To characterize the type-III secreted effector (T3E) repertoire of Psa, we cloned effectors from two Psa strains, V13 and LV5. Psa V13 and LV5 strains carry 38 and 26 T3Es, respectively, based on the computational prediction using their genome sequences (McCann et al., 2013; Templeton et al., 2015). Based on this, we selected a total of 48 T3Es for further study using the following criteria: (i) all V13 or LV5 specific T3Es, (ii) only V13 alleles for T3Es in both V13 and LV5 that shared more than 90% amino acid identity, or (iii) both V13 and LV5 alleles for T3Es that shared less than 90% amino acid identity (Table 1). Based on the protein sequence identity to P. syringae homologs, Psa V13 effectors that are predicted to be significantly truncated (hopA1, hopW1, hopAA1-1, and hopAA1-2) were excluded (Table 1). Only one allele of Psa V13 for duplicated effectors hopBB1 (hopBB1-1/hopBB1-2, 93.6% identity) and hopAM1 (hopAM1-1/hopAM1-2, 100% identity) were functionally analyzed. Additionally, both alleles of hopAY1 were analyzed because the predicted peptide sequence of Psa V13 allele was only 77 amino acids shorter than that of LV5. In order to generate broad host-range plasmid (pBBR 1MCS-5) constructs for Pseudomonas-delivery and binary plasmid constructs for in planta transient expression of the 48 selected Psa T3Es, we used the Golden Gate cloning method (Engler et al., 2008). Briefly, each T3E sequence was divided into several modules roughly of 1 kb size and each module was amplified by polymerase chain reaction (PCR) using primers with flanking BsaI restriction enzyme sites. These amplicons were cloned into the Golden Gate compatible entry vector pICH41021. The number of modules for each effector and their 4 bp BsaI overhangs are listed in Supplementary Table S1. All modules for a given effector were then assembled with a C-terminal 6xHA-tag module into the Golden Gate-compatible derivative of the broad host-range vector pBBR1MCS-5, which carries the bacterial avrRps4 promoter (Jayaraman et al., 2017); or with a C-terminal YFP-tag module into the binary vector pICH86988 under control of a cauliflower mosaic virus (CaMV) 35S promoter for functional analysis.

Determining the subcellular localization of effectors can suggest clues about their mode of action during infection. Several studies have highlighted the importance of localization of effectors in a particular cellular compartment for their in planta function (Lewis et al., 2008; Schornack et al., 2009; Sohn et al., 2014). To investigate the subcellular localization of Psa effectors, we expressed YFP-tagged Psa effectors in N. benthamiana leaf cells using Agrobacterium-mediated transient transformation (hereafter agroinfiltration). To visualize effector-YFP proteins, confocal laser scanning microscopy was undertaken at 48 h after agroinfiltration. The subcellular localization of all tested Psa effectors are shown in Figure 1 (larger images for the localization representatives are in Supplementary Figure S2). Interestingly, approximately half of the Psa effectors (23 effectors) were localized in the nucleus and cytoplasm. Only one effector, HopBN1, exclusively localized in the nucleus (Figure 1 and Supplementary Figure S2). HopBB1-2, localized in the nucleus and cytoplasm as well as subnuclear foci (Figure 1 and Supplementary Figure S2). We identified seven effectors (HopN1, HopR1, HopAB3, HopAG1, HopAH1, HopAM1-1, and HopAU1) that localized to the cytoplasm but were excluded from the nucleus (Figure 1 and Supplementary Figure S2). Among these, HopAG1 showed punctate localization. HopM1 localized to the chloroplasts, while HopAV1 and HopAZ1 localized to the cytoplasm but in strands resembling the cytoskeleton (Figure 1 and Supplementary Figure S2). Eleven effectors localized to the cell periphery (largely absent from cytoplasmic strands), and two of them (HopT1 and AvrE1) showed punctate localization (Figure 1) (Jayaraman et al., 2017). However, we could not determine the subcellular localization of HopAS1, HopW1, and HopX1 in our experimental conditions. Notably, these results suggest that Psa bacterial effectors are localized and, therefore, function at several distinct cellular compartments.

