Molecular insights into the biochemical functions and signalling mechanisms of plant NLRs

Abstract Plant intracellular immune receptors known as NLR (nucleotide‐binding leucine‐rich repeat) proteins confer immunity and cause cell death. Plant NLR proteins that directly or indirectly recognize pathogen effector proteins to initiate immune signalling are regarded as sensor NLRs. Some NLR protein families function downstream of sensor NLRs to transduce immune signalling and are known as helper NLRs. Recent breakthrough studies on plant NLR protein structures and biochemical functions greatly advanced our understanding of NLR biology. Comprehensive and detailed knowledge on NLR biology requires future efforts to solve more NLR protein structures and investigate the signalling events between sensor and helper NLRs, and downstream of helper NLRs.

N-terminal TIR, CC, or CC R domains are known as TNLs, CNLs, or RNLs, respectively (Duxbury et al., 2021). TNLs and CNLs recognize pathogen effectors directly or indirectly to activate cell death and defence responses, thus are regarded as sensor NLRs (Jubic et al., 2019). TNLs or CNLs also function as pairs to perceive pathogen effectors and confer resistance. For instance, in Arabidopsis two TNLs, RPS4 and RRS1, located in a head-to-head orientation in the genome, cooperate to resist infection by at least three different pathogens (Narusaka et al., 2009). In rice, two CNLs, RGA4 (Resistance Gene Analogue 4) and RGA5 (Resistance Gene Analogue 5), function together to resist Magnaporthe oryzae carrying the effector Avr-Pia or Avr1-CO39 (Cesari et al., 2013). In the NLR pair system, one NLR possesses an additional domain called the integrated decoy (ID) domain for direct effector interaction functioning as a sensor, while the other NLR functions as an executor for immune signalling (Le Roux et al., 2015;Williams et al., 2014). Downstream of sensor NLRs or sensor/executor NLR pairs, higher plants also use helper NLRs to transduce signals. There are three described families of helper NLRs: NLR Required for Cell death (NRC), Activated Disease Resistance 1 (ADR1), and N Required Gene 1 (NRG1) (Jubic et al., 2019). In terms of domain structure, NRCs belong to the CNL class, while ADR1s and NRG1s belong to the RNL class. In solanaceous species such as Nicotiana benthamiana, NRCs are required for phylogenetically related CNLs to activate cell death and immune responses (Wu et al., 2017). In N. benthamiana, NRC2, NRC3, and NRC4 function downstream of different CNLs to positively regulate immunity, while NRCX negatively regulates NRC2 and NRC3-mediated immunity (Adachi et al., 2021). In Arabidopsis thaliana, three ADR1s, NRG1.1 and NRG1.2 are full-length RNLs (CC R -NB-LRR) functioning as positive regulators, while the truncated NRG1.3 (NB-LRR) negatively regulates AtNRG1.1-and AtNRG1.2-mediated immunity . In dicots, ADR1s and NRG1s are required downstream of TIRonly receptors, TNLs and TNL pairs to activate defence responses, and ADR1s are also required for the full function of CNLs (Jubic et al., 2019). ADR1s, but not NRG1s, are also present in monocots with uncharacterized function. Some NLRs that function as both sensors of effectors and helpers directly activating cell death are regarded as NLR singletons (Adachi et al., 2019) (Figure 1).

| NLR RECOG NITI ON OF EFFEC TOR S
Plant sensor NLRs recognize pathogen effectors via direct interaction or perceiving the modifications by effectors (Duxbury et al., 2021 (Ma et al., 2020;Martin et al., 2020). Interestingly, a previously undefined C-terminal jelly roll/Ig-like domain (C-JID) was revealed in the post-LRR domain regions of RPP1 and ROQ1. Both C-JID and LRR domains are involved in effector recognition (Figure 2a,b). The structure-conserved but sequence-diversified C-JID is present in F I G U R E 1 Schematic diagram depicting the domain structures of sensor NLRs and helper NLRs and their interdependence for function. TIR, toll/interleukin-1 receptor; CC, coiled-coil; CC R , (Resistance to Powdery Mildew 8)-like CC; NB, nucleotide-binding; LRR, leucine-rich repeat; ID, integrated decoy many other TNLs from dicotyledonous plant species, but not in CNLs from diverse plant species (Ma et al., 2020).
In addition, well-studied NLR pairs use the ID domain of sensor NLRs to physically associate with effectors for activation (Duxbury et al., 2021). In the TNL pair of RPS4 (Resistance to Pseudomonas syringae 4) and RRS1 (Resistance to Ralstonia solanacearum 1), the integrated WRKY domain of RRS1 detects effectors AvrRps4 from Pseudomonas syringae pv. pisi and PopP2 from Ralstonia solanacearum (Le Roux et al., 2015;Narusaka et al., 2009). Interestingly, crystal structures of RRS1 WRKY in complex with AvrRps4 or PopP2 revealed that the two effectors target the same region of RRS1 WRKY (Mukhi et al., 2021, Zhang et al., 2017b (Figure 2c). In rice, two CNL pairs, RGA4/RGA5 and Pik-1/Pik-2, use the integrated heavy metal-associated (HMA) domain in the sensors RGA5 and Pik-1, respectively, to directly recognize cognate effectors. A few structures of the Pik-1 HMA domain in complex with the effector Avr-Pik highlighting the recognition specificities and key residues for activation have been reported (Bialas et al., 2021;De la Concepcion et al., 2018 (Figure 2d).
The structures of the inactive RKS1-ZAR1 complex (ADP bound) and the intermediate PBL2 UMP -RKS1-ZAR1 complex (ADP free) revealed effector-triggered allosteric ADP release as a key step to prime NLR for ATP binding and activation (Wang et al., 2019b) (Figure 2e).

