The blast pathogen effector AVR-Pik binds and stabilizes rice heavy metal-associated (HMA) proteins to co-opt their function in immunity

Intracellular nucleotide-binding domain and leucine-rich repeat-containing (NLR) receptors play crucial roles in immunity across multiple domains of life. In plants, a subset of NLRs contain noncanonical integrated domains that are thought to have evolved from host targets of pathogen effectors to serve as pathogen baits. However, the functions of host proteins with similarity to NLR integrated domains and the extent to which they are targeted by pathogen effectors remain largely unknown. Here, we show that the blast fungus effector AVR-Pik binds a subset of related rice proteins containing a heavy metal-associated (HMA) domain, one of the domains that has repeatedly integrated into plant NLR immune receptors. We find that AVR-Pik binding stabilizes the rice small HMA (sHMA) proteins OsHIPP19 and OsHIPP20. Knockout of OsHIPP20 causes enhanced disease resistance towards the blast pathogen, indicating that OsHIPP20 is a susceptibility gene (S-gene). We propose that AVR-Pik has evolved to bind HMA domain proteins and co-opt their function to suppress immunity. Yet this binding carries a trade-off, it triggers immunity in plants carrying NLR receptors with integrated HMA domains. Significance Statement Rice blast disease, caused by the fungus Magnaporthe oryzae, is one of the most devastating diseases of rice. Therefore, understanding the mechanisms of blast fungus infection and resistance of rice against the disease is important for global food security. In this study, we show that the M. oryzae effector protein AVR-PikD binds rice sHMA proteins and stabilizes them, presumably to enhance pathogen infection. We show that loss-of-function mutants in one rice sHMA, OsHIPP20, reduced the level of susceptibility against a compatible isolate of M. oryzae, suggesting that M. oryzae requires host sHMA to facilitate invasion. Remarkably, OsHIPP20 knockout rice line showed no growth defect, suggesting editing sHMA genes may present a novel source of resistance against blast disease.


