Structural Insights into the Plant Immune Receptors PRRs and NLRs1[OPEN]

Recent progresses made in structural analysis of plant PRRs and NLRs show the advancements in cryo-EM structural biology.

The plant immune system is composed of two classes of receptors-the membrane-anchored pattern recognition receptors (PRRs) and the intracellular nucleotidebinding and leucine-rich repeat receptors (NLRs; Jones and Dangl, 2006;Dangl et al., 2013). The PRR family consists of receptor kinases (RKs) and receptor-like proteins (RLPs) acting in the first tier of the plant immune system. PRRs contain a variable ectodomain that usually function to recognize either conserved microbial signatures known as pathogen-associated molecular patterns (PAMPs) or damage indicators known as danger-associated molecular patterns (DAMPs). PRR activation induces immune responses known as PAMP-triggered immunity (PTI) including expression of immune-related genes ( Fig. 1) to ward off microbes (Macho and Zipfel, 2014). Many pathogenic microbes, however, can successfully deliver effector proteins into the plant cell to dampen PTI signaling by manipulating host targets. Plants hence have evolved NLRs as intracellular immune receptors to mediate the second level of surveillance ( Fig. 1) through specific recognition of pathogen effectors (Chisholm et al., 2006). Upon perception of effectors, NLRs coordinate a rapid and robust immune signaling response termed effectortriggered immunity, which often leads to hypersensitive response (HR, local cell death at the infection site) and limitation of pathogenic microbes ( Fig. 1; Jones et al., 2016).
In the last two decades, tremendous advances have been made in functional and mechanistic dissection of plant PRRs and NLRs. There are many excellent reviews on these exciting achievements, mainly from the point view of genetics and physiology (Cui et al., 2015;Boutrot and Zipfel, 2017;Tang et al., 2017;Kourelis and van der Hoorn, 2018;Wan et al., 2019b;Zhang et al., 2017b). In this review, we highlight some of recent structural studies of PRRs and NLRs and discuss how they provided insights into their acting mechanisms.
Ligand sensing by ECDs activates KDs for immune signaling. Because of the lack of KDs, RLP-PRRs generally function together with RKs. PAMP perception by single RK-PRRs such as Arabidopsis (Arabidopsis thaliana) AtCERK1 has been reported (Miya et al., 2007;Wan et al., 2008), but many RK-PRRs require a coreceptor for signaling. Several recent studies (Jaillais et al., 2011;Smakowska-Luzan et al., 2018;Xi et al., 2019) suggested that the size of ECD is crucial for ligand perception by LRR-RKs and could be used to predict whether they function as receptors or coreceptors. For instance, when the LRR-RKs containing large size of ECDs act as the ligand-binding receptors, another small LRR-RK is preferred for coreceptor, but not itself. The LRR-RK SOMATIC EMBRYOGENESIS RECEP-TOR KINASE3 (SERK3, also called BRI1-ASSOCIATED KINASE1 [BAK1]) and its orthologs are commonly shared coreceptors by LRR-RKs involved in diverse signaling pathways including immunity . Additionally, BAK1/SERKs together with the LRR-RK SUPPRESSOR OF BIR-1 (SOBIR1) also function as coreceptors of multiple LRR-RLPs (Liebrand et al., 2014). Given their critical role in plant immunity, BAK1/SERKs are subjected to negative regulation by pathogen effectors  and by host components like the LRR-RK BAK1-INTERACTING RECEPTOR KINASEs (BIRs; Gao et al., 2009;Halter et al., 2014) in Arabidopsis. More recently, the Arabidopsis malectin-like RK FERONIA (FER) was shown to act as a scaffold for regulation of assembly of PRRcontaining complexes (Shen et al., 2017;Stegmann et al., 2017).

