Research Article

Ligand-triggered allosteric ADP release primes a plant NLR complex

See allHide authors and affiliations

Science  05 Apr 2019:
Vol. 364, Issue 6435, eaav5868
DOI: 10.1126/science.aav5868

The plant resistosome comes into focus

Nucleotide-binding, leucine-rich repeat receptors (NLRs) initiate immune responses when they sense a pathogen-associated effector. In animals, oligomerization of NLRs upon binding their effectors is key to downstream activity, but plant systems differ in many ways and their activation mechanisms have been less clear. In two papers, Wang et al. studied the composition and structure of an NLR called ZAR1 in the small mustard plant Arabidopsis (see the Perspective by Dangl and Jones). They determined cryo–electron microscopy structures that illustrate differences between inactive and intermediate states. The active, intermediate state of ZAR1 forms a wheel-like pentamer, called the resistosome. In this activated complex, a set of helices come together to form a funnel-shaped structure required for immune responsiveness and association with the plasma membrane.

Science, this issue p. eaav5868, p. eaav5870; see also p. 31

Structured Abstract

INTRODUCTION

Nucleotide-binding (NB), leucine-rich repeat (LRR) receptor (NLR) proteins constitute a family of intracellular immune receptors in both animals and plants that detect the presence of pathogen molecules or host-derived signals. NLRs share a conserved tripartite domain structure with a conserved central NB and oligomerization domain (NOD), a C-terminal LRR domain, and a variable N-terminal domain. The NOD module can be further divided into an NB domain (NBD), a helical domain (HD1), and a winged-helix domain (WHD). In plants, direct or indirect recognition of pathogen effectors by NLRs induces numerous defenses, including programmed cell death called hypersensitive response, and restricts pathogens to the infection site. For instance, the coiled-coil (CC)–NLR HOPZ-ACTIVATED RESISTANCE 1 (ZAR1) of the small mustard plant Arabidopsis thaliana forms a preactivation complex with resistance-related kinase 1 (RKS1, a pseudokinase belonging to receptor-like cytoplasmic kinase subfamily XII-2) and recognizes the uridylyltransferase effector AvrAC from the pathogen Xanthomonas campestris pv. campestris that is responsible for the black rot disease of crucifiers. AvrAC uridylates a number of host protein kinases, including the PBS1-like protein 2 (PBL2) kinase. PBL2UMP, the version of the Arabidopsis protein uridylated by AvrAC, then acts as a ligand of the preformed ZAR1-RKS1 complex. NLRs are believed to function as a nucleotide [adenosine diphosphate (ADP) or adenosine triphosphate (ATP)]–operated molecular switch, with ADP- and ATP-bound forms corresponding to the “off” and “on” states, respectively, but the mechanism of how ADP is released from an NLR for exchange with ATP remains elusive. Structural elucidation of a full-length plant NLR protein and its recognition of modified self is lacking.

RATONALE

We reconstituted a ZAR1-RKS1 and a ZAR1-RKS1-PBL2UMP complex and determined their cryo–electron microscopy (cryo-EM) structures at resolutions of 3.7 and 4.3 Å, respectively. The structures were verified by biochemical, cell-based, and functional data. We determined how PBL2UMP affects the ADP-binding activity of the ZAR1-RKS1 complex by radiolabeled assays. Structural comparison of the ZAR1-RKS1 and ZAR1-RKS1-PBL2UMP complexes was used to probe the mechanism of PBL2UMP-induced ADP release from ZAR1, which was further validated by biochemical assays.

RESULTS

The cryo-EM structure of the ZAR1-RKS1 complex revealed that intramolecular interactions within ZAR1 maintain the NLR protein in an inactive state. The inactive state is further stabilized by an ADP. The LRR domain of ZAR1 (ZAR1LRR) is positioned differently from LRR domains of animal NLRs but functions similarly to sequester ZAR1 in a monomeric state. ZAR1CC appears to be kept in an inactive state via contacts with ZAR1LRR, ZAR1HD1, and ZAR1WHD. This contrasts with the flexible N-terminal domain of inactive apoptotic protease-activating factor 1 (Apaf-1). ZAR1LRR mediates interaction with RKS1 in the preformed ZAR1-RKS1 complex. The ZAR1-RKS1-PBL2UMP structure shows that RKS1 is exclusively responsible for the binding of PBL2UMP. The two uridylyl moieties of PBL2UMP interact with and consequently stabilize the activation segment of RKS1. Comparison of the two cryo-EM structures shows that the stabilized activation segment of RKS1 sterically clashes with the ADP-bound ZAR1NBD from the ZAR1-RKS1 complex, resulting in conformational changes in the NBD but not other domains of ZAR1: ZAR1NBD is rotated outward about 60° compared with that from the inactive ZAR1. Thus, PBL2UMP allosterically induces release of ADP from the ZAR1-RKS1-PBL2UMP complex. Indeed, radiolabeling assays showed that PBL2UMP, but not PBL2, reduced the ADP-binding activity of the ZAR1-RKS1 complex.

CONCLUSION

Our study revealed the mechanisms of PBL2UMP recognition by ZAR1-RKS1 and PBL2UMP-induced priming of ZAR1, providing a structural template for understanding NLR proteins.

PBL2UMP-induced ADP release from ZAR1.

ZAR1 is maintained in an inactive state through contacts of multiple domains and an ADP molecule (in stick representation). ZAR1LRR mediates ZAR1 interaction with RKS1. The AvrAC-uridylated PBL2 (PBL2UMP, blue) as a ligand is exclusively recognized by the ZAR1-bound RKS1. The activation segment of RKS1, which is flexible in the inactive ZAR1-RKS1 complex (red mesh), becomes stabilized (red surface) after interaction with the two uridylyl moieties (in sphere representation) of PBL2UMP and clashes with ZAR1NBD. The steric interference then causes ZAR1NBD to rotate outward and, consequently, ADP release. The ZAR1-RKS1-PBL2UMP complex thus represents an intermediate state.