Introduction
Plant pathogens target host proteins to promote disease by secreting effector proteins (Hogenhout et al. 2009). Some of these host targets have been co-opted by plant intracellular nucleotide-binding leucine rich repeat (NLR) immune receptors to act as baits to detect pathogens, and are then known as integrated domains (IDs) (Cesari et al. 2014;Wu et al. 2015;Sarris et al. 2016;Kroj et al. 2016).
Genome-wide bioinformatics searches have found these domains in diverse immune receptors from multiple plant families (Sarris et al. 2016;Kroj et al. 2016;Baggs et al. 2017;Bailey et al., 2018). We hypothesize that such widespread NLR-IDs modulate basic immune responses that are conserved among plants. One such domain is the heavy metal-associated (HMA) domain found in four botanical families (Sarris et al. 2016). In rice, HMA domains have been integrated into two different NLRs, RGA5 (Okuyama et al. 2011) and Pik-1 (Ashikawa et al. 2008). In addition to rice (a member of Poaceae), HMA domains have also integrated into NLR immune receptors of plant species in the Brassicaceae, Fabaceae and Rosaceae (Sarris et al. 2016;Kroj et al. 2016). This indicates that HMAcontaining proteins have probably been repeatedly targeted by pathogens across a diversity of flowering plants. Therefore, understanding the endogenous function of HMA-containing proteins has the potential to reveal important basic features of plant disease susceptibility and immunity. In this study, we identify rice HMA-containing proteins as targets of an effector of the blast pathogen Magnaporthe (syn. Pyricularia) oryzae and address their potential function.
The rice Pik locus comprises two NLR genes, Pik-1 and Pik-2, and recognizes the M. oryzae effector AVR-Pik (Ashikawa et al. 2008). This recognition triggers plant immunity mediated by the hypersensitive response (HR). AVR-Pik is a 113-amino-acid protein originally defined as having no sequence similarity to known protein domains (Yoshida et al. 2009). More recently, structureinformed similarity searches showed that AVR-Pik belongs to the MAX (Magnaporthe AVRs and ToxB-like) family of fungal effectors (De Guillen et al. 2015), which adopt a six β-sandwich fold stabilized by buried hydrophobic residues, and commonly but not always, a disulphide bond. Pik-1 recognition of AVR-Pik is mediated by direct binding of the effector to an HMA domain (Maqbool et al. 2015) located between the N-terminal coiled-coil (CC) and nucleotide binding (NB) domains of Pik-1 (Kanzaki et al. 2012;Cesari et al., 2013). AVR-Pik and Pik-1 are described as being involved in a coevolutionary arms race that has resulted in the emergence of allelic series of both effector genes in the pathogen and NLR genes in the host (Kanzaki et al. 2012;De la Concepcion et al. 2018;Białas et al. 2018).
Biochemical and structural analysis of complexes between AVR-Pik variants and HMA domains of different Pik-1 alleles revealed the molecular interactions between the effector and NLR-ID (Maqbool et al. 2015;De la Concepcion et al. 2018;De la Concepcion et al. 2019). This knowledge recently allowed structure-guided protein engineering to expand the recognition profile of a Pik NLR to different AVR-Pik variants (De la Concepcion et al. 2019). The Pik-1 HMA domains exhibit a four b-sheets and two a-strands (babbab) topology similar to the yeast copper transporter domain Ccc2A (Banci et al. 2001), even though the characteristic MxCxxC metal-binding motif is degenerate in Pik-1. The integrated HMA domain of RGA5 also adopts the classical HMA domain fold but, intriguingly, uses a different interface to interact with the M. oryzae effectors AVR-Pia and AVR1-CO39 (Guo et al. 2018).
HMA domains are also found in other plant proteins that are unrelated to NLRs (De Abreu-Neto et al. 2013). These proteins form large and complex families known as heavy metal-associated plant proteins (HPPs) and heavy metal-associated isoprenylated plant proteins (HIPPs), here collectively referred to as small proteins containing an HMA domain (abbreviated as sHMA proteins).
One such sHMA protein is the product of the rice blast partial resistance gene pi21 (Fukuoka et al. 2009). The recessive allele pi21, a presumed loss-of-function allele with a deletion mutation, confers partial broad-spectrum resistance to rice against compatible isolates of M. oryzae. This finding implicates HMA domain-containing proteins in rice defense (Fukuoka et al. 2009). However, the molecular function of Pi21 and other rice sHMA proteins have not been characterized to date.
Unlike other M. oryzae effectors, AVR-Pik does not show extensive presence/absence polymorphisms within the rice-infecting lineage, and its evolution in natural pathogen populations is mainly driven by nonsynonymous amino acid substitutions (Kanzaki et al. 2012;Shi et al. 2018). This suggests that AVR-Pik encodes an activity of benefit to the pathogen that is maintained in resistanceevading forms of the effector. To address the virulence function of AVR-Pik, we set out to identify rice proteins other than the Pik NLRs that interact with this effector. We found that AVR-Pik binds and stabilizes sHMA proteins. Knockout of one sHMA gene (OsHIPP20) conferred enhanced resistance to infection by the blast pathogen, suggesting OsHIPP20 is a susceptibility gene (S-gene).
Our model is that AVR-Pik effectors interfere with sHMA function by stabilizing these proteins to support pathogen invasion.

