Transgenic Expression of the Dicotyledonous Pattern Recognition Receptor EFR in Rice Leads to Ligand-Dependent Activation of Defense Responses

Plant plasma membrane localized pattern recognition receptors (PRRs) detect extracellular pathogen-associated molecules. PRRs such as Arabidopsis EFR and rice XA21 are taxonomically restricted and are absent from most plant genomes. Here we show that rice plants expressing EFR or the chimeric receptor EFR::XA21, containing the EFR ectodomain and the XA21 intracellular domain, sense both Escherichia coli- and Xanthomonas oryzae pv. oryzae (Xoo)-derived elf18 peptides at sub-nanomolar concentrations. Treatment of EFR and EFR::XA21 rice leaf tissue with elf18 leads to MAP kinase activation, reactive oxygen production and defense gene expression. Although expression of EFR does not lead to robust enhanced resistance to fully virulent Xoo isolates, it does lead to quantitatively enhanced resistance to weakly virulent Xoo isolates. EFR interacts with OsSERK2 and the XA21 binding protein 24 (XB24), two key components of the rice XA21-mediated immune response. Rice-EFR plants silenced for OsSERK2, or overexpressing rice XB24 are compromised in elf18-induced reactive oxygen production and defense gene expression indicating that these proteins are also important for EFR-mediated signaling in transgenic rice. Taken together, our results demonstrate the potential feasibility of enhancing disease resistance in rice and possibly other monocotyledonous crop species by expression of dicotyledonous PRRs. Our results also suggest that Arabidopsis EFR utilizes at least a subset of the known endogenous rice XA21 signaling components.


Introduction
Plants possess multi-layered immune systems enabling them to fend off most pathogens. Plasma membrane localized pattern recognition receptors (PRRs) sense danger-associated molecules, including pathogen/microbe-associated molecular patterns (P/MAMPs) and endogenous elicitors released during the infections process. The activation of PRRs triggers a rapid intracellular signaling cascade [1][2][3]. In most cases, PRR-triggered immunity (PTI) is sufficient to halt microbial replication and disease development [4,5]. Successful pathogens are able to suppress PTI by employing effector molecules in the apoplast or inside the plant cell and thereby enable plant colonization. Plants, in turn, recognize specific effector molecules either in the apoplast via transmembrane receptors or inside the cell via intracellular immune receptors, which are part of the nucleotide binding site-leucine rich repeat (NBS-LRR) protein family. This recognition event is often referred to as effector triggered immunity (ETI) [6,7,4,8].
In contrast to the wide conservation of the flagellin-FLS2 perception system, other PRRs and their respective PAMP recognition specificity are restricted to limited plant families or species [1,3,15]. For example, EF-TU RECEPTOR (EFR; At5g20480), the receptor that recognizes the highly conserved N-terminal 18 amino acids of EF-Tu, originally isolated from Escherichia coli (from here on referred to as elf18 E.coli ), is restricted to the plant family Brassicaceae [26,27]. Similarly, XA21 (U37133), which recognizes the bacterial pathogen Xanthomonas oryzae pv. oryzae (Xoo), appears to be restricted to the wild rice species Oryza longistaminata [28][29][30]. These taxonomically restricted PRRs are prime candidates for inter-species transfer, because they may be able to confer resistance to a wide range of pathogens for which there is presently no disease control measures. Indeed, it was recently shown that the inter-family transfer of EFR and XA21 are phylogenetically closely related (S1 Fig) [60] and share some, but not all, orthologous signaling components. For example, both receptor require the ER quality machinery for proper folding and function [32][33][34][35][36]51,52]. Orthologous SERK family members interact with both receptors and are required for downstream signaling initiation in both Arabidopsis and rice [41,42,54]. In both cases, WRKY transcription factors are involved in transcriptional reprogramming [50,58]. However, it is not yet known if EFR activity in Arabidopsis is down regulated by dephosphorylation, if an ATPase is required for its inactivation in the absence of the ligand or if an E3 ligase is important for its stability. It is therefore very difficult to predict if EFR would be functional when transgenically expressed in rice and if it would employ the same signaling components as the endogenous rice PRR XA21.
Here we report that the expression of EFR or the chimeric receptor EFR::XA21 makes rice receptive to elf18 Ecoli at sub-nanomolar concentrations, inducing MAP kinase activation, ROS production, and defense gene expression. We show that EF-Tu is highly conserved in over twenty different Xoo isolates and that elf18 Xoo , which carries two amino acid substitutions at position 2 and 4 in comparison with elf18 E.coli , is fully recognized by rice plants expressing EFR at similar sub-nanomolar concentrations. The recognition of elf18 Xoo in rice plants expressing EFR leads to moderate quantitatively enhanced resistance to weakly virulent isolates of Xoo. In contrast, the EFR-rice plants were only slightly more resistant to fully virulent isolates of Xoo in 3 out of 6 experiments. Surprisingly, rice-EFR::XA21 plants did not become fully resistant to Xoo instead displaying a weak resistance profile similar to rice-EFR plants. We further demonstrate that EFR directly interacts with two signaling components of XA21, OsSERK2 and XB24, and both are required for EFR signaling in rice.