| N -TERMINAL S IG NALLING DOMAIN: DEFINING THE B I O CHEMI C AL FUN C TI ON S OF NLR S
The N-terminal TIR, CC, or CC R domains of NLRs are the signalling domains, overexpression of which in N. benthamiana is sufficient to trigger cell death (Bernoux et al., 2011;Collier et al., 2011;Maekawa et al., 2011;Swiderski et al., 2009). Structural biology studies on plant NLRs were limited to the N-terminal signalling domains before the advancement of cryo-EM as a technique to solve full-length NLR structures.   (Bentham et al., 2018;Daskalov et al., 2016;Jubic et al., 2019;Mahdi et al., 2020).

| CONS ERVED FOUR-HELIC AL-BUNDLE FOLD OF CC AND CC R DOMAIN S FOR CELL DE ATH FUN C TION
The crystal structure of AtNRG1.1 CC R confirmed the 4HB fold of CC R (Jacob et al., 2021) (Figure 4a). AtNRG1.1 CC R harbours an extra region N-terminal to the 4HB fold compared to canonical CC domains, which is consistent with previous structural predictions of CC R and RPW8. The extra region N-terminal to the 4HB fold of CC R and RPW8 is predicted to be responsible for membrane anchoring (Zhong & Cheng, 2016). A recent study showed that deletion of the extra region N-terminal to the 4HB fold impairs the cell death function of auto-active AtNRG1.1 but not its membrane localization (Jacob et al., 2021). It has also been shown that RNLs interact with anionic plasma membrane (PM) phospholipids, and depletion of phosphatidylinositol-4-phosphate from the PM causes mislocalization of RNLs and abolishes their cell death activities .
MLKL is a well-studied membrane pore-forming protein (Murphy et al., 2013). Current modelling suggests that MLKL remains cytosolic and monomeric in the resting state, but starts to oligomerize, move to the PM, and form pores at the PM, leading to loss of membrane integrity and cell death (Huang et al., 2017). The exact oligomerization state and pore size of activated MLKL are still unknown.
It has also been shown that MLKL functions as a cation channel to induce cell death, implicating the possible channel activities of plant CNLs and RNLs (Xia et al., 2016).

| NLR RE S IS TOSOME S AND S IG NALLING
Self-association is known to be required for NLR function. The NB domain is responsible for the self-association and oligomerization of NLRs and functions as a molecular switch bridging immune stimulus perception by the LRR domain and immune signalling by the Nterminal domain Saur et al., 2021).

| CNL Z AR1 PENTAMERIC RE S IS TOSOME FUN C TI ONING A S A C a 2+ CHANNEL
In 2019, cryo-EM strudies of full-length ZAR1 in both resting and active states for the very first time revealed the structure of interdomain interactions for autoinhibition and oligomerization for immune activation of a plant NLR (Wang et al., 2019a(Wang et al., , 2019b. The structure of monomeric ADP-bound ZAR1 in complex with RKS1 represents the resting state. RKS1 interacts exclusively with the ZAR1 LRR domain to maintain ZAR1 in the resting state (Wang et al., 2019b). The structure of dATP-bound ZAR1 in complex with RKS1 and effectorproduced PBL2 UMP represents the active state. On effector recognition, ZAR1 undergoes dramatic conformational changes from an inactive monomeric state to an active pentameric state with the first helix in the 4HB fold of its CC domain being flipped, potentially forming a pore at the PM. In addition to the first helix α1 being flipped, the fourth helix α4A becomes disordered in the active state. Previously identified sequences C-terminal to the 4HB domain that are critical for self-association and cell death function are largely disordered in the resting state but become a helix, termed α4B, in the active state (Figure 4b,c). These conformational changes in the CC domain and pentamerization driven by the NB domain feature the dynamics of ZAR1 resistosome formation. The ZAR1 resistosome displays a channel pore of 1 nm in diameter (Wang et al., 2019a) (Figure 4d). A follow-up study confirmed that the effector-activated ZAR1 resistosome protrudes into the PM and forms a calcium-permeable cation channel leading to calcium influx and further activation of cell death and defence responses   (Figure S1a).