In order to validate the subcellular localization of effector-YFP fusion proteins, immunoblot analysis was conducted using total protein extracts from agroinfiltrated N. benthamiana leaves (Supplementary Figure S1). Out of 48 effectors, 38 were confirmed for protein expression by anti-GFP immunoblots (Supplementary Figure S1) (Jayaraman et al., 2017). Of the 10 effectors that could not be detected, six (AvrE1, HopR1, HopT1, HopW1, HopAM1-1, and HopAS1) triggered cell death in N. benthamiana (Figure 2). The remaining 4 effectors (HopI1, HopBN1, AvrD1, and HopAV1) could not be detected despite the lack of strong cell death response. In contrast, HopX1, which also triggered cell death, was detected in the immunoblot but could not be localized via confocal microscopy. The subcellular localization of AvrE1, HopR1, HopBN1, HopI1, HopAM1-1, HopAV1, and HopT1 was detected by confocal microscopy but protein expression was not validated by immunoblot analysis.

Multiple P. syringae effectors have been shown to induce HR-like cell death (HCD) in Nicotiana (Vinatzer et al., 2006; Wei et al., 2007; Wroblewski et al., 2009). To identify Psa T3Es that induce HCD in N. benthamiana or N. tabacum, we transiently expressed each Psa effector using agroinfiltration. Surprisingly, 17 Psa T3Es induced HCD in N. tabacum. Of these, 10 Psa T3Es also triggered HCD in N. benthamiana (Figure 2 and Supplementary Figures S3, S4) (Jayaraman et al., 2017). Interestingly, none of the tested Psa T3Es triggered N. benthamiana-specific HCD. In N. tabacum, 17 T3Es [AvrE1, HopA1, HopF4b, HopH1, HopQ1, HopR1, HopT1, HopW1, HopX1, HopZ3, HopAB3, HopAE1, HopAI1 (LV5), HopBN1, HopAM1-1, HopZ5 and HopAS1] induced HCD in 3–4 days. The intensity of HCD was quantified by ion conductivity measurement at different time points (Supplementary Figure S3). Fourteen Psa T3Es triggered a full collapse of infiltrated leaf tissue while 3 effectors (AvrE1, HopX1, and HopA1) triggered a slower/weaker HCD (Figure 2 and Supplementary Figure S3). Psa-effector triggered HCD was assessed similarly for N. benthamiana, and 6 T3Es (HopR1, HopW1, HopX1, HopAB3, HopZ5, and HopT1) induced strong HCD while 4 T3Es (AvrE1, HopAM1-1, HopAE1, and HopAS1) induced a slower/weaker response (Figure 2 and Supplementary Figure S4).

Natural sequence diversity in the T3E repertoire of P. syringae strains can cause variation in their ability to induce HCD (Ma et al., 2006). In order to identify if allelic sequence variation between Psa T3Es and their homologs found in other P. syringae strains caused an altered HCD phenotype in Nicotiana, we compared their amino acid sequence identity and HCD-inducing ability based on published literature (Table 2). In total, 15 P. syringae homologs of Psa T3Es that were previously tested for HCD induction in N. tabacum and/or N. benthamiana were included in our comparison. Phylogenetic analysis showed that Pto DC3000 is closely related to Psa and 23 T3Es are conserved between these strains (Butler et al., 2013). We selected 10 Pto DC3000 T3Es that triggered HCD, or had a Psa homolog that triggered HCD in Nicotiana spp., for comparison (Table 2) (Wei et al., 2007; Wroblewski et al., 2009). In addition to Pto DC3000, 3 T3E homologs from P. syringae pv. syringae (Psy) B728a, one from P. cannabina pv. alisalensis ES4326 and one from P. syringae pv. tomato (Pto) T1 were analyzed (Lin et al., 2006; Robert-Seilaniantz et al., 2006; Vinatzer et al., 2006; Wroblewski et al., 2009).