Results
Expression of EFR or the chimeric receptor EFR::XA21 in rice confers responsiveness to elf18 E.coli We generated two constructs to test if the expression of the EFR ectodomain enables rice to sense and respond to elf18 E.coli . The first construct expresses full-length Arabidopsis EFR with a carboxyl-terminal green fluorescent protein (GFP) fusion under the control of the maize ubiquitin promoter (Ubi::EFR::GFP) [26]. The second construct expresses the EFR ectodomain (EFR 1-649 aa) fused to the XA21 transmembrane, juxtamembrane and intracellular domains (XA21 651-1025 aa) with a carboxyl-terminal GFP fusion under the control of the maize ubiquitin promoter (Ubi::EFR::XA21::GFP) (S2 Fig) [26,28]. We reasoned that the XA21 kinase domain might be more adapted to intracellular signal initiation in rice. We obtained six independent PCR positive T 0 rice transformants for Ubi::EFR::GFP and five in the case of Ubi:: EFR::XA21::GFP. Four of the T 0 lines for Ubi::EFR::GFP (-4, -6, -7 and -9) and three of the T 0 lines for Ubi::EFR::XA21::GFP (-1, -3 and -4) expressed detectable full length protein in the T 1 generation (S3 Fig). The molecular weight of both GFP fusion proteins was similar to that of XA21::GFP of~175kDa. This is well above the predicted molecular weight of~140kDa and suggests that the ectodomain of EFR undergoes post-translation modification when expressed in rice. This is similar to the observation made in Arabidopsis where EFR also migrates well above its predicted molecular weight due to the glycosylation of its extracellular domain. It was previously shown that proper glycosylation of EFR is essential for its function [26,32,[34][35][36]. We next chose three PCR-positive lines for each construct (T 0 lines Ubi::EFR::GFP -1, -7, -9 and Ubi::EFR::XA21::GFP -2, -3, -4) and performed in depth functional analyses of the T 1 progeny. Based on the transgene segregation analysis in the T 1 , all T 0 lines carry a single T-DNA insertion (S1 Table). First, we assessed the transcript level of EFR::GFP or EFR::XA21::GFP with primers annealing to the sequence corresponding to the EFR ectodomain. As shown in Fig 1A, all lines except for Ubi::EFR::GFP-1 expressed the transgene to comparable levels. We also tested the expression of both constructs at the protein level. Consistent with the qRT-PCR results, we were not able to detect any protein in line Ubi::EFR::GFP-1. All other lines expressed fulllength EFR::GFP or EFR::XA21::GFP (Figs 1B and S4). To assess if the expression of EFR::GFP or EFR::XA21::GFP enables rice to sense and respond to elf18 Ecoli , we measured the expression of two well-established rice defense marker genes PR10b and Os04g10010 [54] in response to 500 nM elf18 Ecoli in mature leaves of 4-week old T 1 plants derived from Ubi::EFR::GFP-1, -7 and -9 and Ubi::EFR::XA21::GFP -2, -3 and -4. Only lines expressing EFR::GFP or EFR::XA21:: GFP showed induction of PR10b and Os04g10010 expression in response to elf18 E.coli treatment (Fig 1C and 1D). The absence of elf18 Ecoli -triggered defense gene expression in Kitaake and Ubi::EFR::GFP-1 plants clearly demonstrates that expression of the EFR-ectodomain is required to confer elf18 Ecoli responsiveness. PR10b (Os12g36850) expression was significantly higher in all three lines expressing EFR::XA21::GFP when compared to EFR::GFP expressing lines. This observation is supported by the fact that even the high expression level of EFR::GFP in line 7 does not lead to induction of PR10b expression to the same level as in Ubi::EFR:: XA21::GFP plants (Fig 1B and 1C).
Based on these initial observations, we focused on one line per construct, Ubi::EFR::GFP-9 and Ubi::EFR::XA21::GFP-4, for our next set of experiments (Figs 2 and 3) with the aim of assessing the signaling capacity of both receptor proteins. These and all other transgenic lines Os04g10010 (D) in response to elf18 E.coli (500 nM) in mature leaves of the three independent Ubi::EFR::GFP and Ubi::EFR::XA21::GFP lines. Expression levels were measured by qRT-PCR and normalized to actin reference gene expression. Data shown is normalized to the Kitaake mock treated (2 hour) sample. Bars depict average expression level ± SE of three technical replicates. This experiment was repeated twice with similar results. Statistical analysis was performed using the Tukey-Kramer HSD test. Different letters indicate significant differences (p < 0.05).
used in this study arose from a single T-DNA insertion event as shown by segregation analysis of individual T1 populations (S1 Table). Because these lines were still segregating in the T 1 and T 2 generation we confirmed the presence of the transgene by PCR and performed experiments on PCR positive individuals only. In addition, transgene expression was confirmed by qRT-PCR and found to be stable over multiple generations.
In plants, the kinase domains of several PRRs, including EFR in Arabidopsis, induce MAP kinase activation within minutes of ligand perception [2]. The activation of MAP kinases in Arabidopsis is often measured by an increase of the doubly phosphorylated isoforms detected by an anti-phospho antibody recognizing the two highly conserved activation loop phosphorylation sites pTXpY (where pT or pY represents a phosphorylated threonine or tyrosine, respectively, and x any amino acid) [41]. Recently, this assay has also been established for rice and it was shown that MAP kinases are activated upon treatment with chitin, flg22 Pta , and other MAMPs [61][62][63][64]. Before testing the effect of elf18 E.coli on MAP kinase activation, we first attempted to reproduce MAP kinase activation by flg22 Pta and chitin in mature leaves of 4-weekold Kitaake rice plants treated with 1 μM flg22 Pta or 50 μg/ml chitin for 0, 5, 10, 15, 30 and 60 minutes. Using the anti-phospho p44/42-antibody, we detected an increase of two distinct bands of an approximate molecular weight of~47 kDa and 40 kDa after treatment for at least 15 minutes to 30 minutes, respectively (S5 Fig). The higher molecular weight MAP kinase band was already visible without treatment at 0 minutes, which is in agreement with previous reports [61][62][63][64].
Next we tested if elf18 E.coli treatment of rice leaves expressing EFR::GFP or EFR::XA21::GFP would lead to activation of MAP kinases using the same phosphosite-specific antibody. We treated mature leaves of 4-week-old Kitaake, Ubi::EFR::GFP-9 and Ubi::EFR::XA21::GFP-4 rice plants with 1 μM elf18 E.coli for 0, 5, 10, 15, 30 and 60 minutes. We observed a similar activation pattern as observed for chitin and flg22 treatment for both MAP kinases in plants expressing EFR::GFP or EFR::XA21::GFP but not in the Kitaake wild-type plants (Fig 2). The lack of an endogenous receptor for elf18 E.coli makes Kitaake rice plants insensitive to this elicitor. These The perception of elf18 E.coli in EFR and EFR::XA21 rice plants activates several MAP kinases and ROS production. Fully mature leaves of Ubi:: EFR::GFP-9-11-12 (A), Ubi::EFR::XA21::GFP-3-6-7 (B) and Kitaake (C), lines were treated with 1 μM elf18 E.coli for the indicated time. Upper panel anti-p42/44 MAP kinase western blot on total protein extracts, lower panel CBB stain of membrane as loading control. (D) Elf18 E.coli -triggered ROS production in fully mature leaves of Ubi::EFR::GFP-9-11-12, Ubi::EFR::XA21::GFP-3-6-7 but not Kitaake. Leaves of the indicated genotypes were treated with water (mock) or elf18 E.coli at a final concentration of 100 nM. Data points depict average relative light production ± SE of at least six biological replicates. These experiment were repeated at least three times with similar results.
observations indicate that the kinase domains of EFR and of XA21 are able to activate MAP kinases in rice in a temporal and ligand-dependent manner.
Similarly to MAP kinase activation, the production of ROS mediated by plasma membranelocalized NADPH oxidases is a frequently utilized measurement to assess PTI signaling [2]. We therefore tested if the expression of EFR and EFR::XA21 in fully mature rice leaves leads to ROS production after treatment with 100 nM elf18 E.coli . We found that only rice plants expressing EFR::GFP or EFR::XA21::GFP triggered ligand-dependent ROS production whereas Kitaake control plants did not respond to elf18 E.coli treatment (Figs 2D and S6). The ROS production peaked around 45 minutes and lasted for about 150 minutes with both lines producing the same amount of total ROS (Figs 2D and S6). Using the ROS production assay, we determined the EC50 value of EFR::GFP and EFR::XA21::GFP expressing rice plants. The expression of either receptors confers a highly sensitive chemoperception system for elf18 E.coli in rice with an EC 50 value of~300 pM (Table 1, S7 Fig), which is very similar to the sensitivity observed in Arabidopsis [27].
Rice plants expressing EFR or EFR::XA21 recognize elf18 Xoo from Xoo Many bacterial species carry two highly similar copies of the tuf gene that encodes EF-Tu. In the case of Xoo, three full genome sequences are publicly available (PXO99A, KACC10331 and MAFF311018) [65][66][67]. We investigated the coding sequence for both tuf genes, PXO_04524 Expression levels were measured by qRT-PCR and normalized to actin reference gene expression. Data shown is normalized to the Kitaake mock treated (2 hour) sample. Bars depict average expression level ± SE of three technical replicates. Fully mature leaves of Ubi::EFR::GFP-9-11-12 (C) and Ubi::EFR:: XA21::GFP-3-6-7 (D) lines were treated with 1 μM elf18 E.coli for the indicated time. Upper panel anti-p42/44 MAP kinase western blot on total protein extracts, lower panel CBB stain of membrane as loading control. (E) Total ROS production over 3 hours in response to elf18 E.coli or elf18 Xoo at a concentration of 100 nM in mature leaves of Kitaake, Ubi::EFR::GFP-9-11-12 and Ubi::EFR::XA21::GFP-3-6-7 lines. Bars depict average relative light production ± SE of at least six biological replicates. Statistical analysis was performed using the Tukey-Kramer HSD test. Different letters indicate significant differences (p < 0.05). These experiments were repeated at least twice with similar results. and PXO_04538, in the Xoo PXO99A genome, and for non-synonymous mutations that lead to changes in the 18 N-terminal amino acids (elf18), when compared to the elf18 E.coli amino acid sequence. The elf18 Xoo sequence carried two substitutions when compared to elf18 E.coli (S 2 -> A 2 and E 4 -> A 4 ) in both gene copies in all 3 genomes. In addition to the publically available sequences, we analyzed the first~700 bp of both EF-Tu genes in 20 Xoo isolates from our laboratory collection by standard Sanger sequencing (S2 Table). The first 230 N-terminal amino acids of both EF-Tu proteins were 100% conserved in all 23 Xoo isolates giving rise to a single EF-Tu Xoo sequence as shown in S8 Fig [10]. We next tested if elf18 Xoo would be recognized by the EFR ectodomain. We measured defense gene expression in mature leaves of 4-week-old rice plants after treatment with 500 nM elf18 Xoo or elf18 E.coli for 2 and 12 hours. PR10b and Os04g10010 were up-regulated only in lines expressing EFR::GFP and EFR::XA21:: GFP but not the Kitaake control (Fig 3A and 3B). Both elf18 Xoo and elf18 E.coli peptides induced a similar defense gene expression levels at all time points and in both lines (Fig 3A and 3B). As observed previously (Fig 1), the Ubi::EFR::XA21::GFP plants induces a higher PR10b expression at 12 hours when compared to Ubi::EFR::GFP plants. The recognition of elf18 Xoo in mature leaves of 4-week-old rice plants expressing EFR::GFP and EFR::XA21::GFP also induced MAP kinase activation (Fig 3C and 3D) and ligand dependent ROS production ( Fig 3E). The sensitivity of EFR and EFR::XA21::GFP plants towards elf18 Xoo was nearly identical with the one observed for elf18 E.coli with an EC 50 value of~300 pM and~400 pM, respectively (Table 1 and  These results suggest that the EFR ectodomain may be able to sense EF-Tu from Xoo during the infection process. We therefore tested if EF-Tu Xoo is present in cell free Xoo supernatants. We detected full-length EF-Tu Xoo in cell free supernatants using a commercially available antibody and by mass-spectrometry analysis (S9 Fig). This observation is consistent with a recent report that identified EF-Tu Xoo in the cell-free xylem sap of rice plants infected with Xoo [68]. Therefore, we hypothesize that EF-Tu from Xoo is readily available for detection by EFR and EFR::XA21 during the infection process.

