A copper-dependent redox-based hydrogen peroxide perception in plants

Abstract

Redox-related molecules, such as quinones and reactive oxygen species (ROS), are important signaling molecules for all living organisms. A plant-specific leucine-rich repeat receptor-like kinase (LRR-RLK) CANNOT RESPOND TO DMBQ 1 (CARD1), also known as HYDROGEN-PEROXIDE-INDUCED Ca²⁺ INCREASES (HPCA1), perceives both quinones and ROS, but the mechanism by which it distinguishes between these two types of signals remains unclear. Here, we determine the structure of the CARD1 ectodomain and uncover its unique features. Structural studies, coupled with genetics and biochemical analysis, demonstrate that previously identified unique cysteine residues are not essential for signal perception in CARD1. Interestingly, CARD1 harbors a copper ion on the surface of the ectodomain via histidine-coordination that is crucial for hydrogen peroxide signaling. Our work reports a unique copper-dependent redox perception in plants and provides insight into interactions between receptors and non-peptide stimuli during perception.

Plants use CARD1/HPCA1 to detect signals such as quinones and ROS, but how they distinguish them was unclear. Here, the authors show CARD1 uses a copper-based mechanism for H₂O₂ sensing, revealing a metal-dependent mode of ROS perception in plants.

Introduction

Plants’ ability to perceive external and internal signals via an extensive number of cell surface receptors reinforces their adaptation to a fluctuating environment. Among these receptors, the leucine-rich repeat receptor-like kinases (LRR-RLKs) represent one of the largest and most rapidly expanding gene families in the plant kingdom¹. These LRR-RLKs are highly diversified and are responsible for perceiving a wide range of signals, from developmental cues to pathogen recognition². While many plant LRR-RLKs perceive their cognate proteinaceous ligands, a notable exception is the brassinosteroid receptor BRI1, which perceives a steroid ligand³.

The CANNOT RESPOND TO DMBQ 1 (CARD1) receptor (also known as HYDROGEN-PEROXIDE-INDUCED Ca²⁺ INCREASES – HPCA1) was previously identified to perceive quinone compounds⁴, particularly the 2,6-dimethoxy-1,4- benzoquinone (DMBQ), as well as reactive oxygen species (ROS), such as hydrogen peroxide (H₂O₂)⁵. CARD1 features a unique cluster of cysteine residues (Cys395, Cys405, Cys421, Cys424, Cys434, and Cys436) at the C-terminal part of its ectodomain, raising the possibility that these residues may be involved in the perception of both quinones and ROS^4,5. Previous studies have demonstrated that mutation of cysteine residues Cys395 and Cys405 resulted in a loss of Ca²⁺ signaling in response to DMBQ. However, it remains unclear whether such an effect is due to the structural role of cysteine residues or their direct involvement in quinone perception⁴. In contrast, cysteine residues Cys421, Cys424, Cys434, and Cys436 have been implicated in H₂O₂ sensing, with disulfide bond formation between Cys421-Cys424 and Cys434-Cys436 proposed as crucial for H₂O₂ perception⁵. Notably, these residues do not appear to affect quinone sensing⁴.

To elucidate the distinct molecular mechanisms underlying quinone and H₂O₂ perception by CARD1, we determined the cryo-electron microscopy (cryo-EM) structure of CARD1 ectodomain (EctoCARD1) at 3.26 Å resolution. Our analysis revealed previously unrecognized features of EctoCARD1 and, contrary to previous analysis, demonstrated that cysteine residues are not mechanistically required for the perception of either quinone or H₂O₂. Instead, we identified a distinct copper (Cu) binding histidine cluster in the EctoCARD1 that is necessary for H₂O₂-induced signaling, but is not essential for quinone perception. Overall, this study delineates how CARD1 distinguishes between quinone and H₂O₂ signal perception, and offers structural insights into the LRR-RLK clade VIII-1 subfamily. It also provides evidence of metal ion coordination – specifically copper – in the LRR region of a plant receptor. Rather than relying on a typical cysteine-mediated, redox-based H₂O₂ perception, CARD1 employs a unique copper-dependent mechanism for H₂O₂ perception in plants.

Results

H₂O₂ perception by CARD1 and its phylogeny

We first validated the previous finding that CARD1 is required for H₂O₂-induced Ca²⁺ elevation, using mutants in Arabidopsis that express aequorin – a Ca²⁺ indicator based on bioluminescence (hereafter, WT-AEQ)⁴. Indeed, all card1 mutant alleles displayed abolished Ca²⁺ signaling in response to H₂O₂ (Supplementary Fig. 1a) while complementation assays restored the phenotype for low H₂O₂ (100 µM) and high H₂O₂ (5 mM) (Fig. 1a, b), confirming its functional requirement. Kinetic analysis revealed that H₂O₂ exhibited a higher apparent K_m (58.6 µM) compared to DMBQ (1.68 µM)⁴, highlighting different sensitivity profiles for ROS and quinones (Supplementary Fig. 1b). We also found that CARD1 undergoes signal desensitization after activation (Supplementary Fig. 1c, d). Subsequent stimulation with either DMBQ or H₂O₂ (Supplementary Fig. 1c) resulted in reduced responsiveness, whereas stimulation with ligands activating different receptors, such as DMBQ and flg22, did not (Supplementary Fig. 1d). Flg22 peptide is a well-known pathogen-associated molecular pattern (PAMP) molecule that is perceived by another LRR-RLK in Arabidopsis – FLAGELLIN SENSING 2 (FLS2) receptor. This further supports that CARD1 can perceive both DMBQ and H₂O₂. Next, we tested if ROS generated in the apoplast, such as by the NADPH oxidases (RBOHD and RBOHF), could be perceived by CARD1. Flg22 peptide elicits PAMP-triggered immunity that includes NADPH oxidase-mediated ROS production. WT-AEQ seedlings were challenged with 1 µM flg22 peptide and showed two distinct Ca²⁺ peaks (Fig. 1c) as previously reported⁶. Interestingly, in all card1 alleles, the second Ca²⁺ peak was absent (Supplementary Fig. 1e). If CARD1 sensed the ROS generated from RBOHD, flg22-induced Ca²⁺ phenotype in card1 should phenocopy rbohd. Indeed, the second peak was abolished in both card1 and rbohd mutants, implying that ROS generated by RBOHD during flg22 signaling is perceived by CARD1 in the apoplast (Fig. 1d), further highlighting that CARD1, in addition to quinones, is capable of perceiving ROS generated during plant immune signaling.

Fig. 1. CARD1 perceives apoplastic ROS and CARD1 phylogeny.

a–c [Ca²⁺]_cyt response of Arabidopsis seedlings WT-AEQ, card1-2 and complemented line (card1-2/pCARD1:CARD1) in response to (a) 100 µM H₂O₂, (b) 5 mM H₂O₂, and (c) 1 µM flg22. Stimuli were applied at 60 s. Data are mean ± SE (n = 4) for H₂O₂ treatments and (n = 6) for flg22. d [Ca²⁺]_cyt response of Arabidopsis seedlings WT-AEQ, card1-2, rbohd, and rbohf in response to 1 µM flg22. Stimuli were applied at 60 s. Data are mean ± SE (n = 6). rbohf was included as a negative control. e An unrooted phylogenetic tree of the CARD1 protein in land plants. Amino acid sequences were mined from the Phytozome database version 12.1 and full protein sequences were aligned using MUSCLE (default parameters). See methods for full descriptions. f,g [Ca²⁺]_cyt response of Arabidopsis seedlings WT-AEQ, card1-2, card1-2 lines expressing AtCADL1 and AtCADL3 under the control of CARD1 promoter (card1-2/pCARD1:AtCADL1 and card1-2/pCARD1:AtCADL3) in response to (f) 5 µM DMBQ, and (g) 5 mM H₂O₂. Stimuli were applied at 60 s. Data are mean ± SE (n = 6).