Abstract

Pathogen recognition by nucleotide-binding (NB), leucine-rich repeat (LRR) receptors (NLRs) plays roles in plant immunity. The Xanthomonas campestris pv. campestris effector AvrAC uridylylates the Arabidopsis PBL2 kinase, and the latter (PBL2UMP) acts as a ligand to activate the NLR ZAR1 precomplexed with the RKS1 pseudokinase. Here we report the cryo–electron microscopy structures of ZAR1-RKS1 and ZAR1-RKS1-PBL2UMP in an inactive and intermediate state, respectively. The ZAR1LRR domain, compared with animal NLRLRR domains, is differently positioned to sequester ZAR1 in an inactive state. Recognition of PBL2UMP is exclusively through RKS1, which interacts with ZAR1LRR. PBL2UMP binding stabilizes the RKS1 activation segment, which sterically blocks ZAR1 adenosine diphosphate (ADP) binding. This engenders a more flexible NB domain without conformational changes in the other ZAR1 domains. Our study provides a structural template for understanding plant NLRs.

Perception of microbial pathogens by immune receptors activates plant defense responses. Whereas cell surface–localized immune receptors perceive extracellular molecular patterns associated with pathogenesis (1, 2), cytoplasmic immune receptors directly or indirectly perceive pathogen effectors that are secreted into the cytosol of plant cells (3, 4). The latter class of immune receptors are primarily nucleotide-binding (NB), leucine-rich repeat (LRR) receptors (NLRs), which constitute the majority of plant disease resistance (R) proteins. NLRs are shared by both plants and animals and are characterized by a conserved central NB and oligomerization domain (NOD), a C-terminal LRR domain, and a variable N-terminal domain (5). NOD is related to AAA+ adenosine triphosphatases (ATPases), which are defined by the structurally conserved adenosine diphosphate or adenosine triphosphate (ADP or ATP) binding motif and belong to the signal-transducing ATPase with numerous domains (STAND) subfamily, including the pro-apoptotic proteins Apaf-1 and CED-4 (6). NOD in plant NLRs is thus referred to as NB-ARC for the shared sequence in Apaf-1, R, and CED-4 proteins. Depending on their N termini, plant NLR proteins are classified into two major categories: coiled coil (CC)–NLRs and Toll interleukin-1 receptor (TIR)–NLRs (5). Activation of plant NLRs typically leads to an array of immune responses, including hypersensitive response (HR), a form of programmed cell death believed to limit pathogens to the infection site (7).

During the past two decades, efforts have been made toward understanding the mechanisms of action of plant NLRs (8). However, the proposed models of plant NLR action concerning autoinhibition, ligand recognition, and activation are largely inferred from structures of Apaf-1 and animal NLRs. On the basis of NOD in Apaf-1 (911) and animal NLRs (1215), NB-ARC of plant NLRs is believed to function as a molecular switch with ADP- and ATP-bound forms dictating the “off” and “on” states of NLR signaling (1618), respectively, but the underlying mechanism remains enigmatic. Intramolecular interactions such as the interaction between the CC domain and NB-ARC-LRR act to keep the CC-NLR protein Rx in an inactive state (19). Evidence from several studies supports a role of the C-terminal LRR domain in the perception of effectors (2025). Direct or indirect recognition of an effector is thought to trigger conformational changes in the LRR domain, relieving its inhibition, enabling exchange of ADP for ATP, and consequently activating the NLR protein (3, 4, 8).

The Arabidopsis CC-NLR HOPZ-ACTIVATED RESISTANCE 1 (ZAR1) indirectly recognizes three unrelated bacterial effector proteins, all through an association with closely related pseudokinases ZED1 (26), resistance-related kinase 1 (RKS1) (27), and ZRK3 (28) that belong to receptor-like cytoplasmic kinase subfamily XII-2 (RLCK XII-2). Thus, the ZAR1-ZED1, ZAR1-RKS1, and ZAR1-ZRK3 complexes perceive the Pseudomonas syringae effector protein HopZ1a (26), an acetyl transferase; the Xanthomonas campestris pv. campestris effector AvrAC (27), a uridylyl transferase; and the P. syringae effector HopF2 (28), a ribosyl transferase, respectively. ZAR1 is an ancient NLR that is also present in Nicotiana benthamiana (29). A recent study shows that N. benthamiana ZAR1 (NbZAR1) associates with another RLCK XII member, JIM2, to recognize the X. campestris perforans effector XopJ4, an acetyl transferase distantly related to HopZ1a (30). AvrAC uridylylates receptor-like cytoplasmic kinases belonging to family VII (RLCK-VII), including PBS1-like protein 2 (PBL2) and BIK1. BIK1 is a true virulence target of AvrAC, which is a key component of immune signaling pathways governed by transmembrane receptor kinases, and the uridylylation by AvrAC inhibits its kinase activity and dampens host defenses that are otherwise activated by BIK1. By contrast, PBL2 is a decoy which, upon uridylylation by AvrAC (referred to as PBL2UMP), is recruited to the ZAR1-RKS1 complex through a direct interaction with RKS1 to trigger ZAR1 activation and disease resistance. Thus PBL2UMP, a modified self, is the ligand triggering ZAR1 activation (27). Therefore, ZAR1 represents a model not only for studying indirect recognition of effectors by NLRs but also for understanding how an NLR expands its recognition specificity by association with multiple “adapter” proteins.

To understand the mechanisms of plant NLR autoinhibition and activation, we sought to solve the cryo–electron microscopy (cryo-EM) structures of the inactive ZAR1-RKS1 and intermediate ZAR1-RKS1-PBL2UMP complexes. Supported by biochemical and functional data, the structures revealed extensive intramolecular interactions within ZAR1 that are further stabilized by ADP to maintain ZAR1 in an inactive state. The LRR domain plays a key role in the autoinhibition of ZAR1 and mediates the interaction with RKS1. Recognition of PBL2UMP is exclusively mediated by RKS1 from the preformed ZAR1-RKS1 complex. Structural comparison revealed that PBL2UMP binding allosterically facilitates ADP release from inactive ZAR1 by inducing conformational changes in the NB domain (NBD), which likely results in a nucleotide-depleted ZAR1. ZAR1 from the ZAR1-RKS1-PBL2UMP complex, however, is monomeric and assumes a conformation similar to that of the inactive ZAR1 except NBD, suggesting that the monomeric ZAR1-RKS1-PBL2UMP complex is in an intermediate state. Taken together, our data unveil the mechanisms of autoinhibition and AvrAC-induced nucleotide exchange of ZAR1 and suggest the existence of a second signal required for activation of the NLR protein.