Transgenic expression of EFR in rice does not negatively affect growth and yield
In some instances, transgenic expression of defense related genes has a negative impact on the plant's growth and yield. To test whether the transgenic expression of EFR or EFR::XA21 has a negative impact on rice growth and yield, we grew wild type Kitaake plants next to Ubi::EFR:: GFP and Ubi::EFR::XA21::GFP lines until maturity and measured total dry biomass and yield. S10 Fig shows that two independent transgenic rice lines expressing EFR (Ubi::EFR::GFP-7-8-8 and Ubi::EFR::GFP-9-4-3-13) do not differ in total biomass or yield compared to the wild type Ubi::EFR::GFP~300 pM~300 pM Ubi::EFR::XA21::GFP~300 pM~400 pM parent. In contrast, EFR::XA21::GFP expressing plants (Ubi::EFR::XA21::GFP-3-8-7-20 and Ubi::EFR::XA21::GFP-4-5-4) do suffer from growth defects such as necrosis of older leaves and stunting starting at the 5-week stage under our greenhouse conditions. Although the overall biomass and yield of EFR::XA21::GFP plants was reduced at maturity (S10 Fig Table). This differential gene expression list includes 115 up-regulated and 26 down-regulated genes. The defense marker gene Os04g10010 was not included in the set of up-regulated genes in Ubi::EFR::XA21:: GFP plants in the absence of the ligand treatment consistent with our previous observations (Figs 1 and 3). In contrast, PR10b expression was up-regulated by 2.17-fold, which is just above our 2.00-fold log fold-change cut-off (S3 Table). We observed a similar slight up-regulation of Transgenic rice plants expressing the EFR receptor are more resistant to two weakly virulent Xoo isolates The expression of EFR and EFR::XA21 enables rice to sense and respond to elf18 Xoo at subnanolmolar concentrations (Fig 3, Table 1). Moreover, EF-Tu Xoo is most likely readily available for EFR recognition during the infection process (S9 Fig) [69]. To determine whether the transgenic expression of EFR or EFR::XA21 in rice confers enhanced resistance to Xoo, we inoculated Ubi::EFR::GFP and Ubi::EFR::XA21::GFP transgenic plants and compared the length of disease lesions with that of Kitaake plants. EFR expression did not confer robust resistance to the fully virulent PXO99A isolate. We tested three independent EFR lines: 3-6 (T1), 7-8-8 (T2) and 9-11-2 (T2)/9-4-3-13 (T3), referred to as lines 3, 7 and 9, respectively, as detailed in Table 2. In 5 out of 8 infections of Ubi::EFR::GFP plants with PXO99A, we observed a moderate, but statistically significant, reduction in lesion length as shown in Fig 4A. In planta bacterial growth curve analysis revealed no statistical difference in PXO99A populations between EFR and Kitaake lines in three independent experiments (S13 Fig). When we attempted to perform similar inoculation assays with Ubi::EFR::XA21::GFP lines, we had difficulties maintaining healthy plants throughout the course of inoculation starting at the 6-week-old stage (S10 Fig). However, in two experiments we did obtain healthy plants and carried out the inoculation assays in full. In these assays (see experiment number II and number VI in Table 2) Ubi:: EFR::XA21::GFP plants were as susceptible to Xoo PXO99A infection as Kitaake control plants.
These results indicate that despite the ability of the EFR and EFR::XA21 chimeric receptors to detect EF-Tu Xoo in the detached leaf assay (Fig 3), this recognition does not confer robust resistance to Xoo PXO99A in whole rice plants.
Because the leaf-clipping infection assay might mask subtle differences in resistance, we sought to use less aggressive infection assays. We took two approaches to address this issue. In our first approach we inoculated plants with a lower dose of the fully virulent PXO99A isolate. In our second approach we inoculated with weakly virulent Xoo isolates. Inoculation with a lower concentration of the PXO99A isolate (10 6 CFU/mL instead of 10 8 CFU/mL) did not result in statistically significant differences in disease lesion length between wild type and EFR transgenic plants (S4 Table). We next screened 10 different Xoo isolates from our lab collection for their level of virulence on Kitaake plants. We identified three isolates that were significantly less virulent than PXO99A and other fully virulent isolates (S4 Table). Two of these weakly virulent isolates (NXO256 and MXO90), were significantly less virulent on rice lines expressing EFR when compared with wild-type Kitaake plants ( Fig 4B, Table 2). Ubi::EFR::GFP lines (7  Table 2). Plants were infected using the leafclipping assay at the 5-to 6-week-old stage. Lesions were measured at 14 dpi. On the right hand side of each panel two representative inoculated leaves at the time of lesion scoring for either Kitaake or Ubi::EFR::GFP transgenic line. Statistical analysis was performed using the Tukey-Kramer HSD test. Different letters indicate significant differences within each Isolate. These experiments were repeated at least twice with similar results.
doi:10.1371/journal.ppat.1004809.g004 and 9) were statistically significantly more resistant to isolate MXO90 (shorter lesions) in 8 out of 8 inoculations and more resistant to isolate NXO256 in 5 out of 6 inoculations ( Table 2). These moderately enhanced resistance phenotypes were caused by the expression of EFR in the Ubi::EFR::GFP lines. The rice line Ubi::EFR::GFP-1, which does not express EFR, is not responsive to elf18 E.coli (Fig 1) and is not resistant to Xoo (Fig 4, Table 2 experiment V). The moderate resistance conferred by the Ubi::EFR::GFP lines, as measured by lesion lengths, was further supported by in planta growth curves (S13 Fig). These results indicate that the recognition of elf18-Xoo by EFR leads to a reduction of bacterial populations of these two weakly virulent isolates. When we tested the Ubi::EFR::XA21::GFP lines with these two isolates (see experiment number II and number VI in Table 2), the lesion lengths were between those obtained with Kitaake and those obtained with Ubi::EFR::GFP transgenic lines. In the case of the Xoo isolate NXO256, no significant differences in lesion length could be observed on Ubi::EFR::GFP transgenic lines. For the Xoo isolate MXO90, statistically significant differences between Kitaake and Ubi::EFR:: GFP transgenic lines were observed 1 out 2 experiments (Table 2). In summary, while the expression of EFR and EFR::XA21 does not confer robust resistance to fully virulent Xoo isolate PXO99A, expression of EFR provides quantitatively enhanced resistance to two weakly virulent Xoo isolates.
We found that the XA21 ID (668-1025aa) interacted with all four proteins (S14 Fig) as previously reported [53,54,56,59]. Next, we tested the interaction of EFR ID with the same four proteins. In these experiments, the EFR ID interacted with XB24 but not XB3, XB15 or OsSERK2 ID (Fig 5A). The expression of all fusion proteins in yeast was confirmed by western blot analysis (S15 Fig). The interaction of XA21 with XB24 is dependent on the catalytic activity of the XA21 kinase [53]. We tested if this is also the case for the EFR/XB24 interaction by yeast-two hybrid analysis between catalytically inactive EFR(D848N) ID and XB24. For this purpose, we mutated the conserved aspartate at position 848 in EFR, which was previously shown to be required for catalytic activity [41], to an asparagine. In the yeast two-hybrid system, EFR(D848N) was still able to directly interact with XB24 ( Fig 5A). This suggests that the interaction between EFR and XB24 is independent of the kinase catalytic activity of EFR. Next, we aimed to confirm the interaction between XB24 and EFR in planta. In the absence of a suitable antibody for XB24, we decided to test this interaction in N. benthamiana after co-expression of tagged versions of each protein using Agrobacterium tumefaciens-mediated transient expression. As shown in Fig 5B, XB24::FLAG specifically co-immunoprecipitated with EFR:: GFP. The association was unaltered by treatment with 100 nM elf18 E.coli for 10 minutes.
The absence of a direct interaction between the EFR ID and OsSERK2 ID in the yeast twohybrid system is surprising because orthologous SERK family members in rice and Arabidopsis interact with XA21 or EFR in vivo and are required for XA21-and EFR-mediated immune responses [37,[40][41][42]54]. We therefore hypothesized that we may be able to detect the interaction in planta. For this purpose, we used our recently developed specific anti-OsSERK2 antibody  (1-198aa). EFR(D-N) indicates a mutation of the catalytic aspartate (D) 848aa to asparagine (N). The blue color indicates nuclear interaction between the two coexpressed proteins. Expression of all fusion proteins was confirmed by western blot analysis as shown in S15 Fig. (B) In planta interaction of EFR::GFP and XB24::FLAG in N. benthamiana. EFR::GFP and XB24::FLAG were transiently co-expressed in N. benthamiana leaves. Leaves were mock treated with water (-) or with 100 [54], to test for the interaction between EFR::GFP and EFR::XA21::GFP in leaf strips of 4-week-old plants after treatment with 1 μM elf18 E.coli or elf18 Xoo for 0, 5 and 15 minutes. We choose these time points based on previous interaction data reported for EFR-AtSERK3/BAK1 (At4g33430) at 5 minutes [38,41,42] and on in vivo data demonstrating initial MAP kinase activation within 15 minutes of elf18 E.coli treatment (Fig 2). In immunoprecipitation experiments with anti-GFP agarose, we detected proteins at the expected size of 175kDa for full-length EFR::GFP and EFR::XA21::GFP using anti-GFP antibody only in transgenic plants and not in Kitaake control plants ( Fig 5C). Next, we tested for the presence of OsSERK2 in the anti-GFP immunoprecipitates using anti-OsSERK2 antibody. OsSERK2 was readily detectable in all immunoprecpitates from EFR::GFP or EFR::XA21::GFP expressing plants even in the absence of elf18 treatment but not in immunoprecipitates from Kitaake control plants (Fig 5C). No increase in co-immunoprecipitated OsSERK2 could be observed within 15 minutes of elf18 E.coli treatment (Fig 5C). This is consistent with our previous observation that XA21 and OsSERK2 form constitutive heteromeric complexes in the same plant tissue [54]. In contrast, the interaction between EFR and AtSERK3/BAK1 is clearly ligand-induced in Arabidopsis and after transient co-expression in N. benthamiana [41,42]. These interaction studies indicate that EFR in rice utilizes at least a subset of the XA21-signaling components.