Many P. syringae homologs showed similar phenotypes to their Psa counterparts in triggering HCD in N. benthamiana such as AvrE1, HopF4b, HopT1, HopW1, HopX1, HopAA1, and HopAE1 (Table 2). In Pto DC3000, there are two copies of hopT1 (indicated as hopT1-1 and hopT1-2). Psa HopT1 shares the highest amino acid sequence identity with Pto DC3000 HopT1-2 (96.9%). Despite this, we compared Psa HopT1 to Pto DC3000 HopT1-1 (70.4%) because the HCD phenotype was only tested for HopT1-1 (Wei et al., 2007; Wroblewski et al., 2009). Regardless, both Psa HopT1 and Pto DC3000 HopT1-1 triggered HCD in N. benthamiana. Conversely, there are several T3E homolog pairs that show differing phenotypes despite high amino acid sequence identity. For example, Pto T1 HopAB3 shares high amino acid identity with Psa HopAB3 (84.5%) (Lin et al., 2006). However, unlike HopAB3 from Pto T1, Psa HopAB3 induced HCD in both N. benthamiana and N. tabacum. In addition, Pto DC3000 T3Es HopAA1-1 and HopR1 are highly similar to Psa HopAA1 and HopR1 (93.8 and 95.9% amino acid identity, respectively) but only Pto DC3000 HopAA1-1 and Psa HopR1 induced HCD in N. benthamiana (Figure 3). Pto DC3000 HopQ1-1 and Psa HopQ1 share high amino acid sequence identity (99.1%) and they both trigger strong HCD in N. tabacum (Table 2). In contrast, despite relatively low amino acid sequence identity (57.6%), Pto DC3000 HopF2 and its Psa homolog HopF4b triggered HCD in N. tabacum while Psa HopF1 did not (Table 2). Taken together, our data indicate that natural sequence variation might have caused loss or gain of HCD-inducing activity of Psa effectors.

Knocking down gene expression of key components of plant immunity by VIGS has been widely used to investigate the involvement of defense components in plant immunity (Liu et al., 2002; Peart et al., 2002; Oh et al., 2014). To test the requirement of general immunity regulators for Psa effector-triggered HCD, we applied VIGS to generate N. benthamiana plants silenced for EDS1 (pTRV2:NbEDS1), NDR1 (pTRV2:NbNDR1), or SGT1 (pTRV2:NbSGT1) together with control plants (pTRV2:EV). Two alleles of NbEDS1 have been reported, NbEDS1a and NbEDS1b, and our VIGS construct was designed to target NbEDS1a as NbEDS1b is likely a pseudogene (Adlung et al., 2016; Ordon et al., 2017). Hereafter, NbEDS1a is referred to as NbEDS1. First, in order to validate our experimental conditions, HCD induced by agroinfiltration of the RPS4 TIR domain (RPS4^TIR) was tested in N. benthamiana plants VIGSed for EV, NbEDS1, NbNDR1, or NbSGT1. RPS4^TIR induced HCD in EV– or NbNDR1–silenced but not in NbEDS1– or NbSGT1–silenced plants, consistent with previous results (Figure 3A and Table 3) (Zhang et al., 2004; Swiderski et al., 2009). Next, 10 T3Es (HopAM1-1, HopAS1, HopAE1, AvrE1, HopR1, HopT1, HopW1, HopX1, HopZ5, and HopAB3) that triggered HCD in N. benthamiana were agroinfiltrated in NbEDS1–, NbNDR1– or NbSGT1–silenced plants. Interestingly, Psa effector-triggered HCD in NbEDS1– or NbNDR1–silenced plants was comparable to EV-silenced plants (Figure 3B). However, in NbSGT1–silenced plants, 6 T3Es, HopAM1-1, HopAE1, HopT1, HopW1, HopZ5, and HopAB3 were significantly affected in their ability to trigger HCD (Figure 3B and Table 3) (Jayaraman et al., 2017). In contrast, HopAS1-, HopX1-, AvrE1-, and HopR1-triggered HCD was not affected by NbSGT1–silencing. To further analyze the efficiency of VIGS, semi-quantitative RT-PCR was conducted. As a result, the level of NbEDS1, NbNDR1, and NbSGT1 transcripts were significantly reduced in the respective silenced plants compared to EV-silenced control plants indicating that these genes were sufficiently silenced (Figure 3C). However, we noticed that unlike NbNDR1, NbEDS1-, or NbSGT1-silenced plants showed significantly reduced but not completely eliminated transcript levels of silenced genes. Taken together, these results suggest that SGT1 plays a key role in HCD induced by several Psa effectors.