Results

Autoinhibition of ZAR1

To probe the autoinhibition mechanism of ZAR1, His-SUMO–tagged RKS1 was coexpressed with the full-length ZAR1 in insect cells. Pull-down and gel filtration assays showed that these two proteins strongly interacted with each other when coexpressed (fig. S1, A and B). The protein complex thus purified was then used for structural analysis with cryo-EM. After three-dimensional (3D) classification, a subset of 148,718 particles was used for image reconstruction, generating a map with a global resolution of 3.7 Å (Fig. 1A and fig. S1, C to F) as determined with a gold-standard Fourier shell correlation.

Fig. 1 Structure of inactive ZAR1.

(A) Two orientations of the final EM density map of the ZAR1-RKS1 complex color coded to show the local resolution in angstroms. (B) The autoinhibition mechanisms of ZAR1, NLRC4, and Apaf-1. Shown on the top left is a cartoon representation of ZAR1 from ZAR1-RKS1. Subdomains of ZAR1 are shown in different colors, and their boundaries are indicated by residue numbers in parentheses. ADP is shown in stick representation. Shown on the top right is the structure of the inactive NLRC4 [Protein Data Bank (PDB) 4KXF] with its NBD and HD1 aligned with those of ZAR1 (top left). HD2, helical domain 2; C, C terminus; N, N terminus. Shown at the bottom left is a lateral dimer of Apaf-1 from the Apaf-1 apoptosome (PDB 3JBT). HD1 and NBD from the left protomer of the lateral dimer were aligned with those from ZAR1 (top left). (C) Binding of ADP by ZAR1. Residues of ZAR1 involved in the interaction with ADP are shown in stick representation. Cryo-EM density around the ADP binding site is shown in green mesh. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

The RKS1-bound ZAR1 contains an NB-ARC module, including a canonical NBD (ZAR1NBD, residues 145 to 317), a helical domain 1 (ZAR1HD1, residues 318 to 394), and a winged-helix domain (ZAR1WHD, residues 395 to 514) (Fig. 1B, left, and table S1). The three structural domains of ZAR1 are similarly positioned to those of the inactive NLRC4 (Fig. 1B, right) and Apaf-1 (fig. S2A), indicating that the structure of ZAR1 represents an inactive state. This conclusion is consistent with the monomeric ZAR1-RKS1 complex protein indicated by gel filtration (fig. S1B). Compared with NOD in the inactive NLRC4, however, ZAR1NB-ARC varies in packing against the other domains. The LRR domain sequesters NLRC4 in a monomeric state through structural coupling with the opposite side of the NBD where the WHD packs (Fig. 1B, right). Similar positioning of the C-terminal WD40 repeats (W, Trp; D, Asp) is also found in the inactive Apaf-1 (fig. S2A). In stark contrast, ZAR1WHD interacts extensively with one lateral side of the LRR domain (ZAR1LRR; Fig. 1B, left), presenting both ZAR1WHD and ZAR1LRR on the same side of ZAR1NBD. The oligomerization of NLRC4 (13, 14), Apaf-1 (10), and other AAA+ ATPases (31, 32) is mediated by stacking of one side of the NBD of one protomer against the opposite side of the NBD of the other protomer in a lateral dimer. Structural superposition of the inactive ZAR1 (Fig. 1B, left) with a protomer from one lateral Apaf-1 (Fig. 1B, bottom), NLRC4, or CED-4 (fig. S2B) dimer suggested that such a dorsal-ventral stacking of NBDs is completely blocked by ZAR1LRR in the inactive state. Thus, ZAR1LRR can also play a role in sequestering ZAR1 in a monomeric state, although differently positioned compared with NLRC4LRR.

In further support of the inactive conformation of the RKS1-bound ZAR1, an ADP molecule is well defined by the 3D reconstruction of the ZAR1-RKS1 complex (Fig. 1C). Like that in the inactive NLRC4 (12) and Apaf-1 (9, 11), the ADP in the inactive ZAR1 also binds the joint interface formed by ZAR1HD1, ZAR1NBD, and ZAR1WHD via multiple polar interactions and van der Waals contacts (Fig. 1C). An inhibitory role of the WHD-ADP interaction formed between His443 and the β-phosphate group of ADP was demonstrated in NLRC4 activation in cell-based assays and autoinflammatory diseases (33, 34). A similar WHD-ADP interaction in the inactive ZAR1 is also established through a hydrogen bond between His488 and the β-phosphate group of the ZAR1-bound ADP (Fig. 1C). ZAR1His488 corresponds to “H” of the “MHD” motif that is highly conserved among plant NLRs, and mutations of this residue result in constitutive activation of plant NLRs in several instances (17, 35, 36). These data suggest that the WHD-ADP interaction can also have an inhibitory role in ZAR1 regulation.

ZAR1CC forms a four-helix bundle that is similar to the structures of the CC domains of the CC-NLRs Rx (37) and Sr33 (38) (fig. S3). In addition to ZAR1LRR, ZAR1CC also packs against ZAR1HD1 and ZAR1WHD (Fig. 1B, left). This is consistent with a previous study showing that RPM1CC interacts with multiple domains of RPM1 (39). The intramolecular interactions within ZAR1 may in turn keep ZAR1CC in an inactive state. This conclusion is consistent with the observation that overexpression of ZAR1CC, but not ZAR1CC-NBD, induces HR in plants (29). The simultaneous interactions of ZAR1WHD with the other domains result in a WHD-organized overall structure of the inactive ZAR1. ZAR1WHD is therefore much more buried as compared with that in the inactive NLRC4 or Apaf-1 (Fig. 1B and fig. S2A). ZAR1WHD is further buried by the N-terminal loop region of ZAR1NBD, which interacts with the interface between ZAR1WHD and ZAR1HD on one side (Fig. 1B, left). These intramolecular interactions collectively act to further stabilize the LRR-sequestered inactive conformation of ZAR1.

