Protein engineering expands the effector recognition profile of a rice NLR immune receptor

Plant nucleotide binding, leucine-rich repeat (NLR) receptors detect pathogen effectors and initiate an immune response. Since their discovery, NLRs have been the focus of protein engineering to improve disease resistance. However, this approach has proven challenging, in part due to their narrow response specificity. Previously, we revealed the structural basis of pathogen recognition by the integrated heavy metal associated (HMA) domain of the rice NLR Pikp (Maqbool et al., 2015). Here, we used structure-guided engineering to expand the response profile of Pikp to variants of the rice blast pathogen effector AVR-Pik. A mutation located within an effector-binding interface of the integrated Pikp–HMA domain increased the binding affinity for AVR-Pik variants in vitro and in vivo. This translates to an expanded cell-death response to AVR-Pik variants previously unrecognized by Pikp in planta. The structures of the engineered Pikp–HMA in complex with AVR-Pik variants revealed the mechanism of expanded recognition. These results provide a proof-of-concept that protein engineering can improve the utility of plant NLR receptors where direct interaction between effectors and NLRs is established, particularly where this interaction occurs via integrated domains.


Introduction
Protein engineering offers opportunities to develop new or improved molecular recognition capabilities that have applications in basic research, health and agricultural settings. Protein resurfacing, where the properties of solvent-exposed regions are changed (often by mutation), has been used extensively in diverse areas from antibody engineering for clinical use to the production of more stable, soluble proteins for biotechnology applications (Chapman and McNaughton, 2016).
Intracellular nucleotide binding, leucine-rich repeat (NLR) receptors are key components of plant innate immunity pathways. They recognize the presence or activity of virulence-associated, hosttranslocated pathogen effector proteins and initiate an immune response (Kourelis and van der Hoorn, 2018;Cesari, 2018). Because they confer resistance to disease, plant NLRs are widely used in crop breeding programs (Dangl et al., 2013). However, the recognition spectra of plant NLRs tend to be very specific, and pathogens may delete detected effectors from their genome or evolve novel effector variants that are not detected by the NLRs to re-establish disease (Yoshida et al., 2016).
The potential of engineering NLRs to overcome these limitations, or to detect new effector activities, is emerging (Rodriguez-Moreno et al., 2017). Gain-of-function random mutagenesis has achieved some success in expanding the activation sensitivity or effector recognition profiles of NLRs (Segretin et al., 2014;Giannakopoulou et al., 2015;Harris et al., 2013). An alternative strategy, in which NLRs perceive protease effectors through their activity on engineered host proteins, can lead to expanded recognition profiles (Carter et al., 2018;Helm et al., 2019;Kim et al., 2016). When detailed knowledge of the direct binding interfaces between an effector and an NLR are known, there is potential for protein resurfacing to modify interactions and thereby impact immune signaling.
Plant NLRs are modular proteins that are defined by their nucleotide-binding (NB-ARC) and leucine-rich repeat (LRR) domains, but that also have either an N-terminal coiled-coil (CC) or a Toll/ Interleukin-1/Resistance-protein (TIR) signaling domain (Takken and Goverse, 2012). Many NLRs also contain non-canonical integrated domains (Kroj et al., 2016;Sarris et al., 2016;Bailey et al., 2018). Integrated domains are thought to be derived from ancestral virulence-associated effector targets that directly bind pathogen effectors (or host proteins; Fujisaki et al., 2017) or are modified by them, thereby initiating an immune response (Cesari et al., 2014;Le Roux et al., 2015;Maqbool et al., 2015;Sarris et al., 2015;De la Concepcion et al., 2018). These domains present an exciting target for protein engineering approaches to improve NLR activities. NLRs containing integrated domains (often called the 'sensor') typically function in pairs, requiring a second genetically linked NLR (the 'helper') for immune signaling (Białas et al., 2018;Eitas and Dangl, 2010).
Two rice NLR pairs, Pik and Pia, contain an integrated Heavy Metal Associated (HMA) domain in their sensor NLR that directly binds effectors from the rice blast pathogen Magnaporthe oryzae (also known as Pycularia oryzae) (Maqbool et al., 2015;Cesari et al., 2013;Ortiz et al., 2017;Guo et al., 2018). The integrated HMA domain in the sensor NLR Pik-1 directly binds the effector AVR-Pik. Co-evolutionary dynamics has driven the emergence of polymorphic Pik-1 HMA domains and AVR-Pik effectors in natural populations, and these display differential disease-resistance phenotypes (Costanzo and Jia, 2010;Kanzaki et al., 2012;Yoshida et al., 2009). The Pikp NLR allele only responds to the effector variant AVR-PikD, but the Pikm allele responds to AVR-PikD, AVR-PikE, and AVR-PikA. These phenotypes can be recapitulated in the model plant Nicotiana benthamiana using a cell-death assay, and are underpinned by differences in effector-receptor binding interfaces that lead to different affinities in vitro (Maqbool et al., 2015;De la Concepcion et al., 2018).
We hypothesized that by combining naturally occurring favorable interactions observed across different interfaces, as defined in different Pik-HMA/AVR-Pik structures (Maqbool et al., 2015;De la Concepcion et al., 2018), we could generate a Pik NLR with improved recognition profiles. Here, we graft an interface from Pikm onto Pikp by mutating two residues in Pikp (Asn261Lys, Lys262Glu), forming Pikp NK-KE . This single-site mutation strengthens the cell-death response in N. benthamiana to AVR-PikD, and gains a Pikm-like response to AVR-PikE and AVR-PikA. We show that this gain-offunction phenotype correlates with increased binding affinity of the effectors by the Pikp NK-KE-HMA domain in vitro and in vivo, and demonstrate that this mutation results in a Pikm-like structure for Pikp NK-KE when in complex with AVR-Pik effectors. Finally, we confirm that the newly engineered interface is responsible for the expanded response of Pikp NK-KE by mutation of the effectors.
This study serves as a proof-of-concept for the use of protein resurfacing by targeted mutation to develop plant NLR immune receptors with new capabilities. In the future, such approaches have the potential to improve disease resistance in crops.