AVR-PikD interacts with small heavy metal associated proteins (sHMAs) of rice
To identify rice proteins that may be putative targets of AVR-PikD, we performed a yeast 2-hybrid screen (Y2H) with the effector as bait and a cDNA library prepared from leaves of rice cultivar Sasanishiki inoculated with M. oryzae as the prey. From this screen we identified four HMAcontaining proteins, named OsHIPP19 (LOC_Os04g39350), OsHIPP20 (LOC_Os04g39010), OsHPP04 (LOC_Os02g37300) and OsHPP03 (LOC_Os02g37290) (De Abreu-Neto et al. 2013), as interactors of AVR-PikD, amongst other proteins ( Table S1). The sizes of AVR-PikD interacting HMAs ranged from 118 (OsHPP03) to 123 (OsHIPP19) amino acids. Rice sHMA proteins typically comprise a conserved N-terminal HMA domain followed by a variable proline-rich domain (Fig. 1), and may contain a C-terminal "CaaX" isoprenylation motif (where "a" represents an aliphatic amino acid and X represents any amino acid). They form a large protein family with 87 members in the rice genome (cultivar Nipponbare) as annotated by Rice Genome Annotation Project (Kawahara et al. 2013) (Fig. 2). Phylogenetic analyses of the aligned HMA domains of rice sHMA proteins revealed two clades supported by high bootstrap values (> 90%) with relatively large number of members that we designate here as Clades A and B. All four sHMA proteins interacting with AVR-PikD belong to Clade A (Fig. 2). Interestingly, the HMA domains of RGA5 and three alleles of the Pik-1 NLRs also cluster in Clade A. However, the integrated HMA domains of Pik-1 (Pik*-HMA, Pikm-HMA and Pikp-HMA) and RGA5 (RGA5-HMA) are on separate branches in the tree, indicating distinct lineages and diversification patterns.
To determine if AVR-PikD interacts with other Clade A sHMA proteins, we selected 9 that are expressed in rice leaves with FPKM (fragments per kilobase of exon per million reads mapped) value > 2 ( Fig. 2) and tested pairwise interactions by Y2H (Fig. 3). This experiment showed that AVR-PikD binds around half of the tested Clade A sHMA proteins (Fig. 2, 3, Fig S1). We also tested binding of AVR-PikD with three sHMA proteins from Clade B, including OsHIPP05 (Pi21). These sHMA proteins did not bind AVR-PikD (Fig. 2, 3, Fig S1). These data reveal that AVR-PikD shows specific binding to Clade A sHMA proteins.
Further, we also tested the interaction of Clade A sHMA proteins OsHIPP19, OsHIPP20, OsHPP04, OsHPP03 and LOC_Os04g39380 with the M. oryzae effectors AVR-Pia and AVR1-CO39, which interact with the HMA domain of RGA5 (Cesari et al. 2013). We found that AVR-Pia and AVR1-CO39 did not bind any of the sHMAs tested (Fig. S2). Also, none of the three AVRs interacted with the Pi21 HMA protein (Fig S2).

Co-immunoprecipitation confirms AVR-PikD binding to OsHIPP19 and OsHIPP20
Interactions between AVR-PikD and OsHIPP19 and OsHIPP20 were further tested by coimmunoprecipitation using proteins expressed in N. benthamiana (Fig. 4). OsHIPP19 and OsHIPP20 were fused with a FLAG-epitope at their N-termini (FLAG:OsHIPP19 and FLAG:OsHIPP20), and AVR-PikD was tagged with the hemagglutinin (HA) epitope at the C-terminus (AVR-PikD:HA).
These proteins were separately expressed in N. benthamiana leaves by Agrobacterium tumefaciens mediated transformation (agroinfiltration) and resulting leaf crude extracts were mixed and subjected to a pull-down experiment. Upon pull-down of FLAG:OsHIPP19 from leaf cell extracts with an anti-FLAG antibody, AVR-PikD was co-immunoprecipitated and detected following incubation with the anti-HA antibody (Fig. 4). A similar result was obtained for the interaction between OsHIPP20 and AVR-PikD. These results confirm that AVR-PikD binds OsHIPP19 and OsHIPP20 in plant extracts as well as in yeast.

All AVR-Pik variants tested bind rice sHMA proteins OsHIPP19 and OsHIPP20
Naturally occurring AVR-Pik variants are differentially recognized by allelic Pik NLRs. These recognition specificities correlate with the binding affinity of AVR-Pik variants to the integrated HMA domain of the Pik-1 NLR (Kanzaki et al. 2012;Maqbool et al. 2015;De la Concepcion et al. 2018;De la Concepcion et al., 2019). We tested whether the AVR-Pik variants AVR-PikA, C, or E, interact with the rice sHMA proteins OsHIPP19 and OsHIPP20 in Y2H. The results of this experiment (Fig.   5, Fig. S3) showed that similar to AVR-PikD, all AVR-Pik variants tested interacted with OsHIPP19 and OsHIPP20. This result suggests that all the tested AVR-Pik variants bind sHMAs, the possible host target proteins, whereas they vary in the recognition by different alleles of Pik NLRs.

