Acclimation of Photosynthesis to the Environment 1 acts as a copper-binding superoxide detoxification enzyme

Abstract

Plant acclimation is a growing scientific concept, at molecular, cellular and global scales. All photosynthetic organisms that created an oxic atmosphere on earth possess a gene of unknown function “Acclimation of Photosynthesis to the Environment 1”. Here we show that APE1 encodes a thylakoid-bound protein with a unique motif that binds copper and detoxifies the superoxide anion radical, O₂^•−. Maturation of the recombinant APE1 protein from Chlamydomonas reinhardtii requires formation of cysteine disulfide bonds after copper binding or via a high affinity interaction with a copper chaperone (Plastid Copper Chaperone 1) that boosts its scavenging capacity for O₂^•−. APE1 co-occurs in evolution with Photosystem II oxygen evolving proteins and it is the archaic O₂^•− detoxifying enzyme for acclimating photosynthesis to an oxygenic environment.

APE1 co-evolved with Photosystem II as an ancestral enzyme that contributed to the Great Oxidation Event. This thylakoid-bound copper protein detoxifies O₂^•⁻, playing an essential role in photosynthetic acclimation to high light in oxic environments.

Introduction

Over time, vegetation responds to environmental changes, such as light, temperature and CO₂ availability, by adjustments to its physiology. This is referred to as plant acclimation, and it is a growing concept in science, in various fields, from the molecular scale¹ to the global level². Acclimation of photosynthesis is now implemented in land surface models (LSM) with an aim at deriving a more general response of vegetation across plant functional types, thereby simplifying the parametrization of the models^3–5. The trait “Acclimation of Photosynthesis to the Environment 1” (APE1) was identified in genetic screens in the angiosperm Arabidopsis thaliana and the green algae Chlamydomonas reinhardtii^6,7 where mutants could not maintain Photosystem II (PSII) quantum yield when exposed to a higher light intensity. Various proteomics studies have found APE1 linked to the D2 module during PSII biogenesis^8–10, but its Domain of Unknown Function (DUF) 2854 has remained undefined. Nevertheless, the strict conservation of APE1 in all oxygenic phototrophs⁶ and its involvement in light acclimation suggest an ancestral role in PSII maintenance and photoprotection.

Oxygenic photosynthesis produces dioxygen (O₂) from water splitting at the level of PSII. Along the photosynthetic electron transport chain, strongly reducing reaction intermediates can reduce O₂ into the superoxide anion radical (O₂^•−), at the level of Photosystem I¹¹, Photosystem II^12,13 or the cytochrome b₆ f complex¹⁴. O₂^•− is a reactive oxygen species (ROS) that is toxic to living cells. Superoxide dismutases (SOD) detoxify O₂^•−, working at diffusion-limited rates to produce O₂ and H₂O₂ from 2 molecules of O₂^•−15. The soluble Fe-SOD isoforms in the stroma of chloroplasts, conserved across species, are the first step of the “water-water” cycle to detoxify the O₂^•− produced by PSI¹⁶. However, detoxification at other O₂^•− production sites along the thylakoid membrane remains virtually unknown. Enzymatic O₂^•− scavenging via the protein Antioxidant 1 (ATX1) and related homologs^17–19 has also been reported in yeast, plant cytosol and animals.

In this work, we show that APE1 has its origins at the beginnings of oxygenic photosynthesis. To find the function of the DUF2854, we produced the recombinant versions of the APE1 orthologues from three very different photosynthetic organisms, the cyanobacterium Synechocystis sp. PCC 6803 (slr0575, SyAPE1), the green algae Chlamydomonas reinhardtii (Cre16g665250, CrAPE1), and the angiosperm Arabidopsis thaliana (At5g38660, AtAPE1). We show that APE1 binds copper in vitro. Via site-directed mutagenesis of the active and regulatory sites of APE1 from Chlamydomonas (CrAPE1), we reveal its complex dual maturation pathway. We show that the Plastid Cu Chaperone 1 (PCC1) transfers copper to CrAPE1, and provide evidence that the two proteins act as plastid antioxidants, cooperatively participating in acclimation of the photosynthetic apparatus to high light via their O₂^•− scavenging activity. Our results suggest that APE1 would have played an important role in the early steps of the evolution of the oxygenic photosynthetic apparatus, protecting it from its potentially harmful by-product.

Results

APE1 has an evolutionary link to PSII and is found in all oxygenic phototrophs

APE1 is part of the GreenCut group of genes, which are exclusively conserved in all oxygenic photosynthetic organisms, from cyanobacteria to plastid-containing eukaryotes^20,21. Comparative genomic analyses also identified APE1 as part of the cyanobacterial core genome, which consists of only 63 genes shared amongst all oxygenic photosynthetic organisms²². We extended this comparison using 77 cyanobacterial genomes, including those that have lost PSII and the genes required for carbon fixation^23,24. Through this cross-analysis, APE1 was restricted to PSII-containing cyanobacteria in a group of nine genes with a similar co-occurrence that included genes coding for PSII core subunits and the O₂-evolving complex (Table S1). This early appearance and conservation of APE1 in photosynthetic organisms suggests that APE1 played a pivotal role in the emergence of O₂ evolution.

The membrane-bound APE1 protein has a domain of unknown function

In all species, APE1 has 2 transmembrane helices (TMH). In eukaryotes it has a chloroplast transit peptide (Fig. 1A). The soluble stromal 15–17 kDa domain annotated as unknown function (DUF2854) has a conserved motif DVTR(Y/H)RYGDE(A/Q)HL(D/E) (inset, Fig. 1A). It is rich in charged amino acids (His, Asp and Glu), reminiscent of a metal-binding site or an enzymatically active site²⁵. D152, Y158, and H163 (CrAPE1 numbering) are conserved in all the orthologues. The majority of prokaryotes (Sy) have no cysteines, while unicellular eukaryotes (Cr) have up to 2 Cys and more complex, multicellular organisms (At) up to 3 Cys (Fig. S1). Like for many other proteins, an increase in Cys number in APE1 over the course of evolution suggests additional roles related to function, regulation, oligomerization or interactions²⁶. The cysteine C128 is conserved throughout the green lineage and in some diatoms (Fig. S1).