Results
Structure-informed engineering expands Pikp-mediated effector recognition in N. benthamiana By comparing protein interfaces in the structures of Pikp-HMA and Pikm-HMA bound to different AVR-Pik effectors (Maqbool et al., 2015;De la Concepcion et al., 2018), we hypothesized that we could engineer expanded effector recognition capabilities by point mutation of Pikp. We constructed a series of mutations in the previously identified interface 2 and interface 3 regions of Pik-HMA-AVR-Pik structures , swapping residues found in Pikm into Pikp ( Figure 1A, Figure 1-figure supplement 1). We then screened these mutations for expanded Pikm-1 HMA domains. Secondary structure features of the HMA fold are shown above, and the residues that are located at binding interfaces are as colored. Key residues from interface 2 and interface 3 involved in this study are highlighted in red. (B) Representative leaf images showing Pikp-(left) or Pikp-1 NK-KE (right)-mediated cell death in response to AVR-Pik variants as autofluorescence under UV light. (C) Autofluorescence intensity is scored as previously described (Maqbool et al., 2015;De la Concepcion et al., 2018). Cell death assay scores are represented as dot plots for Pikp and Pikp NK-KE (blue and purple, respectively). For each sample, all of the data points are represented as dots with a distinct color for each of the three biological replicates; these dots are plotted around the cell death score for visualization purposes. The size of the centre dot at each cell death value is directly proportional to the number of replicates in the sample with that score. The total number of repeats was 80. Data for Pikp have been previously shown , but was acquired at the same time as those for Pikp NK-KE . The estimation methods used to visualize differences in the data sets are shown in Figure 1-figure supplement 3. (D) Conductivity measurements showing ion leakage as a quantitative measure of cell death. The centre line represents the median, the box limits are the upper and lower quartiles, the whiskers extend to the largest value within Q1 -1.5x the interquartile range (IQR) and the smallest value within Q3 + 1.5x IQR. All the data points are shown as dots with distinct colors for each biological replicate. For each experiment, six biological replicates with 5 or 10 internal repeats were performed (total data points = 40). 'p' is the p-value obtained from statistical analysis and Tukey's HSD (honestly significant difference) test. DOI: https://doi.org/10.7554/eLife.47713.002 The following source data and figure supplements are available for figure 1: effector recognition by monitoring cell death in a well-established N. benthamiana assay (Maqbool et al., 2015;De la Concepcion et al., 2018). We found that one double mutation in two adjacent amino-acid residues contained within interface 3, Asn261Lys and Lys262Glu (henceforth We subsequently focused on this double mutant, and independently repeated the cell-death assay to ensure its robustness ( Figure 1B,C, Figure 1-figure supplement 3). When looking at the intensity of cell death mediated by Pikp NK-KE , as for the Pikm allele (De la Concepcion et al., 2018), we observed an effector hierarchy of AVR-PikD > AVR-PikE > AVR-PikA . Pikp NK-KE shows a comparable, but elevated (not statistically significant in the case of AVR-PikE), response to effector variants when compared to that of Pikm ( Figure 1-figure supplements 2 and 4). The cell-death response to AVR-PikD mediated by Pikp NK-KE is similar but elevated when compared to that mediated by Pikp ( Figure 1B,C). Pikp NK-KE does not show a response to the stealthy AVR-PikC variant. To obtain a quantitative measure of cell death, we performed ion-leakage assays ( Figure 1D, Figure 1-figure supplement 2, Supplementary file 1). The results of these assays correlate well with the cell-death index scores based on autofluorescence described above.
We conclude that the single Asn261Lys/Lys262Glu (Pikp NK-KE ) mutation at interface 3 in the Pikp NLR expands this protein's recognition profile to include the effector variants AVR-PikE and AVR-PikA, similar to that observed for Pikm.