We next clarified the phylogeny of CARD1 in land plants and surveyed for CARD1 and CARD1-like (CADLs) homologs across land plants and the late divergent charophyte groups. Following manual curation, 405 CARD1/CADLs sequences, across 57 species, were identified (Supplementary Data 1) and are conserved throughout embryophytes. Phylogenetic analysis revealed two major subfamilies within Group VIII-1 of LRR-RLK family (VIII-1a and VIII-1b) which diverged early in embryophytes’ evolution (Fig. 1e). We further identified two distinct subgroups within LRR-RLK Group VIII-1a (VIII-1a1 and VIII-1a2). The split of the LRR-RLK VIII-1a to a1 and a2 subgroups did not occur until after the bryophytes (Fig. 1e). CARD1, along with CADL1 and CADL2, belongs to Group VIII-1a1; CADL3 is in Group VIII-1a2; while CADL4 through CADL7 are part of Group LRR-RLK VIII-1b (Fig. 1e). To establish whether the perception of quinones and ROS occurs in other Arabidopsis CARD1 homologs in Group VIII-1a, we performed complementation assays by expressing other Arabidopsis CARD1 homologs (AtCADL1 and AtCADL3) in card1-2 plants under CARD1 promoter. Both homologs could restore DMBQ and H₂O₂-induced Ca²⁺ signaling (Fig. 1f, g). These findings support the idea that quinone and H₂O₂ perception capabilities in CARD1 and CADL homologs are likely not unique to CARD1 alone, and that DMBQ and H₂O₂ perception could be performed by other homologs, depending on when and where each gene is expressed.

Overall EctoCARD1 protein structure

To gain further insights into how CARD1 perceives quinones and ROS, such as DMBQ or H₂O₂, respectively, we sought to determine the structure of CARD1 protein by employing previously optimized expression and purification strategies in Nicotiana benthamiana⁷. We also fused EctoCARD1 to maltose-binding protein (MBP) to increase the size of the EctoCARD1 to a respectable size (Supplementary Fig. 2, Supplementary Data 2) for cryo-EM analysis (Supplementary Fig. 3). The MBP region did not appear in the 2D classification particles, allowing us to reconstruct a cryo-EM map corresponding solely to the EctoCARD1 region (Supplementary Fig. 3, 4). We determined the structure of EctoCARD1 (PDB 9VBD) at 3.26 Å resolution (Fig. 2a, b, Supplementary Table 1). EctoCARD1 has two distinct regions: the N-terminal LRR region and the C-terminal domain, which is connected by a loop region containing conserved Cys residues (Cys421, Cys424, Cys434, and Cys436) (Fig. 2c, Supplementary Fig. 5). The fold of the C-terminal domain is highly similar to the SEA (Sperm protein, Enterokinase, and Agrin domain) (Supplementary Data 3, 4)⁸. The LRR and SEA-like domains interact extensively through hydrogen bonds and electrostatic interactions (Fig. 2d). Notably, the interaction surface is reminiscent of the ligand-binding site in other known LRR-RLKs, such as FLS2 or HAESA receptors³, suggesting that the SEA-like domain may play a role in modulating receptor activity. The LRR domain consists of 14 LRRs, adopting a characteristic twisted solenoid configuration resembling a horseshoe structure³. This region also includes a highly conserved cysteine-pair capping domain at both the N terminus (Cys53 and Cys62; S-S1 in Fig. 2b) and C-Terminus (Cys395 and Cys405; S-S2 in Fig. 2b)⁹.

Fig. 2. Cryo-EM structure of EctoCARD1.

a Overall structure of EctoCARD1 shown in four different orientations based on the cryo-EM map. b Structural model of EctoCARD1, highlighting N-glycans, disulfide bonds, and the histidine residues involved in copper binding (shown as stick models). The copper atom is shown as a brown sphere. c Schematic of EctoCARD1 domain architecture, with glycosylation sites, disulfide bond sites, and corresponding residue numbers indicated. d Molecular interface between the LRR and SEA-like domains with possible hydrogen bonds and salt bridges depicted as black and yellow dotted lines, respectively. e N-glycosylation at Asn298, showing possible hydrogen bonds between the sugar chain and the protein (black dashed lines). Cryo-EM maps are shown as gray mesh surfaces.

The role of glycosylation in LRR-RLKs has been well-documented, with notable examples including the tomato Cf-9 receptor¹⁰ and immune receptors in Arabidopsis¹¹ and recently for MIK2 receptor^12,13. Similarly, the cryo-EM map shows that at least eight asparagine residues in EctoCARD1 including Asn298 are modified by N-glycans. Notably, the card1-7 allele, which carries a mutation at the neighboring Ser297, fails to exhibit the DMBQ-induced calcium elevation phenotype⁴. This impairment may be linked to disruptions in local glycosylation patterns on the CARD1 protein, which could compromise its function. Indeed, the carbohydrate chain on Asn298 interacts with the SEA-like domain (Fig. 2e), indicating that glycosylation at this site plays an important role in retaining and stabilizing the EctoCARD1 structure.

Cys421-Cys434 and Cys424-Cys436 pairs are not required for signaling

The reactivity of cysteine residues is influenced by two key factors: (i) the accessibility of the cysteine residue within the protein structure, and (ii) the local electrostatic environment surrounding the cysteine, including factors, such as local pH. Higher pH (more alkaline), for instance, favors the thiolate anion form, thereby increasing susceptibility to oxidation^14–17. Previous studies have suggested that specific cysteine residues play roles in sensing DMBQ (Cys395 and Cys405)⁴ and H₂O₂ (Cys421, Cys424, Cys434, and Cys436)⁵.

We thus investigated the physical properties of these cysteines within the EctoCARD1 structure. Our structural analysis revealed that Cys395 and Cys405 are located in the C-terminal end of the LRR solenoid structure, while Cys421, Cys424, Cys434, and Cys436 are located in the loop region of the EctoCARD1 protein (Fig. 3a). Importantly, these cysteines form disulfide bonds in the structure of EctoCARD1. The disulfide bond between Cys395 and Cys405 likely forms a C-terminal cysteine-pair capping domain—a feature commonly seen in plant LRRs. Cysteine-pair capping domains are often associated with protein stability and function (Fig. 3a)^9,18. To examine the role of Cys395 and Cys405 in CARD1 signaling, we generated card1 complemented lines with C395S/C405S substitutions (CARD1^C395S/C405S). These mutants exhibited variable protein levels (Supplementary Fig. 6a). Similar to the single mutants of Cys395 and Cys405, the CARD1^C395S/C405S mutant showed additional bands on protein blots, suggesting possible protein degradation⁴ (Supplementary Fig. 6a). Transformants with higher protein expression exhibited weak responses to both DMBQ and H₂O₂ compared to WT-AEQ (Fig. 3b–d), suggesting that Cys395 and Cys405 substitutions may disrupt CARD1 structural integrity. This would in turn reduce the ability of CARD1^C395S/C405S to perceive quinones and H₂O₂, while increasing its susceptibility to endogenous proteases, leading to the accumulation of truncated CARD1^C395S/C405S.