Fig. 1. Proteins APE1 and PCC1 are required for light acclimation of photosynthesis.

A CrAPE1 protein model (Chlamydomonas) contains a chloroplast target peptide (CTP), two thylakoid transmembrane helices (TMH) and the DUF2854 (gray shading), notably containing cysteines at positions 128 and 247 and the consensus sequence around histidine 163. B Plastid Copper Chaperone 1 (PCC1) protein model (Chlamydomonas) shows the heavy metal binding domain HMBD (Pfam PF19335; gray shading) and its consensus sequence between PCC1 orthologues. C Growth tests and imaging of photosynthetic activity (Photosystem II quantum yield, ϕPSII) under low light and high light in phototrophic conditions. Mutants of pcc1 (bold) and copper transporters ctp2 and ctp4 are compared against their corresponding WT strains and two ape1 mutants (bold).

Plastid Copper Chaperone 1 (PCC1) is co-expressed with CrAPE1

To provide context for APE1 function, data-mining of the top 250 genes co-expressed with Cre16.g665250 (obtained from PhytoMine against Chlamydomonas genome V5.6) showed high representation from GreenCut proteins targeted to the plastid²⁰. These factors could be grouped into functional clusters involved in assembly or repair of photosynthetic complexes, thylakoid remodeling, alternative electron pathways, carbon metabolism, redox and antioxidant regulation, pigment biosynthesis, and cofactor/chaperone activity (Fig. S2). These same functional groups could be found in APE1 co-expression data from Synechocystis sp. PCC 6803 (ALCOdbCyano) and Arabidopsis (ATTED-II v12.0), establishing the case for a similar function across very diverse photosynthetic organisms (Fig. S2). Among the genes co-expressed with CrAPE1, we found Cre05.g248600, annotated as Plastid Copper Chaperone 1 (PCC1)²⁷. This gene model contains a predicted plastid target sequence and the Pfam Heavy Metal Binding Domain (HMBD, PF19335), a conserved Cu-binding motif MXCXXC (Fig. 1B). Orthologues of CrPCC1 may only be retained in the Chlamydomonales (Chlamydomonas and Volvox), but other chlorophytes and diverse bacteria have a protein with significant sequence similarity (Fig. S3). PCC1, as well as its closely related proteins, showed high structural homology to the ATX1-like Cu chaperones, such as CopZ, ATX1, and Domain 1 of the Copper Chaperone to CuZnSOD (CCS), containing the typical ferredoxin-like fold (βαββαβ) (Fig. S4), the HMBD and a number of conserved lysines (K).

CrAPE1 and PCC1 maintain high PSII quantum yield in high light

We measured the acclimation from low light to high light (Fig. 1C) of Chlamydomonas mutant strains ape1 and pcc1, and of ctp2 and ctp4 that are devoid of P-type ATPase Cu transporters localized to the chloroplast envelope and the thylakoid membrane, respectively. All strains have insertions in exons of the genes (Fig. S5). The ctp2 and ctp4 mutants have low amounts of the Cu-containing electron carrier plastocyanin (as inferred by flash-dependent P₇₀₀ oxidation kinetics in Fig. S6), resulting in a lower PSII yield due to a restriction in downstream electron transport in both low and high light. While the residual electron transport is sufficient to sustain photoautotrophic growth in low light, over-reduction of the intersystem chain downstream of PSII likely causes ROS production and hence photodamage in high light, impairing cell growth (Fig. 1C). Interestingly, plastocyanin levels are only slightly decreased in pcc1 (Fig. S6), with a minor effect on PSII yield in low light, but its growth and PSII yield are still strongly impaired in high light. Hence, while CTP2 and CTP4 are required for plastocyanin accumulation, PCC1 is likely involved in a different pathway in the chloroplast, such as Cu sequestration, Cu transport or antioxidant activity, as reported for ATX1¹⁹. The ape1 mutant also presents normal plastocyanin levels and normal growth in both light intensities, but a strongly reduced PSII yield in high light. The co-expression profiles and the similar responses of the mutants to high light suggested that PCC1 and APE1 might be involved in the same stress acclimation mechanism.

APE1 is a copper-binding protein

We expressed the soluble region of the domain of unknown function, DUF2854, of Synechocystis (Sy), Chlamydomonas (Cr) and Arabidopsis (At) as recombinant proteins in E. coli (Figs. S1, S7 and Table S2). We tested APE1’s interaction with divalent cations using nano differential scanning fluorimetry (nanoDSF). Protein fluorescence was quenched and red-shifted upon Cu addition (Fig. S8). This was observed as an increase in the 350/330 nm fluorescence ratio for CrAPE1 (Fig. 2A), AtAPE1 and SyAPE1(Fig. S9A–F), and lowered the melting temperature (T_m) (Figs. 2B, S9G–I). The changes in protein fluorescence were specific to Cu and dose-dependent (Fig. S9). It suggested that Cu-binding to the protein induced conformational changes leading to a different exposure of aromatic amino acids. Indeed, the putative metal-binding motif in DUF2854 contains a conserved Tyr residue (Fig. 1A). Cu affinity measured at lower CrAPE1 concentrations was high, with 128 < K_D < 580 nM (Fig. S10). Notably, the addition of reductant (TCEP) inhibited Cu-binding for CrAPE1 (Fig. S8B) either because of the reduction of Cu(II) to Cu(I) or due to the reduction of cysteine disulfide bonds (S-S) to free thiols (-SH), pointing to a possible involvement of Cys residues in Cu-binding for CrAPE1. Both CrAPE1 and AtAPE1 have a conserved cysteine (Cys 128 residue in the CrAPE1 sequence) located upstream (N-terminal) of the putative metal binding site (Figs. 1A, S1). The possible involvement of this Cys in disulfide bonds modified by metal binding could explain the stronger effect of Cu on the T_m of CrAPE1 and AtAPE1 than SyAPE1, where Cys is absent. Of all the divalent cations tested, Zn also lowered the T_m in all APE1 orthologues without affecting protein fluorescence, possibly suggesting binding to a different site further from aromatic residues. The most common residues binding Cu are His and Cys (MetalPDB²⁵), thus we used EPR and site-directed mutagenesis in CrAPE1 to identify the Cu coordination.