The engineered Pikp NK-KE -HMA mutant shows increased binding to effector variants in vivo and in vitro
We used yeast-2-hybrid (Y2H) and surface plasmon resonance (SPR) to determine whether the expanded Pikp NK-KE cell-death response in N. benthamiana correlates with increased binding affinity of the Pikp NK-KE -HMA domain for AVR-Pik effectors.
As AVR-PikE and AVR-PikA showed some interaction with Pikp-HMA using these approaches (Maqbool et al., 2015;De la Concepcion et al., 2018), we tested interactions with Pikp NK-KE -HMA side-by-side with interactions with wild-type Pikp-HMA. Using Y2H, we observed increases in growth and in blue coloration (both indicative of protein-protein interaction) for Pikp-HMA NK-KE with effectors AVR-PikE and, particularly, AVR-PikA when compared with Pikp-HMA ( Figure 2A). This was accentuated with more stringent conditions (imposed by increasing concentration of Aureobasidin A). Unexpectedly, we also observed limited yeast growth for Pikp-HMA NK-KE with AVR-PikC at the lower stringency ( Figure 2A). The unrelated M. oryzae effector AVR-Pii was used as a negative control. Expression of all proteins was confirmed in yeast (Figure 2-figure supplement 1).

Figure 1 continued
Source data 1. Cell death scoring data used in the preparation of Figure 1C. DOI: https://doi.org/10.7554/eLife.47713.009 Source data 2. Conductivity measurements used in the preparation of Figure 1D and  respectively. The centre line within each box represents the median, the box limits are the upper and lower quartiles, the whiskers extend to the largest value within Q1 -1.5 Â the interquartile range (IQR) and the smallest value within Q3 + 1.5 Â IQR. All of the data points are represented as dots with distinct colors for each biological replicate. For each experiment, three biological replicates with three internal repeats were performed. 'p' is the p-value obtained from statistical analysis and Tukey's HSD. For results of experiments with 4 nM and 100 nM HMA protein concentration, see These results show that the Pikp-HMA NK-KE mutant has a higher binding affinity for effectors AVR-PikE and AVR-PikA than does the wild-type protein. This suggests that the increased binding affinity to the HMA domain correlates with the expanded cell-death response in planta ( Figure 1B).

The engineered Pikp NK-KE mutant expands the association of full-length Pik-1 with effector variants in planta
In addition to interaction with the isolated HMA domain, we tested whether the Asn261Lys/Lys262-Glu mutant could expand effector variant binding in the context of the full-length NLR. After generating the mutant in the full-length protein, we co-expressed either Pikp-1 or Pikp-1 NK-KE with the AVR-Pik effector variants in N. benthamiana, followed by immunoprecipitation and western blotting to determine effector association.
AVR-PikD shows a robust association with Pikp-1. However, although we observe limited binding to the isolated Pikp-HMA domain in Y2H and SPR screens, we did not detect association of AVR-PikE and AVR-PikA with the full-length Pikp-1 in planta ( Figure 2C). By contrast, we observe clear association of AVR-PikE and AVR-PikA with the Pikp-1 NK-KE mutant, albeit with reduced intensity compared to AVR-PikD, correlating with the hierarchical cell-death response observed in planta ( Figure 1C). We also detect a very low level of association between full-length Pikp-1 NK-KE and AVR-PikC ( Figure 2C). However, co-expression of Pikp-1 NK-KE and AVR-PikC does not result in macroscopic cell death in N. benthamiana ( Figure 1C).
These results show that effector variant association with full-length Pikp-1 and Pikp-1 NK-KE correlates with the in planta cell-death response ( Figure 1C).