AVR-PikD stabilizes OsHIPP19 and OsHIPP20 in plant cells
Next, we aimed to determine the effect of AVR-PikD binding to sHMA proteins. For this we coexpressed OsHIPP19 and OsHIPP20 with AVR-PikD in N. benthamiana leaves by agroinfiltration.
Following expression, the leaf extract was separated into supernatant and pellet fractions by centrifugation, and each fraction was analyzed by western blot (Fig. 6). We used expression of GUS protein and AVR-Pii as controls that do not bind OsHIPP19 and OsHIPP20. In the supernatant fraction we observed that the OsHIPP19 and OsHIPP20 proteins were degraded to smaller fragments when coexpressed with GUS or AVR-Pii. However, when co-expressed with AVR-PikD, we detected stronger signals of intact molecules of OsHIPP19 and OsHIPP20. As an additional control we also tested OsHIPP17, an sHMA that does not bind AVR-PikD (Fig. 2, 3, Fig. S4). We found that OsHIPP17 was degraded to a smaller fragment even in the presence of AVR-PikD. OsHIPP19 and OsHIPP20 proteins in the pellet fraction were not degraded irrespective of the presence or absence of AVR-PikD, whereas OsHIPP17 in the pellet was degraded to a smaller fragment. These results show that the AVR-PikD effector stabilizes sHMA proteins OsHIPP19 and OsHIPP20 in the plant cytosol, and this stabilization is specific to sHMA proteins that interact with this effector. We also noted a lower accumulation of AVR-PikD in the supernatant when co-expressed with OsHIPP17. This suggests the possibility that AVR-PikD may require interaction with other proteins in the plant cytosol for full stability.

N. benthamiana leaves
After observing the enhanced stability of OsHIPP19 and OsHIPP20 on co-expression with AVR-PikD, we tested whether binding of AVR-PikD affected the subcellular localization of these sHMA proteins. For this, we transiently co-expressed GFP-tagged OsHIPP19 and OsHIPP20 (GFP:OsHIPP19 and GFP:OsHIPP20) in N. benthamiana leaves together with either GUS, AVR-Pii or AVR-PikD, and performed confocal microscopy. In the presence of GUS or AVR-Pii we found that GFP:OsHIPP19 and GFP:OsHIPP20 showed nucleo-cytoplasmic localization and accumulated in punctate structures with varying sizes (Fig. 7). Interestingly, when OsHIPP19 or OsHIPP20 was co-expressed with AVR-PikD, these punctae-like structures were not observed and the sHMA proteins were diffused in the cytoplasm and nucleus. This finding indicates that AVR-PikD binding alters the subcellular distribution of the sHMA proteins.

Knockout of the OsHIPP20 gene reduces rice susceptibility to Magnaporthe oryzae
Of the seven sHMA proteins that interacted with AVR-PikD, we selected OsHIPP19, OsHIPP20 and OsHPP04 for further study because these were identified most frequently in the initial Y2H screen (Table S1) and showed strong interaction profiles with AVR-PikD ( Fig. 2, 3). To explore the function of these sHMA proteins in rice, we generated knockout (KO) mutants by CRISPR/Cas9-mediated mutagenesis in the rice cultivar Sasanishiki (Piks/Pia), which is easy to transform and susceptible to the blast fungus isolate Sasa2 without AVR-Pik or AVR-Pia. We targeted OsHIPP19 and OsHPP04 genes individually, as well as OsHIPP19 and OsHIPP20 together (Fig. 8, Fig. S5). The resulting KO lines were challenged with the compatible M. oryzae isolate Sasa2. The OsHIPP19 (two independent lines) and OsHPP04 individual KO lines showed a similar level of infection as the wild-type control (Sasanishiki). However, the OsHIPP20 KO line, as well as OsHIPP19+OsHIPP20 double KO line, showed a reduction in lesion size caused by M. oryzae infection (Fig. 8, Fig. S5). These results indicate that OsHIPP20 is a susceptibility (S-) gene that is required for full infection of rice (cultivar Sasanishiki) by M. oryzae.

