The Bgh genome encodes several hundreds of potential effectors, and ∼491 effector-like proteins were initially identified to be CSEPs (Spanu et al., 2010; Pedersen et al., 2012). We selected a hundred of these CSEP genes for further characterization based on their expression levels and abundance in haustoria (Godfrey et al., 2010; Pedersen et al., 2012). The cDNA sequences of 101 CSEPs from 34 families were amplified with specific primers using RNA samples derived from barley leaf materials infected with the compatible isolate BghA6 (Supplementary Table 1). All CSEP cDNA sequences excluding the predicted signal peptide (ΔSP) were subcloned into vector pGR107 for Agrobacterium tumefaciens-mediated transient expression in N. benthamiana (Wang et al., 2011). We identified several CSEPs suppressing cell death in plants (Li et al., 2021), but much fewer CSEPs inducing cell death. As shown in Figure 1A, CSEP0027 is one of the CSEPs inducing clear water-soaked-like cell death phenotype in N. benthamiana, as compared to GFP alone, which serves as a negative control. The AVR_a13 effector and its cognate receptor MLA13 were also coexpressed and triggered cell death in N. benthamiana (Lu et al., 2016), which severed as a positive and technique control here (Supplementary Figure 1). Trypan blue staining confirmed the localized cell death and immunoblotting verified the expression of the HA-tagged fusion proteins (Figure 1A), and DAB (3, 3′-diaminobenzidine) staining also revealed H₂O₂ accumulation in the infiltrated area (Supplementary Figure 2).

The CSEP0027, CSEP0028, and CSEP0340 are the three members from the same Bgh CSEP family 41 (Pedersen et al., 2012), in addition, BgtE-10117 and BgtE-20000 are the two potential Bgt homologs being identified as highly related sequences to CSEP0027 (Supplementary Figure 3A) (Praz et al., 2017). All these five CSEPs harbor a predicted SP, a Y/FxC motif, and a conserved C-terminal cysteine, with some conserved residues in the middle (Figure 1C and Supplementary Figure 3A). We tested if any of the other four CSEPs trigger cell death, unexpectedly none of them induced cell death in N. benthamiana (Figure 1B and Supplementary Figure 3B). CSEP0027, thus represents a unique Bgh effector protein to induce cell death in N. benthamiana.

To validate the secretory function of the CSEP0027 signal peptide, we used a yeast secretion assay based on invertase secretion and yeast growth on sucrose or raffinose media (Lee et al., 2006; Oh et al., 2009). The predicted SPs were fused in frame to the mature sequence of yeast invertase in the vector pSUC2 and expressed in the invertase mutant yeast strain YTK12 that otherwise cannot grow on YPRAA medium (Gu et al., 2011). CSEP0027-SP derived construct enabled transformed yeast cells to grow on YPRAA plate (with raffinose instead of sucrose as the carbon source), and so did the PsAvr1b-SP from the oomycete Avr1b effector as a positive control (Figure 2, middle panel). The first 25 amino acids of Mg87, a Magnaporthe grisea cytoplasmic protein as a negative control, did not enable yeast to grow (Figure 2). In addition, the secretion of the invertase was confirmed by the conversion of 2, 3, 5-triphenyltetrazolium chloride (TTC) to the insoluble red-colored triphenylformazan (Figure 2, bottom panel). These results suggest that CSEP0027 is a secreted protein carrying a functional SP.

To investigate the function of CSEP0027 in fungal virulence, we first overexpressed CSEP0027 in barley epidermal cells through single-cell transient gene expression followed by BghA6 infection in a compatible interaction (Bai et al., 2012). The expression of mature CSEP0027 (CSEP0027^ΔSP) in barley cells led to markedly increased haustorial formation rate (i.e., haustorium index) to ∼68%, as compared to ∼52% in the empty vector control (EV) (Figure 3A). By contrast, silencing CSEP0027 through HIGS significantly decreased haustorium formation rate by ∼40%, relative to the EV control (Figure 3B). Similarly, the silencing of CSEP0105, an effector gene used as a positive control (Nowara et al., 2010; Ahmed et al., 2015), led to a stronger effect on the reduction of haustorium index by ∼60%, also relative to the EV (Figure 3B). These data indicate that CSEP0027 contributes to Bgh virulence.

