Structural Basis for flg22-Induced Activation of the Arabidopsis FLS2-BAK1 Immune Complex

See allHide authors and affiliations

Science  01 Nov 2013:
Vol. 342, Issue 6158, pp. 624-628
DOI: 10.1126/science.1243825

First Defense

In defense against bacterial infection, plants carry a cell-surface receptor, known as FLS2, that can bind to a fragment of bacterial flagellin and trigger defense responses. Y. Sun et al. (p. 624, published online 10 October) investigated the structural details that govern the binding between FLS2, its co-receptor BAK1, and the flagellin fragment flg22. The assembled complex initiates signals to activate the plant's innate immune response.


Flagellin perception in Arabidopsis is through recognition of its highly conserved N-terminal epitope (flg22) by flagellin-sensitive 2 (FLS2). Flg22 binding induces FLS2 heteromerization with BRASSINOSTEROID INSENSITIVE 1–associated kinase 1 (BAK1) and their reciprocal activation followed by plant immunity. Here, we report the crystal structure of FLS2 and BAK1 ectodomains complexed with flg22 at 3.06 angstroms. A conserved and a nonconserved site from the inner surface of the FLS2 solenoid recognize the C- and N-terminal segment of flg22, respectively, without oligomerization or conformational changes in the FLS2 ectodomain. Besides directly interacting with FLS2, BAK1 acts as a co-receptor by recognizing the C terminus of the FLS2-bound flg22. Our data reveal the molecular mechanisms underlying FLS2-BAK1 complex recognition of flg22 and provide insight into the immune receptor complex activation.

Innate immunity in higher eukaryotes relies on the perception of conserved signature components of pathogens, termed pathogen-associated molecular patterns (PAMPs), by plasma membrane–localized pattern recognition receptors (PRRs). In plants, PRRs are mainly receptor kinases (RKs) or receptor-like proteins, and several of them carry leucine-rich repeats (LRRs) in their ectodomains for PAMP recognition. Upon recognition of PAMPs, PRRs initiate an array of shared immune responses, leading to PAMP-triggered immunity (1).

Present in most higher plant species and critical for antibacterial immunity (1), flagellin-sensitive 2 (FLS2) is an LRR-RK and acts as the PRR for bacterial flagellin by recognizing the epitope flg22 (26). Direct recognition of flg22 by FLS2 is sufficient for inducing immune responses, establishing FLS2 as a flagellin receptor (7). Flg22 binding nearly instantly triggers FLS2 association with the LRR-RK BRI1-associated kinase 1 (BAK1) (8, 9). BAK1 also interacts with the LRR-RK BR INSENSITIVE 1 (BRI1) to positively regulate brassinosteroid (BR) signaling (10, 11). BAK1 is also called SERK3, a member of the subfamily of SERK LRR-RKs (12).

BAK1 also forms heteromers with several other PRRs and is a major component of plant immunity (13, 14). The flg22-induced FLS2-BAK1 heteromerization results in their trans-phosphorylation (8, 9, 15). Flg22 also induces FLS2- and BAK1-dependent phosphorylation of BIK1 (BOTRYTIS-INDUCED KINASE 1, a receptor-like cytoplasmic kinase) and dissociation of BIK1 from FLS2 for plant immunity (16, 17).

To confirm that the ectodomains of FLS2 and BAK1 are sufficient to form an flg22-induced complex, we used glutathione S-transferase (GST) precipitation, gel filtration, and coimmunoprecipitation (Co-IP) to assay their interaction. Collectively, the data from these assays (fig. S1) showed that the extracellular LRR domains of Arabidopsis FLS2 (residues 25 to 800, FLS2LRR) and BAK1 (residues 1 to 220, BAK1LRR) formed a monomeric heterodimer induced by flg22. But it remains possible that full-length FLS2 forms homo-oligomers (18).