Interaction of RKS1 with ZAR1

The ZAR1-RKS1 interaction is mainly mediated by contacts of RKS1 with one lateral side of ZAR1LRR (Fig. 2A). The long N-terminal α helix of RKS1 tightly packs against ZAR1LRR mainly through hydrophobic contacts (Fig. 2B). In the middle of the ZAR1-RKS1 interface is the loop region C-terminal to the long α helix of RKS1, which makes polar and hydrophobic interactions with ZAR1 (Fig. 2C). At the distal side of the ZAR1-RKS1 interface, two α helices from the C-lobe of RKS1 pack against the very C-terminal side of ZAR1 via hydrophobic and van der Waals contacts (Fig. 2D). Sequence alignment indicates that the ZAR1-interacting residues of RKS1 are highly conserved among pseudokinases ZED1, ZRK3, and RKS1 and other members of the XII-2 subfamily (fig. S4). This result provides an explanation for the observations that ZAR1 formed complexes with these three pseudokinases as well as other members of this RLCK subfamily (27).

Fig. 2 The C-terminal LRR domain mediates ZAR1 interaction with RKS1.

(A) Cartoon showing the overall structure of the ZAR1-RKS1 complex. The interacting regions between the two proteins are highlighted with open frames. (B) Detailed interactions of the N-terminal helix of RKS1 with ZAR1LRR for the red-framed region in (A). Cryo-EM density is shown in green mesh. (C) Detailed interactions of a loop region of RKS1 with ZAR1LRR for the blue-framed region in (A). (D) Detailed interactions of RKS1 with the last LRR of ZAR1 for the green-framed region in (A). (E) RKS1 (left) or ZAR1 (right) mutations reduce ZAR1-RKS1 interaction in vitro. N-terminal 6×His-SUMO tagged RKS1 was coexpressed with full-length ZAR1 (left) or glutathione S-transferase (GST)–tagged ZAR1 was coexpressed with full-length RKS1 (right) in Sf21 insect cells. The proteins were purified using Ni-NTA or glutathione sepharose 4B (GS4B) beads, and the proteins were visualized by SDS–polyacrylamide gel electrophoresis (PAGE) with Coomassie brilliant blue staining. (F) RKS1 (left) or ZAR1 (right) mutations diminish ZAR1-RKS1 interaction in protoplasts. Co-IP was performed using agarose-conjugated anti-FLAG (α-FLAG) antibodies, and the resulting protein was subjected to Western blot analysis. The experiments were repeated three times with similar results. HA, hemagglutinin tag; –, transfected ZAR1 construct alone as control. (G) RKS1 (left) or ZAR1 (right) mutations impair AvrAC-induced cell death in protoplasts. Protoplasts of rks1 (left) or zar1 (right) background were transfected with the indicated constructs, and a protoplast viability assay was performed. Data are represented as mean ± SEM (n = 6). Different letters indicate significant difference (P < 0.05, Tukey post hoc test). The experiments were performed three times with similar results. –, mock control. (H) RKS1 (left) or ZAR1 (right) mutations abolish avrAC-specified disease resistance in Arabidopsis plants. Plants of rks1 (left) or zar1 (right) background complemented with the indicated constructs were inoculated with WT strain (Xcc8004) or a strain lacking avrACavrAC). Disease symptoms were scored 7 days after inoculation. Numbers indicate the ratio of leaves developing chlorosis to the total number of inoculated leaves. The experiments were repeated twice with similar results. –, rks1 and zar1 control plants.

Supporting our structural observations, deletion of the last LRR (29 amino acids) that interacts with the C-lobe of RKS1 (Fig. 2D) abrogated ZAR1 interaction with ZED1 (27, 29). Although not directly involved in interaction with RKS1, the ZAR1P816L and ZAR1S831F mutations (P816L, Pro816→Leu; S831F, Ser831→Phe) can perturb the conformation of the last LRR repeat, resulting in loss of ZAR1 interaction with RKS1 (27). The ZAR1G645E mutation (Gly645→Glu) predicted to disrupt the middle ZAR1-RKS1 interface (Fig. 2C) is known to abolish ZAR1-ZED1 interaction, HopZ1a-induced HR, and disease resistance to P. syringae (29). To further verify our structural observations, we first introduced amino acid substitutions in RKS1 and ZAR1 and evaluated their impact on ZAR1-RKS1 interaction using coexpression and pull-down assays. As predicted by the structure (Fig. 2, B to D), the RKS1 G27A (Gly27→Ala), L31E (Leu31→Glu), and V35E (Val35→Glu) mutations greatly impaired ZAR1-RKS1 interaction in pull-down and coimmunoprecipitation (co-IP) assays (Fig. 2, E and F, and fig. S5). Similarly, the ZAR1 V544E (Val544→Glu), H597E (His597→Glu), and W825A/F839A (Trp825→Ala/Phe839→Ala) mutations severely diminished or abolished ZAR1-RKS1 interaction in both pull-down and co-IP assays (Fig. 2, E and F, and fig. S5). The bulkier and negatively charged glutamic acid residue is more effective than an alanine residue in damaging the middle ZAR1-RKS1 interface (Fig. 2C), explaining the more pronounced effect generated by ZAR1G645E on ZAR1-ZED1 interaction (29).

We then investigated the effect of RKS1 or ZAR1 mutations on AvrAC-induced cell death in Arabidopsis protoplasts. As expected, coexpression of AvrAC with PBL2, RKS1, and ZAR1 in protoplasts led to cell death (Fig. 2G). By contrast, when the RKS1 or ZAR1 mutants with reduced ZAR1-RKS1 interaction were used for the assays, the cell death activity was substantially reduced, albeit to varied degrees. To further test the observed interactions, we introduced the RKS1 and ZAR1 variants as transgenes under control of their native promoters into rks1 and zar1 mutant plants, respectively. Transgenic plants of the T1 generation were inoculated with the wild-type (WT) X. campestris pv. campestris (Xcc8004), which carries functional avrAC and a strain lacking avrACavrAC). As expected, the WT Col-0 plants, rks1 plants complemented with WT RKS1, and zar1 plants complemented with WT ZAR1 all displayed complete resistance to Xcc8004 and showed no chlorotic disease symptoms (Fig. 2H and fig. S6). This resistance is dependent on the recognition of AvrAC, because the ΔavrAC mutant strain caused disease on all genotypes. By contrast, the RKS1G27A and RKS1V35E variants failed to complement the rks1 mutation, and the ZAR1V544E, ZAR1H597E, and ZAR1W825A/F839A variants failed to complement the zar1 mutation. These transgenic plants were completely susceptible to Xcc8004 and indistinguishable from nontransgenic (–) rks1 and zar1 plants (Fig. 2H and fig. S6). Taken together, our biochemical and functional data support the ZAR1-RKS1 interaction observed in the structure.