The expression of many predicted or functionally confirmed CSEP genes is induced during barley infection (Godfrey et al., 2009; Spanu et al., 2010; Pedersen et al., 2012; Hacquard et al., 2013; Schmidt et al., 2014). To further analyze the expression pattern of CSEP0027, we conducted a time course experiment (Figure 3C). The transcript level of CSEP0027 remained low from 0 to12 hpi and was markedly induced at 24 and 48 hpi in both the haustorial containing samples (H) and epiphytic structures (E), with highly enriched transcripts in H sample but not in E sample at 48 hpi (Figure 3C). This expression pattern supports CSEP0027 functioning during barley infection and likely at the post-penetration stages.

To identify host targets of CSEP0027, we performed a Y2H screening of a cDNA prey library derived from Bgh infected barley leaves. Using a bait of CSEP0027 without the SP, we identified two independent clones harboring the fragments of a barley catalase gene, HvCAT1. The targeted Y2H analysis showed that CSEP0027 interacted with full-length HvCAT1 but not with HvCAT2 (Figure 4A), another reported barley catalase that shares more than 70% amino acid identity with HvCAT1 (Supplementary Figure 4; Skadsen et al., 1995). Further interaction analysis indicated that HvCAT1 interacts with CSEP0027 likely through the N-terminal catalase domain but not the C-terminal domain (Supplementary Figure 5).

The interaction between CSEP0027 and HvCAT1 was further verified by in vitro and in vivo assays (Figures 4B–D). For glutathione S-transferase (GST) pull-down assay, GST-CSEP0027 fusion or GST alone derived from E. coli was incubated with HvCAT1-HA containing crude lysate of N. benthamiana. An immunoblotting analysis indicated that GST-CSEP0027 pulled down HvCAT1-HA whereas GST did not (Figure 4B). In luciferase complementation imaging (LCI) assays, CSEP0027-nLuc interacted with cLuc-HvCAT1, thus generated luminescence signal, the reciprocal pair HvCAT1-nLuc and cLuc-CSEP0027 also generated strong luminescence signal in N. benthamiana (Figures 2, 4), while two pairs of negative control did not produce any detectable signal (Figures 1, 3, 4C). In addition, the HvCAT2-nLuc and cLuc-CSEP0027 did not generate detectable signal (Figures 4C, 5). In co-immunoprecipitation (co-IP) analysis, the HvCAT1-Flag fusion did immuno-precipitate with CSEP0027-HA in N. benthamiana, whereas GFP-Flag did not (Figure 4D).

Together, these results indicate that CSEP0027 specifically interacts with barley HvCAT1.

Since CSEP0027 interacts with HvCAT1, we examined the subcellular localization of CSEP0027 and catalases in barley cells. The plasmids expressing CSEP0027^ΔSP-CFP (Cyan Fluorescent Protein), YFP (Yellow Fluorescent Protein)-HvCAT1, and YFP-HvCAT2 fusions were constructed and delivered into barley cells by particle bombardment. Confocal imaging indicated that CSEP0027^ΔSP-CFP was localized in both cytosol and nucleus, similar to YFP alone (Figure 5, the top panels), while YFP-HvCAT1 was localized in many small dots in the cytoplasm, totally different from that of CFP alone (Figure 5, 2nd panels). Since many plant catalases are localized to peroxisomes, we tested the localization of YFP-HvCAT1 in peroxisomes by coexpression of YFP-HvCAT1 with a peroxisomal marker, PST1-RFP (Red Fluorescent Protein). As expected, YFP-HvCAT1 was almost fully co-localized with PST1-RFP in many cytoplasmic foci in the same cells (Figure 5, 3rd panels). Interestingly, YFP-HvCAT2 was also co-localized with PST1-RFP in most of the cytoplasmic dots (Figure 5, 4th panels). Next, we tested the localization of CSEP0027^ΔSP-CFP and YFP-HvCAT1 in barley cells by coexpression analysis. Remarkably, confocal imaging indicated that YFP-HvCAT1 was detected not only in the peroxisomal dots but also in the nucleus, and CSEP0027^ΔSP-CFP appeared to co-localize with YFP-HvCAT1 in some of the cytoplasmic dots but fully overlapped with YFP-HvCAT1 in the nucleus (Figure 5, 5th panels). Interestingly, when YFP-HvCAT2 was coexpressed with CSEP0027^ΔSP-CFP in barley cells, YFP-HvCAT2 remained to localize in the peroxisomal dots and some dots appeared to overlap with CSEP0027^ΔSP-CFP in the cytoplasm (Figure 5, the bottom panels). These localization analyses suggest that HvCAT1 and CSEP0027 have overlapped subcellular localization in the cytosol and CSEP0027 specifically induces the nuclear localization of HvCAT1.