To understand the molecular mechanism underlying FLS2 recognition of flg22, we solved the crystal structure of the FLS2LRR-flg22-BAK1LRR complex at 3.06 Å (Fig. 1A and table S1). None of the dimeric packing related by crystallographic symmetry can be biologically relevant (fig. S2), further supporting the gel filtration data (fig. S1B). This is in contrast with flagellin-induced Toll-like receptor 5 (TLR5) homodimerization (19). The structure of FLS2LRR is superhelical (fig. S3) and resembles that of BRI1LRR (20, 21). Flg22, which is well defined by electron density but the last residue (fig. S4), binds to the concave surface of FLS2LRR by running across 14 LRRs (LRR3 to LRR16) (Fig. 1, A and B), confirming previous hypotheses (22, 23). The flg22 binding groove is largely conserved in tomato FLS2 (fig. S5).The FLS2LRR-BAK1LRR heterodimerization is both flg22- and receptor-mediated. The C terminus of flg22 is sandwiched between FLS2LRR and BAK1LRR, whereas direct FLS2LRR-BAK1LRR interactions stem from anchoring of BAK1LRR to the C-terminal portion of FLS2LRR (Fig. 1A). The structural organization of FLS2LRR-BAK1LRR differs from the m-shaped homo- or heterodimeric TLRs (24). Nonetheless, the C-termini of BAK1LRR and FLS2LRR are similarly oriented, presumably pointing to the membrane surface.

Fig. 1 Ectodomains mediate the flg22-induced heterodimerization of FLS2 and BAK1.

(A) Overall structure of FLS2LRR-flg22-BAK1LRR. The positions of LRR3 and LRR16 are indicated by blue numbers. “N” and “C” represent the N and C terminus, respectively. Color codes are indicated. (B) Flg22 binds to a shallow groove at the inner surface of the FLS2LRR solenoid. FLS2LRR is shown in electrostatic surface (in transparency). The FLS2LRR-interacting residues in flg22 are shown stick. I, Ile; Q, Gln. White, blue, and red indicate neutral, positive, and negative surfaces, respectively. Detailed interactions of the left and right highlighted regions are shown in Fig. 2, A and B, respectively. (C) Structural comparison of the ligand-bound FLS2LRR with the free FLS2△LRR2-6. For clarity, the N- and C-terminal sides of the flg22-bound FLS2LRR are not shown. Numbers in blue indicate the positions of LRRs.

Interactions of flg22 with FLS2LRR can be divided into two parts separated by a kink (flg22 Asn10 and Ser11) in the central region of the peptide (Fig. 1B). Before the kink, the N-terminal seven residues bind to FLS2 LRR2 to LRR6 (FLS2LRR2-6) (Fig. 2A). Thus, deletion of these four LRRs would negate FLS2 interaction with the N- but not the C-terminal segment of flg22, phenocopying an flg22 variant with the N-terminal seven residues deleted (flg15) (4). Indeed, an FLS2LRR mutant with five LRRs deleted, FLS2ΔLRR2-6, still formed an flg22-induced complex with BAK1LRR (fig. S6). Structural superposition of the FLS2ΔLRR2-6 mutant with FLS2LRR bound by flg22 and BAK1LRR showed that the two structures are nearly identical (Fig. 1C), with a root mean square deviation = 0.43 Å over 543 Cα-aligned atoms, suggesting that, in the cellular milieu as well, conformational changes in FLS2LRR may not be necessary for flg22 binding and heterodimerization with BAK1LRR.

Fig. 2 Mechanism of flg22 recognition by FLS2.

(A) Interaction of the N-terminal portion (residues 1 to 7) of flg22 with FLS2LRR. The side chains FLS2LRR and flg22 are labeled in cream white and yellow, respectively. (B) Interaction of the C-terminal side (residues 8 to 21) of flg22 with FLS2LRR. FLS2 Gly318 is indicated in red. K, Lys; S, Ser; T, Thr. (C) FLS2LRR mutations reduce interaction with GST-flg22. GST-flg22 bound to GS4B agarose was used to precipitate various FLS2LRR wild-type (WT) and mutant proteins. The bound proteins were visualized by SDS–polyacrylamide gel electrophoresis (PAGE) with Coomassie blue staining. The assay was repeated three times. (D) FLS2 mutations compromise MPK phosphorylation. Null fls2 mutant mesophyll protoplasts were transfected with plasmids as indicated. The samples were separated into two and treated with water (-) or 1 μM flg22 (+). Immunoblots were analyzed by using antibodies against FLAG, BAK1, or pMPK. CBB, Coomassie brilliant blue. (E) FLS2 mutations attenuate flg22-induced FRK1::LUC expression. Null fls2 mutant Arabidopsis mesophyll protoplasts were transfected with plasmids as indicated along with 35S::R-LUC and FRK1::LUC. The FRK1::LUC activity was determined after protoplasts were treated with 1.0 μM flg22 for 10 min. IB, immunoblot.