Structure of the ZAR1-RKS1-PBL2UMP complex

To understand how PBL2UMP binding to RKS1 activates ZAR1, we first reconstituted an RKS1-ZAR1-PBL2UMP complex using ZAR1-RKS1 purified from insect cells and PBL2 purified from Escherichia coli. PBL2 coexpressed with AvrAC, but not when expressed alone, interacted with the RKS1-ZAR1 complex in gel filtration and pull-down assays (Fig. 3A and fig. S7). The lack of AvrAC in the tertiary protein complex confirms that the AvrAC-mediated uridylylation of PBL2, but not AvrAC protein per se, is recognized by the ZAR1-RKS1 receptor complex. We then solved a cryo-EM structure of the ZAR1-RKS1-PBL2UMP tertiary complex at a resolution of 4.3 Å (Fig. 3B and fig. S8). However, ZAR1NBD became much less well defined compared with that in the ZAR1-RKS1 complex after PBL2UMP binding, as indicated by the 3D reconstructions of the tertiary complex (Fig. 3C and fig. S8). Exclusion of this domain generated a 3D reconstruction with a resolution of 3.9 Å (Fig. 3B and fig. S8).

Fig. 3 PBL2UMP interaction with ZAR1-RKS1 enhances ZAR1NBD flexibility.

(A) ZAR1-RKS1 and PBL2UMP form a monomeric tertiary complex in gel filtration. Shown on the left are gel filtration profiles of ZAR1-RKS1 (red), PBL2UMP (blue), and ZAR1-RKS1+PBL2UMP (cyan) proteins. Positions of standard molecular mass are indicated by arrows. A280, absorbance at 280 nm; mAU, milli–absorbance units. Peak fractions in the left were visualized by SDS-PAGE followed by Coomassie blue staining and are shown on the right. MMM, molecular mass marker. (B) 3D reconstructions of the ZAR1-RKS1-PBL2UMP complex. Two orientations of the final EM density maps of the ZAR1-RKS1-PBL2UMP complex with ZAR1NBD unmasked (the two on the left) and masked (the two on the right). Shown in the middle is the color scale, in angstroms, for local resolution of the density maps. (C) ZAR1NBD is much less well defined than the remaining parts of the ZAR1-RKS1-PBL2UMP complex. Shown on the left and right are EM densities of the ZAR1-RKS1-PBL2UMP complex contoured at 4.0 σ and 2.0 σ, respectively. ZAR1NBD is highlighted in the red frames.

The model built into the higher resolution (3.9 Å) reconstruction of ZAR1-RKS1-PBL2UMP (table S1) is used for discussion of the RKS1-PBL2UMP interaction. Consistent with the data from gel filtration (Fig. 3A), the structure of ZAR1-RKS1-PBL2UMP is monomeric (Fig. 4A). PBL2UMP interacts exclusively with RKS1 (Fig. 4A), providing an explanation for AvrAC-induced recruitment of PBL2 to the ZAR1-RKS1 complex (27). The loop region harboring the two uridylylated residues of PBL2UMP is sandwiched between the N- and C-lobes of RKS1, mainly interacting with the activation segment of RKS1 (residues 213 to 243) (Fig. 4A). Additionally, the short α helix C-terminal to the uridylylated loop of PBL2 establishes contacts with the C-lobe of RKS1. The uridylyl moieties on PBL2Ser253, Thr254 form extensive polar and van der Waals interactions with residues 226 to 232 from the activation segment of RKS1 (Fig. 4B, top). On the opposite side, residues including Asp69 and Val70 make additional contacts with the uridylyl group of Ser253. These structural observations explain an essential role of uridylylation of these two residues in PBL2 interaction with RKS1 and the avrAC-specified disease resistance (27). The RKS1-PBL2UMP interaction is further strengthened largely by hydrophobic packing of the C-lobe of RKS1 against the short α helix C-terminal to the uridylylated loop of PBL2 (Fig. 4B, bottom). Contrary to the ZAR1-interacting residues of RKS1, the PBL2UMP-interacting residues of RKS1 are not conserved among RLCK XII-2 subfamily proteins (fig. S4), explaining the specific recognition of RKS1 by PBL2.

Fig. 4 RKS1-mediated PBL2UMP recognition by the ZAR1-RKS1.

(A) Overall structure of the ZAR1-RKS1-PBL2UMP complex. The red and green frames highlight the interacting regions between RKS1 and PBL2UMP. (B) Detailed interactions of uridylated PBL2Ser253, Thr254 with RKS1 (top) and an α helix of PBL2UMP with RKS1 (bottom), corresponding to the red- and green-framed regions in (A). Cryo-EM density is shown in green mesh. (C) Mutations of RKS1 around the two interfaces shown in (B) affect RKS1 interaction with PBL2UMP in vitro. GST-tagged RKS1 bound to GS4B beads was incubated with an excess amount of PBL2UMP. After extensive washing, the beads were analyzed by SDS-PAGE and Coomassie brilliant blue staining. (D) Mutations of RKS1 around the two interfaces shown in (B) diminish RKS1 interaction with PBL2UMP in protoplasts. Protoplasts isolated from rks1 plants were transfected with the indicated constructs for co-IP assays as in Fig. 2F. (E) Mutations of RKS1 around the two interfaces shown in (B) reduce AvrAC-induced cell death in protoplasts. The assays were performed as in Fig. 2G. Different letters indicate significant difference (P < 0.05, Tukey post hoc test). (F) Mutations of RKS1 around the two interfaces shown in (B) impair avrAC-specified resistance in Arabidopsis plants. rks1 mutant plants were complemented with the indicated variants of RKS1. Transgenic plants of the T1 generation were inoculated with the indicated bacterial strains and scored for disease symptoms as in Fig. 2H.

