poorly known. A recent screen showed the potential of P. infestans RXLR effectors to suppress early PTI signaling [88]. Although several RXLR effectors suppressed flg22-mediated PTI responses, AVR3a was not able to suppress the activation of the flg22-responsive FRK1 gene in tomato or Arabidopsis protoplasts [88]. The difference with our results could be due to intrinsic experimental differences, since the reporter screen was performed on protoplasts and the mechanisms of PTI suppression may vary under those conditions. Alternatively, it may indicate a level of specificity for the AVR3a suppression activity. Indeed, here we showed that AVR3a equally suppresses the activation of two marker genes after flg22 treatment but differentially suppressed the same marker genes set upon INF1 elicitation (Fig 1 and S1 Fig). Nevertheless, our work is consistent with the previous finding that AVR3a blocks signal transduction cascades initiated at the plasma membrane after pathogen perception [69]. Importantly, we found that interference with FLS2 signaling and endocytosis does not involve the interaction with the E3 ligase CMPG1 since AVR3a KI-Y147del , a variant that is unable to bind and stabilize CMPG1, is still able to suppress flg22-triggered ROS production and endocytosis (Fig 1A, Fig 3B and 3C). Therefore, AVR3a is a multifunctional effector that can suppress BAK1/SERK3-mediated immunity through at least two different pathways, possibly by acting at different sites or at different time points during immune signaling.
The step at which AVR3a interferes with FLS2 signaling is unclear. AVR3a did not alter subcellular localization of FLS2 or BAK1/SERK3 (Fig 3A and S3 Fig) and it did not affect ligand induced complex formation between FLS2 and BAK1/SERK3 (Fig 2). However, the receptor complex somehow remained inactive in the presence of AVR3a given that downstream signaling cascades were not activated. Given that elicitor-induced internalization of FLS2 occurs in a BAK1-dependent manner [42,43] it is possible that the effect of AVR3a in FLS2 internalization occurs via BAK1/SERK3. However, BAK1/SERK3 C-terminal fusions are partially impaired in early PTI responses [89], and assays evaluating the extent to which BAK1/SERK3 undergoes endocytosis after flg22 perception can be difficult to interpret. Interestingly, AVR3a exhibits some degree of specificity in suppressing PAMP-triggered immunity. Unlike the P. syringae effector AvrPtoB, which blocks the responses elicited by bacterial flagellin and chitin [90][91][92], none of the AVR3a variants suppressed chitin-induced ROS burst suggesting that AVR3a specifically suppresses BAK1/SERK3-dependent immune responses. At this stage, we propose that AVR3a interferes with BAK1/SERK3-mediated cellular trafficking involved in the perception of pathogens at the cell periphery but additional work is needed to clarify the underlying mechanisms.
Our finding that AVR3a co-immunoprecipitates with DRP2 does not necessarily imply that these two proteins directly bind in-planta nor that DRP2 is a target of the effector. DRP2 could be a helper (facilitator) of AVR3a that functions as a cofactor or enables localization of AVR3a to particular subcellular compartments as defined by Win et al. [4]. Whether AVR3a affects the biochemical activity or regulation of DRP2 remains to be studied. However, AVR3a is unlikely to completely inhibit DRP2 activity since DRP2 silencing resulted in plant lethality unlike transient or stable in planta expression of AVR3a. Nonetheless, our results place AVR3a in proximity to a membrane-remodeling complex that is implicated in endocytosis of the cell surface pattern-recognition receptor FLS2.
Although we cannot rule out that PTI suppression and inhibition of FLS2 endocytosis by AVR3a are two unrelated processes, AVR3a affected FLS2 but not BRI1 endocytosis, suggesting that AVR3a may specifically affect PRRs activities and that the mechanisms for FLS2 and BRI1 internalization may be different. Indeed, previous reports have shown that the regulatory role of BAK1/SERK3 in BR and PTI signaling are distinct and can be mechanistically uncoupled [71,89]. In addition, there are different pools of BAK1/SERK3 that are not interchangeable between BRI1 and FLS2 and activation by BR or flg22 does not cross activate these signaling pathways [93]. Consistent with the specific activity of AVR3a on FLS2 internalization, DRP2 is specifically required for FLS2-endocytosis but not for BRI1. This supports a model in which different internalization pathways for receptor endocytosis may occur following flg22 and brassinosteroid perception. For instance, AVR3a may perturb DRP2 functions only at sites where the activated FLS2 receptor complex accumulates preventing downstream endocytosis. Our results suggest that BAK1/SERK3-associated membrane-bound receptors may be initially internalized via different endocytic pathways, which later on can converge at late endosomal compartments.