PAMP-induced Homodimerization of PRRs for Activation
Chitin, a polymeric N-acetyl-glucosamine (NAG), is a well-characterized fungal PAMP. Different mechanisms are employed by Arabidopsis and rice (Oryza sativa) for perception of the PAMP. The LysM-RKs AtCERK1 (Miya et al., 2007;Wan et al., 2008;Liu et al., 2012b) and LYSIN MOTIF RECEPTOR KINASE5  are direct receptors of chitin in Arabidopsis. In contrast, direct chitin recognition in rice is through the LysM-RLP OsCEBiP to activate OsCERK1 (Kaku et al., 2006;Shimizu et al., 2010;Liu et al., 2016a). The crystal structures of AtCERK1 ECD and OsCEBiP ECD in complex with a chito-pentamer and a chito-tetramer, respectively, revealed a conserved chitin recognition mechanism (Fig. 3, A and B;Liu et al., 2016aLiu et al., , 2012b. The structures of AtCERK1 ECD and OsCEBiP ECD are conserved in their LysMs, but only LysM2 was found to bind chitin. In the structures, chito-oligomers anchor to a shallow groove created by two loops of LysM2. Recognition of the chitin oligomers by AtCERK1 and OsCEBiP is mainly through a conserved set of residues of LysM2 that interacts with three NAG units. Interaction of the N-acetyl groups of chitin with AtCERK1 and OsCEBiP can allow the two proteins to distinguish the PAMP from b-1,3-glucan (Glc; Liu et al., 2016aLiu et al., , 2012b. However, in addition to chitin, PGN (Willmann et al., 2011) and nonbranched Glc (Melida et al., 2018) also trigger AtCERK1-mediated immune responses in Arabidopsis. PGN perception by AtCERK1 is through the LysM-RLPs LYM1 and LYM3, although how AtCERK1 is activated remains elusive. Modeling studies suggested that b-1,3-Glc 6 also binds to LysM2 of AtCERK1 but with a different orientation from the chitopentamer or the chitotetramer (Melida et al., 2018). AtCERK1 was suggested to recognize chitosan, a partially deacetylated chitin, to elicit immune responses in Arabidopsis (Cabrera et al., 2010;Petutschnig et al., 2010), which is further confirmed by a more recent study (Gubaeva et al., 2018).
Biochemical and functional data support chitininduced AtCERK1 homodimerization for activation (Liu et al., 2012b). A longer chitin chain was proposed to act as a cross linker along which two AtCERK1 molecules bind for dimerization, known as the crosslinking model. A later modeling study of OsCEBiP dimerization suggested that hexachitin mediates homodimerization of OsCEBiP, with each LysM2 binding three NAGs in a "sliding mode" (Liu et al., 2016a), further supporting the cross-linking model. Two alternative models were recently suggested on chitininduced OsCEBiP/AtCERK1 dimerization. One is called the sandwich-like model, where two CEBiP molecules simultaneously bind to one chitin chain from opposite sides (Hayafune et al., 2014). Analyses of chitin/chitosan oligosaccharides with varying degrees of polymerization or acetylation led to the slipped sandwich model (Gubaeva et al., 2018), in which two AtCERK1 molecules form an off-set chitin-binding groove for chitin or chitosan binding. This new model is a combination of the above models and provides an explanation for inhibition of chito-octamer-induced immunity by a chitosan octamer consisting of alternating GlcN and GlcNAc (Hayafune et al., 2014). Regardless of the mechanisms involved, chitin-induced homodimerization is required for AtCERK1 and OsCEBiP activation.

PAMPs and DAMPs Act as Molecular Glue To Induce Heterodimerization of PRRs with Their Co-receptors
The LRR-RK PRR FLS2 perceives flagellin by recognizing its highly conserved N-terminal epitope, a 22residue peptide called flg22. BAK1 as a coreceptor is required for FLS2-mediated immune signaling (Chinchilla et al., 2007;Sun et al., 2013). The crystal structure of the ecto-LRR domain of FLS2 (FLS2 LRR ) in complex with flg22 and BAK1 LRR revealed that flg22 adopts an elongated conformation interacting with the inner surface of FLS2 LRR (Sun et al., 2013;Fig. 3C). Flg22 binding creates a novel surface on FLS2 LRR for interaction with BAK1 LRR . The C-terminal side of flg22 is sandwiched between FLS2 LRR and BAK1 LRR , indicating that flg22 acts as molecular glue to connect FLS2 with BAK1. In addition to the flg22-mediated interaction, BAK1 LRR also anchors to the C-terminal portion of FLS2 LRR . Both flg22-mediated and direct FLS2 LRR -BAK1 LRR contacts are important to form the FLS2-flg22-BAK2 complex.