Both hydrogen bonds and hydrophobic contacts mediate flg22 interaction with FLS2LRR. Flg22 Leu3 inserts into a hydrophobic pocket of FLS2 (Fig. 2A). In addition to hydrophobic contacts, FLS2 Arg152 and FLS2 Tyr148 also engage hydrogen bonds with flg22 Gln1 and flg22 Leu3, respectively. The two residues are highly varied in tomato FLS2 (fig. S5), which can adversely affect recognition of the N-terminal part of flg22 by the latter. This may explain the fact that flg15 displays a low activity in Arabidopsis but is fully active in tomato cells (23). The C-terminal 14 amino acids, particularly those after the kink, form denser interactions with FLS2LRR, burying a surface area of 1817 Å2 compared with 373 Å2 by the N-terminal seven amino acids. Consistently, flg15 still bound FLS2LRR (fig. S7A), agreeing with previous cell-based assays (3). Flg22 Asp14 and Asp15, important for FLS2 interacting (fig. S7A) and immunogenic activities (3), bind to two positively charged pockets (Figs. 1B and 2B). The hydrogen bonds formed between FLS2 Tyr272 and Tyr296 and flg22 Lys13 also contribute to the interactions around this interface. In contrast, flg22 Asn10 and flg22 Lys13 (whose side chain is not involved in interaction with FLS2LRR) are solvent-exposed (fig. S4), and their mutations generated little effect on the flg22 activity (3). Flg22 Leu19 and Ile21 bind to two neighboring hydrophobic pockets. Mutation of flg22 Ala17 that contacts FLS2 Thr366 underneath to tyrosine (Fig. 2B) reduced flg22 interaction with FLS2LRR (fig. S7A). All the amino acids critical for flg22 binding to FLS2LRR are conserved among FLS2-activating bacteria (fig. S8).

Further supporting our structural observations, a precipitation assay showed that FLS2 with two mutations, Arg294 and His316 to Ala (24), R294A/H316A, resulted in no FLS2LRR-flg22 interaction (Fig. 2C), phenocopying flg22 D14A (D, Asp) (3). FLS2 Thr342 is located immediately underneath the peptide (Fig. 2B). Mutation of this residue, but not the unrelated FLS2 Thr434, to the bulkier tyrosine abolished the interaction with GST-flg22. Consistently, FLS2 G318R (G, Gly) (Fig. 2B) caused by ethyl methanesulfonate–induced mutation fls2-24 does not bind flg22 (4). Mutations of other FLS2 residues from the interface also compromised interaction with GST-flg22 (Fig. 2C). The mutations affecting FLS2LRR recognition of flg22 disrupted flg22-induced FLS2-BAK1 interaction in Arabidopsis protoplasts (fig. S9). Furthermore, flg22-induced FLS2 and BAK1 phosphorylation (fig. S9), mitogen-activated protein kinase (MPK) activation (Fig. 2D) and expression of FRK1::LUC (Fig. 2E), a reporter gene induced by multiple PAMPs, were also attenuated when the FLS2 mutants were transiently expressed in null fls2 mesophyll protoplasts. Together, these results indicate that these residues are functionally important in the plant cell.