Fig. 2. Recombinant CrAPE1 specifically interacts with Cu via His163.

A Cu significantly changes CrAPE1 protein conformation, measured as an increase in the 350/330 nm fluorescence ratio of aromatic residues. B Cu decreases the melting temperature of CrAPE1, and so does Zn to a lesser extent. For A and B, the data are means ± S.D. of three values originating from independent experiments. The statistical relevance was calculated by one-way Analysis of variance (ANOVA), and shown as pairwise comparisons (p ≤ 0.0001 ****) C X-field EPR spectra of CrAPE1_WT-Cu and CrAPE1_C128S-Cu mutants are different from free Cu and the CrAPE1_H163A-Cu mutant, showing that His163 binds copper.

Cu binds to the conserved histidine 163

The Cu(II) EPR spectrum of CrAPE1_WT recombinant protein showed an anisotropic shape (thick black line in Figs. 2C, S11), different from free copper (gray). To define the binding site, we analyzed the Cu(II) EPR spectra of CrAPE1, where the conserved histidine was substituted with alanine (CrAPE1_H163A) and in the mutant of the most conserved of the two cysteines (CrAPE1_C128S). The spectrum of the CrAPE1_C128S mutant (red) closely resembled that of the wild type, while the spectrum of the CrAPE1_H163A mutant (blue) was significantly different, although still distinct from that of Cu(II) in solution (gray). These results show that His163 binds Cu(II), but Cys128 does not. Specific binding of Cu(II) to His163 was confirmed by mass spectrometry (Table S3, Fig. S12), showing only low-affinity binding to CrAPE1_H163A (or possibly non-specific interaction with the Zn binding site).

Regulation of CrAPE1 through disulfide bonds

In size-exclusion chromatography (SEC), CrAPE1-Cu eluted earlier than the apoprotein (Fig. 3A), supporting a change in protein conformation suggested from the protein fluorescence studies (see above). CrAPE1-Cu was predominantly a monomer as confirmed by multi-angle light scattering (MALS) (Fig. 3E), with <5% of the molecules forming dimers (Table S4). Any changes to mobility or oligomerization were suppressed when the protein was pretreated with TCEP before Cu addition, further substantiating the absence of fluorescence quenching in this condition (Fig. S8B). In order to discriminate between the possible redox effect of TCEP on Cu(II) / Cu(I) or on the two cysteines Cys128 and Cys247, we compared CrAPE1_WT to single mutants CrAPE1_C128S and CrAPE1_C247S, and to double mutant CrAPE1_C128SC247S (Fig. 3B, C, D). In all of these single and double Cys mutants, the monomer did not shift upon Cu addition, and no differences were observed in the TCEP-treated sample. This confirmed the intramolecular disulfide bond formation (S–S) in CrAPE1_WT. However, CrAPE1_C128S-Cu and CrAPE1_C247S-Cu formed a larger molecular complex (Fig. 3B, C, Table S4), identified as a protein dimer by MALS (Fig. 3F, G). TCEP fully dissociated this dimer (seen in Fig. 3B, C).

Fig. 3. The cysteines of CrAPE1 control monomerization through Cu-binding to His163.

Analysis of CrAPE1 oligomerisation: size exclusion chromatography (SEC) profile of A CrAPE1_WT, B CrAPE1_C128S, C CrAPE1_C247S and D CrAPE1_C128SC247S incubated without an additive (black), with Cu addition in a ratio of 1/5 (+Cu, dotted) or with 1 mM TCEP followed by Cu addition in a ratio of 1/5 ( + TCEP+Cu, gray). Molecular mass was estimated by multi-angle light scattering (MALS) for E CrAPE1_WT, F CrAPE1_C128S. G CrAPE1_C247S and H CrAPE1_C128SC247S. The masses and molar ratios are reported in Table S4. I to L. Alkylation with mPEG-maleimide-2000 of CrAPE1_WT and cysteine mutants denatured after incubation without or with Cu (protein:Cu ratio of 1:2, 1:5 and 1:10). The number of cysteine thiols is indicated by asterisks (*). N.A. refers to non-alkylated protein. Treated and untreated samples were separated electrophoretically in reducing conditions. Profils and gels are representative of three independent experiments.