Discussion
In this paper, we set out to identify the host targets of the M. oryzae effector AVR-Pik, and study the virulence function of this effector. We found that AVR-Pik binds multiple sHMA proteins of rice that belong to the same phylogenetic clade (Clade A), which also contains the integrated HMA of Pik-1 and RGA5 (Fig 2, 3). These findings support the view that NLR integrated domains have evolved from the host targets of pathogen effectors and that HMA-containing proteins are a major host target of plant pathogen effectors. In an independent study, Maidment et al. (2020) demonstrate that AVR-Pik binds to OsHIPP19 with nanomolar affinity in vitro and show the interaction of the effector with this sHMA is via an interface conserved with the Pik-1 integrated HMA domains providing further evidence that this effector targets host sHMA proteins.
Heavy metal-associated (HMA) domains were first defined in metal binding domains of Ptype ATPase family copper transport proteins, including human MNK and WND proteins, mutations of which cause Menkes disease and Wilson disease, respectively (Bull & Cox. 1994). HMA domains are also found in a number of heavy metal transport or detoxification proteins both in bacteria and eukaryotes. The yeast metallochaperone Atx1 was shown to deliver monovalent copper ions to the Ptype ATPase Ccc2 that transports copper to trans-Golgi vesicle where it is taken up by the multicopper oxidase Fet3 (Askwith et al. 1994;Pufahl et al. 1997;Rosenzweig & O'Halloran. 2000). A typical HMA domain contains two conserved cysteine residues involved in metal binding in a MxCxxC motif that is located towards the N-termini of the domain (Bull & Cox, 1994).
In most organisms, only a small number of HMA-containing proteins have been reported. By contrast, in plants, proteins containing HMA-like domains have massively expanded (Dykema et al. 1999; Barth et al. 2009;De Abreu-Neto et al. 2013). For example, Barth et al. (2009)  HMA-containing small protein (abbreviated as sHMA) genes (Fig. 2).
The biological functions of plant sHMA proteins reported so far are diverse. Two Arabidopsis HMA-containing proteins, CCH and ATX1, complemented yeast atx1 mutant, and are presumed to be involved in copper transport (Himelbrau et al. 1998;Puig et al. 2007;Shin et al. 2012). Barth et al. (2009) showed that the Arabidopsis HMA-containing protein HIPP26 localizes to nuclei and interacts with a zinc-finger transcription factor ATHB29, while Gao et al. (2009) reported the same protein (with an alternative name, ATFP6) was localized to plasma membrane and interacted with acyl-CoA-binding protein ACBP2, which was hypothesized to be involved in membrane repair after oxidative stress. Zhu et al. (2016) reported that the Arabidopsis HMA-containing protein NaKR1 interacts with Flowering Locus T (FT) and mediates its translocation from leaves to shoot apices. Cowan et al. (2018) reported that potato mop-top virus (PMTV) movement protein TGB1 interacts with Nicotiana benthamiana sHMA protein HIPP26 and relocalizes this protein from the plasma membrane to the nucleus, thus contributing to PMTV long-distance movement by altering transcriptional regulation.
Genetic studies have also revealed roles of specific plant sHMA proteins in defense and susceptibility towards pathogens. Deletion in the proline-rich domain of Pi21, a rice sHMA, conferred a partial resistance against compatible isolates of M. oryzae (Fukuoka et al. 2009). Virus-induced gene silencing of wheat TaHIPP1 enhanced resistance against stripe rust caused by Puccinia striiformis f. However, it remains unclear how these sHMA proteins impact interactions with these diverse pathogens. Nonetheless, given that HMA domains have integrated into NLR immune receptors in at least four botanical families, it is likely that HMA containing proteins have repeatedly been targeted by pathogens across a diversity of flowering plant species and are thus important components in plantpathogen interactions.
In addition to Pik-1, the NLR RGA5 also carries an integrated HMA domain that binds two M. oryzae effectors, AVR-Pia and AVR1-CO39. However, in our Y2H assays with a high stringency conditions, we didn't detect any interaction between AVR-Pia and AVR1-CO39 and the tested sHMA proteins. We hypothesize that these two effectors may weakly bind the tested sHMAs or bind other rice sHMA proteins among the >80 members of this family.
In this study, we revealed that gene knockout of OsHIPP20 confers enhanced resistance to rice against a compatible isolate of M. oryzae (Fig. 8). Therefore, like Pi21, OsHIPP20 is a susceptibility gene (S-gene), whose activity is required for full invasion of the M. oryzae pathogen in rice. We also found that AVR-PikD binds and stabilizes OsHIPP19 and OsHIPP20 (Fig. 6). We hypothesize that AVR-Pik-mediated stabilization of sHMA proteins suppresses host defenses, resulting in enhanced M. oryzae invasion of rice cells (Fig. 9). The next steps in this research are to determine the roles of the extended family of sHMA proteins in rice and other plants to understand the interplay between effector-mediated protein stabilization and disease.  Grund et al. 2020). Here, we show that AVR-Pik interacts with sHMA proteins that belong to the same phylogenetic clade as the HMA domains integrated into the rice NLRs Pik-1 and RGA5. Therefore, throughout evolution, the Pik-1 NLR immune receptor has co-opted sHMA proteins through the integration of an HMA domain and neofunctionalization of this domain as a bait for the effector (Fig. 9). This has launched a coevolutionary arms race between Pik-1 and AVR-Pik. Given that binding of AVR-Pik to Pik-1 HMA domains is necessary for triggering the hypersensitive response (HR) cell death and disease resistance in rice, new variants of AVR-Pik have arisen in M. oryzae populations that evade binding the integrated Pik-1 HMA but maintain their virulence activity (Kanzaki et al. 2012;Maqbool et al. 2015;Białas et al. 2018). Here we show that each of the AVR-Pik variants tested retain their binding to OsHIPP19 and OsHIPP20 (Fig. 5) consistent with the view that these stealthy effectors have retained their virulence activities. This demonstrates that effector variation can affect the phenotypic outcomes of disease susceptibility and resistance independently through mediating bespoke interactions with different HMA domains. This elegant model highlights a surprisingly intimate relationship between plant disease susceptibility and resistance, as well as pathogen virulence and avirulence activities, driven by complex coevolutionary dynamics between pathogen and host.