Some pathogen effectors suppress ETI to enable pathogen proliferation. In order to investigate if Psa effectors can suppress ETI, we conducted an agroinfiltration assay of HCD-inducing proteins with or without Psa effectors in N. benthamiana. Bcl-2-associated X (BAX) is an animal pro-apoptotic regulator that induces HCD in plant cells when overexpressed (Lacomme and Santa Cruz, 1999). As previously shown, agroinfiltration of BAX with a GFP control induced strong HCD in N. benthamiana leaf cells (Figure 4A). Interestingly, several Psa effectors, HopF4b, HopQ1, HopF1, and HopAR1, suppressed BAX-induced HCD when coexpressed in N. benthamiana, indicating that these effectors may interfere with cell death signaling. The P. syringae effector AvrPto is recognized by tomato kinase Pto and its cognate R protein Prf (Tobias et al., 1999). Transient expression of AvrPto and Pto trigger HCD in N. benthamiana, which carries a Prf homolog (Scofield et al., 1996). Another P. syringae effector AvrPtoB interferes with Pto-mediated recognition of AvrPto due to its E3-ligase activity (Abramovitch, 2003; Abramovitch et al., 2006). As expected, agroinfiltration of AvrPto and Pto induced strong HCD whereas coexpression of AvrPtoB did not (Figure 4B). We identified only one Psa effector, HopQ1, that suppressed AvrPto/Pto-induced HCD. In addition, since some Psa effectors induced HCD in N. benthamiana (Figure 2), we conducted an agroinfiltration assay to test if other Psa effectors have the ability to suppress Psa effector-induced HCD. Among 8 tested, 5 effectors showed suppression activity on Psa effector-induced HCD (Figure 4C). Interestingly, HopQ1 suppressed HCD triggered by multiple Psa effectors whereas AvrB4-1, HopF4b, HopAR1, and HopA1 suppressed HCD induced by one or two effectors. None of the Psa effectors were able to suppress HopAM1-1 or HopZ5-triggered HCD (Figure 4C). Overall, these results illustrate the specificity of HCD suppression by multiple Psa effectors.

Many bacterial effectors were shown to interfere with PTI to enhance bacterial virulence. PTI triggered by a non-pathogenic bacterial strain can suppress ETI-associated HR elicited by a subsequent infiltration of another bacterial strain (Oh and Collmer, 2005; Crabill et al., 2010). The presence of a PTI-suppressing effector in the non-pathogenic first strain can suppress PTI sufficiently to allow the second ETI-triggering strain to trigger an HR (Le Roux et al., 2015). Pto DC3000 delivers effectors via the T3S and induces HCD in N. benthamiana. Activation of PTI prior to Pto DC3000 infection inhibits effector secretion by the T3S, resulting in significantly reduced HR triggered by Pto DC3000. We performed a PTI inhibition assay by delivery of each Psa effector from P. fluorescens Pf0-1(T3S) prior to Pto DC3000 infection in N. benthamiana (Le Roux et al., 2015). As shown previously, delivery of the wild-type Ralstonia solanacearum effector PopP2 from Pf0-1(T3S) interfered with Pf0-1(T3S)-mediated inhibition of Pto DC3000-induced HCD (Figure 5 and Supplementary Figure S5) (Le Roux et al., 2015). In contrast, the enzymatically inactive PopP2-C321A variant did not affect Pf0-1(T3S)-mediated inhibition of Pto DC3000-induced HCD. Among 35 Psa effectors tested, we found that only HopD1 showed robust PTI suppression activity in our assay (Figure 5 and Supplementary Figure S5). However, this result does not exclude the possibility that other Psa effectors carry PTI-suppression activity.

Localization of a T3E in planta provides an indication of the host target protein location. For example, several plasma membrane-localized T3Es including AvrRpm1, AvrB, and AvrRpt2 target membrane-associated RIN4 while nuclear-localized T3E PopP2 binds its corresponding NLR, RRS1, inside the nucleus (Nimchuk et al., 2000; Axtell and Staskawicz, 2003; Deslandes et al., 2003). Through our large-scale screening for the sub-cellular localization of transiently expressed Psa T3Es, we concluded that the nucleus, cytoplasm and plasma membrane are major target compartments for Psa T3Es. Psa T3Es of a multitude of sizes that localize to the nucleus make up a large proportion of the T3Es (23/44). Proteins between 90 and 110 kDa were found to passively diffuse through nuclear pores (Wang and Brattain, 2007). This may be one of the reasons for nucleocytoplasmic localization of a large number of effectors. Notably, since the in planta subcellular localization was assessed by transient expression of effector-YFP proteins, we cannot rule out the possibility that effector localization may be affected by over-accumulation of proteins or C-terminally tagged YFP. Nonetheless, it appears that Psa effectors localized in diverse cell compartments.