Fig. 3. Disulfide bonds between Cys421-Cys434 and Cys424-Cys436 are dispensable for CARD1 H₂O₂ signaling.

a Structural model of EctoCARD1, highlighting disulfide bonds between cysteine pairs at the N- and C-terminal ends of the LRR domain (green) and between cysteines in the loop region (blue). Cryo-EM maps surrounding these disulfide bonds are shown as gray mesh surfaces. b–d [Ca²⁺]_cyt response of Arabidopsis seedlings WT-AEQ, card1-2 and three independent transformants of card1-2/CARD1^C39S5/C405S in response to (b) DMBQ, (c) 100 µM H₂O₂ and (d) 5 mM H₂O₂. Stimuli were applied at 60 s. Data are mean ± SE with the number of seedlings indicated in each panel. e–g [Ca²⁺]_cyt response of Arabidopsis seedlings WT-AEQ, card1-2 and 2 independent transformants of card1-2/CARD1^quadCYS in response to (e) DMBQ, (f) 100 µM H₂O₂ and (g) 5 mM H₂O₂. Stimuli were applied at 60 s. Data are mean ± SE (number of seedlings are indicated in each panel). h Possible redox state of cysteine residues in CARD1 and cysteine mutants, as predicted based on the determined EctoCARD1 structure and the CARD1 kinase domain structure predicted by AlphaFold. i Schematic of cysteine labeling workflow using PEG-Mal labeling. j CARD1 and cysteine mutant migration patterns after PEG-Mal labeling. Protein extracts from N. benthamiana leaves expressing a kinase-dead AtCARD1 protein (CARD1^K659R), either alone or in combination with cysteine mutations (C421S/C434S and C424S/C436S). Protein blot was probed with anti-CARD1 antibody. The experiments were repeated three independent times with similar results obtained.

Next, we investigated the role of Cys421, Cys424, Cys434, and Cys436 in quinone and ROS perception. In the EctoCARD1 structure, all four cysteines form disulfide bonds (Cys421-Cys434 (S-S3) and Cys424-Cys436 (S-S4); Figs. 2b, 3a). This disulfide bonding pattern is intriguing as it was previously suggested that the formation of disulfide bonds between Cys421-Cys424 and Cys434-Cys436 is important for ROS sensing⁵. To clarify the role of these cysteine residues, we initially checked their involvement in H₂O₂ signaling in the single cysteine mutants. However, no effect was observed on ROS perception at either low (100 µM) or high (5 mM) concentrations (Supplementary Fig. 6b, c). We then generated a quadruple cysteine mutant in the card1-2 background (card1-2/pCARD1:CARD1^quadCYS) in which Cys421, Cys424, Cys434, and Cys436 were mutated to serine, and tested its ability to complement Ca²⁺ signaling phenotype in response to H₂O₂ or DMBQ. The CARD1^quadCYS-expressing plants responded to quinone and low H₂O₂ (100 µM) stimuli, albeit at a slight reduction in Ca²⁺ bursts compared to WT-AEQ plants (Fig. 3e, f). Since the CARD1^quadCYS protein also displayed a faint additional band on the blot, it is likely that the protein stability is affected, although not as much as the CARD1^C395S/C405S protein (Supplementary Fig. 6a, d). At high H₂O₂ concentration (5 mM), Ca²⁺ signaling was similar between the transformants and WT-AEQ (Fig. 3g). Overall, genetic data suggest that Cys421, Cys424, Cys434, and Cys436 are unlikely to play an important role in CARD1 signaling upon the perception of quinone and H₂O₂.

Our structural analysis reveals that EctoCARD1 contains one reduced cysteine (Cys124) and four disulfide bonds (Cys53-Cys62, Cys395-Cys405, Cys421-Cys434, and Cys424-Cys436) (Fig. 3a). The predicted structure of CARD1 kinase domain by AlphaFold contains 3 reduced cysteine residues (Cys627, Cys694, and Cys883) (Supplementary Fig. 7a). Thus, CARD1 is estimated to have 4 reduced and 8 oxidized cysteine residues (Fig. 3h). To validate disulfide pairing in EctoCARD1 in vivo, we used a thiol-specific labeling approach using a 2 kDa polyethylene glycol-conjugated maleimide (PEG-Mal), which selectively reacts with free thiol groups and adds a PEG moiety per modified cysteine (Fig. 3i). To prevent further oxidation of cysteine thiols during sample preparation, proteins were acidified and denatured using a trichloroacetic acid (TCA)/acetone solution before labeling with PEG-Mal^19,20. PEGylation induces an SDS-PAGE mobility shift proportional to the number of PEG molecules added²¹. Unlabeled native CARD1 from untreated Arabidopsis showed a band at estimated molecular weight (102 kDa) while PEGylated CARD1 shifted to higher molecular weight (120 kDa) (Supplementary Fig. 7b). In N. benthamiana, we used the kinase-dead mutant of Arabidopsis CARD1 (CARD1^K659R) that does not induce cell death phenotype associated with CARD1 overaccumulation (Supplementary Fig. 7c) and the protein migrated similarly to native CARD1 from Arabidopsis (Supplementary Fig. 7b). Protein bands from CARD1^K659R with substitutions of Cys421 and Cys434 to Ser (CARD1^{K659R/C421S/C434S}) or Cys424 and Cys436 to Ser (CARD1^{K659R/C424S/C436S}) showed comparable migration to that of CARD1^K659R in unlabeled samples (Fig. 3j). However, in PEG-Mal labeling samples after TCEP reduction (+ TCEP + PEG-Mal), both PEGylated CARD1^{K659R/C421S/C434S} and CARD1^{K659R/C424S/C436S} migrated further than PEGylated CARD1^K659R, confirming that the PEG-Mal labeling assay provides sufficient resolution to distinguish one disulfide bond formation (or reduction) in CARD1 (Fig. 3j). Importantly, in samples without TCEP reduction prior to PEG-Mal labeling (- TCEP + PEG-Mal), both PEGylated CARD1^{K659R/C421S/C434S} and CARD1^{K659R/C424S/C436S} migrated at a similar distance to CARD1^K659R, implying that neither C421S/C434S nor C424S/C436S mutations influence the number of reduced cysteine residues in EctoCARD1 (Fig. 3j). Overall, our structural and biochemical data collectively support the idea that Cys421-Cys434 and Cys424-Cys436 pairs form disulfide bonds in vivo in untreated plants.

Copper ion coordination by the histidine triad is essential for CARD1 signaling

Given that cysteines do not appear to play an essential role in CARD1 signaling, we carefully searched for other structural features in EctoCARD1 that may explain how CARD1 can perceive ROS or quinones. Further inspection revealed a connected cryo-EM map between His199 and His222, indicating the presence of a metal ion (Fig. 4a). This possible metal-binding site, composed of histidine residues His197, His199, and His222 (forming a His-triad), appears to adopt a slightly distorted T-shaped coordination geometry (Fig. 4a). T-shaped metal sites involving histidine residues are commonly found in copper-containing proteins, such as lytic polysaccharide monooxygenase (LPMO)²², copper amine oxidase²³, and peptidylglycine α-hydroxylating monooxygenase²⁴, which utilize dioxygen (O₂) or H₂O₂ to perform enzymatic reactions.