Construction of the maximum likelihood tree of HMA family genes
The protein sequences of the HMA domains were aligned by MAFFT (Katoh et al. 2013) with the following method parameter set: --maxiterate 1000 --localpair. Then, the maximum likelihood tree was constructed by IQ-TREE (Nguyen et al. 2015) with 1,000 bootstrap replicates (Hoang et al. 2018).
MATCHMAKER Library Construction & Screening kit was used to construct the rice cDNA library from leaf tissues of rice cultivar Sasanishiki 4, 24 and 48 h after inoculation with Magnaporthe oryzae strain Sasa2 (race 037.1). Yeast strain AH109 competent cells were transformed with pGBKT7/AVR-PikD pGADT7-Rec and the rice cDNA library by using the polyethylene glycol/lithium acetate (PEG/LiAc) method, and plated on selective agar plates containing minimal medium without Trp, Leu, Ade and His, and supplemented with 20 mg/L of 5-Bromo-4-Chloro-3-indolyl a-D-galactopyranoside (X-α-gal) and 10 mM 3-amino-1,2,4-triazole (3-AT). cDNAs in the library were transferred to pGAD-Rec vector harboring GAL4 activation domain (AD) by homologous recombination in yeast cells.
Positive yeast transformants were streaked onto a minimal medium agar plate without Trp and Leu and used for sequence analysis.
To examine the protein-protein interactions between sHMAs and AVR-Pia, AVR-Pii, AVR1-CO39 and AVR-Pik alleles, yeast two-hybrid assay was performed as described previously (Kanzaki et al. 2012). Bait and prey plasmid vectors were constructed as described in Table S2. Signal peptidetruncated cDNA fragments of AVRs were amplified by PCR by using primer set (Table S2) and inserted into EcoRI and BamHI sites of pGADT7 (prey) or pGBKT7 (bait) vectors (Clontech). sHMA cDNAs were synthesized from total RNAs of rice leaves (cultivar Sasanishiki) and inserted into pGADT7 and pGBKT7 by using SpeI and BamHI sites as described in Table S2. In the case of sHMAs containing SpeI or BamHI site, In-Fusion HD Cloning Kit (Takara) was utilized to construct plasmid vectors. The various combinations of bait and prey vectors were transformed into yeast strain AH109 by using the PEG/LiAc method. To detect the protein-protein interactions, ten-fold dilution series (×1, ×10 -1 , ×10 -2 ) of yeast cells (×1 : OD600=1.0) were spotted onto on basal medium lacking Trp, Leu, Ade and His but containing X-α-gal (Clontech) and 10 mM 3-amino-1,2,4-triazole (3-AT). Positive signals were evaluated by blue coloration and growth of the diluted yeast. As a control, yeast growth on basal medium lacking Trp, Leu was also checked. Details of plasmids used are indicated in Table   S2.

Generation of rice mutants of OsHIPPs by CRISPR/Cas9 system
Rice plants with mutated OsHIPP19, OsHIPP20 or OsHPP04 were generated using the  Table S2. The rice cultivar 'Sasanishiki' was used for Agrobacterium-mediated transformation following the methods of Toki et al. (2005), with a sequence mutation. Thereafter, regenerated T0 plants were sequenced and the mutation type was confirmed the mutation type by using primers listed in Table S2. T3 lines from the selected T2 plants were used for the experiments.