OsSERK2 and XB24 regulate EFR signaling in rice
Based on the interaction of EFR with OsSERK2 and XB24, we tested if OsSERK2 and XB24 are also involved in EFR-signaling in rice by assessing double transgenic lines of Ubi::EFR::GFP with altered expression of OsSERK2 and XB24. Because OsSERK2 directly interacts with EFR, we tested if OsSERK2 is also a positive regulator of EFR signaling in rice. We crossed Ubi::EFR:: GFP-9-2 expressing lines with previously characterized OsSERK2RNAi silencing lines [54]. In the F2 generation, we isolated double transgenic lines from two independent F1 plants (67 and 71) by PCR using primers specific for each transgene and confirmed stable expression of the EFR transgene by qRT-PCR (S16A Fig). We compared elf18 E.coli -induced defense gene expression in plants expressing EFR::GFP and silenced for OsSERK2 (Ubi::EFR::GFP x OsSERK2R-NAi) with plants expressing only EFR::GFP (Ubi::EFR::GFP). We treated leaf strips of 4-weekold plants from both genotypes with water or 500 nM elf18 E.coli for 2 and 12 hours. As shown in Fig 6A and 6B, elf18 E.coli -induced PR10b and Os04g10010 expression was significantly reduced in both independent double transgenic lines Ubi::EFR::GFP x OsSERK2RNAi-67 and -71 expressing EFR::GFP and silenced for OsSERK2 at both time points when compared with EFR:: GFP expressing controls. Similarly, the ROS production triggered by 100 nM elf18 E. coli application was significantly reduced and delayed in EFR::GFP expressing lines silenced for OsSERK2 (Fig 6C and 6D). These results demonstrate that silencing of OsSERK2 interferes with elf18 E.coli -induced EFR signaling in rice. Similar to AtSERK3/BAK1 and AtSERK4/BKK1 nM elf18 E.coli for 10 min. Immuno-complexes were precipitated (IP) with GFP-Trap agarose beads. Proteins were separated and detected by SDS-PAGE electrophoresis followed by western blot analysis with HRPconjugated anti-GFP or anti-FLAG. (C) EFR::GFP or EFR::XA21::GFP form constitutive ligand-independent complexes with OsSERK2 in vivo without quantitative changes of the interaction within 15 min of elf18 E.coli treatment. Immuno-complexes were precipitated from leaf material of Ubi::EFR::GFP-9-11-12 and Ubi::EFR:: XA21::GFP-3-6-7 expressing rice plants treated with 1 μM elf18 E.coli for the indicated time using GFP-trap beads. Kitaake rice leaves were used as negative control. Components of the immuno-precipitated complexes were separated by SDS-PAGE gel followed by immuno-detection with anti-GFP (for EFR::GFP and EFR::XA21::GFP) and anti-OsSERK2 (for OsSERK2). EFR::GFP and EFR::XA21::GFP gives rise to a signal at about 175 kDa. OsSERK2 (~70 kDa) was co-immunoprecipitated with EFR::GFP and EFR::XA21:: GFP in the absence of elf18 E.coli treatment. The lower panel shows equal amounts of OsSERK2 in both total protein fractions before immunoprecipitation. This experiment was repeated twice with similar results.  in Arabidopsis, OsSERK2 appears to be a positive regulator of EFR signaling in rice orthologous to its role in XA21 signaling [42,54].
In contrast to OsSERK2, XB24 is a negative regulator of XA21-mediated immunity [53]. To test if XB24 is also involved in EFR signaling in rice, we crossed previously described XB24 overexpressing lines (XB24OE A109-6-5-1) [53] with lines expressing EFR::GFP (Ubi::EFR:: GFP-9-2). In the F2 generation, we isolated double transgenic lines from two independent crosses (14 and 18) by PCR with primers specific for each transgene and confirmed stable expression of the EFR transgene by qRT-PCR (S16B Fig). We assessed the impact of XB24 overexpression on elf18 E.coli -induced EFR-signaling in defense gene expression and ROS assays as describe above. Elf18 E.coli -triggered defense gene expression was significantly reduced at 12 hours post treatment for both marker genes in both lines overexpressing XB24 in the EFR::GFP background (Ubi::EFR::GFP x XB24OE-14 and -18) when compared to EFR::GFP expressing controls (Fig 6E and 6F). Similarly, ROS production was significantly altered with both lines showing an extended ROS production ( Fig 6G). This effect was more pronounced in the line Ubi::EFR::GFP x XB24OE-14 and led to a significant increase in total ROS production in this line only (Fig 6H). This is consistent with the more severe defect of elf18 E.coli-triggered defense gene expression in this line (Fig 6E and 6F). This suggests that XB24 is involved in EFR signaling in rice when overexpressed, similarly to its negative role in XA21-signaling.