To confirm the redox state of the cysteine residues in CrAPE1, we used an SDS-PAGE mobility-shift assay by alkylation with mPEG. In CrAPE1_WT, containing two Cys, three bands were found after alkylation in the absence of Cu (Fig. 3I): CrAPE1 alkylated with two PEG tags (**), with one PEG-alkylation (*), and non-alkylated (N.A.). After the addition of Cu, most of CrAPE1_WT was non-alkylated, showing that cysteines were oxidized, forming an intramolecular disulfide bridge, with only a faint band of alkylated protein suggesting incomplete Cu binding. Expectedly, CrAPE1_C128SC247S-Cu could not be alkylated, while CrAPE1_C128S and CrAPE1_C247S formed two bands, non-alkylated and alkylated, as they contain only one Cys. The alkylation of the sole Cys was maintained upon addition of Cu, but also caused a relative increase in the abundance of the non-alkylated fraction. This is consistent with single Cys mutants forming homodimers, stabilized by an inter-molecular disulfide bond between the remaining Cys of each monomer (see above, Fig. 3B, C). In its oxidized state, the remaining Cys is thus unavailable for alkylation. In all cases, addition of Cu to the WT and single Cys mutants decreased the intensity of alkylated protein bands (Fig. 3I, J, K), suggesting a competition between Cu and the alkylating agent. Similarly, in SEC analysis, the TCEP-treated WT-CrAPE1-Cu retained a shoulder, overlapping with the WT-CrAPE1-Cu peak (Fig. 3A), which suggests a mixed population of oxidized and reduced cysteines. The relative height of the monomer/dimer peaks in SEC (Fig. 3B, C) is quantified in supplemental Table 4 and is comparable to the relative intensities of the bands of alkylated proteins (Fig. 3J, K). Our interpretation is that C128S (Fig. 3B, J) retains more alkylated monomers than C247S (Fig. 3C, K), suggesting that the conserved Cys128 is the more accessible cysteine for CrAPE1 dimerization.

Nevertheless, the incomplete alkylation that we observe may also stem from issues related to protein biochemistry under in vitro conditions: reduced accessibility of PEG to the cysteines and incomplete protein denaturation on SDS-PAGE, a mixed holoprotein population with associated disulfides, a high-pH mediated maleimide lability/exchange or a non-physiological protein oxidation after denaturation.

In summary, when both Cys are missing, there is no alkylation (Fig. 3L) and no protein dimer is formed (Fig. 3H). Altogether, these results suggest that APE1 homodimers are formed by an inter-molecular disulfide bond between either one or the other Cys of the two monomers, while upon Cu binding, both Cys are normally engaged in an intra-molecular disulfide bond in the WT.

CrAPE1 has high affinity for Plastid Copper Chaperone 1 (PCC1)

100 µM addition of free Cu to the growth medium during protein expression did not enhance Cu loading in the purified protein. The zeptomolar cytoplasmic concentrations of free Cu, and the abundance of other chaperones sequestering Cu may well explain the absence of Cu bound to APE1 after Cu supplementation during growth. This observation also suggested the requirement for a Cu chaperone specific to APE1. We thus produced recombinant PCC1 (in the presence of Cu). We identified Cu-PCC1 as a 16 kDa homodimer, with a small fraction of 8 kDa monomer (Fig. S4C, D). Metal analysis by MP-AES showed it contained 0.589 ± 0.003 Cu per PCC1, suggesting a stoichiometry of 1 Cu atom per PCC1 dimer (Table S5). This resembles ATX1 copper chaperones, where Cu is coordinated by 4 cysteine residues, 2 from each monomer²⁸. Microscale Thermophoresis (MST) confirmed an interaction between CuPCC1 and CrAPE1_WT with a dissociation constant of K_D = 40 nM (Fig. 4A), unchanged in CrAPE1_C128S. Similarly to ATX1, PCC1 likely interacts with APE1 through electrostatic interactions via a conserved set of 6 lysines¹⁹ (Fig. S3).

Fig. 4. CrAPE1 and PCC1 interact and detoxify superoxide.

A PCC1 binds CrAPE1 (K_D = 40 nM), and this is independent of the Cys128 residue: CrAPE1_WT (blue), CrAPE1_C128S (red). B Denaturing SDS-PAGE separation of CrAPE1_WT, CrAPE1_C128S and PCC1 alone, in complex with Cu, and co-purified. C O₂^•− scavenging activity of CrAPE1 and PCC1. O₂^•− was generated by xanthine/xanthine-oxidase and reported by formazan formation from the dye WST-1, causing an increase in the absorption at 450 nm (orange trace). Negative controls are CrAPE1 without Cu (150 µM, gray) and Cu alone (black), which only cause a slight positive shift of the 450 nm absorption without modifying the slope. CrAPE1 (122 µM) with Cu added (green), has sustained activity; CuPCC1 (50 µM) (purple) only has an initial scavenging activity, resulting in a biphasic slope. D O₂^•− scavenging by CuPCC1-CrAPE1 wildtype (blue) and C128S mutant (red) complexes, tested at higher O₂^•− production rates than in (C) (orange). The initial rate of superoxide detoxification exceeds the rate of production, read as 100% detoxification, then slows with different kinetics for the two complexes. All proteins in (D) are at concentrations of 75 µM. For A and C, the data are means ± S.D. of three values originating from independent experiments.

CrPCC1 acts as a Cu metallo-chaperone for CrAPE1

Such high affinity between these two proteins suggested that PCC1 was involved in delivering copper to CrAPE1. To test this, we expressed the two proteins in two separate E. coli cultures, induced in the presence of Cu for PCC1 or in its absence for CrAPE1. We co-purified CrAPE1_WT and CuPCC1 by lysing E. coli cultures together, then the proteins were dialyzed and separated by cation exchange chromatography (Figs. 4B; S13). MP-AES detected Cu in all analyzed fractions (Table S5). The pre-separated CuPCC1-CrAPE1_WT complex contained almost 1 Cu per complex (0.865 ± 0.003). After separation, all CrAPE1_WT fractions contained Cu and some fractions up to 0.840 ± 0.010 Cu per CrAPE1. As mentioned in the previous paragraph, a similar Cu loading was not observed for CrAPE1 when purified alone from E. coli cells grown in the presence of Cu. This demonstrates that CuPCC1 can transfer Cu to CrAPE1 via a mechanism that has not been previously described.