In animal systems, canonical dynamin proteins mediate pinching off of clathrin-coated vesicles from the membrane during constitutive endocytosis [52,55]. In Arabidopsis, the two canonical dynamins DRP2A and DRP2B have been shown to be genetically and functionally redundant and to play a role in endocytosis and gametophyte development [79,80,85]. However, Smith et al., [84] recently showed that DRP2B but not DRP2A, has distinct effects on flg22-signaling. DRP2B acts as a positive regulator of plant immunity to Pseudomonas syringae pv. tomato DC3000 while it functions as a negative regulator of RbohD/Ca 2+ -dependent responses. In addition, FLS2 endocytosis is partially dependent on DRP2B but not DRP2A [84]. In this study, we also found that DRP2 is involved in FLS2 internalization and that it affects the accumulation of ROS upon PAMP treatment potentially placing this plant protein along the flg22-signaling pathway. However, these results do not indicate that FLS2 internalization is required for flg22 immune signaling responses. Indeed, suppression of ROS accumulation by DRP2 overexpression could be an indirect consequence of overexpressing an important component of plant cellular trafficking or may reflect pleiotropic effects on the PRRs or other components involved in ROS accumulation. Interestingly, the DRP2 family has expanded in Solanaceae compared to Arabidopsis, with no apparent orthologs of DRP2A and DRP2B (S7 Fig). Consistent with Smith et al., [84], our results also indicate that the DRP2 family has evolved multiple activities, with the Solanaceous DRP2 being involved in FLS2 but not BRI1 internalization (Fig 5 and panel C in S5 Fig). Possibly, plant DRPs have diversified to enable increased plasticity in response to biotic and abiotic stimuli. Further studies are clearly needed to fully understand the precise contributions of different plant DRPs to vesicle trafficking and flg22 responses.
In summary, we link the P. infestans effector AVR3a to a membrane complex that includes the vesicle trafficking protein DRP2. We found evidence supporting an active role of DRP2 during endocytosis of the classic pattern recognition receptor FLS2 following activation by the flagellin-derived peptide flg22. Although the FLS2 co-receptor BAK1/SERK3 is required for basal immunity against the oomycete P. infestans, there is no evidence that FLS2, a receptor for bacterial flagellin, is activated and internalized during infection by this oomycete pathogen [39]. Therefore, AVR3a suppression of FLS2 responses and endocytosis may indicate that this effector targets a common node shared by FLS2 and a yet to be discovered PRR involved in oomycete immunity. Future studies are required to address this possibility and further determine the various mechanisms by which AVR3a perturbs PTI signaling. The Sainsbury Laboratory and the John Innes Centre are registered with the Health and Safety Executive (site reference GM38) to use the laboratory premises, including growth rooms and controlled environment rooms (CERs) under containment level 1 and 2 for work involving genetically modified organisms. We confirm that our study did not involve any endangered or protected species.

A. tumefaciens-mediated transient gene expression assays in N. benthamiana
Agrobacterium tumefaciens (strain GV3101) carrying the desired T-DNA construct was grown overnight at 28°C in Luria-Bertani culture medium with the appropriate antibiotics. Cells were harvested by centrifugation at 8000 g and resuspended in agro-infiltration media [5 mM MES, 10 mM MgCl 2 , pH 5.6] prior to syringe infiltration into leaves of 3-4 week-old N. benthamiana plants. Bacteria carrying each construct were infiltrated at a final OD 600nm of 0.3. Acetosyringone was added to the resuspended cultures at a final concentration of 150 μM and bacterial cultures were left at room temperature for 2 hours before infiltration.