The effector-binding interface in the Pikp NK-KE mutant adopts a Pikmlike conformation
Having established that the Pikp NK-KE mutant displays an expanded effector recognition profile compared to wild-type Pikp, we sort to determine the structural basis of this activity. To this end, we determined the crystal structures of Pikp-HMA NK-KE bound to AVR-PikD, and to AVR-PikE. We obtained samples of Pikp-HMA NK-KE /AVR-PikD and Pikp-HMA NK-KE /AVR-PikE complexes by coexpression in E. coli (described in the Materials and methods and by De la Concepcion et al., 2018). Each complex was crystallized (see Materials and methods) and X-ray diffraction data were collected at the Diamond Light Source (Oxford, UK) to 1.6 Å and 1.85 Å resolution, respectively. The details of X-ray data collection, structure solution, and completion are given in the Materials and methods and in Table 1.
The overall architecture of these complexes is the same as that observed for all Pik-HMA/AVR-Pik effector structures. A key interaction at interface 3, one of the previously defined Pik-HMA-AVR-Pik interfaces , involves a lysine residue (Lys262 in Pikp and Pikm) that forms intimate contacts within a pocket on the effector surface ( Figure 3). In order to position this lysine in the effector pocket, Pikp has to loop-out regions adjacent to this residue, compromising the packing at the interface (De la Concepcion et  and D). By contrast, in Pikm, where the position of the lysine is shifted one residue to the N-terminus, no looping-out is required to locate the lysine within the pocket ( Figure 3A [right panel], B [right panel], C and D). In the Pikp NK-KE mutant, the position of this key lysine is shifted one residue to the N-terminus compared to that in the wild-type, so that it occupies the same position in the Pikp NK-KE sequence as in the Pikm sequence. In the crystal structures of Pikp-HMA NK-KE in complex with either AVR-PikD or AVR-PikE, we see that this region of the HMA adopts a Pikm-like conformation ( Figure 3A [middle panel], B [middle panel], C and D), with no looping-out of the preceding structure. This confirms that in making the Pikp NK-KE mutant, we have resurfaced Pikp to have a more robust, Pikm-like interface in this region. We found only limited structural perturbations at either of the other previously defined interfaces (interface 1 or 2 ) resulting from the binding of the AVR-PikD or AVR-PikE effectors to Pikp-HMA or Pikp-HMA NK-KE (Figure 3-figure supplement 1). We therefore conclude that the effects of the Pikp NK-KE mutant on protein function are mediated by altered interactions at interface 3.

Mutation in AVR-Pik effectors at the engineered binding interface impacts the in planta response and in vivo binding
To further confirm that the engineered binding interface is responsible for the expanded recognition of AVR-PikE and AVR-PikA by Pikp NK-KE , we used mutants in the effectors at interface 2 (AVR-PikD H46E ) and interface 3 (AVR-PikD,E,A E53R ), which have previously been shown to impact the interactions and in planta responses of wild-type NLR alleles (De la Concepcion et al., 2018). We tested whether these mutants affected the cell-death response in N. benthamiana, the interactions between effectors and Pikp-HMA NK-KE (using Y2H), and the interactions between effectors and full-length Pikp NK-KE (using in planta co-immunoprecipitation [co-IP]).
First, we investigated the impact of mutation at interface 2 using the AVR-PikD H46E mutant. We found that cell death in N. benthamiana is essentially blocked when co-expressing either Pikp or Pikp NK-KE with this mutant (AVR-PikD H46E ), suggesting that the engineered NLR is still reliant on this interface for response ( Figure 4A, Figure 4-figure supplement 1). Intriguingly, Y2H shows that the AVR-PikD H46E mutant displays some interaction with Pikp-HMA NK-KE ( Figure 4B), similar to this mutant's interaction with Pikm-HMA (De la Concepcion et al., 2018), although this interaction is barely observed by co-IP with the full-length NLR ( Figure 4C).
Second, we investigated the impact of mutations at interface 3 using the Glu53Arg (E53R) mutant in AVR-PikD, AVR-PikE and AVR-PikA. We found that the AVR-PikD E53R mutant has essentially no effect on recognition of the effector by Pikp NK-KE in N. benthamiana, and little or no effect on the interaction with Pikp-HMA NK-KE or full-length Pikp NK-KE ( Figure 4A,B,C, Figure 4-figure supplement 1). By contrast, the equivalent mutation in AVR-PikE and AVR-PikA restricted the cell-death response in N. benthamiana, reduced the binding to Pikp-HMA NK-KE in Y2H (as shown by the reduced blue coloration) and produced a less intense band for the effector following Pikp NK-KE co-IP ( Figure 4A,B,C). The expression of all proteins in yeast was confirmed by western blot (Figure 4figure supplement 2).
These results provide evidence to show that while interactions across interface 2 remain important for the Pikp NK-KE interaction with AVR-Pik effectors, it is the altered interaction at interface 3, as observed in the structures, that is responsible for the expanded recognition profile of this engineered mutant.