| TNL RPP1 AND RO Q1 TE TR AMERI C RE S IS TOSOME S FUN C TI ONING A S NADase
Recent cryo-EM strudies of TNL/effector pairs ROQ1/XopQ and RPP1/ATR1 structures, followed by structure-guided biochemical studies in vitro and functional analyses in planta, greatly advanced our understanding of TNL activation mechanisms and functioning as a NAD + cleavage enzyme (Ma et al., 2020;Martin et al., 2020). On effector recognition, ROQ1 and RPP1 assemble into homotetrameric resistosomes. An apparent difference is that the ROQ1 resistosome binds an ATP molecule in the NB domain as expected, while the RPP1 resistosome binds an ADP molecule although ATP was supplemented in the purification process. Further analyses indicated that plant CNLs and some TNLs, such as ROQ1, use a "TT/SR" motif for ATP binding, while some TNLs, such as RPP1, have a "TTE/Q" motif, losing the ATP binding ability and bind ADP molecules (Ma et al., 2020). Tetramerization in the NB domain drives the formation of the TIR tetramer as an active holoenzyme. The TIR tetramer contains two previously characterized symmetrical AE interfaces and two asymmetric interfaces that create the potential active sites for NAD + cleavage catalysis (Figure 3b).
Interestingly, each of the two active sites in the RPP1 resistosome binds an ATP molecule, which was supplemented during purification.
The bound ATP probably acts as an analogue of NAD + at the active site. More interestingly, structural analyses revealed that NADP + binds to the TIR domain of RUN1 at a similar position to the ATP-occupied active site (Horsefield et al., 2019;Ma et al., 2020). These observations support the idea that active TNL resistosomes function as a holoenzyme cleaving NAD(P) + to transduce the signal in the absence of NAD(P) + -bound TNL resistosome structures. However, the identity of the TNL-produced signalling molecules remains elusive.

| N UCLE A S E AC TIVITIE S OF PL ANT TIR FIL AMENTS
Previous crystal structures of plant TIR domains identified two dimerization interfaces required for function (Zhang et al., 2017a). The functional relevance of the AE dimerization interface was demonstrated in TNL tetrameric resistosomes for NADase activity. However, the DE interface was not observed in TNL tetrameric resistosomes and its functional relevance remains unclear. Interestingly, a recent study showed that the TIR domain of flax TNL L7 can assemble into a filament using the AE and DE interfaces, as proposed previously (Figure 3a). Moreover, the L7 TIR domain was purified in complex with DNA. Cryo-EM studies of the structure of the L7 TIR filament wrapping DNA showed that filament-DNA interaction is critical for TIR nuclease activity and production of 2′,3′-cAMP . 2′,3′-cAMP molecules were identified in the NAD + -binding pocket of L7 TIR domains in the TIR filament structure, suggesting that NADase and nuclease activities of TIR domains use similar catalytical sites. Further mutational studies confirmed that 2′,3′-cAMP synthetase activity correlates with TIR cell death function . It is not clear whether 2′,3′-cAMP functions as a signalling molecule. Thus, plant TIR domains could assemble into filaments and function as nucleases, but the possibility of TIR filament formation in the context of full-length TNLs and the distinct roles of TIR NADase and nuclease activities in TIR signalling remain unknown.

| UNDEFINED FUN C TI ON S OF LIPA S E-LIK E PROTEIN S AND RNL S DOWN S TRE AM OF TNL
All tested TNLs require lipase-like proteins, including Enhanced Disease Susceptibility1 (EDS1), Senescence-Associated Gene101 (SAG101), and Phytoalexin Deficient4 (PAD4), for function (Gantner et al., 2019;Lapin et al., 2019;Wagner et al., 2013). In Arabidopsis, AtEDS1 and AtSAG101 form a stable heterodimer and cooperate with the AtNRG1s to control cell death, while AtEDS1 and AtPAD4 physically associate and function together with the AtADR1s to mediate bacterial growth restriction and resistance. EDS1, SAG101, and PAD4 contain a lipase-like domain, whose enzymatic activity is dispensable for function, and a plant-unique EDS1-PAD4 (EP) domain. In EDS1/SAG101 and EDS1/PAD4 heterodimers, the pockets between the two EP domains are critical for TNL function and are proposed to be the potential binding sites of TIR-generated small molecules, but are not validated (Sun et al., 2021). More interestingly, activation of the TIR signalling pathway has been shown to trigger the association between EDS1/SAG101and NRG1s as well as the interaction between EDS1/PAD4 and ADR1s (Sun et al., 2021;Wu et al., 2021). The activation mechanisms of EDS1/SAG101/ NRG1 and EDS1/PAD4/ADR1 to initiate cell death and resistance responses, respectively, are currently unknown.  Saile et al., 2020). In the context of ETI, RNLs function cooperatively with lipase-like proteins. Autoactive RNLs or their CC R domains trigger EDS1-independent cell death in N. benthamiana (Collier et al., 2011;Jacob et al., 2021). The AtNRG1.1 CC R domain adopts a 4HB structure fold reminiscent of the cell death domain of MLKL that has been shown to function as cation channel. Further investigations demonstrated that autoactive NRG1 mutant and ADR1 function as a Ca 2+ channel when expressed in planta and in human cells to trigger cell death (Jacob et al., 2021) (Figure S2a).

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable to this article as no new data were created or analysed.