Stabilization of the transient CuPCC1-CrAPE1 complex formed upon Cu transfer

We co-purified CrAPE1_C128S with CuPCC1 in an identical manner. When the wildtype and mutant complexes were separated by SDS-PAGE with the addition of DTT, the CuPCC1-CrAPE1_WT complex dissociated as 8 kDa and 17 kDa bands, but CuPCC1-CrAPE1_C128S did not, running as a single band at 25 kDa (Fig. 4B). The complex formed by CuPCC1-CrAPE1_C128S was thus resistant to dissociation by reducing or denaturing agents. This suggested that co-purified CuPCC1 and CrAPE1_C128S were trapped in the maturation process, providing us with a snapshot of an intermediate. The most likely explanation is that CuPCC1-CrAPE1_C128S formed a Cu-dithiolate complex through the Cys247 residue. When in interaction with CrAPE1_WT, the complex was only transient because it possibly dissociated through competition with His163 for the Cu and with the thiol group of Cys128 for the protonation and release of PCC1 (a model is presented in Fig. 5). This hypothetical model, that relies on disulfide isomerization and conservation of protons, is based on similarities with the maturation pathway of CuZnSOD1 and ATX1 or CCS1. The nature of the electron acceptor needed to oxidize Cu(I) to Cu(II) remains unknown, as it does, to our knowledge, for the SOD maturation pathway. Although all the data reported in the present work fit this model, some uncertainties remain to be tested; like the nature of the copper-dithiolate complexes, the interchangeability of the cysteines in the interaction or if the incomplete formation of disulfide bonds noted in vitro also occurs in vivo.

Fig. 5. Hypothetical model for APE1 copper incorporation and enzyme activity.

Copper uptake by APE1 may vary, from species to species, as for CuZn SOD1. From cyanobacteria to green algae and higher plants, nascent APE1 spontaneously binds Cu. The redox state of Cu was not assessed in the present study. In Chlamydomonas, the PCC1 copper chaperone, homolog to ATX1 or domain 1 from human CCS, may similarly bind Cu(I) and transfer it to APE1 as Cu(II). In CrAPE1, Cu binding to His163 induces protein conformational changes, possibly through disulfide bond formation between Cys128 and Cys247. This provides APE1 with a low but sustained O₂^•− detoxification activity. CrAPE1 maturation potentially occurs with the PCC1 Cu-dithiolate complex, homolog to ATX1 or domain 1 from human CCS. Upon interaction, the Cys247 thiol may attack PCC1 Cu-thiolate, forming a PCC1-CrAPE1 Cu-dithiolate heterodimeric complex. Then Cys128 thiol may attack the second PCC1 Cu-thiolate, thereby possibly transferring Cu to the His-binding site in the matured APE1 and forming an intramolecular disulfide bond. This may dissociate the CuPCC1-CrAPE1 complex and make PCC1 available for subsequent Cu binding. CuPCC1 is a low activity O₂^•− scavenger. The C128S substitution would prevent the second nucleophilic attack of the Cu-thiolate and stabilize the CuPCC1-CrAPE1 heterodimer. This complex is a high activity O₂^•− scavenger.

CrAPE1-Cu and CuPCC1 have superoxide detoxification activity

Based on the uncanny similarities to the CCS1 maturation pathway of CuZnSOD²⁹, including copper transfer through possible disulfide isomerization, we tested the reactivity of the recombinant proteins with O₂^•−. We generated O₂^•− by xanthine/xanthine oxidase and used the dye WST-1 formazan to report its formation as an absorbance increase at 450 nm (Fig. 4C, orange). While apoAPE1 (150 µM, gray) or free copper (black) did not modify the rates of O₂^•− formation, CrAPE1-Cu (122 µM, green) provided a 50% inhibition, sustained over the 20-min assay. CuPCC1 (purple) initially inhibited 50% of O₂^•− production with less protein (50 µM), but its detoxification was not sustained and resembled the rates of scavenging of O₂^•− previously described for ATX1^18,19,30. The enzymatic activity for both CuPCC1 and CrAPE1-Cu was around 2-3 orders of magnitude lower than that of CuZnSOD¹⁵. In Fig. 4D higher rates of O₂^•− production were used, CuPCC1-CrAPE1_WT (75 µM, blue) initially showed high activity, fully detoxifying O₂^•− over 3 min (Fig. 4D). As for PCC1, the kinetics were strongly bi-phasic, but in this case with a second phase of lower sustained activity similar to that of CrAPE1-Cu (Fig. 4C). At variance, the phase of full O₂^•− detoxification was extended to 20 min in CuPCC1-CrAPE1_C128S (red). For both complexes, the initial phase of O₂^•− detoxification was significantly stronger and longer when compared to CuPCC1 alone (Fig. 4C), and the scavenging capacity of CuPCC1-CrAPE1_C128S correlated with its increased stability in comparison to CuPCC1-CrAPE1_WT (Fig. 4B). This suggests that PCC1 has a dual function both in maturation and in transiently boosting O₂^•− detoxification of CrAPE1 (see Fig. 5).

Discussion

Here we identified the function of the molecular factor involved in acclimation of photosynthesis to the environment, APE1, a chloroplast thylakoid copper-binding enzyme detoxifying the superoxide anion radical O₂^•−at the vicinity of photosynthetic complexes.

The evolution of an expanding repertoire of ROS detoxification enzymes enabled life on Earth to transition from an anoxic to an oxic environment. Superoxide dismutases and reductases use various elements, Ni, Fe, Mn or Cu/Zn, as cofactors for electron transfer¹⁵. Although these metals seem equally effective for catalysis, their availability as micronutrients changed dramatically during earth history³¹. During the Great Oxygenation Event, oxygenic photosynthesis precipitated Fe(II) into Fe(III) and solubilized copper sulfides, making Cu(II) more available than Fe(III)^31,32. It seems consistent that APE1, which dates back to oxygen production as far as we can trace it, selected Cu as a protein cofactor³². Later in evolution, in a green alga, it recruited the Cu chaperone PCC1, an atypical, chloroplastic, ATX-like protein whose transcription decreases in response to Cu deficiency²⁷.