Discussion
The main aim of this study was to investigate the feasibility of inter-class transfer of dicotolydenous PRRs, such as EFR from Arabidopsis, into the model monocotyledonous species, rice. We aimed to determine if PRRs, when transgenically expressed, from evolutionarily distant plant species can enhance resistance and if they employ the same signaling components as endogenous PRRs. We demonstrate that transgenic expression of EFR in rice confers a high-level of sensitivity to the previously unrecognized elf18 E.coli and elf18 Xoo . Ligand-dependent activation of EFR elicits well-characterized defense responses such as defense gene expression, ROS production and MAP kinase activation (Figs 1-3, Table 1). While expression of EFR does not lead to robust enhanced resistance to fully virulent Xoo isolates, it does lead to quantitatively enhanced resistance to two weakly virulent Xoo isolates (Fig 4, Table 2). We made similar observations of full defense response activation at sub-nanomolar concentrations of elf18 E.coli and elf18 Xoo (Table 1) but limited enhanced resistance to Xoo for rice plants expressing the chimeric receptor EFR::XA21, which consists of the EFR ectodomain and the intracellular XA21 kinase domain (Figs 1-4, Table 1 and 2). These results indicate that ligand activated XA21 kinase alone is not sufficient to induce the robust resistance response observed with the full-length XA21.
EFR in rice utilizes at least two well-described XA21-signaling components OsSERK2 and XB24, most likely via direct protein interactions (Figs 5 and 6).

Signaling downstream of EFR and EFR::XA21 in rice
In recent years, tremendous advances have been made in deciphering the signaling events occurring at the PRR level within seconds to minutes of ligand perception [2,15,71]. These advances have been mainly driven by studies in Arabidopsis involving EFR, FLS2 and CERK1 relative light production ± SE of at least six biological replicates. Statistical analysis was performed using the Tukey-Kramer HSD test. Different letters indicate significant differences (p < 0.05). These experiments were repeated at least three times with similar results. and in rice involving chitin perception by CEBiP [2,15,71]. Most of the recent progress on rice receptor kinase PRRs, including XA21 and XA3, has relied on the characterization of much later phenotypic read-outs such as disease progression, which is recorded over 1 week after inoculation [72,73]. This is mostly caused by the paucity of well-defined ligands for most rice receptor kinase PRRs such as XA21, Pi-d2 (FJ915121.1) and XA3 (DQ426646.1) [74][75][76][77]. In Arabidopsis, several well-defined defense read-outs are readily available to assess signaling activation post-ligand treatment including defense marker genes, ROS burst and MAP kinase activation [2]. Very little is known about the immediate signaling activated by peptide ligands in the absence of infectious agents in fully mature rice leaves.
We used EFR and the chimeric receptor EFR::XA21 to probe rice responses using the welldefined peptide ligand elf18 E.coli . Both receptors elicit qualitatively similar defense signaling pathways including the activation of several MAP kinases, ROS production, and up-regulation of two defense maker genes, PR10b and Os04g10010 (Figs 1-3, Table 1). Indeed expression of EFR::GFP and EFR::XA21::GFP confers similar, high sensitivity to elf18 E.coli and elf18 Xoo as was previously reported for EFR in Arabidopsis with an EC 50 value of~300 pM (Table 1). While both receptors elicited both marker genes, the kinase domain of XA21 appears to consistently lead to a higher up-regulation of PR10b (Figs 1C and 3B). This suggests that the XA21 kinase might be better adapted to defense signaling in rice as compared with the EFR kinase domain, which is derived from a dicotyledonous plant species. This increased signal capacity of the XA21 kinase domain might also be the reason why older Ubi::EFR::XA21::GFP plants appear to be necrotic, tend to senesce earlier and accumulate lower biomass at full maturity (S10 Fig). These phenotypes of Ubi::EFR::XA21::GFP plants might be caused by the abundance of EF-Tu in the rhizosphere and phyllosphere. This detection might lead to a stronger, more severe continuous defense activation in Ubi::EFR::XA21::GFP plants when compared with Ubi::EFR:: GFP plants.
The observed severe phenotypic differences between Ubi::EFR::XA21::GFP versus Ubi::EFR:: GFP and Kitaake controls were clearly age-dependent and only observable from the 5-weekold stage onwards. When we attempted to identify underlying signaling pathways that may be activated at the 4-week-old stage, before macroscopic necrosis developed, we only detected 131 differentially expressed genes when comparing Ubi::EFR::XA21::GFP with Kitaake control in the absence of ligand treatment (S3 Table). No specific GO terms were enriched in the up-regulated gene set (S12A Fig). However, we identified a significant enrichment for GO terms associated with oxidoreductase activity in genes down-regulated in EFR::XA21 (S12B Fig). This suggests that Ubi::EFR::XA21::GFP plants might be more susceptible to oxidative stress at older developmental stages. Consistent with these observations, SA signaling appears to be activated in these 5-week-old plants as two SA marker genes are upregulated in EFR::XA21::GFP expressing plants in the absence of any treatment (S17 Fig). However, the overall transcriptomic comparison between Ubi::EFR::XA21::GFP and Kitaake suggests that at 4-week-old stage Ubi:: EFR::XA21::GFP plants are very similar to wild-type plants and do not overexpress stress related genes. These results indicate that the Ubi::EFR::XA21::GFP plants serve as a useful surrogate system to investigate the transcriptional reprogramming induced by ligand activated XA21 kinase in rice leaf tissue.
Our studies to determine if EFR in rice utilizes similar signaling components as XA21 [2,[70][71][72] identified OsSERK2 and XB24 but not XB3 and XB15 as interaction partners of EFR (Fig 5). We therefore focused our further genetic studies on OsSERK2 and XB24. We crossed EFR-expressing rice lines to OsSERK2-silenced lines and XB24-overexpressing lines. We found that OsSERK2 is a positive regulator of EFR-mediated defense signaling in rice, similar to its role in the XA21-mediated immune response [54]. EFR lines silenced for OsSerk2 are significantly impaired in elf18-induced defense gene expression and ROS production (Fig 6A-6D). In Arabidopsis, EFR requires several SERK proteins for its function including SERK3/BAK1 and SERK4/BKK1 (At2g13790) [42]. EFR signaling is not strongly inhibited in single bak1 or bkk1 null mutant. Only when using the hypomorphic allele bak1-5, which is strongly inhibited in PTI signaling, and the bak1-5 bkk1-1 double mutant, a clear contribution of both SERK proteins to EFR signaling is detectable [41,42]. In Arabidopsis, the SERK family underwent an expansion and contains 5 members, which might have led to functional redundancy and diversification [78]. In rice, the SERK family contains only two members, OsSERK1 (Os08g07760) and OsSERK2 [54]. Only OsSERK2-silenced lines, but not OsSERK1-silenced lines are impaired in XA21-mediated immunity [54,79]. Rice expressing EFR and silenced for OsSERK2 are impaired in elf18 E.coli -triggered signaling. This suggests that rice SERK2 is the functional ortholog of Arabidopsis SERK3/BAK1 and SERK4/BKK1. Curiously, OsSERK2 is phylogenetically more closely related to Arabidopsis SERK1 (At1g71830) and SERK2 (At1g34210) [54]. Single mutants of Arabidopsis serk1 and serk2 in Arabidopsis are not impaired in elf18-triggered signaling, indicating that they do not play a role in the responses tested despite forming a ligand-induced complex with EFR [42]. The recruitment of phylogenetically distinct SERK proteins into the EFR plasma membrane signaling complex when comparing rice and Arabidopsis might also explain the different kinetics observed for the interaction of OsSERK2 and EFR in rice and BAK1 and EFR in Arabidopsis. In rice, OsSERK2 and EFR form ligand independent heteromers that do not appear to change within 15 min of ligand treatment (Fig 5B). In Arabidopsis, the interaction between SERK3 and EFR is only observable after ligand treatment within seconds to minutes [41,42].
EFR also directly interacts in a kinase activity independent manner with XB24, an enzymatically active ATPase. EFR rice lines overexpressing XB24 were slightly impaired in elf18 E.colimediated defense signaling, especially at later time points (Fig 6E-6H). XB24 therefore appears to be a negative regulator of EFR signaling in rice similar to its involvement in XA21-mediated immunity [53]. It remains to be determined if XB24 also enhances the autophosphorylation activity of EFR and employs identical mechanisms of regulation of EFR signaling as it does for XA21 signaling in rice.
A related study by Zipfel and colleagues shows that Arabidopsis XB24 also interacts with EFR [80]. Yet Arabidopsis xb24 mutants are not impaired in elf18-triggered signaling. This might be due to the fact that Arabidopsis XB24 lacks several amino acids important for ATPase activity [80]. EFR and EFR::XA21 confer quantitatively enhanced resistance to two weakly virulent Xoo isolates EFR and EFR::XA21 rice plants are fully able to recognize the elf18 sequence derived from EF-Tu of E. coli in fully mature leaf tissue at sub-nanomolar concentrations (Figs 1-3, Table 1). Sequence analysis of over 20 Xoo isolates (S2 Table) revealed that the elf18 sequence in Xoo is highly conserved and contains two single amino acid changes at the second and fourth position when compared with elf18 E.coli sequence (S8 Fig). The resulting elf18 Xoo sequence was previously shown to be as active as elf18 E.coli with an EC50 value of~200 pM when used as double alanine substitution control in the medium alkalization assay of Arabidopsis cell cultures [27]. This observation also holds true for rice plants expressing EFR and EFR::XA21, because elf18 E. coli and elf18 Xoo elicited similar defense responses in these plants (Fig 3, Table 1). These observations led us to hypothesize that EF-Tu from Xoo would be fully recognized by EFR and EFR:: XA21 expressing rice plants. Full length EF-Tu Xoo protein is most likely also readily available for recognition at the infection site of the xylem pathogen Xoo (S9 Fig) [68]. Based on this observation, we hypothesized that EFR, and especially EFR::XA21, expressing rice plants would be more resistant to Xoo. In the initial infection experiments, we used our fully virulent Xoo isolate PXO99A, to which rice lines expressing the rice immune receptor XA21 are fully resistant (Table 2) [28]. EFR-expressing plants were slightly more resistant to the fully virulent Xoo isolate PXO99A in 5 out of 8 infection assays (Fig 4A, Table 2). This partial disease resistant phenotype of EFR expressing rice plants is similar to the contribution of EFR to the resistance against the fully virulent Pseudomonas syringae pv. tomato (Pto) DC3000 isolate on Arabidopsis [35]. Several Pto DC3000 effectors have been shown to suppress EFR signaling during the infection process including AvrPto and HopAO1 [43,81]. Xoo PXO99A does not encode orthologs for any of these effectors but might be able to secrete Xoo specific effectors into rice cells to suppress PTI signaling initiated by EFR and other endogenous rice PRRs.
A screen of our diverse collection of Xoo isolates identified several Xoo isolates that are less virulent on Kitaake plants (S4 Table). EFR expressing rice plants show an enhanced resistance to the weakly virulent Xoo isolate MXO90 in 6 out of 6 experiments and to isolate Xoo NXO256 in 5 out 6 experiments for (Fig 4, Table 2). This quantitatively enhanced resistance phenotype in Ubi::EFR::GFP plants requires the expression of full-length EFR, because Ubi:: EFR::GFP-1, which does not express EFR to any detectable level when measured by qRT-PCR and western blot analyses (Fig 1A and 1B), is not responsive to exogenous elf18 E.coli application (Fig 1C and 1B) and does not show an enhanced resistance phenotype for any Xoo isolate tested ( Fig 4B, Table 2). Therefore, it is very likely that the recognition of EF-Tu from Xoo by EFR during the infection process leads to the quantitatively enhanced resistance phenotype of Ubi:: EFR::GFP plants expressing EFR (Fig 4B, Table 2). This again is similar to the observation in Arabidopsis where the contribution of EFR towards disease resistance is more readily accessible when using hypo-virulent isolates of Pto such as Pto ΔCorand Pto ΔAvrPto/ΔAvrPtoB [35]. This is in contrast to previous gain-of-function experiments showing that the transgenic expression of EFR in tomato and N. benthamiana leads to a strong resistance response to taxonomically diverse bacterial pathogens under laboratory conditions [10]. The strength of the immune response conferred by the expression of EFR might be defined by the specific plant pathogen interaction and the infection methods used. We cannot completely rule out the possibility that the expression of EFR in rice does not lead to a fully functional immune receptor, because EFR might not interact appropriately with all required downstream immune signaling components. This defect in full signaling capacity might be the reason why the expression of EFR does not confer a qualitative resistance response to Xoo in rice. However, this is unlikely as the expression of EFR confers sensitivity to elf18 Xoo (Figs 1-3, Table 1) similar to that observed for elf18 E.coli in Arabidopsis [26,27].
We also attempted to assess the resistance phenotype of Ubi::EFR::XA21::GFP plants, which was extremely difficult due to the observed early senescence phenotype at the infection stage using 6-week-old plants (S10 Fig). Nonetheless, we were able to perform two full experiments in which we infected fully mature leaves of Ubi::EFR::XA21::GFP plants that did not show any necrosis or early senescence phenotypes. In these experiments, Ubi::EFR::XA21::GFP plants were slightly more resistant to NXO256 in 1 out of 2 experiments, but not to MXO90 and PXO99A, however, they were less resistant than Ubi::EFR::GFP plants ( Table 2). Although we did not perform as many experiments with the Ubi::EFR::XA21::GFP lines as we did with Ubi:: EFR::GFP lines, it was clearly evident that the EFR::XA21 chimera receptor does not confer robust resistance to Xoo infection, despite its ability to detect and respond to elf18 Xoo at subnamolar concentrations (Fig 3, Table 1). These results are somewhat surprising because we and others previously hypothesized that the XA21 intracellular kinase domain would define the strong disease resistance phenotype mediated by XA21 [82,83]. For example, Kishimito et al. demonstrated that the expression of the chimeric receptor CeBIP::XA21 in rice confers enhanced resistance to the fungal pathogen Magnaporthe oryzae by increasing chitin perception and signaling [84]. However in our experimental system, expression of EFR::XA21::GFP in rice is not sufficient to trigger robust resistant to Xoo. This might be caused by inadequate immune signaling activation by this chimeric receptor, as suggested for EFR (see Discussion above). Alternatively, these results suggest that the extracellular XA21 LRR, or the to-date unidentified ligand of XA21, contribute significantly to the strong disease phenotype of XA21 rice plants against the bacterial pathogen Xoo. In the future, it will be important to assess the disease phenotype of rice plants expressing the reverse chimera XA21::EFR, in which the extracellular domain of XA21 is fused to the intracellular kinase of EFR. The analyses of the Ubi::XA21::EFR genotype and the knowledge of the ligand of XA21 will enable us to assess the role of the extracellular LRR of XA21 towards the strong disease phenotype of XA21 plants. If the expression of XA21::EFR confers robust resistance to Xoo it will demonstrate that the ligand and the ectodomain of PRRs plays a more important role for the disease resistance response than previously anticipated.