The Arabidopsis PLANT ELICITOR PEPTIDES are classic DAMPs recognized by the LRR-RKs PEPR1 and PEPR2 (Yamaguchi et al., 2006(Yamaguchi et al., , 2010. The crystal structure of PEPR1 LRR bound by AtPep1 showed that their recognition mechanism is remarkably conserved with that for FLS2 LRR recognition of flg22 (Tang et al., 2015;Fig. 3, C and D), although flg22 and AtPep1 are sequence-unrelated. Similarly, AtPep1 induced a heterodimeric PEPR1 LRR -BAK1 LRR complex. Modeling and binding studies indicated that the C-terminal side of AtPep1 is required for PEPR1 LRR interaction with BAK1 LRR , supporting AtPep1 as molecular glue to induce PEPR1 LRR -BAK1 LRR heterodimerization. Later biochemical and structural studies showed that many plant growth-promoting peptides such as CLAV-ATA3/ENDOSPERM SURROUNDING REGION-RE-LATED41   Successful pathogens deliver effector proteins (red dots) into the plant cell to dampen PTI. In some host plants, effector proteins are specifically recognized by the intracellular NLR immune receptors via different strategies, inducing effector-triggered immunity (ETI) that includes expression of immune-related genes and localized cell death referred to as HR. X, a host molecule guarded by NLRs; ID, integrated domain. and ROOT MERISTEM GROWTH FACTOR1 (Song et al., 2016) employ a similar mechanism to induce heterodimerization of their respective receptors with the BAK1/SERKs coreceptors.

Sequestering of BAK/SERKs by BIRs Negatively Regulates Plant Immunity
BIR1 was initially identified as a BAK1-interacting protein (Gao et al., 2009). Loss of BIR1 led to SOBIR1dependent autoimmunity and cell death. There are four BIR members (BIR1, BIR2, BIR3, and BIR4) in Arabidopsis, and all of them interacted with BAK1 when expressed in Nicotiana benthamiana (Halter et al., 2014). Recent structural and biochemical studies Hohmann et al., 2018) underlined that the ectodomains of BAK1 and BIR1-4 are sufficient for their interaction. BIR1 LRR -BAK1 LRR interaction is mediated by packing of one lateral side of BIR1 LRR against the C-terminal inner surface and the C-terminal capping domain of BAK1 LRR (Fig. 4A). The BAK1-interacting residues are highly conserved among BIR1-BIR4, suggesting a conserved mechanism of BIR-BAK1 interaction as further confirmed by the structure of BIR3 LRR -SERK1 LRR ( Fig. 4A; Hohmann et al., 2018). Importantly, structural comparison showed that the BIR1-contacting surface of BAK1 LRR or the BIR3-contacting surface of SERK1 LRR is also involved in interaction with other LRR-RKs such as FLS2 (Fig. 4B), suggesting that a BIR and these LRR-RKs may compete for interaction with BAK1/ SERKs. Indeed, the FLS2 LRR -flg22 complex efficiently outcompeted BIR1 LRR for BAK1 LRR binding . A similar observation was made for BRI1 LRRbrassinolide (BRASSINOSTEROID INSENSITIVE1 [BRI1]) with BIR2 LRR and BAK1 LRR (Hohmann et al., 2018). These data support the idea that a BIR can negatively regulate BAK1/SERK signaling by sequestering them from their paired RKs, as suggested by Halter et al. (2014). A similar mechanism is applied to negative regulation of BR (brassinosteroid) signaling by BIR3 (Imkampe et al., 2017).
Loss of BIR1 promotes BAK1-SOBIR1 interaction (Liu et al., 2016c), suggesting that BIR1 and SOBIR1 may interact with BAK1 in a competitive manner. This mechanism is consistent with the observation that overexpression of full-length or the ECD-TM of BAK1 in plants generated SOBIR1-dependent autoimmunity (Domínguez-Ferreras et al., 2015). On the other hand, overexpression of ECD-TM of BAK1 can interfere with immune signaling mediated by BAK1 and SOBIR1. Consistently, plants overexpressing the ECD-TM of BAK1 developed better than those overexpressing fulllength BAK1 (Domínguez-Ferreras et al., 2015). Although both BIR1 and SOBIR1 interact with BAK1, the BIR1-binding region of BAK1 is less likely to completely overlap with the SOBIR1-interacting domain of BAK1, as transgenic plants expressing a BAK1 mutant protein with compromised binding to BIR1 were constitutively active in inducing immune responses . It currently remains unknown what signals relieve BIR1inhibited SOBIR1 signaling when needed. Given the fact that cell death in bir-1 occurs even under sterile conditions (Gao et al., 2009), such signals, if present, appear to be endogenous.