Despite not being the main site of superoxide production in the photosynthetic chain, PSII can reduce O₂ at the level of cytochrome b₅₅₉ (especially in its low potential form, like in the absence of an intact manganese cluster)^12,33, or at its reduced primary quinone acceptor (Q_A⁻) when forward electron transfer to the second quinone Q_B is blocked and the non-heme Fe is devoid of bicarbonate^12,13,34. Hence, O₂^•− production might especially occur in PSII during assembly or repair when these events have been observed. We suppose APE1 alone, found in nascent PSII assembly intermediates in complex with D2^8,9, would have sufficient activity to protect PSII from O₂^•− production.

We observed that the interaction of CrAPE1 with PCC1 results in a transiently augmented O₂^•− scavenging capacity that may be relevant under stress conditions. Indeed, in high light, when photosynthesis is limited by CO₂ availability, the resilience of PSII depends on APE1^6,7. There is also strong evidence of functionalization of APE1 with other proteins and pathways, with its inclusion in autophagy stromules and plastoglobules under light stress^35,36.

These interactions reveal a pivotal role of APE1’s function in O₂^•− detoxification in chloroplast acclimation to stress. APE1’s substrate, O₂^•−, is not only a harmful byproduct of photosynthesis, toxic for living cells, but it is also used as a signal molecule for biological acclimation^37,38. In this context, APE1 stands as the primordial regulatory control point of O₂^•− at the intersection of photosynthesis, light and CO₂ levels, fully justifying its identification as Acclimation of Photosynthesis to the Environment 1.

Methods

Chlamydomonas cell culture

The wild-type strain (T222 mt⁺) used for generating the ape1 knock-out mutants is a progeny of 137c backcrosses³⁹; CC-5325 background wild-type strain and pcc1 (LMJ.RY0402.059810), ctp2 (LMJ.RY0402.149111), and ctp4 (LMJ.RY0402.047261) mutants were obtained from the CLiP mutant collection background⁴⁰. Maintained cultures were always cultivated at 20 μmol photons‧m⁻²‧s⁻¹ in Tris-Acetate-Phosphate (TAP) + 2% agar media plates. For experimental setups, liquid cell cultures were grown in TAP medium at 25 °C under ambient air at continuous 25 μmol photons‧m⁻²‧s⁻¹ illumination in incubation shakers (photomixotrophic conditions) for 3 days, and then shifted to photoautotrophic conditions by centrifugation and resuspension in minimal (MIN) media under ambient air before being plated for spot tests, photos were taken at 7 days. Standard recipes for media preparation were used as in⁴¹.

Generation of Chlamydomonas CRISPR/Cas9 mutants

Specific gRNA and Alt-R^TM S.p. Cas9 Nuclease V3 were ordered from IDT-DNA and used to assemble in vitro RNP particles; the sequence TCTGCTCCCGCTAAGCCGGCTGG in the exon 2 of Cre16.g665250 (encoding APE1) was selected as the PAM site. Chlamydomonas wild-type cells were harvested at an early exponential stage (2 × 10⁶ cells‧ml⁻¹) via centrifugation (1600 g, 5 min, RT), and re-suspended in “MAX Efficiency™ Transformation Reagent for Algae” (Thermo Fisher Scientific), supplemented with 20% sucrose, to a final concentration of 2 × 10⁸ cells‧ml⁻¹. A heat-shock (40 °C, 30 min, 450 rpm shaking) was then applied to the cells, followed by a 20-min recovery at RT, before 120 µl of cells were added to a pre-cooled 2-mm gap electroporation cuvette together with 15 µl of RNP particles and 1 µg of paromomycin-resistance plasmid for electroporation (600 V, 50 µF, infinite external resistance). Cells were allowed to recover overnight in TAP medium supplemented with 20% sucrose and 1 mg‧l⁻¹ Vitamin B12 at 33 °C in the dark before being plated in TAP + 2% agar + 10 mg‧l⁻¹ paromomycin and incubated under 20 μmol photons‧m⁻²‧s⁻¹ illumination until colonies could be picked.

DNA isolation and Polymerase Chain Reaction (PCR)

Total genomic DNA and PCR were done using the “Phire Plant Direct PCR Master Mix” kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Briefly, a small amount of cells was solubilized with 20 µl of “Dilution buffer” (included in the PCR kit) via agitation, before centrifugation (4000 g, 10 min, RT) to obtain the DNA in the soluble fraction. The PCR master mix was prepared by mixing 5 µl of 2X reagent with 3 µl of 3 M betaine, 1 µl of each primer, and 1 µl of DNA template (or ddH2O as a negative control), and amplification was done during 35 cycles. PCR products were loaded on a 2% TAE-agarose gel, and the DNA was stained for visualization with ClearSight (Euromedex). The primers used in this manuscript are listed in Supplemental Table 5.

Immunoblot analysis

Western blot analysis was performed as in⁷ with minor modifications. Briefly, proteins were separated under denaturing conditions on 15% SDS-PAGE gels supplemented with 4 M Urea. Samples were loaded based on chlorophyll content (100% equals 1 μg Chl a + b). Proteins were transferred onto nitrocellulose membranes (BioTrace NT, Pall Corporation) using liquid transfer. Primary anti-APE1 antibody was generated in rabbit against the recombinant soluble domain of CrAPE1 (i.e., from residue Q122 to residue E276 of Cre16.g665250) produced in E. coli (ProteoGenix) and used at a 1:5000 ratio; primary PsaD antibody was a kind gift from S. Bujaldon and used at a 1:5000 ratio. Secondary antibodies used were always HRP-conjugated anti-rabbit (Invitrogen) used at a 1:10000. HRP-peroxidase chemiluminescent substrate (SuperSignal™ West Pico PLUS, Thermo Fisher Scientific) was used to reveal the antibody signal, using a charge-coupled device (CCD)-camera based imaging system (Cytiva).