Fig. 4. The His-triad in CARD1 protein is essential for proper quinone and ROS perception.

a Structural model of EctoCARD1 showing the copper (Cu)-binding site. Cryo-EM maps are shown at high and low thresholds (burgundy surface and gray meshes, respectively). The Cu-binding site is shown in two orientations. The inset provides a close-up view of the Cu-binding site, with copper depicted as a brown sphere. b Quantification of metal elements bound to CARD1 by ICP-MS. Data represent the technical mean ± SE (n = 5). The experiments were repeated three times using different batches of purified EctoCARD1, yielding similar results. c Sequence logos generated with WebLogo server, showing the conservation of histidine residues in the His-triad of land plants CARD1/CADLs. d Time evolution of the distance between the copper ion (either Cu⁺ or Cu²⁺) and the geometric center of the His-triad (His197, His199, and His222) in CARD1 during MD simulations. The Cu²⁺ ion rapidly dissociates from the His-triad shortly after the simulation begins. In contrast, the Cu⁺ ion remains stably coordinated to the triad throughout the simulation window, maintaining an average distance of 0.74 Å with a standard deviation of 0.12 Å over 300 ns. e–g [Ca²⁺]_cyt response of Arabidopsis seedlings WT-AEQ, card1-2, and T3 lines (line 1-1-1 to line 1-1-3) of card1-2/CARD1^HtriA in response to (e) DMBQ, (f) 100 µM H₂O₂ and (g) 5 mM H₂O₂. Stimuli were applied at 60 s. For DMBQ treatment, data are mean ± SE (n = 6 for WT and card1-2, n = 12 for transformant lines). For H₂O₂ treatments, n = 6 for all genotypes. h–j [Ca²⁺]_cyt response of Arabidopsis seedlings WT-AEQ, card1-2, and T3 lines (line 2-1-1 to line 2-1-3) of card1-2/CARD1^HtriA in response to (h) DMBQ, (i) 100 µM H₂O₂, and (j) 5 mM H₂O₂. Stimuli were applied at 60 s. For DMBQ treatment, data are mean ± SE (n = 6 for WT and card1-2, n = 12 for transformant lines). For H₂O₂ treatments, n = 6 for all genotypes. k [Ca²⁺]_cyt response of Arabidopsis seedlings WT-AEQ, card1-2, and T3 lines of card1-2/CARD1^HtriA in response to 1 µM flg22. Data are mean ± SE (n = 6). l Sequence conservation of CARD1 across angiosperms. Conservation scores of 19 CARD1 proteins from angiosperms are calculated by the CONSURF program and mapped onto the EctoCARD1 structure. m Surface view of the His-triad region (highlighted box in (l)). Histidine residues forming the His-triad are indicated. n Ribbon diagram of the His-triad region. The side chains around the His-triad are shown in stick representation.

The metal ion is positioned on the plane formed by three ligand nitrogen atoms from the His-triad (Fig. 4a). The cryo-EM map around the His-triad indicates that His197 is more flexible than the other two His ligands, potentially flipping out and altering the coordination geometry of the metal site to a linear configuration. Additionally, His199 forms a hydrogen bond with nearby Asp168, which electrostatically interacts with Arg447 in the SEA-like domain. Arg447 further engages in a cation-π interaction with His199’s imidazole ring. Further, His222 and Pro534 in the SEA-like domain show a van der Waals contact as do Trp166 with His197 and His199 (Fig. 4a). The surface of EctoCARD1 forms an exposed pocket that could accommodate small redox-active molecules, such as quinones or H₂O₂ (Supplementary Fig. 8a). To confirm that the copper ions are coordinated by the His-triad, we performed inductively coupled plasma mass spectrometry on purified EctoCARD1 protein. The results consistently showed copper as the most abundant metal in the sample (Fig. 4b, Supplementary Fig. 8b, c). We also confirmed that histidine residues within the triad are highly conserved in LRR-RLK Group VIII-1 (Fig. 4c) and that it is absent in other LRR-RLK groups (Supplementary Fig. 8d). Altogether, these data indicate that copper is the primary metal coordinated by the surface-exposed His-triad, which is highly conserved in CARD1.

The T-shaped geometry and the copper position on the plane formed by ligand N atoms imply that the Cu ion is likely in the reduced (Cu⁺) state, which may be oxidized to Cu²⁺ by H₂O₂. To characterize the binding affinity of a Cu ion to CARD1, we performed molecular dynamics (MD) simulations using the cryo-EM structure of EctoCARD1. Simulations were conducted for both Cu⁺- and Cu²⁺-bound EctoCARD1 systems in explicit solvent under periodic boundary conditions (Supplementary Table 2, Supplementary Data 5). The results revealed that the Cu⁺ ion remained stably bound to the His-triad throughout the extended 1 μs simulation period, whereas the Cu²⁺ ion rapidly dissociated from the binding site upon simulation initiation, even under explicit solvation (Supplementary Movie 1, 2). Time for Cu⁺ and Cu²⁺ localization within the His-triad was quantified and Cu⁺ significantly stayed within the His-triad in a longer arbitrary timeframe (Fig. 4d). These findings indicate that, in this His-triad geometry, Cu⁺ binds stably to CARD1, whereas Cu²⁺ is not comparably stabilized. The MD simulations strongly support the role of the His-triad as the Cu⁺ binding site, consistent with experimental observations and the coordination geometry reported in other copper-binding proteins^25,26.

To further substantiate the preference for Cu⁺ over Cu²⁺ from an electronic-structure perspective, we performed simplified quantum-chemical calculations using a cluster model of the His-triad. For each oxidation state, we quantified the strain energy required to constrain the complex to the MD-derived coordination geometry, relative to the corresponding anchored, relaxed minimum-energy structure (Supplementary Table 3, Supplementary Data 6). Enforcing the MD-observed geometry incurs a substantially larger energetic penalty for Cu²⁺ than for Cu⁺, yielding a strain-energy difference of approximately 3.25 kcal/mol in favor of Cu⁺. This preferential electronic-structure–level destabilization of Cu²⁺ within the His-triad geometry offers valuable insight into the mechanistic basis for the selective stabilization of the reduced ion and is consistent with our MD results.

To validate the functional importance of the His-triad and Cu⁺ in CARD1, copper-free CARD1 mutant was generated by substituting histidine residues in His-triad to alanine residues (H197A/H199A/H222A; HtriA). We first experimentally determined the structure of EctoCARD1^HtriA by cryo-EM (PDB 9VBV) at a resolution of 3.61 Å (Supplementary Fig. 9, Supplementary Table 1). Superimposing the structures of EctoCARD1 and EctoCARD1^HtriA revealed no significant structural differences between the two proteins (Supplementary Fig. 10), suggesting that the HtriA substitution did not affect the overall structure. Notably, a clear absence of density at the Cu-binding site in the cryo-EM map of EctoCARD1^HtriA, compared to EctoCARD1, confirms that the mutant protein is devoid of copper (Supplementary Fig. 10b, c). We then checked whether CARD1^HtriA could complement the stimuli-induced calcium elevation phenotypes in card1-2 plants. The His-triad mutant (card1-2/pCARD1:CARD1^HtriA) exhibited a severely impaired Ca²⁺ peak response to DMBQ in all 5 independent lines, irrespective of protein levels (Supplementary Fig. 11a), suggesting that the His-triad mutation significantly impairs DMBQ signaling (Fig. 4e, h, Supplementary Fig. 11b). Importantly, the His-triad mutants were not able to respond to low (100 µM) or high (5 mM) H₂O₂ as they phenocopied card1-2 in their response to H₂O₂, even in transformants with high protein expression (Fig. 4f, g, i, j, Supplementary Fig. 11c, d). Additionally, the mutants also phenocopied card1-2 in response to flg22 (Fig. 4k, Supplementary Fig. 11e). Overall, these data collectively imply that the His-triad, and thus Cu⁺ coordination, is crucial for signaling in CARD1, particularly for H₂O₂ perception.

To identify evolutionarily conserved regions of CARD1 across angiosperms, we mapped sequence conservation onto the EctoCARD1 structure (Fig. 4l, Supplementary Data 7). This analysis revealed that not only the His-triad, but also surrounding residues, are highly conserved (Fig. 4m, n), implying that this region is crucial for CARD1 function and has been slowly evolving in CARD1 homologs. Altogether, these data further support the importance of Cu⁺ ion coordination by the His-triad in CARD1 for H₂O₂ perception.