Despite their conserved biochemical activities, BIR1-BIR4 have diversified functions. BIR1 is important to inhibit immunity mediated by BAK1 and SOBIR1 (Gao et al., 2009), whereas BIR2 and BIR3 have critical roles in negative regulation of PTI (Halter et al., 2014) and negative regulation of BR signaling (Imkampe et al., 2017;Hohmann et al., 2018), respectively. One possibility to reconcile the conserved biochemical activities of BIRs with their signaling specificity may be that they exist in distinct pools that are accessible to different RK-signaling complexes. It is of interest to note that a recent study using live-cell imaging showed that FLS2 and BRI1 localize to distinct plasma membrane (PM) nanodomains (Bücherl et al., 2017). But whether this is the case with BIRs remains undetermined.

GPI-Anchored Proteins as Co-receptors of RKs to Regulate Plant Immunity
FER belongs to the Catharanthus roseus RLK1-like subfamily with 17 members in Arabidopsis (Franck et al., 2018) and plays pleiotropic roles in plant growth, development, and immunity. The endogenous Cys-rich peptides RAPID ALKALINIZATION FAC-TOR (RALFs; Pearce et al., 2001;Escobar-Restrepo et al., 2007;Haruta et al., 2014;Ge et al., 2017;Stegmann et al., 2017) and the glycosyl-phosphatidyl-inositol (GPI)-anchored proteins (GAPs) LORELEI and its homologs LLG1, LLG2, and LLG3 (Capron et al., 2008;Li et al., 2015;Liu et al., 2016b;Shen et al., 2017) are essential for FER-mediated signaling. FER negatively regulates PTI via recognition of RALF23 (Stegmann et al., 2017). The recently solved crystal structure of the RALF23-LLG2-FER ECD complex revealed that RALF23 directly binds to LLG2 (Xiao et al., 2019;Fig. 5). A highly conserved N-terminal region is sufficient for RALF23 recognition by LLG1 and LLG2. Consistently, RALFs containing this region interact with LLG1, LLG2, and LLG3 and induce binding of the three LLG proteins to FER ECD in vitro. Recognition of diverse RALFs via LLGs is consistent with the multitasking FER. Biochemical and functional data showed that recognition of RALF23 by LLG1 results in recruitment of FER through formation of a composite LLG1-RALF23 interface. Structural comparison between apo-LLG1 and the RALF23-LLG2-FER ECD complex suggests that RALF23 binding induces no conformational change in LLG2 (Fig. 5). These and functional data established LLG1 as a coreceptor of FER to modulate plant immunity. Two more recent studies showed that LLG2 and LLG3 also function as coreceptors of the FER orthologs ANXUR/ BUPS to regulate pollen tube growth and development (Feng et al., 2019;Ge et al., 2019) in response to RALF4 and RALF9. The emerging data suggest that LLGs function as coreceptors of different members of the CrRLK1-like subfamily for regulation of diverse signaling pathways.
Around 250 GAPs are encoded in the genome of Arabidopsis (Zhou, 2019). The data discussed above suggest that other GAPs might also function as coreceptors to indirectly transmit signals from the PM by working in concert with RKs. OsCEBiP was initially thought to be an RLP, but was recently determined to be a GAP (Gong et al., 2017). Like RALF23 with LLG1 and FER, chitin binding induces OsCEBiP interaction with the RK OsCERK1 for defense signaling. The LysM-containing proteins LYM1 and LYM2 from Medicago truncatula were also shown to be GAPs . Unlike OsCEBiP and BAK1/ SERKs that act as coreceptors through homotypic interactions with other RKs, however, LLGs form ligand-induced complexes with the phylogenetically unrelated FER family members. Thus, the RALFinduced LLG-FER/ANXUR/BUPS complexes represent a novel type of ones for perception of plant peptides. The GAP GFRa in animals recognizes the glial-cell-line-derived neurotrophic factors and is consequently recruited to the receptor Tyr kinase RET (Paratcha and Ledda, 2008), forming complexes similar to those induced by RALFs.