Cloning

For heterologous expression in E. coli, the region coding for the transmembrane subunits of APE1 was not included. The synthetic DNA sequences of the predicted soluble domain of Chlamydomonas APE1 (corresponding to the residues Q122 to E276 of Cre16.g665250), Arabidopsis APE1 (corresponding to the residues E146 to S286 of At5g38660), and Synechocystis APE1 (corresponding to the residues E54 to P184 of slr0575), flanked by complementary BsaI restriction sites sequences, were ordered from Twist Bioscience and sub-cloned into the plasmid pLIC03 (LIC: ligation-independent cloning⁴² using the GoldenGate cloning technique; the synthetic DNA sequence of Chlamydomonas PCC1 gene without the predicted chloroplast transit peptide (corresponding to the residues A39 to Q118 of Cre05.g248600), flanked by complementary BsaI restriction sites sequences, was ordered from Twist Bioscience and sub-cloned into the plasmid pNIC28-Bsa4 (Addgene #26103) using GoldenGate cloning technique. Both pLIC03 (modified Novagen pET-28a+) and pNIC28-Bsa4 vectors contain the T7 promoter/terminator pair, and an N-terminus 6 × His-tag coupled with a TEV protease-cleavage site followed by the suicide gene SacB (flanked by BsaI restriction sites), which is replaced by our synthetic gene sequences. 20 fmol of plasmid was used with 20 fmol synthetic DNA for three successive rounds of digestion (BsaI, 37 °C, 10 min) and ligation (T4 DNA ligase, 16 °C, 10 min) in the same mix. TOP10 E. coli cells (NEB) were then transformed with the ligation products for vector amplification, screening and Sanger sequencing (Eurofins Genomics). 50 ng of vector was finally used to transform BL21(DE3) cells (Invitrogen), chemically competent E. coli for protein expression.

Site-directed mutagenesis

Site-directed mutagenesis for the different CrAPE1 recombinant versions was based on⁴³. Briefly, pLIC03-APE1 was amplified via PCR using NEB Q5 High-Fidelity 2X Master Mix in 25 µl reaction volumes containing 10 ng DNA template and 0.5 µM of specific primers pairs carrying the new codon (C128S mutation: TGC-AGC; H163A: CAC-GCG; C247S: TGC-AGC); the PCR protocol was set according to the manufacturer’s instructions (amplification was done for 25 cycles with 180 s elongation time each) and annealing temperatures were calculated with the online NEB Tm calculator tool. After amplification, to remove the initial DNA template (i.e., without the added mutation) and ligate the PCR products, 1 µl of PCR product was incubated at RT for 10 min with NEB 10X KLD (Kinase, Ligase, DpnI) Enzyme Mix in a 10 µl reaction volume before being used to transform NEB 5-alpha competent E. coli cells, for plasmid isolation and Sanger sequencing (Eurofins Genomics). Plasmids with the correct sequence were used to transform Rosetta™(DE3)pLysS Competent E. coli Cells (Novagen) for recombinant protein production.

Recombinant protein production

Bacterial cells were cultured at 37 °C in Terrific Broth Media (APE1) or Luria Broth Media supplemented with 50 mg l⁻¹ of kanamycin (and 25 mg l⁻¹ of chloramphenicol for APE1). For PCC1 production, LB media was complemented with 0.1 mM CuSO₄. Expression was induced once cells reached an OD600 of 0.6 by the addition of 0.5 mM IPTG, and the cultures were left overnight at 17 °C. Cells were collected by centrifugation and re-suspended in IMAC buffer (50 mM Sodium-Phosphate buffer, 300 mM NaCl, 10 mM Imidazole pH 8.0) supplemented with EDTA-free protease inhibitor cocktail (Roche) and DNase (Sigma-Aldrich). For the PCC1-CrAPE1 complex, an equal amount of each pellet was mixed before lysis.

Cells were disrupted twice by French Press at 1000 psi and clarified by centrifugation at 40000 x g for 45 min. The 0.45 µM filtered lysates were loaded on a HiTrap Chelating HP 5 ml column (Cytiva), pre-equilibrated with IMAC buffer. Bound protein was eluted with a linear gradient of IMAC buffer containing 125 mM imidazole, and protein-containing fractions were pooled and dialyzed overnight in 50 mM Sodium-Phosphate buffer, 300 mM NaCl, pH 8, with his-tag TEV protease in a mass ratio of 1/40. The resultant was then loaded on another HiTrap Chelating HP 5 ml column (Cytiva), pre-equilibrated with IMAC buffer, and the untagged protein was collected in the flowthrough. This fraction was then loaded on a 120 ml HiLoad 16/60 Superdex 75 column (Cytiva), pre-equilibrated in 20 mM MES, 150 mM NaCl, pH 6.5. Fractions containing the target protein were pooled and concentrated using an AMICON centrifugal filter device (Millipore) with a cut-off size of 3.5 kDa. SDS-PAGE was used to follow each purification step and to verify the purity of the proteins and mutants, see Fig. S7. The mass of the purified proteins was verified by electrospray ionization mass spectrometry. Protein quantification was assayed spectrophotometrically using theoretical extinction coefficient at 280 nm of 8605 M⁻¹ cm^–1 for CrAPE1_WT and CrAPE1_H163A, 8480 M^–1 cm^–1 for CrAPE1_C128S, CrAPE1_C247S, CrAPE1_C128SC247, 1615 M^–1 cm^–1 for PCC1. Metals were removed post-purification by treatment with Chelex or EDTA.