Discussion

CARD1 perceives quinone and H₂O₂ in plants and its homologs are widespread across the land plant lineage. Notably, multiple CARD1 homologs were identified in early-diverging land plants, such as the liverwort Marchantia polymorpha and the moss Physcomitrium patens. While gene family reduction is common in early land plants, the number of CARD1/CADLs homologs remains relatively unchanged, with eight in Arabidopsis and six in M. polymorpha. Paralog complementation by other CARD1-like genes suggests that the CARD1 gene family contains functional redundancy, similar to what has been observed for other LRR-RLKs, such as brassinosteroid receptor BRI1²⁷ and the peptide receptor CLAVATA1²⁸. The functional significance of this family may depend on the expression patterns of individual homologs. For instance, CADL3—capable of perceiving both ROS and quinone (Fig. 1e, f)—is primarily expressed in the vasculature²⁹. This suggests that certain CARD1 homologs may function in perceiving quinones or ROS produced in the vasculatures, potentially playing a role in vascular redox biology. Importantly, here we show that apoplastic ROS generated by RBOHD are perceived by CARD1 during immune signaling (Fig. 1c, d). Consistent with this finding, both card1 and rbohd mutants exhibit a reduction in stomatal immunity³⁰. CARD1 and RBOHD have also been shown to mediate ROS and calcium cell-cell signaling during stresses^31,32, further underscoring their importance in ROS biology and signaling. Interestingly, both CARD1 and NADPH oxidases appear to have emerged concurrently in early land plants³³. This co-evolution suggests that apoplastic ROS production and perception evolved together – a necessary adaptation as plants transitioned to land where ROS can no longer diffuse in the aqueous environment and require specific perception mechanisms.

Our cryo-EM structure of plant-expressed EctoCARD1 provided a reasonable representation of its possible conformational state in planta, and revealed several unique features. First, N-glycosylation at Asn298 stabilizes the intramolecular interaction between the LRR domain and the SEA domain-like fold (Fig. 2e). Second, the ectodomain of CARD1 features a cysteine-pair capping domain at both the N- and C-termini of the LRR domain, with an overall horseshoe shape typical of LRR-RLKs. However, it lacks an island domain, similar to other LRR-RLKs, such as FLS2, PEPR, and HAESA³. Instead, structural homology modeling using the DALI and Foldseek servers^34,35 revealed that the C-terminal end of EctoCARD1 shares significant similarity with SEA domain-containing proteins (Supplementary Data 3, 4). In animals, SEA domains are commonly found in various cell surface receptors and often exhibit autoproteolytic activity^8,36. The bacterial SEA-like domain is also involved in regulating intramembrane proteolysis via its autoproteolytic activity³⁷. Notably, SEA domains are typically located near the C-terminal end of the ectodomain, just before the transmembrane domain³⁶, which is also the case for CARD1 SEA domain-like fold. It is tempting to speculate that the SEA domain-like fold of CARD1 may undergo autoproteolysis, similar to other known mammalian SEA-domain containing receptors, thus modulating protein function in vivo.

Our analyses unequivocally demonstrate that Cys421, Cys424, Cys434, and Cys436 in CARD1 are not essential for signaling function. Further, our structural analysis revealed that these cysteine residues form disulfide bonds between Cys421-Cys434 and Cys424-Cys436, rather than between Cys421-Cys424 and Cys434-Cys436 as previously proposed⁵. The distance between Cys421 and Cys434 is more favorable for bond formation than between Cys421 and Cys424, although one cannot rule out the possibility that disulfide bonding may be exchanged between cysteines during ROS perception. In addition, the PEG-Mal assay validated disulfide bonding between Cys421-Cys434 and Cys424-Cys436 in unstimulated plants (Fig. 3j, Supplementary Fig. 7b). These disulfide bonds in the exposed loop region (Figs. 2a, 3a) may play a role in protecting CARD1 from proteolytic cleavage by endogenous proteases.

We identified a His-triad in the CARD1 ectodomain that coordinates a copper metal ion. This Cu⁺ coordination in plant LRR-RLKs underpins H₂O₂ perception as the loss of Cu⁺ in CARD1^HtriA showed no H₂O₂-induced Ca²⁺ signaling (Fig. 4e–k). Intriguingly, mutation of this site does not abolish DMBQ signaling (Fig. 4e, h) although the calcium signal returns to the resting level faster than in the wild-type. This suggests that while Cu⁺ is likely to subtly affect DMBQ perception, Ca²⁺ homeostasis and signal regulation may also be impaired. The conservation of histidine residues forming the Cu⁺ binding site across angiosperms supports the importance of Cu⁺ for CARD1 function (Fig. 4c, m). While copper is also involved in ethylene receptor function³⁸, the role of copper in CARD1 is distinct. Many plant redox regulation systems or sensors rely on cysteine-based redox couples or methionine-based oxidative post-translational modifications^14–17. The apoplastic environment, which is unique to plant cells and is more oxidized than the cytosol, may require distinct mechanisms for sensing redox-active molecules¹⁶. The Cu⁺-bound EctoCARD1 structure determined in this study likely represents an unstimulated state, as neither H₂O₂ nor quinones were added during sample preparation prior to structural determination. Under aerobic conditions in the apoplast, Cu⁺ is expected to be spontaneously oxidized to Cu²⁺, suggesting that a mechanism must exist to maintain extracellular Cu ions in the reduced Cu⁺ state. A potential source of reducing power for this is ascorbic acid which accumulates in the apoplast at concentrations on the order of 0.1 mM³⁹ and is capable of reducing Cu²⁺ to Cu⁺. Alternatively, plants may employ the ferric reductase oxidases FRO4 or FRO5 to mediate the reduction of Cu²⁺ to Cu⁺ in the apoplast⁴⁰. However, we cannot rule out the possibility that other proteins participate in Cu⁺-based sensing together with CARD1 in vivo. In addition, it remains unclear whether the absence of copper in the His-triad mutant of CARD1 affects post-translational modifications in vivo that could render the receptor nonfunctional.

We have successfully shown that Cu⁺ is necessary for CARD1 to perceive ROS and therefore demonstrate that CARD1 can discriminate between quinone and H₂O₂. However, the exact mechanism by which CARD1 distinctively perceives quinone versus H₂O₂ signals remains to be investigated. Although our in silico analysis indicates that oxidation of Cu⁺ by H₂O₂ results in the release of the Cu ion from the His-triad, this release alone is unlikely to trigger CARD1 activation, as the removal of the Cu ion did not significantly affect the EctoCARD1 structure (Supplementary Fig. 10). Our MD simulations and simple quantum chemical calculations further suggest that Cu⁺ rather than Cu²⁺ is coordinated within the His-triad. To accurately quantify binding affinities and resolve the electronic-structure differences between Cu⁺ and Cu²⁺ within the His-triad, future work should employ advanced quantum chemical calculations. Such calculations will also be essential for modeling the subsequent H₂O₂-driven redox steps at the active site. In addition, direct experimental approaches capable of distinguishing Cu oxidation states within the His-triad will be crucial for advancing our understanding of CARD1-mediated H₂O₂ perception mechanistically.

Based on these data, secondary signaling molecules are likely involved in activating CARD1 following Cu⁺ oxidation and we propose that CARD1 may act as an H₂O₂ signal modifier via its catalytic Cu⁺. We suggest two possible models by which Cu⁺ in CARD1 could mediate H₂O₂ signaling (Fig. 5, Supplementary Fig. 12). In one model, catalytic Cu⁺ facilitates the generation of Fenton reaction-derived hydroxyl radicals. These radicals can then react with apoplastic components, such as lignin or other yet-unidentified molecules, to generate secondary signaling molecules, including quinones. Because copper itself does not participate directly to generate copper intermediates, we refer to this as the ‘off-Cu’ model (Fig. 5). In an alternative model, signal generation may proceed via reactive copper-oxyl (^●O-Cu²⁺) intermediates (Supplementary Fig. 12; on-Cu model). Indeed, Cu ions are well known to function as cofactors for various metalloenzymes⁴¹. For example, LPMOs are copper-dependent enzymes that catalyze the oxidative cleavage of polysaccharide. Notably, LPMO utilizes H₂O₂ as a co-substrate, with Cu⁺-mediated homolytic cleavage of H₂O₂ being a key step in this enzymatic reaction⁴². Similarly, CARD1 may also use H₂O₂ as a co-substrate to generate quinones in the presence of a specific substrate, thereby activating the CARD1 signaling pathway. Whether EctoCARD1 can exhibit LPMO-like mechanisms will need to be further investigated. Overall, our work provides insights into how plasma membrane receptors can integrate metal ions for signal perception and lays the groundwork for further research into ROS sensing mechanisms in plants.