Previously, HopZ3 from P. syringae pv. syringae B728a (approximately 76 kDa including the YFP tag; 72% amino acid identity to Psa HopZ3) showed a nucleocytoplasmic localization to target MPK3 and MPK6, while HopBB1 from P. syringae pv. mori 301020 (approximately 60 kDa including the YFP tag; 93% amino acid identity to Psa HopBB1-2) was localized in the nucleus to target TCP14 and JAZ3 for degradation, despite not possessing a specific nuclear localization signal (Lewis et al., 2014; Yang et al., 2017). However, a nuclear localization may be critical to effector function with studies showing that several nucleus-localized T3Es induce transcriptional reprogramming in the host cell (Nissan et al., 2006; Kay and Bonas, 2009). On the other hand, many PRRs and their associated proteins involved in perception of PAMPs or activation of PTI are localized at the plasma membrane. Thus, many T3Es target these PTI components at the plasma membrane once delivered into the host cell (Kim et al., 2005; Zhou et al., 2014). This is consistent with our finding that some Psa T3Es are localized at the cell periphery. It would be interesting to further investigate if the Psa T3Es localized at the cell periphery enhance bacterial virulence by targeting novel PTI components.

A future application of findings from our study will be to examine Psa T3E function in relation to localization. Promising candidates for this could be HopAZ1 or HopAV1 that both appear to target the host cell cytoskeleton. HopZ1a and HopW1 are two known T3Es from P. syringae that target the cytoskeleton to disrupt PTI (Lee et al., 2012; Kang et al., 2014). Notably, a HopAZ1 homolog from P. savastanoi was recently reported to suppress both ROS production and callose deposition, critical markers of PTI (Matas et al., 2014). hopAZ1 is present in all sequenced strains of Psa and appears to be a ‘Psa core effector’ (McCann et al., 2013). Interestingly, hopAZ1 was reported to have been independently acquired multiple times into the T3E repertoire of pathogens of hazelnut, P. syringae pv. avellenae (O’Brien et al., 2012). These strains that cause hazelnut decline disease are from two different phylogroups of P. syringae strains, suggesting that hopAZ1 is closely associated with a gain of virulence in this woody host. Considering the subcellular localization of HopAZ1, identification of its interacting proteins in planta that also localize to the cytoskeleton could help reveal the role that HopAZ1 plays in promoting Psa virulence.

We note several differences for subcellular localization between Psa effectors and other known Pseudomonas effector homologs. HopM1 from Pto DC3000 is known to target AtMIN7 in Arabidopsis and destabilize AtMIN7 within the trans-golgi network and early endosome (Nomura et al., 2011). However, Psa HopM1 localized to the chloroplast in N. benthamiana. As the Pto DC3000 and Psa homologs of HopM1 share only 66.9% amino acid sequence identity, it is plausible that they target different host proteins in planta. However, previous findings have highlighted an alternative possibility since HopM1 was found to suppress SA-mediated immunity (DebRoy et al., 2004) as well as interact with 14-3-3 proteins to mediate pathogen virulence in planta (Lozano-Duran et al., 2014). This is particularly interesting since SA synthesis is primarily localized within the chloroplast and exported out by the multidrug and toxin extrusion-like transporter, EDS5, and 14-3-3 proteins have been associated with proteins targeted to the chloroplast (Sehnke et al., 2000; Serrano et al., 2013). Taken together, different localization of T3E homologs suggests that diversity of effector functions and targets may exist even in situations where effectors share significant homology.

Multiple Psa effectors and their homologs from other P. syringae strains were compared for their HCD-inducing activity and, despite high amino acid sequence identity, a few effector homologs (HopAA1 and HopR1) showed different activities in Nicotiana spp. (Table 2). On the other hand, a HopF homolog from Psa V13 and Pto DC3000 showed similar HCD-inducing activity in N. tabacum despite relatively low amino acid identity. An R protein-mediated recognition for an effector can occur by direct interaction or indirectly through ‘guardee’ or ‘decoy’ molecules which interact with the effector (Jones and Dangl, 2006; van der Hoorn and Kamoun, 2008). Loss of interaction between the effector and its host target (R protein or guardee/decoy) may cause failure of effector recognition and associated cell death (Jones and Dangl, 2006; van der Hoorn and Kamoun, 2008). The HCD-inducing activities for HopAA1 (Pto DC3000) and HopR1 (Psa V13) may be lost in their corresponding homologs due to evolutionary pressure on the effector active site or recognition determinant to evade recognition while retaining virulence contribution. HopF4b (Psa HopF2), meanwhile, is a member of a wide-spread family of ADP-ribosyltransferases present in P. syringae strains with extensive genetic diversity (Lo et al., 2016). Psa possesses several effectors which share significant amino acid similarity with HopF2: HopF1, HopF4b, HopX3 (a novel member of the HopF family), HopBB1-1, and HopBB1-2. Among these, HopX3, HopBB1-1, and HopBB1-2 are located on the same pathogenicity island, termed the exchangeable effector locus (EEL), sites of significant genome variation between different Psa strains (McCann et al., 2013). This suggests that this family of effectors is under evolutionary pressure and may be a case study in evolutionary functional diversification. Unfortunately, mechanisms of cell death triggered by these effectors are still poorly understood. It would be interesting for future work to identify the recognition mechanisms and to search for the sequence variation in the interacting motifs responsible for variation in the HCD phenotype.