To further verify our structural observations, we made mutations in RKS1 and examined their effect on RKS1 interaction with PBL2UMP using pull-down and co-IP assays. RKS1 itself was sufficient to interact with PBL2UMP in the assays (fig. S9), as shown previously (27). By contrast, the PBL2UMP binding activity was abolished or reduced by RKS1 mutations D69Y (Asp69→Tyr), V70Y (Val70→Tyr), G233A (Gly233→Ala), and T231Y (Thr231→Tyr) (Fig. 4, C and D), a result predicted by the structural data (Fig. 4B, top). A similar result was also obtained for the RKS1 I235E (Ile235→Glu) mutant with a perturbed local conformation of the activation segment (Fig. 4, C and D). Gln68 interacts intramolecularly with Lys73 to stabilize the conformation of the loop carrying Asp69 and Val70. Consistently, the PBL2UMP-binding activity of the RKS1 Q68Y (Gln68→Tyr) mutant was reduced compared with WT RKS1 (Fig. 4, C and D). Similarly, the RKS1 F232A (Phe232→Ala) and H240E (His240→Glu) mutations also decreased the interaction with PBL2UMP in pull-down and co-IP assays (Fig. 4, C and D), supporting a role of Phe232 and His240 in the interaction with the short α helix (Fig. 4B, bottom) of PBL2UMP. We then tested the impact of these RKS1 mutations on AvrAC-induced cell death in protoplasts. When coexpressed with AvrAC, PBL2, and ZAR1, the RKS1 mutants D69Y, T231Y, F232A, I235E, and H240E that are impaired in PBL2 interaction were abolished in cell death activity, whereas those with reduced PBL2UMP-binding activity elicited weaker cell death compared with WT RKS1 (Fig. 4E). To further test these observations, we complemented the rks1 mutant with RKS1 variants by stable transformation and inoculated T1 transgenic plants with Xcc8004 or ΔavrAC strains. rks1 mutant plants complemented with RKS1D69Y, RKS1T231Y, RKS1F232A, and RKS1H240E were fully susceptible to Xcc8004, indicating that these mutant variants were unable to confer avrAC-specified disease resistance in plants (Fig. 4F and fig. S10).

PBL2UMP allosterically promotes the release of ADP from inactive ZAR1

One revelation of the ZAR1-RKS1-PBL2UMP structure is that ZAR1NBD becomes more flexible after PBL2UMP binding. Furthermore, structural comparison between ZAR1-RKS1 and ZAR1-RKS1-PBL2UMP revealed that ZAR1NBD rotates about 60° outward. By contrast, the other domains of ZAR1 retain similar conformations to those observed in the inactive ZAR1-RKS1 complex (Fig. 5A), providing an explanation for the monomeric, but still stable, ZAR1-RKS1-PBL2UMP complex. After PBL2UMP binding, the conformation of RKS1 remains nearly unchanged except for its activation segment, which is flexible in the ZAR1-RKS1 binary complex but becomes well defined in the ZAR1-RKS1-PBL2UMP tertiary complex (Fig. 5A and fig. S11). This result indicates that PBL2UMP binding acts to stabilize the activation segment of RKS1. Structural comparison further showed that the PBL2UMP-stabilized segment of RKS1 collides with one end of the ADP-bound ZAR1NBD from the inactive RKS1-ZAR1 complex (Fig. 5A). The steric clash is expected to dislocate ZAR1NBD of the inactive ZAR1, resulting in conformational changes in it, as observed in the structure. The conformational incompatibility between the PBL2UMP-bound RKS1 and the ADP-bound ZAR1NBD suggests that the PBL2UMP recruitment can indirectly impede the ADP-binding activity of ZAR1, thus releasing ADP from its inactive form. To further support this hypothesis, we tested the ADP-binding activity of the RKS1-ZAR1 complex in the presence of PBL2UMP or PBL2 using the assay previously described (40). Indeed, PBL2UMP induced ADP release from the ZAR1-RKS1 complex with much higher efficiency than PBL2 (Fig. 5B). Taken together, our structural and biochemical data indicate that PBL2UMP binding functions to stabilize the activation segment of RKS1, which sterically hinders ADP binding of ZAR1. This mechanism may also be true with HopZ1a-induced activation of ZAR1. However, in contrast with RKS1, ZED1 was reported to interact with ZAR1CC in addition to ZAR1LRR (29), suggesting that RKS1 and ZED1 might have different roles in ZAR1 activation.

Fig. 5 PBL2UMP binding allosterically promotes ADP release from ZAR1.

(A) PBL2UMP binding stabilizes a segment of RKS1 that is conformationally incompatible with ZAR1NBD in ZAR1-RKS1. Shown on the left and right are structural superpositions of ZAR1-RKS1 and ZAR1-RKS1-PBL2UMP. ZAR1-RKS1-PBL2UMP on the right is shown in both cartoon and transparent surface. ZAR1NBD domains from ZAR1-RKS1 and ZAR1-RKS1-PBL2UMP are shown in gray and pink, respectively. The activation segment of RKS1 (colored in red) is flexible in ZAR1-RKS1, but well defined in ZAR1-RKS1-PBL2UMP. The red frame on the left highlights ZAR1NBD. (B) Recruitment of PBL2UMP greatly reduces ADP-binding activity of the ZAR1-RKS1 complex. An aliquot of [2,8-3H]-ADP–bound ZAR1-RKS1 (6×His fused to the C terminus of ZAR1) was incubated with different concentrations of PBL2 or PBL2UMP at 4°C for 30 min. After flowing the samples through Ni-resins, the [2,8-3H]-ADP bound by ZAR1 was quantified by scintillation counting (left). The data were normalized against the input and are presented as percentages. Data are represented as mean ± SEM (n = 3). The proteins used for the ADP release assay are shown on the right.