Fig. 5. Proposed working model for CARD1 function.

In the ‘off-Cu’ model, unstimulated CARD1 contains Cu⁺ coordinated by the His-triad. Upon stimulation with H₂O₂, electron transfer occurs from Cu⁺ to H₂O₂, oxidizing Cu⁺ to Cu²⁺. Concurrently, hydroxyl radicals (^●OH) and hydroxides (OH^-) are generated through a Fenton reaction between H₂O₂ and Cu⁺ in the apoplastic space. Due to their high reactivity, ^●OH radicals can oxidize various substrates (R-H), including lignin, leading to the formation of quinones. These quinones may be recognized by CARD1 through a distinct region and mechanism, independent of H₂O₂ sensing (uncharacterized binding region). Additionally, other proteins or co-receptors may facilitate quinone binding in vivo (uncharacterized co-receptor). For clarity, only the ectodomain is shown in the diagram. Dotted lines and circles denote hypotheses that remain to be tested in future studies.

Methods

Chemicals and reagents

Chemicals were purchased from Fujifilm Wako Pure Chemical (Fujifilm Wako Pure Chemical Cooperation) unless otherwise mentioned. The flg22 peptide (QRLSTGSRINSAKDDAAGLQIA, 95.3% purity) was custom-synthesized by Medical & Biological Laboratory and dissolved in endotoxin-free water.

Plant materials

All A. thaliana lines are Col-0 background. Transgenic Arabidopsis lines expressing (apo)aequorin (denote as WT-AEQ), card1-2, complemented CARD1 and CARD1 with single cysteine mutations were previously described⁴. Seeds of rbohd and rbohf were previously published⁴³. Seeds were surface-sterilized and stratified for 2 d before growing in liquid medium (1/2 MS, 0.25% sucrose, pH 5.7) at 20 °C–22 °C, 16-h photoperiod, 60–80 µE. Soil grown Arabidopsis plants were grown individually in a 6 cm diameter pot (1:1 mixture of soil and vermiculite) under the same conditions. For Ca²⁺ measurement experiments, liquid grown Arabidopsis seedlings were grown for 1-week post-germination. For experiments with soil-grown Nicotiana benthamiana, plants were grown for 4-5 weeks post-germination in individual pots (10 cm diameter with 1:1 soil and vermiculite) under control conditions (25 °C–26 °C, 16-h photoperiod, 100–150 µE). Where possible, plants were randomly selected from the larger pools that were grown in identical conditions. Different genotypes were randomly positioned in the growth chambers/shelves.

Cloning and transformation

Primers used in this study are provided in Supplementary Table 4. All sequences were verified by Sanger sequencing. To generate pHREAC-EctoCARD1-MBP and pHREAC-EctoCARD1^HtriA-MBP for cryo-EM studies, the CARD1 ectodomain (residues 1–546), containing a GSA linker and a codon-optimized maltose-binding protein (MBP) at its C-terminus (EctoCARD1-MBP), was cloned into the BsaI site of pHREAC vector⁴⁴ using In-Fusion HD cloning kit (Takara Bio). The EctoCARD1 part was cloned from genomic DNA whilst codon-optimized MBP was synthesized (Eurofin). The plasmid was electroporated into Agrobacterium tumefaciens C58C1 strain. See Supplementary Data 2 for the nucleotide sequence of EctoCARD1-MBP.

To generate pHREAC-AtCARD1^K659R, pHREAC-AtCARD1^{K659R/C421S/C434S} or pHREAC-AtCARD1^{K659R/C424S/C436S} for PEG-Mal labeling experiments in N. benthamiana, PCR primers were designed with a 15 bp overlapping sequence at their 5’ ends. Cysteine-coding codons and lysine-coding codon were replaced with serine- and arginine-coding codons, respectively. PCR was performed with full-length CARD1 cDNA sequence to yield mutagenized CARD1 cDNA fragments using KOD one (TOYOBO). These fragments were combined with linearized pHREAC vector using In-Fusion HD cloning kit. The plasmid was electroporated into A. tumefaciens C58C1 strain.

For complementation experiments in Arabidopsis with mutagenized CARD1, PCR primers were designed such that codons of the targeted amino acid were replaced with codons introducing desirable amino acid changes. Further, primers were also designed with a 15 bp overlapping sequence at their 5’ ends. PCR was performed with full-length CARD1 gDNA sequence to yield mutagenized CARD1 gDNA fragments using KOD one. Native CARD1 promoter (2.7 kb 5′ regulatory sequence of CARD1; pCARD1) fragment was amplified from Col-0 gDNA. Fragments were combined with either Bsp1407I-linearized pGWB4 binary vector⁴⁵, or with BamHI/XhoI-linearized pKI1.1 R binary vector⁴⁶. Since the BamHI/XhoI-linearized pKI1.1 R binary vector eliminates Cas9 and guide RNA scaffold coding regions, the vector effectively allows for red fluorescent seed selection in plants. The AtCADL1, AtCADL3, CARD1^C39S/C40S5, and CARD^quadCYS were cloned into Bsp1407I-linearized pGWB4 binary vector with 2-fragments In-Fusion Snap assembly cloning, while the CARD1^HtriA was cloned into BamHI/XhoI-linearized pKI1.1 R vector with 4-fragments In-Fusion Snap assembly cloning.

Protein expression and purification

The expression of CARD1 ectodomain was performed according to the previously described method⁷ with some modifications. Agrobacterium harboring pHREAC-EctoCARD1-MBP was grown overnight and suspended in 10 mM MES-NaOH (pH 5.6) and 10 mM MgCl₂ to OD₆₀₀ = 0.2. The suspension was infiltrated with a needleless syringe into leaves of soil-grown N. benthamiana. From three to five days-post-infiltration, infected leaves were excised and immersed in collection buffer (20 mM Tris-HCl (pH7.5), 0.01% (v/v) Tween 20, and 1 × cOmplete EDTA-free Protease Inhibitor Cocktail (Roche)). The collection buffer was infiltrated into the leaf tissue by reducing the pressure to 60 hPa for 10 min and then slowly releasing the pressure for 10 min. After removing buffer residues, the apoplastic fluid was collected by centrifugation twice (1500 g, 10 min, 4 °C) and subjected to anion exchange chromatography using a HiTrap Q HP column (Cytiva). Proteins were eluted by a linear NaCl gradient (from 50 mM to 300 mM in 20 mM Tris-HCl (pH7.5) and 0.1% (v/v) CHAPS over 15 column volumes). Fractions containing EctoCARD1-MBP were verified by immunoblot using anti-MBP antibody (M091-3; MBL). Peak fractions of EctoCARD1-MBP were pooled and concentrated using a Vivaspin 500 (10 kDa MWCO; Sartorius) and further purified by size exclusion chromatography on a Superdex200 Increase 10/300 GL column (Cytiva) in a buffer containing 24.5 mM Tris-HCl (pH7.5) and 150 mM NaCl. Fractions containing EctoCARD1-MBP were verified by SDS-PAGE followed by Coomassie staining (InstantBlue; abcam), pooled, and concentrated further using Vivaspin 500 (10 kDa MWCO). Protein samples were flash-frozen in liquid nitrogen and stored at −80 °C, in aliquots, until required.