The monomeric ZAR1-RKS1-PBL2 UMP in the absence of (d)ATP is reminiscent of the monomeric Apaf-1-cytochrome c complex . Like Apaf-1 assembly into the apoptosome , (d)ATP induces formation of an oligomeric ZAR1-RKS1-PBL2 UMP complex termed ZAR1 resistosome (Wang et al., 2019a). Cryo-EM analysis revealed a wheel-like pentamer of the ZAR1 resistosome, comparable to the structures of the Apaf-1 apoptosome  and the NLRC4 inflammasome Zhang et al., 2015). Formation of the ZAR1 resistosome is mediated by ZAR1 but not by RKS1 and PBL2 UMP . dATP binding induces structural reorganization between ZAR1 HD1 and ZAR1 WHD (Fig. 6B), as demonstrated in Apaf-1  and NLRC4 Zhang et al., 2015). Structural alignment revealed fold switching of ZAR1 CC after activation. Interestingly, the very N-terminal a-helix (a1) largely buried in the inactive ZAR1 becomes completely exposed after ZAR1 activation, forming a funnel-shaped structure in the ZAR1 resistosome (Fig. 6B). These results support stepwise activation of the ZAR1 resistosome, first primed by AvrAC and then fully activated by (d)ATP (Wang et al., 2019a). ZAR1 CC is sufficient to induce HR cell death when expressed in N. benthamiana (Baudin et al., 2017). The oligomerized ZAR1 CCs , however, are deeply buried in the ZAR1 resistosome except the funnel-shaped structure, suggesting that a1 is important for ZAR1 function. Indeed, N-terminal deletion mutants of ZAR1 lost AvrAC-induced HR cell death in protoplasts and resistance to X. campestris (Wang et al., 2019a). Remarkably, simultaneous mutation of Glu-11 and Glu-18 from the inner surface of the funnel-shaped structure substantially compromised the AvrAC-induced activities of ZAR1. Fractionation and mutagenesis assays showed that ZAR1 became PM-associated upon activation. These results suggest that the ZAR1 resistosome may directly function as a channel or a pore to mediate HR cell death and immune responses. Alternatively, it is also possible that recruitment to the membrane could bring ZAR1 into proximity with other yet-unidentified signaling proteins for further induction of cell death and resistance.

Autoinhibition and Ligand Sensing of NLRs
Although animal and plant NLRs are believed to have evolved independently , arrangement of NBD, HD1, and WHD is highly conserved in the inactive ZAR1, NRC1, NLRC4, and Apaf-1 (Fig. 7). Similar domain positioning is also found in the prototype NLR PH0952 from the hyperthermophilic euryarchaeota Pyrococcus horikoshii (Lisa et al., 2019). The C-terminal domains of ZAR1, NLRC4, Apaf-1, and PH0952 function to sequester these NLRs in a monomeric state, although they are differently positioned in the structures (Reubold et al., 2011;Hu et al., 2013;Lisa et al., 2019;Wang et al., 2019b). These structural observations suggest a conserved autoinhibition mechanism of NLRs.
The C-terminal LRR domain is widely hypothesized to act as the ligand sensor of an NLR. Indeed, some plant NLRs including RECOGNITION OF PER-ONOSPORA PARASITICA1 (Krasileva et al., 2010) Figure 6. Auto-inhibition, priming, and activation of ZAR1. A, Overall structure of the ZAR1-RKS1 complex (PDB: 6J5W) shown in cartoon (left). The a1 is labeled in red. The bound ADP molecule is shown in stick and its detailed interactions with ZAR1 are shown in the middle representation.Overall structure of the PBL2 UMP -RKS1-ZAR1 complex (PDB: 6J5V) shown in the same orientation as that in the left (right). The loop region that is disordered in the ZAR1-RKS1 complex is shown in salmon. B, Structure of the ZAR1 resistosome (PDB: 6J5T; left). Structure of an active ZAR1 protomer (middle). The a1 is labeled in red. The detailed interactions between dATP and ZAR1 NOD (right) from Arabidopsis and MILDEW-A (MLA; Lu et al., 2016) from barley (Hordeum vulgare) have been mapped to recognize their ligands through the variable C-terminal LRR region. While ZAR1 LRR does not directly contact PBL2 UMP , recognition of the AvrACmodified PBL2 is through the LRR-bound RKS1. Thus, ZAR1 LRR is the structural determinant for specific recognition of AvrAC. The integrated domains from several sensor NLRs are responsible for effector binding (Le Maqbool et al., 2015;Sarris et al., 2015;Ortiz et al., 2017). Additionally, the CC and TIR domains can also act as a sensor of effectors. For example, the TIR-only protein RE-SPONSE TO THE BACTERIAL TYPE III EFFECTOR PROTEIN HOPBA1 may act as a receptor of the effector protein HopB1 (Nishimura et al., 2017). Regardless of the recognition mechanisms, effector binding would function to trigger conformational changes in an NLR, promoting exchange of ADP with ATP/dATP to induce structural remodeling for full activation.