A cluster of bulky BAK1 amino acids, including Arg72, Tyr96, Tyr100, Arg143, Phe144, and Arg146, directly contacts with those from FLS2LRR23-26 (Fig. 3A), whereas BAK1Phe60 interacts with residues from FLS2LRR18-20. These FLS2-interacting residues are conserved between Arabidopsis BAK1 and tomato SERK3 (fig. S10). Supporting the structural observations, BAK1 F60A/F144A (F, Phe) and BAK1 Y96A (Y, Tyr), but not the negative control BAK 1D30Y, attenuated flg22-induced FLS2-BAK1 heterodimerization in the precipitation assay and in null mutant bak1-4 mesophyll protoplasts (Fig. 3B). These critical BAK1 amino acids were also important for flg22-induced FLS2 and BAK1 phosphorylation (Fig. 3, C and D).The activation of MPKs by flg22 seems differently affected by the BAK1 mutations, because the activation of MPK3 and MPK6 was slightly diminished 10 min after treatment with the peptide, whereas MPK4/MPK11 was not activated to a detectable level at the same time point (Fig. 3C). Moreover, unlike wild-type BAK1, these critical BAK1 mutants (Fig. 3E) only partially restored the expression of FRK1::LUC to null mutant bak1-4 mesophyll protoplasts. Consistently, the BAK1 L53A, V54Y, and Y96A (L, Leu; V, Val) transgenic lines were abolished or severely compromised in production of reactive oxygen species (fig. S11).

Fig. 3 Direct FLS2LRR-BAK1LRR interactions.

(A) The C-terminal side of FLS2LRR mediates its interaction with BAK1. M, Met; N, Asn. (B) Direct contacts of FLS2 and BAK1 are required for their interaction. (Top) The assays were performed as described in Fig. 2C except that FLS2LRR and BAK1LRR were analyzed by antibodies against His (anti-His). The assay was repeated for three times. (Bottom) Mutagenesis assays for BAK1 mutants in null bak1-4 mutant mesophyll protoplasts. FLAG- and hemagglutinin (HA)–tagged FLS2 and BAK1 constructs were coexpressed in WT Arabidopsis protoplasts. Co-IP assay was performed to detect FLS2-BAK1 interaction after treatment with (+) or without (-) 1.0 μM flg22. (C) Mutations of the FLS2-interacting residues in BAK1 compromise MPK phosphorylation. The assays were performed as described in Fig. 2D. (D) Mutations of the FLS2-interacting residues in BAK1 compromise FLS2 and BAK1 phosphorylation. Immunoprecipitated FLS2-FLAG was incubated in the presence of radioactive [32P]γ-ATP (adenosine triphosphate). Immunoblots were analyzed by using antibodies against FLAG or BAK1. In vitro phosphorylation is revealed by autoradiography (i.e., 32P). (E) FLS2-BAK1 interactions are important for flg22-induced FRK1 expression. The assays were performed as described in Fig. 2E. Error bars indicate SEM.

The C-terminal segment of flg22 bridges FLS2LRR and BAK1LRR (Fig. 1A), reminiscent of auxin that act as a molecular “glue” to connect its receptor with a signaling partner (25). Flg22 Gly18, conserved among the FLS2-activating bacterial flagellins but not in non–FLS2-eliciting bacteria (fig. S8), fits in the inner-curved loop (residues 52 to 54) of BAK1 but makes no contact with FLS2LRR (Fig. 4A). Flg22-BAK1LRR interaction is further stabilized by two hydrogen bonds formed between Flg22 Leu19 and BAK1 Thr52 and Val54. Limited by space, any other amino acids at flg22 Gly18 would generate steric clashes with the BAK1 loop and consequently attenuate their interaction. Supporting this hypothesis, the mutant peptides flg22 G18A and flg22 G18Y, although they had a comparable FLS2LRR binding activity with the wild-type peptide, exhibited a compromised and no activity of inducing FLS2LRR-BAK1LRR interaction, respectively, in the GST precipitation and Co-IP assay in protoplasts (Fig. 4B and fig. S7B). The mutant peptide flg22 G18A seemed to activate MPKs as efficiently as the wild-type flg22 (Fig. 4C), although it failed to induce an interaction and activation between FLS2 and BAK1 (Fig. 4D). Nevertheless, the generation of reactive oxygen species (ROS) in Arabidopsis wild-type Col-0 leaves induced by flg22 G18A and flg22 G18Y was modestly and strongly reduced, respectively (Fig. 4E and fig. S12). The phenotypes generated by the mutations of flg22G18 are reminiscent of the bik1 mutant that is substantially compromised in PAMP-induced resistance but not the flg22-induced MPK activation (26), indicating that downstream signaling is differentially affected by perturbations to the receptor complex. An flg22 mutant lacking the C-terminal two residues acts antagonistically with the wild-type peptide (3, 23). This deletion, although not disrupting interaction with FLS2LRR (Fig. 4B and fig. S7), exposes a free carboxylic acid that may perturb the flg22 Gly18–mediated FLS2LRR-BAK1LRR interaction.

