Structural insights into plant NLR immune receptor function

Plants and animals use intracellular innate immune receptors known as nucleotide-binding oligomerization domain-like receptors (NLRs) to detect the presence of pathogens and activate defense. Although genetic studies have defined an increasing number of receptors and the pathogen triggers that activate them, the structural and mechanistic underpinnings of the events at, and downstream of, perception have lagged behind. There are no full-length structures available for NLRs, so researchers have had to content themselves with partial structures, often individual domains. The functional relevance of these partial structures can be difficult to parse. Nevertheless, in a pair of articles in PNAS, Cesari et al. (1) and Casey et al. (2) have provided a clearer view into exactly how NLR proteins function.

Microbes become pathogens by defeating some aspect of the host immune system. Plant pathogens from all kingdoms have evolved a set of molecular tools: protein virulence effectors that are delivered into the host cytoplasm. Once inside the host, pathogen effectors target a variety of host processes to promote virulence. These targets can be conserved aspects of the immune system, or can be more specialized targets with pathogen-specific outcomes (3, 4). Plant genomes have responded by evolving specific NLRs that recognize the presence of pathogen virulence effectors and respond appropriately to defeat pathogens (often with programmed cell death) (5). A detailed structural and mechanistic understanding of exactly how NLR proteins respond to pathogens is critical for ensuring food security by rationally engineering disease resistance in crops. We need to know both how the immune response is suppressed in the absence of pathogens and how it is appropriately activated.

Models of NLR immune function are informed by their multidomain architecture. In plants, NLRs are characterized by three domains: (i) an N-terminal coiled-coil (CC) or Toll/interleukin-1 receptor (TIR) domain, (ii) a central nucleotide-binding (NB) domain, and (iii) a C-terminal leucine-rich repeat (LRR) domain. These domains are proposed to have distinct functions both pre- and postactivation. NLRs are thought of as molecular switches that flip between two conformations: a closed “off-state” and an open “on-state.” The central NB domain serves as the hinge of this switch, closed when ADP-bound and open when bound to ATP. Structural modeling with the NB domains of animal NLRs has served as a guide to understand plant NBs, allowing the generation of both precise and predictable loss- and gain-of-function mutants (6) and providing models for NLR activation (7, 8). However, how this conformational switch is triggered and the downstream consequences are more poorly understood. Generally, current models suggest that interactions between the LRR domain and the N-terminal or NB domain can promote a resting state in the ADP-bound “off” conformation. Pathogen effectors, directly or indirectly, disrupt negative regulation of the NB, and allow the open conformation to form. However, there are no plant full-length or truncated NB-LRR or CC/TIR-NB structures to inform how this molecular switch turns on and off.

Studies of several plant NLRs indicate that the N-terminal domain (either a CC or TIR domain) dimerizes as the NB switches to an open, active confirmation (9–11). Truncated CC or TIR domains are often sufficient to activate cell death; they become “autoactive” in the absence of pathogen triggers. Mutants that lose dimerization show that self-association is required for function. Thus, N-terminal dimerization is proposed to be the signaling event that activates downstream defense responses. Stymied by the lack of full-length structures, structural biologists have been more successful with understanding how these autoactive N-terminal domains fold and interact. However, these structures do not always tell the same stories. Both plant and animal TIR domains have been crystalized in more than one orientation (12). Before Casey et al. (2), there were two dramatically different CC domain structures (Fig. 1A). The first was the CC domain of MLA10 (mildew resistance locus A), a barley NLR that recognizes a powdery mildew pathogen (9). It forms a long rod-shaped CC dimer. The dimer is composed of two antiparallel helix–turn–helix monomers, stabilized by hydrophobic interactions across the dimer. The second solved structure, the CC domain of the potato NLR Rx (resistance to potato virus X), had a much more compact shape: a monomer of four helices folded back on themselves (13). One important difference between the structures is that Rx was crystalized as a heterodimer with an interacting partner, Ran GTPase-activating protein 2 (RanGAP2). At the time, it was unclear if the major differences between MLA10 and Rx were due to differences in sequences or if the presence of the interacting protein promoted a different conformation for Rx (potentially the off-state, unperturbed by a pathogen trigger). Now, Casey et al. (2) have provided a third option by generating an NMR solution structure for the CC domain of Sr33. Sr33, an ortholog of MLA10, confers resistance to the Ug99 strain of wheat stem rust, a rapidly emerging threat to global wheat production (14).

Fig. 1.

(A) MLA10 (Protein Data Bank ID code 3QFL; blue and white) was the first plant NLR CC domain crystalized. It crystalized as an elongated dimer, stabilized by hydrophobic residues. The second plant CC domain structure is the Rx protein (Protein Data Bank ID code 4M70; yellow). Rx crystalized as a compact monomer, with a four-helix bundle. The NMR structure of the Sr33 CC (Protein Data Bank ID code 2NCG; green) indicates that it is folded similar to Rx, despite being more closely related to the MLA10 CC [∼78% identical (Id.) to MLA10, ∼20% identical to Rx]. Several hydrophobic residues shown to be important for MLA dimerization are conserved in Sr33, likely stabilizing the monomer fold (shown in cyan). (B) Hypothetical CC conformation switch between an inactive off monomer and an active on dimer. Helices are labeled from the N terminus to C terminus (red, orange, yellow, and green). Approximately 180° straightening of turns attached to helix 1 (red) and helix 4 (green) might allow an Rx/Sr33-type CC fold to switch into a MLA10-like dimer. Conserved hydrophobic residues are shown in cyan. The top structure is the Sr33 NMR structure, and the bottom structure is the Sr33 CC domain modeled onto one-half of the MLA10 dimer (other monomer in white). MLA10 dimer helices are color-coded as in the Sr33 monomer.