The plant catalases play an important role in biotic stress responses by regulating ROS signaling and homeostasis (Du et al., 2008; Chaouch et al., 2010; Sharma and Ahmad, 2014). To evaluate the function of HvCAT1 in barley immunity, we knocked down the HvCAT1 expression through barley stripe mosaic virus vector (BSMV)-mediated virus-induced gene silencing (VIGS) approach followed by the inoculation of a compatible Bgh isolate. An antisense fragment of HvCAT1 used efficiently silenced HvCAT1 but not HvCAT2 (Figure 6A and Supplementary Figure 3). Scoring of Bgh microcolony formation rate (i.e., microcolony index, MI%) in barley leaf cells at 60–72 hpi indicated that the relative MI% increased by ∼30% in the HvCAT1-silenced leaves as compared to the EV control (Figure 6B). Staining of the Bgh infected barley leaves showed more microcolonies and better hyphae growth on the leaf surface of HvCAT1-silenced barley, as compared with the EV control (Figure 6C). Furthermore, transiently-induced gene silencing (TIGS) technique was used to silence HvCAT1 in barley leaf epidermal cells (Himmelbach et al., 2007; Bai et al., 2012). The RNAi-HvCAT1 construct was delivered into the barley cells by particle bombardment followed by Bgh spores inoculation. Relative haustorium formation rate (i.e., relative haustorium index, HI%) scored at 48 hpi also significantly increased by ∼50% as compared with the EV control (Figure 6D). By contrast, TIGS-silencing of the barley Mlo, a gene required for full susceptibility to Bgh (Kusch and Panstruga, 2017), drastically reduced Bgh HI% in barley cells by ∼80% (Figure 6D). Together, these data indicate that HvCAT1 is involved in barley immunity against Bgh.

The well-established HIGS technique has been used for identifying Bgh CSEPs with functions in promoting fungal virulence (Nowara et al., 2010). So far, a few dozens of Bgh CSEPs have been shown to contribute to Bgh pathogenicity (Nowara et al., 2010; Zhang et al., 2012, 2019; Pliego et al., 2013; Aguilar et al., 2015; Ahmed et al., 2015, 2016; Pennington et al., 2019; Li et al., 2021). In the present study, HIGS of CSEP0027 led to the reduction of HI% by ∼37% in the infected barley cells. Together with the transient overexpression results, our data support the role of CSEP0027 in promoting fungal virulence during barley infection. Our data also suggest that CSEP0027 is most likely a cytoplasmic effector and HvCAT1 is one of its virulence targets. By affecting the subcellular localization of HvCAT1, CSEP0027 may facilitate Bgh infection of host barley.

Bgh CSEP genes are usually induced and/or differentially expressed during the infection of barley. Some CSEP genes are induced at early stages of barley infection, for example, from ∼6 to 12 hpi, whereas others are induced at later stages from 24 to 48 hpi (Godfrey et al., 2009; Zhang et al., 2012; Hackenberg et al., 2013; Schmidt et al., 2014; Aguilar et al., 2015; Ahmed et al., 2015, 2016). CSEP0027 is induced from 24 to 48 hpi and is enriched in fungal haustoria (Figure 3B). We thus believe CSEP0027 functions at later stages of infection, most likely during and after haustorial formation.