Several Psa effectors were shown to suppress effector-triggered HCD in this study (Figure 4). There are a few hypotheses on how effectors suppress ETI: (1) they share a common target with HCD-eliciting effectors and undergo biochemical processes to antagonistically interfere with the recognition, e.g., AvrRpm1/AvrRpt2/HopF2 (Kim et al., 2005; Wilton et al., 2010) and AvrPto/AvrPtoB (Abramovitch, 2003); (2) they may interfere with the downstream signaling of defense responses triggered by the effectors, e.g., HopF2 (Wang et al., 2010) and HopE1 (Guo et al., 2016); (3) they can interfere with the HCD-eliciting effectors through direct interaction, e.g., HopZ3/AvrB (Lee et al., 2015); or (4) they may suppress Agrobacterium-mediated transient protein expression in N. benthamiana, e.g., HopQ1 (Adlung and Bonas, 2017). Indeed, Psa HopQ1 suppressed most effector-triggered HCD tested in this study but notably failed to suppress HCD triggered by HopZ5, HopT1 and HopAM1-1. It would be interesting if HopZ5, HopT1 and/or HopAM1-1 affect HopQ1-mediated cell death suppression activity. Apart from HopQ1, other Psa effector-mediated HCD suppression activities were specific to one or two HCD-eliciting effectors (Figure 4). Nevertheless, several Psa effectors could suppress Psa effector-triggered HCD in N. benthamiana. Further research on these specific suppression mechanisms would help reveal the mechanism of ETI/HCD-suppression specificity.

Not all HCD triggered by agroinfiltration of effectors may be due to activation of immune responses (Vinatzer et al., 2006; Wroblewski et al., 2009). For instance, HopT1-1 triggered HCD but does not appear to be associated with significant growth restriction of virulent bacteria in N. benthamiana (Wei et al., 2007; Wroblewski et al., 2009). Conversely, P. syringae pv. tomato DC3000 lacking hopQ1-1 can cause disease symptoms in the non-host N. benthamiana but agroinfiltration of HopQ1-1 does not result in significant cell death (Wei et al., 2007; Wroblewski et al., 2009; Adlung and Bonas, 2017). Utilization of VIGS to suppress T3E-triggered cell death can offer clues about the downstream mechanism of effector-triggered HCD. The requirement of known immunity-regulator genes would suggest that the effector-triggered HCD is an immune response (Vinatzer et al., 2006). Wei et al. (2007), using VIGS, confirmed that SGT1 was required for cell death triggered by Pto-delivered HopQ1-1. Similarly, the PsyB728a effector, HopAA1, partially contributes to growth restriction of this strain on N. benthamiana and requires EDS1 for its HCD development (Vinatzer et al., 2006).

In addition to HopZ5 (Jayaraman et al., 2017), SGT1 is at least partially required for 5 Psa T3Es: HopAM1-1, HopAE1, HopT1, HopW1, and HopAB3. hopAM1-1 is unique to Psa V13 while hopT1, hopW1, and hopAB3 are unique to Psa LV5. hopAE1 alleles are present in both Psa strains. Interestingly, HopAM1 from Pto DC3000 is associated with resistance in Arabidopsis accession Bur-0 (Iakovidis et al., 2016), while HopAB3, a homolog of AvrPtoB from PtoDC3000, is an avirulence effector in tomato (Lin et al., 2006). Cell death triggered by HopX1, AvrE1, and HopR1 were largely unaffected by SGT1-silencing suggesting limited involvement in immune responses. Support for this notion comes from previous observations suggesting that HopX1, AvrE1, and HopR1 trigger necrosis rather than HR (Lorang and Keen, 1995; Badel et al., 2006; Nimchuk et al., 2007; Kvitko et al., 2009). Further characterization of Psa T3Es in N. benthamiana through gene knockouts in Psa would be useful for identification of avirulence effectors with a view to subsequently identify cognate R genes. Based on our results, the strongest avirulence gene candidates would be HopW1, HopAB3, HopZ5, HopAM1-1, and HopAE1.