Cryo-EM sample preparation

The WT and mutant protein samples were diluted by the purification buffer to be ~ 0.2 mg/mL. A QUANTIFOIL Cu R1.2/1.3 200 mesh grid (Quantifoil Micro Tools GmbH) was glow discharged by Auto Fine Coater JEC-3000FC (JEOL) at 20 mA and 7 Pa for 20 sec. Three µL of samples were applied to the grids. The grids were then blotted for 3 sec and rapidly cryocooled in liquid ethane by a Vitrobot Mark IV (Thermo Fischer Scientific) device. The sample chamber of Vitrobot was kept at 8 °C and 100% humidity.

Cryo-EM data acquisition and processing

All datasets were collected using a CRYO ARM™ 200 (JEOL) equipped with a K3 direct electron detector (Gatan) controlled by SerialEM⁴⁷ at the Graduate School of Pharmaceutical Sciences, The University of Osaka. All movies were collected at 60,000 magnification (pixel size: 0.83 Å) and the defocus range was from −0.7 to −2.2. Total electron exposure was 40 e⁻/Ų. Detailed conditions are shown in Supplementary Table 1.

Collected movies for the WT and the mutant samples were processed with cryoSPARC v3.2 and CryoSPARC v4.5.1, respectively^48,49. Collected movies were imported and motion corrected. Contrast transfer function (CTF) was estimated by a patch CTF estimation job and particles were picked using a Blob Picking job. Selected particles were extracted. 2D classification was performed to remove junk particles. Ab-initio map reconstruction was performed by particles selected by 2D classification. Several refinement jobs including CTF refinement and aberration correction⁵⁰ were performed to estimate accurate particle orientations and to obtain better cryo-EM maps. The detailed workflows are shown in Supplementary Figs. 3 and 9.

Model building and validation

An initial model of EctoCARD1 was predicted by ColabFold⁵¹. The model was manually fitted to the map on ChimeraX⁵². Fitted models were refined by Phenix Real-space Refinement⁵³. Obtained models were further modified and carbohydrate chains were built manually in Coot⁵⁴. During model building, we noticed that the electrostatic potential map beyond residue 546 appears significant. However, we did not construct atomic models in this density for the following reasons. Based on the overall shape and position of this unassigned density, we consider it most likely to originate from the N-glycan attached to Asn510. Nevertheless, given its close proximity to the observed C-terminus, we could not entirely exclude the possibility that part of the density corresponds to the flexible linker connecting the EctoCARD1 domain and the MBP tag. Owing to the limited resolution of the current reconstruction, this region could not be modeled with sufficient confidence. Real-space refinement and manual model building were repeated. Final models were validated by MolProbity⁵⁵. Models and cryo-EM maps were visualized by PyMOL (The PyMOL Molecular Graphics System, Version 2.4.0 Schrödinger, LLC.) and ChimeraX to draw figures.

Whole-seedling aequorin luminometry

The measurement of cytosolic free calcium ([Ca²⁺]_cyt) in response to stimuli was performed as previously described⁴. In brief, liquid-grown seedlings were individually placed in wells of white 96-well plates (Lumitrac flat bottom, Greiner Bio-One) with coelenterazine solution (Biosynth) overnight. The next day, luminescence values were recorded in each sample using either a Tristar² LB942 modular or a Tristar 3 multimode microplate readers (Berthold Technology). After baseline luminescence values were recorded for 60 s, 100 μl of stimuli was injected (two times the final concentration). Luminescence values were recorded every 5 s for 500 s. Discharge solutions (10% (v/v) ethanol, 1 M CaCl₂) were injected and the luminescence was recorded for a further 60 s. Calibration to convert luminescence values to [Ca²⁺]_cyt was performed. Seedlings that were poorly grown and confirmed not to be genetically related to the genotype and/or possibly due to sterilization method used, were omitted from the study. Plants that did not express sufficient amounts of aequorin, possibly due to silencing, were omitted from the study.

CARD1 immunoblot

To detect CARD1 protein in Arabidopsis, seedlings were harvested, snap-frozen in liquid N₂ and stored in –80 °C. Samples were ground to a fine powder with zirconia beads (3-mm diameter) with either a FastPrep-24 or a FastPrep-5G sample shakers (MP Biomedicals). Proteins were extracted in extraction buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 10% glycerol, 2 mM EDTA, 5 mM DTT, 1 × EDTA-free cOmplete protease inhibitor cocktail, 0.1% IGEPAL CA-630, 0.5 mM phenylmethane sulfonyl fluoride) by incubating the samples for 30 min at 4 °C. Insoluble debris was pelleted by centrifugation (10 min, 4 °C, 20,000 g). Equal amounts of protein were loaded on to an acrylamide gel and subjected to SDS-PAGE (Tris-Glycine-based). Proteins were then electroblotted onto a PVDF membrane using a trans-blot turbo transfer system (Bio-Rad) according to the manufacturer’s instructions. Membranes were blocked overnight at 4 °C in TBS-T (20 mM Tris, 150 mM NaCl, 0.1% (v/v) Tween 20, pH 7.5) containing 5% (w/v) skimmed milk powder. To detect CARD1, we incubated the blot with an anti-CARD1 peptide antibody (1:5000)⁴ for 1 h at room temperature with shaking. Blots were then further incubated for 1 h with peroxidase-linked anti-rabbit IgG (1:10,000; NA934, GE Healthcare). Luminescence signal was detected using Clarity Max Western ECL substrate (Bio-rad) according to the manufacturer’s instructions with either a LAS 4000 system or a LAS4010 system (GE Healthcare). Equal protein loading was verified by staining the detected membrane blot with Coomassie blue solution (0.1% (w/v) CBB R250 in 5:4:1 (v:v:v) mixture of H₂O, methanol, acetic acid).

Protein expression in Nicotiana benthamiana and PEG-Mal labeling

Selective labeling of the thiol groups was carried out as described²¹ with minor modifications. Samples were obtained from either Arabidopsis seedlings or from N. benthamiana leaves transiently expressing AtCARD1 variants. Samples were ground into a fine powder with liquid nitrogen and then homogenized in 10% trichloroacetic acid in acetone. The homogenate was incubated at −20 °C for 1 h and centrifuged (20,000 g for 10 min at 4 °C). The supernatant was discarded, and the pellet was washed twice with ice-cold acetone. The pellet was resuspended in resuspension buffer (50 mM HEPES-NaOH (pH 7.0) and 2% SDS) containing either 10 mM 2k PEG-maleimide (2k PEG-Mal; SUNBRIGHT ME-020MA, NOF) or 40 mM tris (2-carboxyethyl) phosphine (TCEP; 203-20153, FUJIFILM Wako Pure Chemicals) for 90 and 30 min, respectively. After centrifugation (20,000 g for 10 min), the supernatant was collected. A 100 µL aliquot from the TCEP-treated sample was transferred to a new tube, and proteins in the sample were precipitated using chloroform/methanol⁵⁶. The pellet was washed twice with ice-cold acetone and then resuspended with resuspension buffer containing 10 mM 2k PEG-Mal as described above. The samples were mixed with 2× SDS-PAGE sample buffer (without reducing agent) and subjected to immunoblot analysis.