Stacking of PRRs as potential strategy to improve broad-spectrum disease resistance
Rice is unable to recognize elf18 E.coli (Figs 1-3) [85], and only the expression of the Arabidopsis PRR EFR enables rice to sense elf18 E.coli (Figs 1-3). It was recently shown that rice is able to detect a distinct part of EF-Tu from the bacterial pathogen Acidovorax avenae isolate N1141, which is located in its central region (Lys176 to Gly225) [85]. The authors hypothesize that rice possesses an alternate EF-Tu PRR that binds to the central region of EF-Tu. However, this central region of EF-Tu has only 66% amino acid identity between A. avenae and Xoo and it is therefore difficult to speculate whether the endogenous EF-Tu receptor of rice would recognize EF-Tu from Xoo. It is currently unknown if Kitaake and other rice varieties such as the japonica rice cultivar Nipponbare are also able to sense this central region of EF-Tu. Generating rice with two independent EF-Tu immune receptors, EFR and the endogenous unknown PRR, would restrict the pathogens ability to mutate both recognition sites on the same protein concomitantly. It is therefore likely that the resistance mediated by both receptors would be more durable than by each single receptor. In the future, it will be important to test if transgenic rice plants expressing EFR provide resistance to Xoo or other bacterial pathogens such as X. oryzae pv. oryzicola under field conditions. The recognition of elf18 Xoo by EFR during the initial low dosage Xoo infection through hydathodes and natural openings may be useful for limiting pathogen spread in the field especially in the presence of a second independent endogenous EF-Tu receptor recognizing another epitope. This stacking of several PRRs that recognize either different moieties of the same highly conserved protein or different PAMPs might be a valuable strategy to generate long-lasting broad-spectrum resistance [9]. Several recent studies, which describe the transfer of PRRs between different plants species, suggest that interspecies transfer of PRRs is feasible [10][11][12][13]. Field studies are needed to assess the full potential of this promising approach of increasing disease resistance by stacking multiple PRRs.
During the review process of this manuscript three newly published manuscripts also describe the successful transfer of EFR to a monocot crop, or reciprocally of XA21 to dicots. Ridout and colleagues demonstrate that the transgenic expression of EFR in wheat confers elf18 E. coli responsiveness and resistance to Pseudomonas syringae pv. oryzae [86]. Additionally, transgenic expression of EFR driven by the 35S promoter in the rice variety Zhonghua 17 conferred responsiveness to elf18 E.coli , slight enhanced resistance to the bacterial pathogen Acidovorax avenae subsp. avenae at the seedlings stage but did not alter the immune response to Xanthomonas oryzae pv. oryzae PXO99A [61]. Conversely, transgenic expression of XA21 and EFR: XA21 in Arabidopsis thaliana, led to increased resistance to the weakly virulent strain Pseudomonas syringae pv. tomato DC3000 CORand in the case of EFR::XA21 also towards Agrobacterium tumefaciens [80]. Taken together, all four publications highlight the interest in and the applicability of inter-class transfers of EFR and XA21, and potentially other plant PRRs, from dicots to monocots, and vice-versa.