Discussion
Plants, including food crops, are under continuous threat from pathogens and pests, and new solutions to control disease are required. Although largely elusive to date, engineered plant NLR-type intracellular immune receptors have the potential to improve disease-resistance breeding (Dangl et al., 2013;Rodriguez-Moreno et al., 2017). NLR-integrated domains are a particularly attractive target for protein engineering because they interact directly with pathogen effectors (or host effector targets). Further, where tested, their binding affinities in vitro correlate with in planta immunity phenotypes (Maqbool et al., 2015;De la Concepcion et al., 2018;Guo et al., 2018), allowing biochemical and structural techniques to inform NLR design directly. Here, we show that the recognition profile of the rice NLR Pikp can be expanded to include different AVR-Pik variants by engineering the binding interface between these proteins. This strengthens the hypothesis that tighter binding affinity between effectors and integrated HMA domains correlates with increased immune signaling in plants. This was previously shown for both natural alleles of Pik (Maqbool et al., 2015;De la Concepcion et al., 2018) and for Pia (Guo et al., 2018), but is now also shown for an engineered NLR. We propose this may be a general model for integrated domains that directly bind effectors.
Natural variation in Pik NLRs has given rise to different effector recognition profiles, and contributions from different binding interfaces were suggested to underpin this phenotype (De la . In particular, a more favorable interaction at one interface (interface 3) in Pikm, compared to that in Pikp, was concluded to have evolved to compensate for changes in binding at a different site (interface 2). Here, through mutation of residues in Pikp (forming Pikp NK-KE ), we have combined favorable interfaces from Pikp and Pikm into a single protein. This has resulted in an expanded recognition phenotype for effector binding and response in planta. In vitro, the binding of Pikp NK-KE -HMA to effector variants is consistently higher than that of Pikm-HMA (Figure 2-figure  supplement 3). This is likely to be underpinned by differences in a cluster of key residues within interface 2 (spatially equivalent residues Val222, Lys228, and Glu230 in Pikp, and Ala223, Gln229, and Val231 in Pikm) that form more energetically favorable contacts, including water-mediated hydrogen bonds. However, despite a positive trend (which is significant in some cases), the levels of cell death mediated by Pikm and Pikp NK-KE in planta are comparable (Figure 1-figure supplements 2 and 4).
Although the Pikp NK-KE mutant did not deliver a cell-death response in N. benthamiana to the stealthy AVR-PikC effector variant, it did show a gain-of-binding in vitro, as well as weak binding in Y2H and in in planta co-IP. It is surprising that the gain-of-binding for AVR-PikC in vitro, unlike that for AVR-PikA, does not correlate directly with gain of in vivo interactions or cell-death response in N. benthamiana. We hypothesize that this gain-of-binding for AVR-PikC observed in vitro is not sufficiently robust, especially in the context of the full-length NLR, for binding or for triggering immune signaling. Nevertheless, this work sets the scene for future interface engineering experiments that may further improve the response profiles of Pik NLRs to currently unrecognized effector variants. It also suggests that future work should test the disease resistance profile of M. oryzae strains carrying the different effector variants in rice expressing the engineered receptor.  Figure 1. (Right) Pikp NK-KE cell-death assay quantification in the form of dot plots. For each sample, the data points are represented as dots with a distinct color for each of the three biological replicates; these dots are plotted around the cell death score for visualization purposes. The size of the central dot at each cell death value is proportional to the number of replicates of the sample with that score. The number of repeats was 90. (B) Yeasttwo-hybrid assay of Pikp-HMA NK-KE with AVR-Pik variants and mutants. Control plate for yeast growth is on the left with quadruple dropout media supplemented with X-a-gal and increasing concentrations of Aureobasidin A on the right for each combination of HMA/AVR-Pik. The unrelated M. oryzae effector AVR-Pii was used as a negative control. Growth and the development of blue coloration in the selection plate are both indicative of proteinprotein interaction. HMA domains were fused to the GAL4 DNA binding domain, and AVR-Pik alleles to the GAL4 activator domain. Each experiment was repeated a minimum of three times, with similar results. (C) Coimmunoprecipitation of full-length Pikp-1 NK-KE with AVR-Pik variants and mutants. N-terminally 4xMyc tagged AVR-Pik effectors were transiently co-expressed with Pikp-1 NK-KE :6xHis3xFLAG in N. benthamiana leaves. Immunoprecipitates (IPs) obtained with anti-FLAG antiserum, and total protein extracts, were probed with appropriate antisera. Each experiment was repeated at least three times, with similar results. The asterisks mark the Pik-1 band. PS = Ponceau Stain. DOI: https://doi.org/10.7554/eLife.47713.022 The following source data and figure supplements are available for figure 4: Source data 1. Cell-death scoring data used in the preparation of Figure 4A. The integrated HMA domain in the NLR RGA5 (the sensor of the Pia NLR pair in rice) binds to M. oryzae effectors AVR1-CO39 and AVR-Pia via a different interface, and it has been suggested that these binding sites are mutually exclusive (Guo et al., 2018). This raises the possibility that an HMA domain could be engineered to bind and respond to multiple effectors (Guo et al., 2018). Recently, the Pikp-HMA domain was shown to interact with AVR-Pia at the same interface as that used by the RGA5-HMA domain, and this probably underpins the partial resistance of rice plants to M. oryzae expressing AVR-Pia (Varden et al., 2019). This presents a starting point for the use of Pikp as a chassis for such studies. Although it remains to be seen whether any such resurfaced HMA domain can bind to multiple effectors, these studies suggest that this strategy has potential as a novel approach.
Plant breeding is required to provide new genetic solutions to disease resistance in crops. This is necessary to limit the environmental and social damage caused by pesticides, and to deal with changes in climate and in the globalization of agriculture that result in the spread of pathogens and pests into new environments (Islam et al., 2016;Michelmore et al., 2017;Savary et al., 2019). Classical breeding for disease resistance has been limited by issues such as linkage drag and hybrid incompatibility, as also seen in model plant species (Bomblies et al., 2007). Novel molecular approaches such as engineering 'decoys' (Kim et al., 2016) and protein resurfacing, as described here, combined with modern transformation (Altpeter et al., 2016) and breeding pipelines (Watson et al., 2018), offer the opportunity for more targeted approaches to breeding for disease resistance. These strategies will complement other emerging technologies in NLR identification (Arora et al., 2019) and NLR stacking (Dangl et al., 2013) as methods to develop improved crops for the future.