For transient silencing experiments, the same 445-nucleotide cDNA fragment of NbDRP2-1 used for VIGS was cloned into pENTR/D-TOPO (Invitrogen) using the following primers: NbDRP2-1_hp_F2, 5'-CACCATCAGCTCTAAAGGCGGTCA and NbDRP2-1_hp_R2, 5'-GCTGTTGGGCTACTTTCTGC. The hairpin-silencing construct was generated by recombination into pHellesgate8 (Gateway LR recombination, Invitrogen) as described by Helliwell and Waterhouse [97]. The same procedure described above was used to generate the control silencing construct pHellsgate8-GUS (574 bp fragment size) using primers pH8_GUS_F1, 5'-CACCCCAGGCAGTTTTAACGATCAG and pH8_GUS_R1, 5'-GATTCACCACTTGC AAAGTCC. The final constructs were transformed into A. tumefaciens GV3101. Four-week-old N. benthamiana leaves were infiltrated with the hairpin silencing construct at a final OD 600 = 0.3 either individually or co-expressed with a construct expressing the plant receptor under assessment. Further experiments (microscopy, analysis of gene silencing by qRT-PCR, PAMP elicitation and others described elsewhere in Materials and Methods) were performed three days after silencing.

Gene expression analysis
Transgenic N. benthamiana leaves expressing the constructs pBinplus::FLAG-AVR3a KI , pBinplus::FLAG-AVR3a EM , pBinplus::FLAG-AVR3a KI-Y147del or pBinplus::ΔGFP were treated for 0 and 180 minutes with flg22 (100 nM) or INF1[Pi] (10 μg/ml) on one side of the leaf and with Milli-Q water on the other side of the leaf as control. Total RNA was extracted using TRI reagent (Invitrogen) following manufacturer's instructions. DNase treatment (Ambion) was performed according to manufacturer's protocol and total RNA was quantified with a Nanodrop spectrophotometer (Thermo). 1.5 μg of DNase treated RNA was used for cDNA synthesis using SuperScript II reverse transcriptase (Invitrogen). qRT-PCR was performed with SYBR Green master mix (SIGMA) in triplicate per sample per gene. NbEF1α was used to normalize transcript abundance for the marker genes NbACRE132, NbCYP71D20 and NbACRE31 [72]. Primers used for amplification of NbEF1α and PAMP-induced marker genes in N. benthamiana have been reported previously [72].

Elicitor preparations
Chitin (crab shell chitin) and flg22 (QRLSTGSRINSAKDDAAGLQIA) peptides were purchased from SIGMA and EzBiolab, respectively and dissolved in ultrapure water. INF1[Pi] was purified from P. infestans 88069 by chromatography and the final working solution was dissolved in ultrapure water [36,98].
Sample preparation and liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis was performed as described in Caillaud et al., [100]. Candidate AVR3a-associated proteins were identified from the peak lists searched on Mascot server v.2.4.1 (Matrix Science) against an in-house N. benthamiana database available upon request. We followed the same criteria in Caillaud et al., [100] for peptide identification. A complete list of all identified peptides for each AVR3a KI -associated protein is provided in (S3 Table).

Cloning of NtDRP2-1/2
The AVR3a KI -associated protein NCBE_074039.1 sequence identified through the LC-MS/MS analysis was used as query to search the TAIR9 database. Arabidopsis thaliana proteins DRP2A and DRP2B were then used as query to perform a TBLASTN in the Nicotiana benthamiana draft genome scaffolds and tomato genome database (Solanum Genomic Network (SGN)). Partial sequences were used to design the primers Dyn_F3, 5'-CACCATGGAAGCAATCGAGGAA TTGGAGCAG and Dyn_R2, 5'-TTATGATCTATAACCAGATCCAGACTGTGGTGG to amplify the open reading frame of the Nicotiana tabacum putative homologs of AtDRP2A/B from cDNA using Phusion proof reading polymerase (New England Biolabs). Amplicons were cloned into pENTR/D-TOPO (Invitrogen) and sequenced. Clones representing NtDRP2-1 and NtDRP2-2 were recombined into pK7WGF2, pK7WG2, and pK7WG2-3xHA to generate fusion proteins GFP-NtDRP2-1/2, non-tagged NtDRP2-1/2, and HA-NtDRP2-1/2, respectively. The final constructs were transformed into A. tumefaciens GV3101 and used for further analysis.