Identification of R genes involved in the recognition of effectors is challenging and laborious. Recently, however, two advances in the field have facilitated this process. Firstly, Brendolise et al. (2017) have developed a hairpin-RNAi library targeting NLRs in N. benthamiana that allows for rapid identification of NLRs required for the recognition of an effector. They have demonstrated that the hairpin library is not only efficient at identifying the sensor NLR involved (Prf in the case of AvrPto) but also helper NLRs as well. This system is highly amenable for our purposes and multiple NLRs may be identified for cloning and downstream analyses without first mapping the resistance loci involved. The second technology is the development of RenSeq (Resistance gene enrichment sequencing) – a tool that capitalizes on the enrichment of specific NLR sequence(s) in a screened population of disease resistant plants (Jupe et al., 2013). This technology uses NLR enrichment and resequencing/reannotation to identify resistant alleles of NLRs even in crop plants that are still in the early stages of study. RenSeq and its derivative technologies have been used to identify candidate NLRs for multiple crop plants, including potato, tomato, and wheat (Jupe et al., 2013; Steuernagel et al., 2016; Witek et al., 2016). Application of RenSeq to identify candidate NLRs in tobacco and N. benthamiana would accelerate the identification of NLRs that recognize Psa effectors and, ultimately, the development of Psa-resistant kiwifruit.

Escherichia coli DH5α was used to clone and maintain effector constructs. Agrobacterium tumefaciens AGL1 was used for transient transformation of N. benthamiana and N. tabacum leaf cells. E. coli DH5α and A. tumefaciens AGL1 were cultured in low salt L-media with appropriate antibiotics at 37 and 28°, respectively. The final concentrations of antibiotics used for bacterial cultures were 100 μg/ml for ampicillin and 50 μg/ml for kanamycin.

Nicotiana tabacum Wisconsin 38 (W38) and N. benthamiana were grown for 4–5 weeks in a controlled plant growth room in long-day conditions (24°C, 16 h light/8 h dark).

Psa strains genomic DNA was extracted using Thermo gDNA extraction kit (GeneJET^TM, Thermo). All effector sequences were extracted from the published Psa genome database (McCann et al., 2013). Briefly, each effector sequence was divided into 1–1.5 kb modules based on length and presence of internal BsaI or SmaI restriction enzyme recognition sequences. Psa effector modules were PCR-amplified with flanking BsaI site-containing primers using high-fidelity polymerase (Phusion HiFi, Thermo). Amplified PCR products were ligated with SmaI-digested entry vector pICH41021. Ligated constructs were verified by BsaI digestion and Sanger sequencing. If present, internal BsaI sites in effector modules were removed by site-directed mutagenesis, as per instructions for the QuickChange II site-directed mutagenesis kit (Agilent, New Zealand). Chaperones for effectors, where present, were amplified with and cloned into the first module and were used exclusively for Pseudomonas-delivery constructs. Modules starting with the first codon of the effector coding sequence alone were used for Agrobacterium constructs. For Pseudomonas-delivery, effectors we cloned into the broad host-range vector pBBR1MCS5B:avrRps4_pro with a C-terminal 6xHA tag using the Golden Gate cloning method (Engler et al., 2008) as described previously (Jayaraman et al., 2017). For agroinfiltration and construction of the Agrobacterium-expression library, effectors were cloned into the Golden Gate compatible binary vector pICH86988 under the CaMV 35S promoter and TMV Ω leader and with a C-terminal YFP tag using the Golden Gate assembly method (Engler et al., 2008).