Gene cloning
For in vitro studies, Pikp-HMA NK-KE (encompassing residues 186 to 263) was amplified from wildtype Pikp-HMA by introducing the mutations in the reverse primer, followed by cloning into pOPINM (Berrow et al., 2007). The wild-type Pikp-HMA, Pikm-HMA, and AVR-Pik expression constructs used in this study are as described in De la .
For Y2H, we cloned Pikp-HMA NK-KE (as above) into pGBKT7 using an In-Fusion cloning kit (Takara Bio USA), following the manufacturer's protocol. The wild-type Pikp-HMA domain in pGBKT7 and AVR-Pik effector variants in pGADT7 were generated as described in De la .
For protein expression in planta, the Pikp-HMA NK-KE domain was generated using site-directed mutagenesis by introducing the mutations in the reverse primer. This domain was then assembled into a full-length construct using Golden Gate cloning (Engler et al., 2008) and into the plasmid pICH47742 with a C-terminal 6xHis/3xFLAG tag. Expression was driven by the Agrobacterium tumefaciens Mas promoter and terminator. Full-length Pikp-1, Pikp-2, and AVR-Pik variants were generated as described in De la . All DNA constructs were verified by sequencing.
Expression and purification of proteins for in vitro binding studies pOPINM constructs encoding Pikp-HMA, Pikm-HMA and Pikp-HMA NK-KE were produced in E. coli SHuffle cells (Lobstein et al., 2012), using the protocol described in De la . Cell cultures were grown in autoinduction media (Studier, 2005) at 30˚C for 5-7 hr and then at 16˚C overnight. Cells were harvested by centrifugation and re-suspended in 50 mM Tris-HCl (pH 7.5), 500 mM NaCl, 50 mM glycine, 5% (vol/vol) glycerol, and 20 mM imidazole supplemented with EDTA-free protease inhibitor tablets (Roche). Cells were sonicated and, following centrifugation at 40,000xg for 30 min, the clarified lysate was applied to a Ni 2+ -NTA column connected to an AKTA Xpress purification system (GE Healthcare). Proteins were step-eluted with elution buffer (50 mM Tris-HCl [pH7.5], 500 mM NaCl, 50 mM glycine, 5% (vol/vol) glycerol, and 500 mM imidazole) and directly injected onto a Superdex 75 26/60 gel filtration column pre-equilibrated with 20 mM HEPES (pH 7.5) and 150 mM NaCl. Purification tags were removed by incubation with 3C protease (10 mg/mg fusion protein) followed by passing through tandem Ni 2+ -NTA and MBP Trap HP columns (GE Healthcare). The flow-through was concentrated as appropriate and loaded onto a Superdex 75 26/60 gel filtration column for final purification and buffer exchange into 20 mM HEPES (pH 7.5) and 150 mM NaCl.
AVR-Pik effectors, with either a 3C protease-cleavable N-terminal SUMO or a MBP tag and with a non-cleavable C-terminal 6xHis tag, were produced in and purified from E. coli SHuffle cells as previously described (Maqbool et al., 2015;De la Concepcion et al., 2018). All protein concentrations were determined using a Direct Detect Infrared Spectrometer (Merck).