Fig. 4 BAK1 recognizes the C-terminal side of the FLS2-bound flg22.

(A) A selective role of flg22 Gly18 in interaction with BAK1. FLS2LRR is shown in surface (blue), with flg22 (salmon) and BAK1LRR (green). The side chains from BAK1LRR and flg22 are shown in purple and yellow, respectively. (B) Gly18 is required for the flg22-induced FLS2LRR-BAK1LRR interaction but not for FLS2 binding in vitro and in null fls2 mutant mesophyll protoplasts. The assays were performed as described in Fig. 3B. (C) The Flg22 G18A mutation has little effect on MPK phosphorylation in mesophyll protoplasts. The assays were performed as described in Fig. 2D. (D) Gly18 is required for flg22-induced FLS2-BAK1 interaction and the complex activation in seedlings. The assays were performed as described in Fig. 3D. GFP, green fluorescent protein. (E) The flg22 G18A mutation modestly affects generation of ROS in planta. WT Arabidopsis leaves were treated with water, 100 nM flg22, or 100 nM flg22 G18A.

As observed for the TLR heterodimers (24), the flg22-induced FLS2-BAK1 complex does not seem to be homo-oligomeric for its activation. Ligand-induced homodimerization has been demonstrated in chitin elicitor receptor kinase 1 activation (27). Thus, ligand-induced homo- or heterodimerization appears to be a common mechanism of plant receptor kinases and TLRs (24) for signaling. TLR4 and its co-receptor, MD-2, homodimerize after lipopolysaccharide binding (28), presumably because of the lack of an intracellular signaling domain in MD-2. Similarly, homodimerization or oligomerization could be important for activation of those receptor kinases that form ligand-induced heteromers with receptor-like proteins lacking an intracellular kinase domain.

The sequential recognition of flg22 by FLS2 and BAK1 is required to form a signaling-active complex (Figs. 2 to 4), indicating that BAK1 acts as a co-receptor with FLS2. One feature of the mammalian co-receptors is their promiscuity in ligand binding (29). This may also hold true for BAK1 as a co-receptor, because it forms ligand-dependent heteromers with several RKs (13, 14). Indeed, a recent study showed that SERK1 and most likely BAK1 are co-receptors with BRI1 (30). The two interfaces between FLS2LRR and BAK1LRR seems collaborative, because mutations in either of them led to a great reduction or loss of FLS2-BAK1 interaction (Figs. 3 and 4). The direct FLS2LRR-BAK1LRR interface could be responsible for formation of the flg22-independent FLS2-BAK1 complex detected in some studies (9, 13, 15), but it remains possible that the interface is induced by BAK1 recognition of the FLS2-bound flg22. Future studies differentiating the two possibilities will help unravel the mechanism of flg22-induced FLS2-BAK1 activation.

Supplementary Materials

Materials and Methods

Figs. S1 to S12

Table S1

References (3138)

References and Notes

  1. Acknowledgments: We thank S. Huang and J. He at Shanghai Synchrotron Radiation Facility (SSRF) for assistance with data collection. This research was funded by State Key Program of National Natural Science of China (31130063) and the National Outstanding Young Scholar Science Foundation of China (31025008) to J.C.; Chinese Natural Science Foundation (31230007) and Chinese Ministry of Science and Technology (2011CB100700) to J.-M.Z.; and the Gatsby Charitable Foundation and the European Research Council to C.Z. A.P.M. is supported by a postdoctoral fellowship from the Federation of European Biochemical Societies. The coordinates and structural factors for FLS2LRR-flg22-BAK1LRR and FLS2ΔLRR2-6 have been deposited in Protein Data Bank with accession codes 4MN8 and 4MNA, respectively.
View Abstract

Navigate This Article