Activation and Oligomerization of NLRs
The conserved positioning of NBDs, HD1s, and WHDs in the inactive (Fig. 7) and active states of ZAR1, NLRC4, and Apaf-1 (Fig. 8) further solidifies the notion that structural remodeling generally accompanies NLR activation. The underlying mechanisms, however, can vary among different types of plant NLRs. Singleton NLRs could follow the mechanism demonstrated in ZAR1 (Wang et al., 2019a) and Apaf-1  for structural reorganization and activation. The NAIP-NLRC4 inflammasomes' NEURONAL APO-PTOSIS INHIBITOR PROTEIN (NAIP; Hu et al., 2015;Zhang et al., 2015) appears to be an attractive model for activation of paired NLRs. This model, however, needs formation of a substoichiometric complex between the sensor and the executor, which share a common promoter in an NLR pair. Furthermore, unlike the ligand-induced NAIP-NLRC4 complexes, constitutive heteromeric complexes have been shown for several paired NLRs ( Sarris et al., 2015). A more recent study showed that knockout of sensor NLRs from several NLR pairs in rice produced HR-like phenotypes (Wang et al., 2019c), supporting an inhibitory role of the sensors in activating the paired NLRs and agreeing with the model on activation of the paired NLRs RRS1/RPS4 (Le Sarris et al., 2015) and RGA5/RGA4 (Césari et al., 2014b).
Less is known about how helper NLRs are activated. Signaling mediated by TIR-NLRs requires ENHANCED DISEASE SUSCEPTIBILITY1 (EDS1)/ PHYTOALEXIN DEFICIENT4 or EDS1/SENESCENCE-ASSOCIATED GENE101 (Wiermer et al., 2005) and the helper NLRs NRG1s and/or ADR1/ADR1-Ls (Peart et al., 2005;Roberts et al., 2013;Qi et al., 2018;Castel et al., 2019;Wu et al., 2019). Two recent studies (Horsefield et al., 2019;Wan et al., 2019a) showed that TIR-NLRs possess NADase activity. Thus, one plausible model on activation of helper NLRs might be that they sense a signaling molecule(s) generated by TIR-NLRs probably through EDS1. Identification of the putative signaling molecule(s) would be a key to understanding how NRG1s and/or ADR1/ADR1-Ls are activated. However, ADR1/ADR1-Ls are also required for some CC-NLRs (Bonardi et al., 2011;Roberts et al., 2013), raising the question of whether these helper NLRs can sense different signals. In contrast with paired NLRs, NRCs and NRC-dependent NLRs were proposed to act through positive regulation (Wu et al., 2017). In line with this model, overexpression of Rx NBD of the NRC-dependent RESISTANCE TO POTATO VIRUS X in N. benthamiana induces HR-like cell death (Rairdan et al., 2008). This appears to suggest that activation of NRCs is comparable to that of NLRC4 in the NAIP-NLRC4 inflammasomes. However, whether the observed phenotype is dependent on an NRC was not tested. Furthermore, there has been no evidence showing effector-induced interaction between a sensor NLR and an NRC.
The active ZAR1, NLRC4, and Apaf-1 form wheellike structures despite their different oligomerization status ( Fig. 8; Hu et al., 2015;Zhang et al., 2015;Zhou et al., 2015;Wang et al., 2019a). Oligomerization functions to bring their N-terminal signaling domains into proximity, forming a funnel-like structure in the ZAR1 resistosome and caspase-recruiting platforms in the NLRC4 inflammasome and the Apaf-1 apoptosome. Oligomerization can be generally important for plant NLR function. Self-association of TIR domains required for effector-triggered immune response (Bernoux et al., 2011;Williams et al., 2014;Zhang et al., 2017a) supports such a model. However, it remains unknown whether TIR-NLRs form a structure similar to that of the ZAR1 resistosome. Notably, unclosed structures have been observed for some NLRs. For example, deletion of the N-terminal CARD domain of NLRC4 resulted in formation of helical NAIP5-NLRC4 structures (Diebolder et al., 2015;Matyszewski et al., 2018). Another example is the NLR-like protein MalT (the central activator of the mal genes) from Escherichia coli, which forms maltotriose-induced curved oligomers (Larquet et al., 2004).