Plant material and methods
Rice seeds were germinated in water-soaked filter paper for 5-7 days at 28°C and then transplanted into either 4.4-liter pots for plant inoculation assays and growth assessment or 3.5 liter pots for all other experiments. Plants were grown in an 80/20 (sand/peat) soil mixture in an environmentally-controlled greenhouse with temperature set to~28-30°C and humidity to 75-85%. During winter months (November-April) artificial light supplementation was applied to obtain a day/night regime of 14/10.

Generation of transgenic plants
Transgenic plants were generated as described previously [87]. Briefly, pC::UBI::EFR::GFP and pC::UBI::EFR::XA21::GFP were transformed into Kitaake calli by Agrobacterium-mediated transformation. Regenerated plants were selected on hygromycin. The presence of the transgene was confirmed in the T 0 and each following generation by PCR using transgene specific primers (S5 Table).

Plasmid construction
We generated two plasmids for plant transformation using the pNC1300 vector for final plant transformation [88]. The chimeric construct EFR::GFP and EFR::XA21::GFP in the pENTR-D/ TOPO vector (Invitrogen) was generated as follows. We amplified two DNA fragments with about 25bp overlap using Phusion polymerase (Thermo). For the full EFR coding sequence we used primer combination of EFF-F and EFR_NOSTR on EFR CDS containing vector [35] and the 3' GFP fusion part the primer combination GFPoverEFRF and GFPSTR on pNC1300::UBI:: Xa21::GFP [89]. For the 5' EFR fragment we used primer combination EFR-F and EFRectR on EFR CDS containing vector [35] and for the 3' XA21::GFP fragment we used primer combination XaTMoverEFRF and GFPSTR on pNC1300::UBI::Xa21::GFP [89] (S5 Table). PCR products of the expected size were gel purified and 2 μl of each purified PCR product combined for a chimeric PCR reaction without primers using the following conditions: Denaturation 95°C for 1 min, Annealing 42°C for 30 seconds, Extension 72°C for 30 sec/kb, 12 cycles. The chimeric PCR reaction was diluted 1:1000 and used as template in a PCR reaction using the flanking primer combination EFR-F and GFPSTR for both chimeric constructs (S5 Table). PCR products of the expected size were gel purified and cloned into pENTR-D/TOPO vector (Invitrogen). The sequences of the chimeric genes EFR::GFP and EFR::XA21::GFP were confirmed by standard Sanger sequencing. Both EFR::GFP and EFR::XA21::GFP were flipped into the pNC1300:: UBI transfer vector [88] by LRII clonase reactions (Invitrogen). Recombination reactions were confirmed by restriction analysis on the final vectors pUbi::EFR::GFP and pUbi::EFR::XA21:: GFP.
For yeast-two hybrid assays we cloned the intracellular domain of EFR into pLexA. The intracellular domain of EFR was cloned into pENTR-D/TOPO vector (Invitrogen) using the primer combination EFR_2037_GW and EFR_stop_R on pUbi::EFR::GFP. We verified DNA sequence by standard Sanger sequencing. We also generated a clone where the aspartate (EFR849) in the catalytic loop of EFR was mutated to an asparagine in order to disrupt kinase activity [41]. The underlying point mutation was introduced by targeted point mutagenesis using the primer combination EFR_D-N_F and EFR_D-N_R on EFR ID in pENTR-D/TOPO (see above) using PCR conditions described previously [41]. We verified DNA sequence by standard Sanger sequencing. Both EFR ID and EFR (D849N) ID were flipped into the pLexA vector by LRII clonase reaction (Invitrogen). Recombination reactions were confirmed by restriction analysis on the final vectors pLexA-EFR-ID and pLexA-EFR(D849N)-ID.
For transient expression assays in N. benthamiana, we cloned the full coding sequence of XB24 without stop codon into pGWB11 [90]. XB24 without stop codon was cloned into pENTR-D/TOPO vector (Invitrogen) using the primer combination XB24_GW_F and XB24_w/o_stop on cDNA. We verified DNA sequence by standard Sanger sequencing. The CDS of XB24 without stop codon was flipped into pGWB11 by LRII clonase reaction (Invitrogen). Recombination reactions were confirmed by restriction analysis and sequencing of the final vector pGWB11-XB24.

Yeast two-hybrid assays
Yeast two-hybrid assays were performed as described previously [54,56] using the Matchmaker LexA two-hybrid system (Clontech). Yeast pEGY48/p8op-lacZ (Clontech) was co-transformed with pLexA and pB42AD vectors containing the indicated inserts by using the Frozen-EZ yeast transformation II kit (Zymo Research).

Rice leaf tissue treatment with elicitors
Rice leaf tissue was treated with elicitors as described previously [54]. Leaves of 4-week-old greenhouse grown rice plants were cut into 2 cm long strips and incubated for at least 12 hours in ddH20 to reduce residual wound signal. Leaf strips were treated with water, 1 μM flg22 Pst peptide, purchased from Pacific Immunology, 500 nM elf18 peptides, purchased from Gene Script, or 50 μg/mL chitin, purchased from Sigma, for the indicated time. Leaf tissue was snapfrozen in liquid nitrogen and processed appropriately.

Bacterial infection assays
To prepare Xanthomonas oryzae pv. oryzae (Xoo) inoculum, Xoo isolates were spread-plated on peptone sucrose agar plates for 3 days, then washed off with water and adjusted to an OD 600 of~0.5, which corresponds to 5x10 8 CFU/mL. Greenhouse-grown plants were transported into controlled growth chambers at the 5-to 6-week-old stage. Chamber conditions were set tõ 28°C, 85% humidity and 14/10 day/night regime. Plants were allowed to acclimate to the chamber conditions for 2-3 days before being clip-inoculated with the Xoo inoculum [28]. In each plant 5-6 tillers were inoculated and in each tiller the two most recent fully developed leaves were clipped about 2 cm from the tip with scissors dipped in the Xoo inoculum. For each treatment 2-4 plants were inoculated yielding 20-40 inoculated leaves per treatment. Plants were incubated for 12-14 days post inoculation before disease lesions were scored.
In planta bacterial growth curves were performed as previously described [77].

Rice biomass and yield assessment
Rice plants were grown as described above, with two seedlings in each 4.4-liter pot. At maturity, irrigation was stopped and plants were dried. Then total dry biomass and grain yield was weighted.