Discussion

Autoinhibition, ligand recognition, and nucleotide exchange mechanisms of ZAR1

Our structural and biochemical understanding of plant NLRs largely comes from studies of animal NLRs, Apaf-1, and CED-4 (41). However, despite their analogous domains, NLRs from animals and plants resulted from convergent evolution after independent origins (42). Furthermore, although nucleotide exchange (ADP with ATP) is widely believed to have roles in NLR-initiated signaling (3), the underlying mechanism remained elusive, particularly because the bound ADP molecule is deeply buried, as demonstrated in the inactive NLRC4 (12) and Apaf-1 (9, 11). In this study, we report the cryo-EM structures of the plant NLR protein ZAR1 not only in its inactive ADP-bound state but also an intermediate state that is likely nucleotide-free in vitro because of the dislodged NBD. The structures revealed that ZAR1 assumes a canonical NOD structure as observed in NLRC4 (12) and Apaf-1 (9, 11) and that ADP binding functions to maintain its inactive conformation (Fig. 1). Despite the conserved NOD structure shared by these proteins, the C-terminal ZAR1LRR is presented in a position different from its counterpart of NLRC4 or Apaf-1. Nonetheless, the specially positioned ZAR1LRR still acts to sequester ZAR1 in a monomeric and ADP-bound state, which is further stabilized by ZAR1CC via interaction with ZAR1LRR, ZAR1WHD, and ZAR1HD1. ZAR1LRR is primarily responsible for ZAR1 interaction with RKS1 (Fig. 2A), whereas recognition of PBL2UMP occurs exclusively through RKS1 from the preformed ZAR1-RKS1 complex (Fig. 4A). The data provide the first structural view of recognition of modified self by NLRs in animals (43, 44) and plants (3, 4, 7, 8). Furthermore, capturing the likely nucleotide-free ZAR1 structure provides insights into the nucleotide exchange mechanism of the NLR. Structural comparison showed that the stabilized activation segment of RKS1 induced by PBL2UMP sterically clashes with the ADP-bound ZAR1NBD (Fig. 5A), which is expected to induce a substantial distortion of the structural domain, resulting in its dislodgement from the inactive conformation and consequently an ADP-depleted ZAR1. Such structural plasticity of the ATPase domain was also observed in the AAA+ protein p97, with its NBD undergoing order-disorder transitions during the hydrolysis cycle (45). The mechanism of PBL2UMP-induced ADP release from inactive ZAR1 resembles that of HspBP1-catalyzed nucleotide exchange of eukaryotic Hsp70 (46). Collectively, our results offer structural insights into the autoinhibition, ligand recognition, and nucleotide exchange mechanisms of ZAR1, providing a template for understanding of other NLRs from plants and animals.

RKS1 functions as a nucleotide exchange factor

Indirect recognition of effectors by plant NLRs can occur through effector-mediated posttranslational modifications of plant sensor proteins, such as the recognition of AvrAC by ZAR1 (27). Thus, PBL2UMP acts as a ligand for the ZAR1-RKS1 complex. Our structural and biochemical data showed that PBL2UMP indirectly triggers the release of ADP from the inactive ZAR1 (Fig. 5). However, structural comparison revealed that ADP release is not accompanied with conformational changes in the C-terminal ZAR1LRR domain, as proposed for activation of plant NLR proteins (47, 48). Furthermore, PBL2UMP induced no oligomerization of ZAR1 in vitro, similar to what has been observed with cytochrome c binding to Apaf-1 in the absence of ATP or dATP (10), though whether the cytochrome c–bound Apaf-1 is nucleotide-free or not remains undetermined. These results collectively indicate that the primary function of PBL2UMP is to prime the release of ADP from the inactive form of ZAR1 for activation. This sharply contrasts with flagellin binding to the NLR NAIP5 in animals (49, 50), in which flagellin functions to stabilize the active conformation of the NLR protein. The PBL2UMP-primed ADP release from the inactive ZAR1 is directly mediated by RKS1. This biochemical function of RKS1 is reminiscent of the nucleotide exchange factors (NEFs) of Hsp70 (51), which have critical roles in the functional cycle of Hsp70s by facilitating the release of ADP from its inactive state. Thus, RKS1 in the preformed ZAR1-RKS1 complex can be understood as an inactive NEF, whose activity is induced by PBL2UMP. NEFs of Hsp70s, despite their conserved biochemical function, are structurally unrelated and mechanistically highly diverse in nucleotide exchange reactions (51). Because nucleotide exchange is likely a general mechanism for plant NLR activation, as further confirmed in the accompanying study (52), it is conceivable that some other effector proteins indirectly recognized by plant NLRs similarly trigger NLR activation via inducing potential NEF activities of NLR-guarded host proteins. The study presented here provides a template for the analyses of other plant NLRs.

Model for ZAR1 priming

The data presented here and previously (10, 13, 14, 31, 32) support a stepwise activation model of ZAR1 (Fig. 6). ADP binding and the intramolecular interactions among multiple domains act collectively to keep ZAR1 in an inactive state. The C-terminal ZAR1LRR domain mediates formation of the preformed ZAR1-RKS1 complex that recognizes the AvrAC-uridylylated PBL2. Binding of PBL2UMP to the preformed ZAR1-RKS1 complex activates the NEF activity of RKS1, thus releasing ADP from ZAR1. ADP release, however, does not result in full activation of ZAR1, because ZAR1 is still monomeric and adopts a similar conformation to the ADP-bound ZAR1 except its NBD. This indicates that the ADP-depleted form of ZAR1 is in an intermediate state. The primed, but not oligomerized, ZAR1-RKS1-PBL2UMP suggests that a second step involving structural reorganizations similar to those for NLRC4 and Apaf-1 activation is required to fully activate ZAR1. ATP or dATP is the best candidate molecule for binding to the ADP-free ZAR1 to trigger the second signaling step, as observed for activation of Apaf-1. ZAR1 activation after ATP or dATP binding likely involves oligomerization of the NLR protein, a hallmark of AAA+ family proteins (53). In this model, a fail-safe mechanism is used by ZAR1 for its activation, wherein the AvrAC-modified PBL2 acts as a key to unlock ZAR1, but the unlocked ZAR1 is not fully activated until it binds to a second signaling molecule.

Fig. 6 Model for the PBL2UMP-induced priming of the ZAR1 complex.

ADP binding and intramolecular interaction of various ZAR1 domains in the preformed ZAR1-RKS1 complex maintains ZAR1 in an inactive conformation. The X. campestris pv. campestris effector AvrAC uridylylates PBL2, allowing the latter to be recruited by ZAR1-RKS1. The binding of PBL2UMP to RKS1 stabilizes the activation segment of RKS1, which causes a steric hindrance with ZAR1NBD. The dislodged ZAR1NBD releases ADP, and the ZAR1-RKS1-PBL2UMP enters into an intermediate state, which is likely competent for ATP binding and full activation of ZAR1.