ATP/dATP-Binding and Hydrolysis of NLRs
Structural studies support the notion that the ADPand (d)ATP-bound forms of an NLR correspond to off-and on-states, respectively. As in the Apaf-1 apoptosome , the g-phosphate group of dATP functions to stabilize the active conformation of ZAR1 for oligomerization (Wang et al., 2019a). Supporting the important role of the g-phosphate group, other triphosphate nucleosides such as GTP or the nonhydrolyzable AMP-PNP also support assembly of the Apaf-1 apoptosome (Reubold et al., 2009). However, activation of an NLR may not necessarily require (d)ATP binding. For example, flagellin binding alone stabilizes the active conformation of NAIP5 (Tenthorey et al., 2017;Yang et al., 2018), explaining the dispensability of the ATP-binding of this NLR (Halff et al., 2012). Similarly, the P-loop of NRG1.1, NRG1.2, and ADR1-L2 from Arabidopsis is dispensable for their helper function (Bonardi et al., 2011;Wu et al., 2019) downstream of the effector-activated sensor NLRs. Interestingly, however, the auto-activity of overexpressed N. benthamiana NRG1 (Peart et al., 2005) and the autoactive mutant ADR1-L2 (D484V) is P-loop-dependent (Roberts et al., 2013). The reason for this remains unclear, but it appears that different NLRs employ distinct mechanisms to stabilize their active conformation for oligomerization, although ATP/dATP binding is likely the most commonly used one.
Many NLRs are predicted to have the catalytic elements of an ATPase. Indeed, NLR proteins including M and L6 Bernoux et al., 2016) from plants, and NLRC4  from animals, exhibit ATP-hydrolyzing activity. ATP hydrolysis may function to switch the (d)ATP-bound active state back to the ADP-bound inactive state. However, whether the proteins tested for ATP hydrolysis were in active states was not reported. Notably, the catalytic pocket of NLRs is formed by an individual monomer and is not, as in the case for the canonical ATPASES ASSOCIATED WITH DIVERSE CELLULAR ACTIVITIES, a composite pocket formed by two neighboring monomers in the oligomer (Erzberger and Berger, 2006). One study appeared to argue against the model above by showing that only inactive Apaf-1 displayed low ATPase activity but not Apaf-1 from the apoptosome (Reubold et al., 2009). This agrees with the idea that activation of Apaf-1 apoptosome represents the point-of-no-return of programmed cell death pathways (Riedl and Salvesen, 2007).

Altered Subcellular Localization of ZAR1 upon Activation
In parallel to MIXED LINEAGE KINASE DOMAIN-LIKE PROTEIN (MLKL) oligomerization and translocation to the PM after activation (Cai et al., 2014;Chen et al., 2014;Wang et al., 2014), ZAR1 activation induced by AvrAC results in relocalization of the NLR from the cytosol to the PM to mediate cell death. Strong evidence for the altered localization of ZAR1 comes from the E11A/E18A mutation, which did not affect assembly of the ZAR1 resistosome but nearly abolished AvrAC-induced cell death in protoplasts. Because of the loss of cell death activity, the PMassociation of the ZAR1 mutant was easily detected (Wang et al., 2019a). Identification of similar mutations in other CC-NLRs is possible, because a more recent study (Adachi et al., 2019) showed that the very N-terminal fragments of many singleton and helper CC-NLRs are also functionally important when tested in tobacco. Such mutations would be valuable in investigating cellular localization of NLRs, particularly because NLRs have been shown to function in different compartments including nucleus, endoplasmic reticulum, and Golgi apparatus (Cui et al., 2015). Additionally, because these mutations can arrest an activated form of NLRs, they might also be used to identify components regulating NLR complexes. Mutations of the conserved catalytically and functionally important glutamic residue in TIR-NLRs can serve similar purposes.