MAP kinase assays
MAP kinase assays protocols were adapted from Arabidopsis [41]. Rice leave were ground to fine powder in liquid nitrogen and solubilised in better lacus buffer [

ROS production measurement in rice
Leaves of 3-to 4-week-old rice plants were cut longitudinal along the mid vein and then transverse into 1 to 1.5 mm thick leaf pieces. These leaf pieces were floated on autoclaved water overnight. The next morning two leaf pieces each were transferred into one well of a 96-well white plate containing 100 μl elicitation solution (20 μM LO-12 [Wako, Japan], 2 μg/ml HRP [Sigma]). Leaf pieces were treated with the indicated concentration of elf18 E.coli or elf18 Xoo and time. ROS production was measured for 0.5s per reading with a high sensitivity plate reader (TriStar, Berthold, Germany).

Western blot analysis
Total protein extracts from yeast, Xoo and rice plants and Western blot analyses were performed as previously described [41,54]. The primary antibodies used were as follows: Anti-OsSERK2 for detection of OsSERK2 [54], anti-GFP (Santa Cruze Biotech) for detection of EFR::GFP and EFR::XA21::GFP, anti-LexA (Clontech) for detection of LexA-fused proteins expressed in yeast from pLexA, anti-HA (Covance) for detection of HA-tagged proteins expressed in yeast from pB42AD and anti-EF-Tu (Thermo Fisher Scientific, PA5-27512) antibody to detect EF-Tu in Xoo protein preparations. The appropriate secondary antibody, anti-mouse (Santa Cruz Biotech) and anti-rabbit (GE Healthcare) coupled to horseradish peroxidase were used in combination with chemiluminescence substrates (Thermo) to detect proteins by exposure to film.

EF-Tu sequencing and alignment
Xoo carry two copies of the tuf gene that are 100% identical based on amino-acid sequence of the sequenced isolate PXO99A. The two tuf copies, PXO_04538 (copy 1) and PXO_04524 (copy 2), 1191 bp-long each, are separated in the PXO99A genome by 18,580 bp. We used the following primer sets to amplify the whole EF-Tu sequence of both copies. Copy 1: EF-Tu-F1: CCTTTCGTGAGCACCATTGC and EF-Tu-R4: AGCACGTAGACTTCGGCTTC; and copy 2: EF-Tu2-F: CCAAGAAGGGCTGAGTTCGT and EF-Tu-R2: CCTTGAAGAACGGGG-TATGA. In both primer sets the forward prime anneal several bp upstream of the tuf gene to allow sequencing of the gene from the beginning without cloning. Phusion high-fidelity DNA polymerase (NEB) was used to PCR-amplify the tuf gene (~1300 bp). PCR amplicons were gelpurified and directly sequenced with the same forward and reverse primes.

Statistical analysis and EC 50 value determination
For statistical analysis of inoculated rice we used either Student's t-test, Dunnet test or Tukey test depending on experimental set up, and as indicated in each experiment, using the JMP software. ROS production dose response curves were modeled on the non-linear logistic 4p formula using the JMP software. EC 50 were calculated at half maximal ROS production using the best fitted model.

RNA isolation and quality assessment for RNAseq
Rice leaf strips of~1.5 cm were collected from greenhouse grown, 4.5-week-old Kitaake and Ubi::EFR::XA21::GFP-3-4 plants. After 12h of equilibration on sterile water, RNA was isolated from leaf strip tissue using the Spectrum Plant Total RNA Kit from Sigma-Aldrich and oncolumn DNAse treated to remove genomic DNA contamination following the manufacturer's instructions. RNA was quantified using the Quant-IT Ribogreen RNA Assay Kit. RNA quality was assessed on an Agilent Technologies Bioanalyzer.

Sequencing
Stranded RNA-seq libraries were generated using the Truseq Stranded mRNA sample preparation kit (Illumina). mRNA was purified from 1 μg of total RNA using magnetic beads containing poly-T oligos. mRNA was fragmented using divalent cations and high temperature. The fragmented RNA was reversed transcribed using random hexamers and SSII (Invitrogen) followed by second strand synthesis. The double stranded cDNA was treated with end-repair, Atailing, adapter ligation, and 10 cycles of PCR amplification. qRT-PCR was used to determine the concentration of the libraries. Libraries were sequenced on the Illumina Hiseq 2x150 bp.

Identification of proteins by LC-MS/MS
The digested secretome samples were analyzed on an Agilent 6550 iFunnel Q-TOF mass spectrometer (Agilent Technologies) coupled to an Agilent 1290 LC system (Agilent). The desalted peptide samples (40 μg) were loaded onto a Ascentis Peptides ES-C18 column (2.1 mm x 100 mm, 2.7 μm particle size; Sigma-Aldrich) via an Infinity Autosampler (Agilent Technologies) with buffer A (2% acetonitrile, 0.1% formic acid) flowing at 0.400 ml/min. The column compartment was set at 60°C. Peptides were eluted into the mass spectrometer via a gradient with initial starting conditions of 5% buffer B (98% acetonitrile, 0.1% formic acid) increasing to 30% buffer B over 30 minutes, then to 50% buffer B in 5 minutes. Subsequently, buffer B concentration was increased to 90% over 1 minute and held for 7 minutes at a flow rate of 0.6 mL/min followed by a ramp back down to 5% buffer B over one minute, where it was held for 6 minutes to re-equilibrate the column. Peptides were introduced to the mass spectrometer from the LC via a Dual Agilent Jet Stream ESI source operating in positive-ion mode. A second nebulizer was utilized for the introduction of reference masses for optimal mass accuracy.

MS/MS data analysis
The acquired data were exported as. mgf files using the Export as MGF function of the Mas-sHunter Workstation Software, Qualitative Analysis (Version B.05.00 Build 5.0.519.13 Service Pack 1, Agilent Technologies) using the following settings: Peak Filters (MS/MS) the Absolute height (! 20 counts), Relative height (! 0.100% of largest peak), Maximum number of peaks (300) by height; for Charge State (MS/MS) the Peak spacing tolerance (0.0025 m/z plus 7.0 ppm), Isotope model (peptides), Charge state Limit assigned to (5) maximum. Resultant data files were interrogated with the Mascot search engine version 2.3.02 (Matrix Science) with a peptide tolerance of ±50 ppm and MS/MS tolerance of ±0.1 Da; variable modifications Acetyl (N-term), Carbamidomethyl (C), Deamidated (NQ), Oxidation (M); up to one missed cleavage for trypsin; Peptide charge 2+, 3+ and 4+; and the instrument type was set to ESI-QUAD-TOF. Data was acquired and exported using MassHunter (Agilent Technologies) and resultant MS/ MS data was analyzed using Mascot (Matrix Sciences) against a custom database comprising the RefSeq PXO99A proteins (ca. 13,500 proteins) and all Viridiplantae proteins (ca. 565,000 proteins) available through NCBI. Thresholds were also set to reduce the false discovery rate (p<0.05) and ensure significant peptide and protein matching. Protein and peptide matches identified after interrogation of MS/MS data by Mascot were filtered and validated using Scaffold (version 4.1.1, Proteome Software Inc., Portland, OR). Peptide identifications were accepted if they could be established at greater than 95.0% probability by the Peptide Prophet algorithm [97] with Scaffold delta-mass correction. Protein identifications were accepted if they could be established at greater than 99.0% probability and contained at least 1 identified peptide (at 95% and greater). Rice lines were inoculated using the leafclipping method as described in materials and methods. Bacterial burden was recorded at 4 time points (0, 3, 8, 12 days post inoculation). Statistical analysis was done using t-test for each time point separately. (PDF) S14 Fig. XA21 interacts with OsSERK2, XB3, XB15 and XB24 in the yeast-two hybrid assay. Yeast two-hybrid assay between XA21K668 (668-1025aa) and OsSERK2 ID (260-628aa), XB3 full-length (FL) (1-450aa), XB15 FL (1-639aa) and XB24 FL (1-198aa). The blue color indicates nuclear interaction between the two co-expressed proteins. (PDF)   Table. Differentially expressed gene list. Differentially expressed genes between Kitaake and Ubi::EFR::XA21::GFP-3-4. Differentially expressed genes were selected using a false discovery rate of 0.05 and an absolute log fold change ! 2. Tables at the bottom of worksheets summarize gene ontology information from AgriGO. (XLSX) S4 Table. Inoculation of rice with different Xoo isolates. Rice inoculation with ten different Xoo isolates to identify weakly virulent Xoo isolates compared with the fully virulent isolates PXO99A. (XLSX) S5