Broad host-range plasmid constructs of each Psa effector were mobilized from E. coli DH5α to P. fluorescens Pf0-1(T3S) strains (Thomas et al., 2009) by triparental mating using E. coli HB101 (pRK2013) as a helper strain. Psa V13, Psa LV5, or Pf0-1(T3S) carrying empty vector or Psa effector constructs, were streaked from glycerol stocks onto King’s B plates with antibiotic selection and grown for 2 days at 28°C. Bacteria were then harvested from plates, resuspended in 10 mM MgCl₂ and diluted to required OD₆₀₀. Infiltrations were carried out on fully expanded leaves of 4- to 5-week-old N. benthamiana with a blunt end syringe. For assessing HR, Psa V13, Psa LV5 or Pto DC3000 (OD₆₀₀ = 0.2; 1 × 10⁸ CFU/mL), or Pf0-1(T3S) (OD₆₀₀ = 0.6; 3 × 10⁸ CFU/mL) were infiltrated in patches on three N. benthamiana leaves (replicates). HR was assayed visually at 1 day post infiltration (dpi) (Psa/Pto) or 2–3 dpi (Pf0-1). For assessing PTI suppression, the method was adapted from that described previously (Le Roux et al., 2015). Briefly, 12 h after Pf0-1 infiltration (for non-HR-triggering Psa effectors), Pto DC3000 (OD₆₀₀ = 0.03; 2 × 10⁷ CFU/mL) was infiltrated in an overlapping area of the leaves. Pto DC3000-triggered tissue collapse was scored at 3 dpi. PTI suppression experiments were conducted four times with effectors demonstrating full suppression of Pto DC3000-triggered tissue collapse in all three reps in at least three out of four experiments scored as strong suppressors of PTI.

Agrobacterium tumefaciens AGL1 strain was transformed with binary vector effector constructs by electroporation (2.2 kV/6ms/Bio-Rad). Transformed cells were grown on L-agar plates with selective antibiotics for 2 days then inoculated and grown in liquid L-media with selective antibiotics. Overnight grown cultures were centrifuged at 5000 rpm for 4 min, resuspended in agroinfiltration solution (10 mM MgCl₂/10 mM MES) and adjusted at OD₆₀₀ = 0.4 before leaf infiltration with a blunt end syringe.

Nicotiana benthamiana leaves were infiltrated with a mixture of Agrobacterium AGL1 harboring the binary vector effector construct (OD₆₀₀ = 0.4) and the silencing suppressor P19 (OD₆₀₀ = 0.1). Five leaf disks (8 mm in diameter) were harvested from the infiltrated patch at 2 days post infiltration and snap frozen in liquid nitrogen. Samples were ground in Laemmli protein loading buffer (Tris-Cl pH 6.8 250 mM, SDS 8%, Bromophenol blue 0.1%, Glycerol 40% and DTT 100 mM) and boiled for 10 min. Total protein extracts were separated by SDS-PAGE, transferred into PVDF membrane (Sigma–Aldrich) and probed with anti-GFP-HRP conjugated antibodies (Santa Cruz Biotech). Super signal West Pico Chemiluminescent sensitivity substrate (Thermo Fisher) and Super signal West Pico Maximum sensitivity substrate (Thermo Fisher) were used for detection.

Short specific fragments (∼400 bp) from NbSGT1, NbNDR1, and NbEDS1 were designed using the Solgenomics VIGS tool (vigs.solgenomics.net). These fragments were PCR-amplified with primers introducing 5′ end EcoRI and 3′ end XhoI sites and were cloned into the pTRV2 vector between EcoRI and XhoI sites (Liu et al., 2002). Two week-old N. benthamiana seedlings were infiltrated with a mixture of A. tumefaciens AGL1 harboring pTRV1 (OD₆₀₀ = 0.5) and pTRV2 : gene to be silenced (OD₆₀₀ = 0.5) into cotyledons and grown for a further 4–5 weeks (Liu et al., 2002) in short day conditions (22°C, 11 h light/13 h dark). Gene silencing was confirmed by semi-quantitative amplification of the silencing target gene. Total RNA of N. benthamiana was extracted from one leaf disk (8 mm in diameter) using the Tri-reagent RNA extraction method (Ambion). cDNA was synthesized with Maxima kit (Thermo) following the manufacturer’s instruction and was used as template to amplify silenced genes with specific primers and Prime Taq polymerase (GenetBio).

Agrobacterium tumefaciens AGL1 harboring each Psa effector C-terminally tagged with YFP was infiltrated in 4-week-old N. benthamiana leaves and YFP signal was detected with a Carl Zeiss LSM700 confocal microscope at 2 days post infiltration. YFP fluorescence was excited at 488 nm with a 20 mW Argon laser and captured by the confocal channel in the emission range 500–530 nm. Images were processed using ImageJ software.