Rice pathogenicity assays
Rice leaf blade punch inoculations were performed using the M. oryzae strains Sasa2 (without AVR-Pik alleles). A conidial suspension (3 × 10 5 conidia mL -1 ) was punch inoculated onto the rice leaf one month after sowing. The inoculated plants were placed in a dew chamber at 27°C for 24 h in the dark and transferred to a growth chamber with a photoperiod of 16 h. Disease lesions were scanned 10 days post-inoculation(dpi) and the lesion size was measured using 'Image J' software (Schneider et al. 2012). The assays were repeated at least 3 times with qualitatively similar results.

Transient gene expression assay in N. benthamiana
For transient protein expression, Agrobacterium tumefaciens strain GV3101 was transformed with the relevant binary constructs (Table S2)

Localization of OsHIPP19 and OsHIPP20
To visualize subcellular localization of OsHIPP19 and OsHIPP20, GFP tagged OsHIPP19 and OsHIPP20 were generated by Golden Gate methods (Engler et al. 2008)      Interactions between AVR-PikD and a subset of sHMA proteins were tested by Y2H. sHMA proteins were used as prey and AVR-PikD as bait (left panels) and AVR-PikD was used as prey and sHMA proteins as bait (right panels). Results of high stringency selection (QDO+3AT: Trp -Leu -Ade -His -Xα gal + ,10 mM 3AT) as well as no selection (DDO) are shown. In the supernatant fraction, OsHIPP19 and OsHIPP20 bound by AVR-PikD remain largely stable, whereas OsHIPP19 and OsHIPP20 expressed with GUS or AVR-Pii were degraded to a lower mass fragment. OsHIPP17 does not bind AVR-PikD and is degraded even in the presence of the effector. OsHIPP19 and OsHIPP20 proteins in the pellet fraction were intact irrespective of the presence or absence of AVR-PikD, whereas OsHIPP17 in the pellet was degraded to a smaller fragment. AVR-PikD seems to accumulate in the pellet fraction when it does not bind an sHMA. We obtained similar results in three independent experiments.  OsHIPP19 (top) and OsHIPP20 (bottom) proteins fused with GFP at their N-termini (GFP:OsHIPP19 and GFP:OsHIPP20, respectively) were transiently expressed in N. benthamiana leaves by agroinfiltration together with GUS (left), AVR-Pii (center) or AVR-PikD (right) and were observed under confocal laser microscopy. GFP:OsHIPP19 and GFP:OsHIPP20 accumulate to punctae-like structures when expressed with GUS or AVR-Pii, whereas these proteins were evenly distributed in the cytoplasm when expressed with AVR-PikD. We obtained similar results in three independent experiments. Scale bar: 50 μm.   In the compatible interaction (susceptible, left), AVR-Pik binds rice sHMA proteins and stabilizes them, possibly enhancing pathogen virulence. In the incompatible interaction (resistant, right), AVR-Pik interacts with integrated HMA domains of the Pik-1 NLRs which, together with Pik-2, triggers disease resistance by the hypersensitive response (HR).

Incompatible interaction
sHMA Host invasion

AVR-Pik sHMA
Pik-2 HR ID-HMA Table S1. Rice proteins that interacted with AVR-PikD in the Y2H screen. We carried out two Y2H screens (1 st and 2 nd ) and the number of positive clones with insert sequences corresponding to the designated proteins are shown. Four Magnaporthe oryzae effectors, AVR-Pia, AVR-PikD, AVR-Pii and AVR1-CO39 were tested for their binding with Clade A sHMAs (OsHIPP19, OsHIPP20, OsHPP04, OsHPP03 and LOC_Os04g39380) as well as Pi21 (OsHIPP05) of Clade B in Y2H assay with high stringency condition (QDO+3AT: Trp -Leu -Ade -His -Xα gal + ,10 mM 3AT) as well as no selection (DDO). The HMA domain of Pikm-1 NLR protein (Pikm-HMA) interacts with AVR-PikD (Kanzaki et al., 2012) and used as a positive control. AVR-Pii was used as a negative control. Top panels show the results when effectors were used as bait and sHMAs as prey, bottom panels show the results when effectors were used as prey and sHMAs as bait. (b) KO mediated by a guide RNA targeting common sequence of OsHIPP19 and OsHIPP20 resulted in the mutants where only OsHIPP20 was knocked out (oshipp20) and both OsHIPP19 and OsHIPP20 were knocked out (oshipp19&20). (c) KO of OsHPP04.