Co-expression and purification of Pik-HMA and AVR-Pik effectors for crystallization
Pikp-HMA NK-KE was co-expressed with AVR-PikD or AVR-PikE effectors in E. coli SHuffle cells following co-transformation of pOPINM:Pikp-HMA NK-KE and pOPINA:AVR-PikD/E (which were prepared as described in De la Concepcion et al., 2018). Cells were grown in autoinduction media (supplemented with both carbenicillin and kanamycin), harvested, and processed as described in De la . Protein concentrations were measured using a Direct Detect Infrared Spectrometer (Merck).

Protein-protein interaction: yeast-two-hybrid analyses
To detect protein-protein interactions between Pik-HMAs and AVR-Pik effectors in a yeast twohybrid system, we used the Matchmaker Gold System (Takara Bio USA). Plasmid DNA encoding Pikp-HMA NK-KE in pGBKT7, generated in this study, was co-transformed into chemically competent Y2HGold cells (Takara Bio, USA) with the individual AVR-Pik variants or mutants in pGADT7 , as described previously . Single colonies grown on selection plates were inoculated in 5 ml of SD -Leu-Trp overnight at 30˚C. Saturated culture was then used to make serial dilutions of OD 600 1, 1 À1 , 1 À2 , and 1 À3 , respectively. 5 ml of each dilution was then spotted on a SD -Leu-Trp plate as a growth control, and on a SD -Leu-Trp-Ade-His plate containing X-a-gal and supplemented with Aureobasidin A (Takara Bio, USA). Plates were imaged after incubation for 60-72 hr at 30˚C. Each experiment was repeated a minimum of three times, with similar results.
To confirm protein expression in yeast, total protein extracts from transformed colonies were produced by boiling the cells for 10 min in LDS Runblue sample buffer. Samples were centrifugated and the supernatant was subjected to SDS-PAGE gels prior to western blotting. The membranes were probed with anti-GAL4 DNA-BD (Sigma) for the HMA domains in pGBKT7 and with the anti-GAL4 activation domain (Sigma) antibodies for the AVR-Pik effectors in pGADT7.