Reactive oxygen species, as major regulatory and signaling molecules, can be generated in different subcellular compartments of plant cells and are regulated by an array of antioxidant systems (Waszczak et al., 2018). During plant–fungus interaction, one of the early events in plant response to fungal penetration is an oxidative burst in the apoplastic space, generated mainly by the phagocyte respiratory burst oxidase homologous nicotinamide adenine dinucleotide phosphate (NADPH) oxidases, cell wall peroxidases, and oxalate oxidases (Hückelhoven, 2007; Lehmann et al., 2015). In barley/wheat and B. graminis interactions, H₂O₂ and some other ROS molecules are generated in plant cells during the early stages of fungal penetration, participating in the cell wall lignification and apposition (Zhang et al., 1995; Thordal-Christensen et al., 1997; Hückelhoven et al., 1999; Hückelhoven et al., 2001, 2003; Hückelhoven, 2007; Schweizer, 2008; Li et al., 2015). Interestingly, Bgh fungus also secrets an extracellular catalase that may function in H₂O₂ scavenging in the apoplastic space of host cells (Zhang et al., 2004). The catalases have been known as a class of ROS scavenging enzymes catalyzing the conversion of H₂O₂ into H₂O and O₂, thereby regulating the homeostasis of the intracellular ROS level (Mhamdi et al., 2010; Sharma and Ahmad, 2014). ROS homeostasis is maintained in a very complex manner, involving different ROS-scavenging enzymes, such as catalases, ascorbate peroxidases, glutathione, superoxide dismutases (Mittler et al., 2004; Torres et al., 2006). The peroxisomal ROS levels are closely regulated by CAT activity, and in Arabidopsis, the primary peroxisomal H₂O₂ scavenger is CAT2 (Mhamdi et al., 2012). Here, we show that barley HvCAT1 and HvCAT2 are also peroxisomal catalases. It is unclear whether and when these two HvCAT1 and HvCAT2 are involved in the H₂O₂ decomposition and signaling in peroxisomes during barley interaction with Bgh fungus. Since CSEP0027 expression is induced and most likely functions post haustorium formation, we speculate that HvCAT1 may play a role in regulating ROS homeostasis at later stages, e.g., during and post haustorium formation. Our preliminary data suggest that CSEP0027 triggered-cell death involves H₂O₂ accumulation in N. benthamiana, however, it is not clear if the expression CSEP0027 induced disturbance of ROS homeostasis thus cell death, or vice versa. On the other hand, it is also not yet clear if CSEP0027 has activity in cell death during barley interaction with Bgh fungus. Undoubtedly, more data are needed for fully understanding the role of CSEP0027 in interacting with barley catalases and in regulating ROS homeostasis during barley interaction with Bgh fungus, particularly, the cell death signaling pathway that might be a primary target of the biotrophic fungal pathogen.

The current data are in line with the notion that ROS, in particular cellular H₂O₂, may play an important role in the barley interactions with the B. graminis fungi. It is not unexpected that peroxisomal ROS signaling/homeostasis and ROS signaling cross-talk among the organelles are integral and important parts of barley defense responses to the biotrophic Bgh fungal pathogen.

Apart from being regulated at transcriptional level, plant catalases are also regulated at post-translational level (Mhamdi et al., 2010). A variety of plant proteins have been reported to affect the activity and stability of plant catalases (Yang and Poovaiah, 2002; Fukamatsu et al., 2003; Verslues et al., 2007; Li et al., 2013, 2015; Zou et al., 2015; Kneeshaw et al., 2017). In addition, some pathogen secreted proteins are also identified to interact with the plant catalases and affect their activity, stability, and subcellular localization (Inaba et al., 2011; Mathioudakis et al., 2013; Zhang et al., 2015; Murota et al., 2017; Sun et al., 2017). In line with these examples, the current study data provide new evidence that biotrophic fungal pathogen also secretes an effector to target and affect host catalase subcellular localization in plants.

The plant catalases are mostly peroxisomal proteins and imported into the peroxisome matrix via the peroxisomal targeting signal 1 (PTS1) pathway, i.e., relying on the C-terminal tripeptide PTS1 signal to interact with a peroxisomal receptor and translocate into the peroxisome (Gatto et al., 2000; Lanyon-Hogg et al., 2010). Barley HvCAT1 and HvCAT2, each contains a typical PTS1 signal, PNM or PSM, respectively (Supplementary Figure 4; Mhamdi et al., 2012), and both are localized to the peroxisomes of barley cells in a transient expression analysis (Figure 5). Although different mechanisms may account for the specific re-localization of HvCAT1 upon co-expression with CSEP0027, one scenario can be that CSEP0027 interacts with HvCAT1 but not HvCAT2 in the cytoplasm thus interferes with the interaction of PTS1 signal of HvCAT1 with the peroxisomal receptor. However, how HvCAT1 is specifically regulated by CSEP0027 is not yet clear. Further investigation of the subcellular localization, trafficking, and post-translational modification of HvCAT1 will help to better understand the functions of the catalase and the virulence strategies of the biotrophic fungus.

Barley (Hordeum vulgare L.) cultivars (cv) in this study include Golden Promise and “P01” (isogenic line from cv Pallas containing Mla1). Barley seedlings were grown in a growth chamber at 20°C with 16 h light and 8 h dark cycles. N. benthamiana plants were grown in greenhouse at 24 ± 1°C with a long-day cycle (16 h light/8 h dark).