Materials and methods summary

ZAR1 and RKS1 (with an N-terminal 6×His-SUMO) were coexpressed in Sf21 insect cells. The complex protein was first purified using Ni-NTA and further cleaned by ion-exchange and gel filtration after removal of SUMO by PreScission. The purified ZAR1-RKS1 was concentrated to ~1.0 mg/ml for cryo-EM. Similar protocols were used to purify the RKS1 mutants in complex with ZAR1. To purify ZAR1 mutants complexed with RKS1, N-terminally GST-tagged ZAR1 and RKS1 were coexpressed in Sf21 insect cells and the mutant complex proteins were purified by Glutathione Sepharose 4B. AvrAC and PBL2 (with a C-terminal 6×His tag) were coexpressed in E. coli. The PBL2UMP protein was purified using the protocols described above. The purified ZAR1-RKS1 and PBL2UMP proteins were mixed together and then subjected to gel filtration to purify the ZAR1-RKS1-PBL2UMP complex. The complex was concentrated to ~1.5 mg/ml for cryo-EM. To assay ZAR1-RKS1 interaction with PBL2UMP or PBL2, the purified ZAR1-RKS1 was incubated with the His-tagged PBL2UMP or PBL2 and bound to Ni resins. After washing, the Ni beads were analyzed by SDS-PAGE and Coomassie brilliant blue staining.

To assay the effect of PBL2 or PBL2UMP on the ADP-binding activity of ZAR-RKS1, the [2,8-3H]-ADP-ZAR1-RKS1 complex protein bound to Ni resins was incubated with different concentrations of PBL2 or PBL2UMP, and then each sample was pelleted by centrifugation. After washing, the pellet was eluted with 250 mM imidazole, and [2,8-3H]-ADP in the eluent was quantified by scintillation counting.

Cryo-EM data of frozen hydrated grids of ZAR1-RKS1 or ZAR1-RKS1-PBL2UMP were collected on a Titan Krios electron microscope operated at 300 kV equipped with VPP (Volta Phase Plate) and a Gatan K2 Summit direct electron detection camera (Gatan) using AutoEMation. The ZAR1-RKS1-PBL2UMP dataset was collected without the insertion of VPP, and the ZAR1-RKS1 dataset was collected using VPP, as described earlier. The raw supersolution dose-fractionated image stacks were binned, aligned, dose-weighted, and summed using MotionCor2. Contrast transfer function (CTF) parameters were estimated using CTFFIND4 and GCTF. Particle picking, 2D classification, 3D classification, and refinement were all performed in RELION. The EM density reconstructed from ZAR1-RKS1 was used for model building in Chimera (54) and COOT. To build the model of ZAR1-RKS1-PBL2UMP, the refined model of ZAR1-RKS1 without ZAR1NBD was fitted into the EM density reconstructed from the former complex in Chimera. The crystal structure of the BAK1 kinase domain was used as the initial model of PBL2UMP. The final models were refined against their corresponding EM maps by PHENIX.

Structure-guided mutagenesis was carried out to assess the importance of various amino acid residues in ZAR1-RKS1 and RKS1-PBL2UMP interactions, cell-death triggering, and resistance in plants. Wild-type and mutant forms of ZAR1, RKS1, PBL2, and AvrAC constructs were transfected into Arabidopsis protoplasts, and cell viability was determined. RKS1 or ZAR1 variants under the control of native promoters were introduced into rks1 or zar1 mutants, respectively, and stable transgenic plants were wound-inoculated with Xanthomonas campestris pv. campestris strains carrying or lacking avrAC. Disease resistance was scored on the basis of presence or absence of disease symptoms.

Supplementary Materials

www.sciencemag.org/content/364/6435/eaav5868/suppl/DC1

Materials and Methods

Figs. S1 to S11

Table S1

References (5572)

References and Notes

Acknowledgments: We thank J. Lei, X. Li, X. Fan, and H. Wu at Tsinghua University for data collection; Z. Zhou at the Institute of Genetics and Developmental Biology for guiding the isotope assay; and P. Schulze-Lefert at the Max Planck Institute for Plant Breeding Research for critical reading of the manuscript. We acknowledge the Tsinghua University Branch of the China National Center for Protein Sciences (Beijing) for providing the cryo-EM facility support and the computational facility support on the cluster of Bio-Computing Platform. Funding: This research was funded by the Strategic Priority Research Program of the Chinese Academy of Sciences (XDB11020200 to J.-M.Z.), the National Natural Science Foundation of China (31421001 to J.C. and 31700660 to Jiz.W.), the Alexander von Humboldt-Foundation (Humboldt Professorship to J.C.), Max Planck–Gesellschaft (Max Planck Fellow to J.C.), the National Key R&D Program of China (grant 2016YFA0501100 to H.-W.W.), the Beijing Municipal Science and Technology Commission (grant Z161100000116034 to H.-W.W.), and the China Postdoctoral Science Foundation (2016M600081 to Jiz.W. and 2017M620746 to Jia.W.). Author contributions: Conceptualization: J.C., J.-M.Z., and H.-W.W.; methodology: J.C., J.-M.Z., H.-W.W., Jiz.W., Jia.W., and M.H.; investigation: Jiz.W., Jia.W., M.H., and J.Q.; validation: Jiz.W., Jia.W., M.H., and G.W.; supervision: J.C., J.-M.Z., H.-W.W., Y.Q., and Z.H.; writing, original draft: J.C.; writing, review and editing: J.C., J.-M.Z., H.-W.W., Jiz.W., Jia.W., and M.H.; funding acquisition: J.C., J.-M.Z., H.-W.W., and Jiz.W. Competing interests: The authors declare no competing interests. Data and materials availability: All data needed to replicate the work is present either in the supplementary materials or in the listed Protein Data Bank files. For the ZAR1-RKS1 complex, the atomic coordinates and EM map have been deposited in the Protein Data Bank and Electron Microscopy Database with accession codes 6J5W and EMD-0683, respectively. For the ZAR1-RKS1-PBL2UMP complex, the atomic coordinates with or without the NBD have been deposited in the Protein Data Bank with accession codes 6J5V and 6J5U, respectively, and the EM maps with or without the NBD have been deposited in the Electron Microscopy Database with accession codes EMD-0682 and EMD-0681, respectively.
View Abstract

Stay Connected to Science

Navigate This Article