Inductively coupled plasma time-of-flight mass spectrometry (ICP-TOF-MS)

Determination of metal ions bound to EctoCARD1 was performed using the OptiMass 9600 (GBC Scientific Equipment Pty Ltd). The instrument was tuned using a 10 ng/mL multi-element standard solution (Agilent). The concentrations from 0.1 to 10 ng/mL of various standard metal solutions diluted in buffer solution (1 mM Tris-HCl, 10 mM NaCl, pH 7.0) were prepared separately by dilution of standard stock solutions (copper (cat. number 033-16201), zinc (cat. number 266-02341), manganese (cat. number 130-19461), cobalt (cat. number 035-26151) and nickel (cat. number 141-10151); all from FUJIFILM Wako Pure Chemical) with 2% HNO₃. The instrument was optimized and calibrated to measure the concentration of different metal ions individually. Optimized operating conditions are described in Supplementary Table 5. EctoCARD1 protein (1.5 mg/mL to 2.1 mg/mL) samples were diluted and acidified with 2% HNO₃ prior to analysis by ICP-TOF-MS. SEC buffer (24.5 mM Tris-HCl pH 7.5, and 150 mM NaCl) were prepared in the same way as EctoCARD1 protein and analyzed by ICP-TOF-MS and used as control. Intensity of the control was subtracted from EctoCARD1 samples to obtain relative metal concentrations in EctoCARD1.

Molecular dynamics simulations

The stability of copper ion binding was assessed through molecular dynamics (MD) simulations. The cryo-EM-derived three-dimensional structure of EctoCARD1 was used as the target protein for the copper ion. Computational models for the MD simulations were created by manually positioning a copper ion (Cu⁺ or Cu²⁺) within the histidine triad composed of His197, His199, and His222 in the EctoCARD1 protein. These models were placed in a periodic boundary water box with dimensions of 65 × 97 × 112 Å⁵⁷. To maintain charge neutrality, sodium counterions were added – nine to the Cu⁺ system and eight to the Cu²⁺ system. The resulting simulation boxes contained 57,966 atoms (Cu⁺) and 57,971 atoms (Cu²⁺), including 16,636 and 16,638 water molecules, respectively, plus the protein and ions. Protonation states were assigned using PROPKA⁵⁸, with standard states applied for all residues: Asp and Glu as anions, His as neutral, and Lys and Arg as cations. The simulations ran for up to 1 μs with a 2 fs time step under constant pressure and temperature (NPT) conditions of 1 atm and 300 K, while maintaining a constant number of particles. Temperature was incrementally increased from 0 to 300 K and maintained at 300 K using a Langevin thermostat⁵⁷. The particle mesh Ewald (PME) method⁵⁹ was used to take into account the non-bonding interactions, while the SHAKE method⁶⁰ enforced distance constraints on hydrogen-containing bonds. TIP3P⁶¹ and ff14SB⁶² were used for water molecule and protein force fields, respectively. The Li/Merz monovalent ions force field⁶³ was applied to the Cu⁺ ion, while the Li/Merz highly charged ions force field⁶⁴ was used for the Cu²⁺ ion. To ensure statistical robustness and minimize sensitivity to starting configurations, three independent MD trajectories were run for each oxidation state (Cu⁺ and Cu²⁺): one 1 μs and two 400 ns, each initiated with distinct random velocity distributions. For Cu⁺, the initial 100 ns of each trajectory were treated as equilibration and excluded from analysis; the remaining segments constituted the production runs. For Cu²⁺, the ion dissociated within the first 100 ns in all trajectories; accordingly, no equilibration segment was defined. All MD simulations were performed using the AMBER 2022 program package⁶⁵.

Quantum chemical calculations

Simplified quantum-chemical calculations were performed on a cluster model comprising the Cu ion, the His-triad (His197, His199, and His222), and one coordinated water molecule, extracted from representative MD snapshots of the Cu⁺-bound coordination geometry. All calculations were carried out at the uB3LYP⁶⁶/def2-SVP level with empirical dispersion corrections (GD3BJ⁶⁷). To assess intrinsic structural stability under protein-like constraints, we computed the strain energy as shown in Eq. (1),

$$E_{\mathrm{strain}}=E_{\mathrm{constrained}}-E_{\mathrm{relaxed}}$$ 1

where $E_{\mathrm{constrained}}$ is the single-point energy of the MD-derived geometry and $E_{\mathrm{relaxed}}$ is the energy after constrained geometry optimization, in which the histidine C_β atoms were fixed to approximate the restrictions imposed by the protein scaffold. All quantum-chemical calculations were performed using the Gaussian16 program package⁶⁸.

CARD1 sequence mining and phylogenetic analysis and bioinformatics

Sequences of CARD1 homologs from other plant species were obtained from Phytozome (database version 12.1) as previously described⁴. Briefly, full-length Arabidopsis CARD1 amino acid sequence (At5g49760) was used as a query for blastP against the database with the cut-off E-value of 1 ×10⁻¹⁰⁰ and hit sequences with ≥ 80% query coverage were selected. In addition to previously published Phtheirospermum japonicum and Striga CADL homologs⁴, a total of 405 CADL sequences were obtained, spanning across 57 land plant species (Supplementary Data 1). Amino acid sequence alignment was performed on full-length sequences using MUSCLE (default parameters)⁶⁹ and the alignment were further processed using trimAl with ‘Gappyout’ automated trimming method⁷⁰. A phylogenetic tree was inferred from a trimmed alignment with RAxML⁷¹, using the PROTGAMMAAUTO model, and rapid bootstrapping with the autoMRE bootstrapping, using a random number seeds for a full maximum-likelihood tree search. The resulting estimated phylogeny was viewed as unrooted tree using iTOL⁷². His-triad conservation in land plant species were illustrated with Weblogo⁷³. For Consurf analysis, CARD1 homologs were identified from the NCBI non-redundant protein sequences database, using Arabidopsis CARD1 amino acid sequence, with a blastP search. Amino acid sequences of CARD1 homologs from 17 plant species (Supplementary Data 7) were aligned using Clustal Omega⁷⁴. Conservation scores were calculated with ConSurf server⁷⁵ and mapped to the EctoCARD1 structure determined in this study.

Data presentation and statistics

Data and graphs are presented and analyzed using GraphPad Prism 10 (Dotmatics) or Excel (Microsoft).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

N.I., Y.F., K.J.F., H.I., M.N. and A.L. conceptualized and designed the experiments; M.N., T.Y., T.I., K.S. and A.L. supervised the project; N.I., Y.F., K.J.F., T.Y., T.I., K.S. and A.L. acquired funding; N.I., Y.F., Y.S., K.T., R.H., K.J.F., H.T., M.N. and A.L. conducted experiments; N.I., K.T. and R.H. performed protein expression, purification, and biochemical analyses; Y.F. conducted cryo-EM sample preparation, analysis, and interpretation; Y.S., R.H. and A.L. performed genetic analyses, calcium measurements, and mutant generation; N.I., Y.F. and A.L. performed bioinformatics analyses; K.J.F. and T.Y. performed computational calculations and interpretation; H.I. and M.N. performed and analyzed ICP-MS; N.I., Y.F., K.J.F. and A.L. wrote the paper; A.L. reviewed and edited the manuscript with all authors’ input.

Peer review

Peer review information

Nature Communications thanks Bin Wu and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data in this study are available in the article or in supplementary information. The coordination files and cryo-EM maps have been deposited in the Protein Data Bank and the Electron Microscopy Data Bank (EMDB) under the accession codes PDB 9VBD / EMD-64920 [https://www.ebi.ac.uk/pdbe/entry/emdb/ EMD-64920] for WT and PDB 9VBV / EMD-64934 for HtriA. Raw data of cryo-EM have been deposited in the Electron Microscopy Public Image Archive (EMPIAR) under the accession codes EMPIAR-13383 (WT) and EMPIAR-13384 (HtriA). MD simulations and QM calculations input files are provided as Supplementary Data. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-72573-8.