The barley powdery mildew (B. graminis f.sp. hordei [Bgh]) isolates A6 (AvrMla6, AvrMla10, and virMla1) and K1 (AvrMla1, virMla6, and virMla10) used in this study were maintained on Golden Promise.

Total RNA was extracted from P01 barley leaves inoculated with Bgh isolate A6 using Trizol solution (Invitrogen; 15596-026) and the cDNA was synthesized using reverse transcriptase M-MLV (Invitrogen; C28025). Candidate CSEP sequences excluding the signal peptide (ΔSP) were amplified using the specific primer pairs (Supplementary Table 2) and subcloned into pGR107 vector through restriction enzyme digestion and ligation for agroinfiltration in N. benthamiana (Wang et al., 2011), all candidates confirmed by sequencing.

The analysis of CSEP0027 expression profile was performed as previously described (Ahmed et al., 2015). In brief, total RNA was isolated from P01 leaves at 0, 3, 6, 12, 24, and 48 hpi inoculated with virulent isolate A6. The epiphytic Bgh tissues and the remaining leaf tissues containing Bgh haustoria were separately collected at 24 and 48 hpi. The epiphytic tissues were collected from leaf surfaces by dipping the Bgh-infected leaves into 10% cellulose acetate according to previously described (Ahmed et al., 2015). A quantitative real-time PCR (qRT-PCR) was performed on Applied Biosystems step-one real time PCR system with indicated primers (Supplementary Table 2). Bgh glyceraldehyde 3-phosphate dehydrogenase was used as the reference gene. The statistical significance was evaluated by Student’s t test. The assays were repeated two times with three replicates each time.

The yeast invertase secretion was previously described (Gu et al., 2011). Briefly, the predicted SP sequence of CSEP0027 and Avr1b, or the first 25 amino acids of Magnaporthe oryzae Mg87 were fused in frame with the yeast invertase lacking its own SP in the vector pSUC2. The pSUC2-derived constructs were transformed into the invertase secretion-deficient yeast strain YTK12, and yeasts were then placed on CMD-W medium (0.67% yeast N base without amino acids, 0.075% tryptophan dropout supplement, 2% sucrose, 0.1% glucose, and 2% agar). The positive yeast clones were transferred onto YPRAA medium (1% yeast extract, 2% peptone, 2% raffinose, 2 μg L^–1 antimycin, and 2% agar) for growth testing. Invertase activity was also detected by monitoring conversion of TTC to the insoluble red-colored triphenylformazan.

Single-cell transient gene expression assay was carried out as previously described (Bai et al., 2012). Briefly, a construct expressing a gene-of-interest was bombarded in barley leaf epidermal cells together with a vector expressing ß-glucuronidase (GUS) reporter. The leaves were inoculated with a compatible Bgh isolate at 4 h after the bombardment and then stained with GUS staining solution at 48 hpi. The fungal haustorium index was scored as previously described in the barley leaves after inoculated with Bgh spores. The statistical significance was evaluated by Student’s t test with data from three replicate experiments that have been repeated for three times.

For transient gene silencing assay, the specific gene fragments were cloned into pIPK007 to form a hairpin structure and expression driven by 35S promoter as previously described (Himmelbach et al., 2007). The remaining steps were the same as the transient gene expression assay, except that leaves were inoculated with Bgh isolates at 48 h after bombardment.

Yeast two-hybrid screening was performed according to the protocols of the manufacturer (Clontech; PT4048-1). In total, 5 × 10⁷ transformants were screened. In brief, yeast strain Y2HGold expressing pGBKT7-CSEP0027 (ΔSP) was used for mating with yeast strain Y187 harboring a cDNA prey library derived from Bgh-infected barley leaves and placed onto SD-Leu-Trp-His-Ade plates at 30°C. After 35 days, the resistant clones were selected for further verification.

For Y2H assay, the corresponding bait and prey vectors were co-transformed into yeast strain Y2HGold and plated onto SD-Leu-Trp plates. The positive interactions were detected by placing the strains onto SD-Leu-Trp-His-Ade plates at 30°C.