Pore-Forming Activity of ZAR1 CC
Structural and biochemical data suggest that the funnel-shaped structure in the ZAR1 resistosome may function as a channel or pore in the PM. As noted in Burdett et al. (2019), the funnel-shaped structure bears striking similarity to the pore-forming protein MITO-CHONDRIAL CALCIUM UNIPORTER from Caenorhabditis elegans (Oxenoid et al., 2016) and the calcium channel Orai from fruit fly (Drosophila melanogaster; Hou et al., 2012). Although many more investigations are needed to test this model, the pore-forming activity of a CC domain was demonstrated in other proteins. For example, the HeLo domain of fungal Het-S (a prion protein encoded by het-s locus of the nine het-loci), which is a four-helix bundle like a canonical CC domain, forms pores in the PM after activation to mediate cell death (Seuring et al., 2012). Induced pore formation by the N-terminal CC domain of MLKL in animals was also demonstrated . The very N-terminal a1 helix forming the funnel-shaped structure in the ZAR1 resistosome is conserved in many distantly related CC-NLRs (Adachi et al., 2019). Assays performed in N. benthamiana showed that the N-terminal fragments are functionally exchangeable among several CC-NLRs. Notably, when fused with Yellow Fluorescent Protein at the C terminus, the N-terminal 29 amino acids of NRC4 were sufficient to induce cell death. But whether the N-terminal fragment of NRC4 associates with PM and the NLR forms a ZAR1 resistosome-like structure remains unknown.
Formation of the funnel-like structure is remarkably similar to that of the hemolytic actinoporin fragaceatoxin (FraC; Tanaka et al., 2015), although ZAR1 CC and FraC share little structural similarity. Interestingly, fold switching occurs to both ZAR1 CC and FraC during assembly of the funnel-like structures. This is also true with the pore-forming protein Het-S (Daskalov et al., 2015). Fold plasticity of the CC domain appears to also exist in other CC-NLRs. The CC domains of the barley NLR Sr33 and wheat NLR MLA10 display different fold topologies when their structures were determined by nuclear magnetic resonance (Casey et al., 2016) and crystallography (Maekawa et al., 2011a;Casey et al., 2016), despite their highly conserved sequences.

FUTURE PERSPECTIVES
Despite the progress in structural studies of PRRs and NLRs, many open questions remain concerning these two families of proteins (see Outstanding Questions). Obtaining structures of full-length signalingcompetent PRR complexes is one challenge for full understanding of how PAMPs/DAMPs activate them. Clustering of receptor Tyr kinases is important for their activation in animals (Kotani et al., 2008) and is now beginning to be appreciated as an important facet of RK activation in plants (Somssich et al., 2015;Bücherl et al., 2017). Structural and biochemical investigations will allow us to understand how this mechanism operates in PRR activation. Although several structures of LRRand LysM-type PRRs have been solved, structural mechanisms of ligand recognition and activation of several types of PRRs remain to be elucidated. Similarly, how ligand recognition by LRR-RLPs, including those as resistance proteins such as the Cladosporium fulvum proteins (Postma et al., 2016;Wan et al., 2019b) activate their coreceptors BAK1/SERKs and/or SOBIR1, is still poorly understood. The fact that BAK1/SERKs function as coreceptors of many LRR-RKs, including PRRs, raises the question of how the loose specificity is achieved. It should be noted that the coreceptor RKs typically have diverse functions and can mediate different signaling than the ligand-binding ones. Assignment of the nonligand binding functions of these RKs would be a direction in the future studies.
Currently it remains unknown whether the ZAR1 resistosome functions as an executor or a trigger of immune responses. Many investigations will be required to test the model on the ZAR1 resistosome as a channel or a pore. Although oligomerization can be ingrained into the model of NLR activation, direct evidence for this from TIR-NLRs is still lacking. Whether other plant NLRs can form resistosome-like structures is another open question. Structural information of an active TIR-NLR is of particular interest, because it will not just help address this question but also may explain whether and why oligomerization is required for its potential NADase activity. Reconstitution of active complexes containing the helper NLRs NRG1s and ADR1/ADR1-Ls may critically depend on the molecule(s) produced by TIR-NLR as NADases or probably even other enzymes. Thus, identification of such a molecule(s) represents one major challenge to dissect the activation mechanisms of helper NLRs. In addition to effector sensing and negative regulation of immune response, RRS1 also contributes to RPS4-mediated signaling (Narusaka et al., 2009;Ma et al., 2018). How effector binding relieves the negative regulation by RRS1, and how RRS1 contributes to the activation of RPS4, remains elusive. Addressing these questions would provide a model on how other paired NLRs are activated. Emerging evidence suggested that NRCs may follow a similar mechanism to ZAR1 for signaling (Adachi et al., 2019). But how NRCs are activated remains enigmatic. The ZAR1 resistosome is just the tip of the NLR iceberg. With the evergrowing advance in cryo-EM, structural biology will reveal many more exciting mechanisms of NLR action.