Protein-protein interaction: surface plasmon resonance
Surface plasmon resonance (SPR) experiments were performed on a Biacore T200 system (GE Healthcare) using an NTA sensor chip (GE Healthcare). The system was maintained at 25˚C, and a flow rate of 30 ml/min was used. All proteins were prepared in SPR running buffer (20 mM HEPES [pH 7.5], 860 mM NaCl, 0.1% Tween 20). C-terminally 6xHis-tag AVR-Pik variants were immobilized on the chip, giving a response of 200 ± 100. The sensor chip was regenerated between each cycle with an injection of 30 ml of 350 mM EDTA.
For all the assays, the level of binding was expressed as a percentage of the theoretical maximum response (R max ) normalized for the amount of ligand immobilized on the chip. The cycling conditions were the same as those used in De la . For each measurement, in addition to subtracting the response in the reference cell, a further buffer-only subtraction was made to correct for bulk refractive index changes or machine effects (Myszka, 1999). SPR data were exported and plotted using R v3.4.3 (https://www.r-project.org/) and the function ggplot2 (Wickham, 2009). Each experiment was repeated a minimum of three times, including internal repeats, with similar results. The proteins used came from three independent preparations for the HMA domains and two independent preparations of the AVR-Pik effectors.
Protein-protein interaction: in planta co-immunoprecipitation (co-IP) Transient gene-expression in planta for Co-IP was performed by delivering T-DNA constructs within the A. tumefaciens GV3101 strain into 4-week-old N. benthamiana plants grown at 22-25˚C with high-light intensity. A. tumefaciens strains carrying Pikp-1 or Pikp À1 NK-KE were mixed with strains carrying the AVR-Pik effector, at OD 600 0.2 each, in agroinfiltration medium (10 mM MgCl 2 and 10 mM 2-(N-morpholine)-ethanesulfonic acid (MES) [pH5.6]), supplemented with 150 mM acetosyringone. For detection of complexes in planta, leaf tissue was collected 3 days post infiltration (dpi), frozen, and ground to fine powder in liquid nitrogen using a pestle and mortar. Leaf powder was mixed with two times weight/volume ice-cold extraction buffer (10% glycerol, 25 mM Tris [pH 7.5], 1 mM EDTA, 150 mM NaCl, 2% w/v PVPP, 10 mM DTT, 1x protease inhibitor cocktail [Sigma], 0.1% Tween 20 [Sigma]), centrifuged at 4,200 g/4˚C for 20-30 min, and the supernatant was passed through a 0.45 mm Minisart syringe filter. The presence of each protein in the input was determined by SDS-PAGE and western blot. Pik-1 and AVR-Pik effectors were detected by probing the membrane with anti-FLAG M2 antibody (SIGMA) and anti c-Myc monoclonal antibody (Santa Cruz), respectively. For immunoprecipitation, 1.5 ml of filtered plant extract was incubated with 30 ml of M2 anti-FLAG resin (Sigma) in a rotatory mixer at 4˚C. After three hours, the resin was pelleted (800 g, 1 min) and the supernatant removed. The pellet was washed and resuspended in 1 ml of IP buffer (10% glycerol, 25 mM Tris [pH 7.5], 1 mM EDTA, 150 mM NaCl, 0.1% Tween 20 [Sigma]) and pelleted again by centrifugation as before. Washing steps were repeated five times. Finally, 30 ml of LDS Runblue sample buffer was added to the agarose and incubated for 10 min at 70˚C. The resin was pelleted again, and the supernatant loaded onto SDS-PAGE gels prior to western blotting. Membranes were probed with anti-FLAG M2 (Sigma) and anti c-Myc (Santa Cruz) monoclonal antibodies.

Cell-death scoring: UV autofluorescence
Detached leaves were imaged at 5 dpi from the abaxial side of the leaves for UV fluorescence images. Photos were taken using a Nikon D4 camera with a 60 mm macro lens, ISO set 1600 and exposure~10 secs at F14. The filter was a Kodak Wratten No.8 and white balance was set to 6250 degrees Kelvin. Blak-Ray longwave (365 nm) B-100AP spot light lamps were moved around the subject during the exposure to give an even illumination. Images shown are representative of three independent experiments, with internal repeats. The cell death index used for scoring was as presented previously (Maqbool et al., 2015). Dotplots were generated using R v3.4.3 (https://www.rproject.org/) and the graphic package ggplot2 (Wickham, H., 2009). The size of the centre dot at each cell death value is directly proportional to the number of replicates in the sample with that score. All individual data points are represented as dots.

Cell-death scoring: ion leakage
For ion-leakage quantification, plants were infiltrated with the relevant constructs on two different leaves. At 5 dpi, leaves were detached and two leaf discs with a diameter of 8 mm (one disc per leaf spot) were collected and floated in 1.5 mL of Milli-Q water. Conductivity (mS/cm) was measured immediately after transferring the leaf disc to water (time zero) using a LAQUAtwin EC-33 conductivity meter (Horiba UK Ltd). Leaf discs were then incubated for 6 h at room temperature with gentle shaking before measuring the final conductivity. The assay was carried out in six biological replicates with a total of 40 technical replicates (2 discs x 40 plants). Conductivity data for each sample were exported and plotted using R v3.4.3 (https://www.r-project.org/) and the function ggplot2 (Wickham, H., 2009). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.