Phylogenetic analysis
In Arabidopsis, GTPases with a PH domain belong to the DRP2 subfamily [81]. Members of this subfamily, AtDRP2-A, AtDRP2-B, and the NtDRP2-1/2 cloned protein were used as queries to search for homologs of DRP2 in Solanaceous and other dicot plants. We performed TBLASTN searches against the Nicotiana benthamiana genome version 0.4.4, the tomato genome International Tomato Annotation Group release 2.3, the potato genome Potato Genome Sequencing Consortium DM 3.4 (solanaceous genomes available at http:// solgenomics.net) and the NCBI database. To reveal the evolutionary relationship of these proteins, an alignment of the conserved GTPase domain was constructed using the multiple alignment software MUSCLE [101] and a maximum likelihood (ML) tree was generated using the RAxML software [102]. To choose the best model for the ML tree, we estimated likelihoods of the trees constructed based on commonly used substitution models. The ML tree in S7 Fig was  constructed with GAMMA model of rate heterogeneity and Whelan and Goldman model (WAG), since the tree based on WAG showed the highest likelihood. We performed 500 nonparametric bootstrap inferences. Bootstrap values over 70% are shown. All Arabidopsis DRPs were included in the analysis to ensure proper estimation of the closest homologs of AtDRP2A/B. Sequences identifiers are available in S3 Table. Statistical analysis Statistical analyses (One-way ANOVA, TukeyHSD test, Wilcoxon-Mann-Whitney Test, and Student's t Test) were performed using the software package R [103].

Confocal microscopy
Standard confocal microscopy was carried out in N. benthamiana epidermal cells 2 to 3 days post infiltration. Cut leaf pieces were mounted in water and analyzed with a Leica DM6000B/ TCS SP5 microscope (Leica Microsystems, Germany) with laser excitation settings of 488-nm for eYFP and GFP and 561-nm for RFP. Fluorescent emissions were taken at 500-550 nm for GFP and 580-620 nm for RFP. The 63x water immersion objective was used to acquire all images. Image analysis was done with the Leica LAS AF software, ImageJ and Adobe Photoshop CS5. For all Z-stacks, a distance of 1 μm was set. For induction of FLS2 internalization, flg22 peptide (EzBiolab) was gently infiltrated into leaves expressing FLS2-GFP at a concentration of 100 μM [44] and imaging was done 80-150 minutes post elicitation. All images were processed with ImageJ (2.0) and endosomes were quantified based on naked-eye detection of distinct fluorescent signal in punctate structures per total image area (136 μm × 136 μm) using the multi-point tool. Propidium iodide (PI, SIGMA) staining was performed as previously described [104] with few modifications: leaf tissue was incubated in a solution of propidium iodide (200 μg/ml) for 10 minutes at room temperature, followed by three washing steps in ultra-pure water. To visualize PI-stained cells, laser excitation was set at 488-nm and fluorescence was detected between 598-650 nm. Results are average ± SE (n = 3 technical replicates). AVR3a variants were transiently expressed in N. benthamiana using the following constructs: FLAG-AVR3a KI (red), FLAG-AVR3a EM (blue), FLAG-AVR3a KI-Y147del (grey) or vector control (ΔGFP) (green). (E, F) Western blots probed with anti-FLAG antibody after flg22 (E) or INF1 (F) treatment detected total protein expression of AVR3a variants. (TIF) S2 Fig. AVR3a suppresses elf18-ROS production but does not affect chitin-triggered ROS accumulation. (A, B) N. benthamiana agro-infiltrated with FLAG-AVR3a KI (red), FLAG-AV-R3a EM (blue), FLAG-AVR3a KI-Y147del (grey) or vector control (ΔGFP) (green). Leaf discs were incubated in an elf18 (A) or chitin (B) containing solution and ROS production was measured in relative light units (RLU) over time. Letters above the graph indicate statistical significant differences at P < 0.05 assessed by one-way ANOVA followed by TukeyHSD test. No statistical significance was found for group b (P = 0.068). Similar results were observed in two independent experiments. HA-NtDRP2-1 was transiently co-expressed with FLAG-AVR3a KI , FLAG-PcAVR3a-4 or FLAG-RFP (control) in N. benthamiana and immunoprecipitated with anti-FLAG antiserum (SIGMA). Immunoprecipitates and total protein extracts were immunoblotted with the appropriate antisera. (B) Oxidative burst triggered by 100 nM flg22 in N. benthamiana agroinfiltrated with members of the Avr3a family FLAG-AVR3a KI , FLAG-PcAVR3a-4 or FLAG-RFP (control). ROS production was measured in relative light units (RLU) over time and depicted relative to the total ROS burst of the control. Values are average ± SE (n = 16). Statistical significance was evaluated in comparison to the control by one-way ANOVA followed by TukeyHSD test. ÃÃÃ P < 0.001. Experiment was repeated 3 times with similar results. (TIF)