Unexpectedly, the structure of the Sr33 CC looks much more like the distantly related Rx than its true ortholog MLA10 (Fig. 1A). Like the Rx CC, the Sr33 CC also adopts a compact bundle of four α-helices packed back upon themselves. The Sr33 CC structure is also monomeric, indicating that CC domains, as purified proteins, can form this tetracoil fold in the absence of associated proteins like RanGAP2. Casey et al. (2) verified that the MLA10 structure is robust to the vagaries of protein purification and crystallization. They were able to reproduce a very similar elongated dimer structure, albeit with slightly different contacts between the monomers. However, the story in solution was different. In solution, when assayed by light scattering assays [size exclusion chromatography (SEC)-coupled multiangle light scattering and SEC-coupled small-angle X-ray scattering], the CC domains of Sr33, Rx, and MLA10 all behaved like monomeric proteins of the compact Rx/Sr33-like fold. The authors used equivalent constructs (MLA10_5–120 and Sr33_6–120), so the length of the construct itself is not generating the structural difference. Then, why is the MLA10 crystal structure so different?

One possibility is that the elongated MLA10 structure is merely an artifact of crystallization. Perhaps MLA10_1–120 is intrinsically less likely to form a monomer than Sr33_1–120. If unable to form the monomer, the MLA10 CC could pack itself in the crystal to bury its exposed hydrophobic surfaces as a dimer. This hypothesis could explain how an incorrect structure identified hydrophobic residues that are both conserved and required for function in MLA10. The fact that the new MLA10 dimer structure generated by Casey et al. (2) is packed slightly offset from the original is perhaps consistent with a hypothesis of expediency.

An intriguing second hypothesis is that the elongated MLA10 structure, rather than being incorrect, just reflects a different functional state for the CC domain. This hypothesis is supported by discrepancies between what we know about the dimerization state in the crystal and the multimerization state in planta. MLA10 CC crystalizes as a dimer. However, this CC truncation is inactive in planta. The longer MLA10_1–160 form is autoactive, and mutational analysis shows that dimerization is required for function. The longer Sr33 CC in vitro is a monomer; however, in planta, it is also a dimer. Hydrophobic residues equivalent to those residues defined in MLA10 are required for both Sr33 function and dimerization. One way to square these results is if the Rx/Sr33 four-helix structure represents the monomeric off-state for the CC domain and the MLA10 structure is revealing the dimeric on-state. The central two helices of the Sr33 structure and one MLA10 monomer overlap reasonably well, with similar positioning of important hydrophobic residues (Fig. 1B, teal). To switch from one state to another requires straightening turn 1 and turn 3 to reposition helix 1 and helix 4, respectively. These structures present a hypothesis for how CC domains could flip back and forth between inactive monomers and active dimers. The surfaces that are stabilized by helix–helix interactions in the compact monomer are protected in the dimer by a helix from the partner molecule. One caveat to this hypothesis is that the MLA10 dimer defined by the crystal structure is nonfunctional in planta, so this hypothetically “active” dimer structure is incomplete.

While analyzing the function of the structurally determined CC domains (i.e., truncations at amino acid 120), Cesari et al. (1) found that CC domain length is critically important. They provide functional data showing that short MLA10_1–120 or Sr33_1–120 constructs are unable to trigger autoactive cell death, whereas longer 160-aa constructs are functional (1). The longer forms also had a higher propensity to self-associate in vitro and in planta. Secondary structure analysis predicted that the short, inactive forms of MLA10 and Sr33 are missing a C-terminal α-helix. How this functionally critical extension impacts the structure and dimerization state of the CC remains to be determined. What does the active, dimeric state of a functional CC-domain look like? Does it look something like the MLA10 structure, or perhaps something entirely different? How does that structure fit into a larger model of NLR conformational changes? It seems plausible to think of NLR activation as composed of a number of openings. Although the master switch of the molecule is the NB domain, perhaps the CC domain itself is also a switch, subject to its own regulation. A series of conformational openings would allow for fine-tuning of NLR function. The importance of tuning NLRs is supported by functional data. For example, Arabidopsis CC NLRs, such as RPM1 and RPS2, respond differently to loss of RIN4, a physically interacting protein that they monitor for pathogen manipulation (their “guardee”). RPS2 is lethally hyperactive in response to loss of RIN4, whereas RPM1 is only very weakly activated (15). Alleles of Pm3, a wheat NLR, indicate that polymorphisms in the NB domain can tune the strength of an NLR immune response once activated, independent of “triggerability” (16). Similarly, CC dimer strength may be a useful phenotype to manipulate genetically.

For hundreds of years, crop breeders have selected genotypes with resistance to important diseases. More than two decades after the identification of NLRs as the molecules responsible for these traits, we still do not understand basic aspects of their function. Both how the NLR molecule opens and closes and what happens after dimer/multimerization remain critical questions. A full understanding of NLR regulation awaits the generation of multidomain structures. Rational engineering of plant NLR proteins will provide us with new tools to address current and emerging diseases that threaten global food security.