Luciferase complementation imaging assays were performed according to previously described by Chen et al. (2008). Briefly, the coding region of CSEP0027 (ΔSP) and HvCAT1 were subcloned into vectors pCAMBIA-Cluc or pCAMBIA-Nluc, respectively, to generate constructs for expressing Cluc-CSEP0027 and Cluc-HvCAT1, or CSEP0027-Nluc and HvCAT1-Nluc. The NLuc-/CLuc-derivative constructs were transformed into the A. tumefaciens strain GV3101. The overnight agrobacteria cultures were resuspended with infiltration buffer (2% sucrose, 0.5% MS, 100 μM acetosyringone, and 10 mM MES) into OD600 = 1.0. Equal volume of agrobacteria resuspension carrying the nLUC and cLUC derivative constructs were mixed and co-infiltrated into the N. benthamiana leaves. The infiltrated area was examined for the luciferase activity 40–50 h post agroinfiltration with a cooled charge-coupled device (CCD) imaging apparatus. For each pair of constructs, at least 10 leaves were co-infiltrated in one experiment, and three independent replicates were conducted.

Pull-down assays were performed according to previously described with some modifications (Chang et al., 2013). Briefly, 500 ng of GST-CSEP0027 and GST proteins purified from Escherichia coli were incubated with 150 μl of Glutathione Sepharose 4B beads for 1 h at 4°C, then, beads were sealed with 100 μg BSA for 1 h and incubated with 1.0 g crude protein extracted from N. benthamiana leaves expressing HvCAT1-HA. After incubation for 2 h, the beads were washed five times with RB buffer, then resuspended with 30 μl of 2 × Laemmli buffer, and loaded for sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting with anti-HA antibody. GST-CSEP0027 and GST proteins were detected by Ponceau staining.

For Co-IP assay, the total proteins extracted from N. benthamiana coexpressing GFP-Flag/CSEP0027-HA or HvCAT1-Flag/CSEP0027-HA were incubated with anti-FLAG antibody-coupled beads for 2 h, then washed five times with extraction buffer, proteins were further eluted from the beads using 0.5 mg ml^–1 3 × Flag peptide and used for immunoblotting with anti-HA antibody, or anti-Flag antibody.

For subcellular localization analysis, the coding sequences of CSEP0027^ΔSP, HvCAT1 and HvCAT2 were subcloned into vectors pUBI-mYFP-GW and pUBI-GW-CFP to generate pUBI-CSEP0027 ^ΔSP-CFP, pUBI-mYFP-HvCAT1, and pUBI-mYFP-HvCAT2 constructs. A pair of constructs was delivered into barley leaf epidermal cells by the particle bombardment for coexpression of the indicated fusion proteins, and confocal imaging was conducted at 48 h post-particle delivery. Laser illumination was set at 405 nm for CFP, 488 nm for YFP, and 561 nm for RFP using a Nikon confocal microscope. This assay was repeated three independent times and at least 20 cells were examined for each coexpression.

Barley stripe mosaic virus-mediated gene silencing in barley was performed as previously described (Yuan et al., 2011). Briefly, an antisense fragment of HvCAT1 was cloned into the pCaBS-γbLIC vector to create pCaBS-γb-HvCAT1 construct with indicated primers (Supplementary Table 2). pCaBS-α, pCaBS-β, and pCaBS-γb-HvCAT1 constructs were transformed into A. tumefaciens strain EHA105, respectively. The agrobacteria were resuspended in infiltration buffer to OD₆₀₀ = 1.0 and mixed at 1:1:1 ratio to infiltrate N. benthamiana. After 12 days, N. benthamiana leaf sap was extracted to inoculate 10-day-old barley leaves. After 2–3 weeks of inoculation, the newly emerged leaves with virus caused symptoms were used for Bgh infection, and microcolony scoring was done at ∼60–72 hpi. For each treatment, at least four barley leaves were chosen for analysis, and three independent replicates were conducted. The statistical significance was evaluated by Student’s t test.

Agrobacterium tumefaciens-mediated transient gene expression in N. benthamiana assays were performed as described previously (Wang et al., 2011). A. tumefaciens strain GV3101 was transformed with indicated constructs. Agrobacteria were cultured overnight at 28°C, at 200 rpm, then resuspended in 10 mM MgCl₂ to a final OD₆₀₀ = 0.5 and infiltrated into 4-week-old N. benthamiana leaves. The cell death symptoms were photographed at 5 days post-infiltration. For trypan blue staining, the leaves were boiled in a 1:1 mixture of ethanol and staining solution for 5 min as described before (Bai et al., 2012). The leaves were de-stained with chloral hydrate solution (2.5 g ml^–1) for 2 days.