Mass spectroscopy of native proteins

Proteins were desalted by dialysis using 50 ml of 20 mM ammonium acetate overnight at 4 °C. A diluted 200 µl volume was injected at 7 µl min⁻¹ in a Quadropole MS QTOF (Bruker Impact II). The data was recorded using the Otof Control Software, then treated using Compass Data Analysis software. After purification, the molecular mass of the proteins corresponded to the predicted mass of the polypeptide for CrAPE1 and its mutants (see Table S3), for SyAPE1 15114.5 Da (theoretical mass: 15115 kDa) and for AtAPE1 16024.3 Da (theoretical mass: 16025 kDa).

Size-exclusion chromatography with multi-angle light scattering

Multi-angle light scattering (MALS) is a technique that determines the precise molecular weight of the species in samples. Knowing that the variation of refractive index related to variation of protein concentration is constant (dn/dC), measuring the refractive index allows one to access the concentration of protein at any point of the elution. 20 μl of protein with a 1.4 mg ml⁻¹ (80 µM) concentration was loaded on an analytical 14 ml Xbridge Premier Protein SEC 250 Å, 2.5 µM column (Waters) equilibrated with 20 mM MES, 150 mM NaCl, pH 6.5 (filtered 0.1 mm) and connected to an Arc^TM premier HPLC system (Waters). Runs were performed at 25 °C with a flow rate of 0.6 ml min⁻¹. Elutions were monitored by using a variable wavelength detector at 280 nm (Waters), a DAWN^®-8 angle light scattering detector and an Optilab^® refractive index detector (Wyatt Technology). Cu was added to a 1:5 (protein:metal) molar ratio and incubated 5 min with the protein prior to injection. In reducing conditions, proteins were incubated at 25 °C for 10 min with 1 mM tris(2-carboxyethyl) phosphine (TCEP). After baseline correction, all samples presented isolated peaks allowing the determination of absolute molecular masses using ASTRA6 software (Wyatt Instruments) and a theoretical dn/dc value of 0.185 ml g⁻¹.

Alkylation of recombinant proteins

Alkylation assays were carried out to determine the number of free thiols in CrAPE1_WT and cysteine point mutants. Chelexed proteins (20 µM) were incubated 5 min with or without Cu (protein:Cu ratio 1:2, 1:5 and 1:10). SDS was added to the mix at 1% final concentration. After 20 min, mPEG-maleimide-2000 (Laysan Bio Arab, AL, USA) was added to the assay at 2.5 mM and the mix was incubated at room temperature for 3 h. Then, 4X reducing loading buffer for SDS-PAGE (Millipore) was added to the reaction mixture, and samples were loaded on a 15% SDS PAGE gel, stained with READYBLUE^® Protein gel stain (Sigma). Uncroppped gels are available in the Source Data file.

Intrinsic protein fluorescence measurement

Fluorescence measurements were performed with a Varian Cary Eclipse luminescence spectrometer with slit widths set to 5/5 nm, in a 1 × 1 cm quartz cuvette (Hellma), thermostated at 25 °C by circulating water from an external water bath. Chelexed CrAPE1_WT was diluted at 10 µM in 20 mM MES, 150 mM NaCl, pH 6.5 buffer, and incubated 10 min with or without 1 mM tris(2-carboxyethyl) phosphine (TCEP) and then titrated with various concentrations of Cu (from 0 to 40 µM). Fluorescence measurements were performed from 290 to 410 nm, with an excitation wavelength of 280 nm in order to specifically excite tryptophanyl residues. Each curve was the result of 5 consecutive scans.

Thermal unfolding experiments by nano differential scanning fluorimetry

We tested the binding of various divalent cations by nano differential scanning fluorimetry. Fluorescence at 330 and 350 nm (with excitation at 280 nm) was recorded over a temperature gradient scan on Prometheus NT.48 (NanoTemper Technologies, GmbH). APE1 proteins were diluted to a final concentration of 40 µM and incubated with 1 molar excess of each divalent metal or with a range of Cu from 1 to 200 µM. For each condition, 10 µl of sample per capillary was required for one thermal unfolding profile. The samples were loaded into Prometheus capillaries, and the temperature ramp was set to an increase of 2 °C min⁻¹ in a range from 25 °C to 90 °C. Protein unfolding was measured by detecting the temperature-dependent change in tryptophan fluorescence at emission wavelengths of 330 and 350 nm. For the calculation of Melting Temperature (Tm), the first derivative data of the F350/F330 fluorescence ratio unfolding curves were used.

Protein interaction by microscale thermophoresis

PCC1 was diluted to 17 µM and labeled using Monolith Protein Labeling Kit RED-NHS 2nd generation (NanoTemper Technologies GmbH). For CrAPE1, a serial 1:1 dilution was performed by transferring 10 µL of 5 µM CrAPE1 solution to an equal volume of working buffer (20 mM MES, 150 mM NaCl, pH 6.5), mixing and repeating this step 16 times. This way, the ligand concentration is reduced by 50% in each dilution step. 10 µl of labeled PCC1 at 34 nM was mixed with 10 µL of each CrAPE1 serial dilution from 2.5 to 7,7.10⁻⁵ µM. For the quantification of copper binding to CrAPE1, a serial dilution of CuCl₂ from 1.6 10⁻⁵ M to 4.88 10⁻¹⁰ M was prepared in 20 mM MES, 150 mM NaCl, pH 6.5. Labeled CrAPE1 was added to a final concentration of 24 nM. Labeling of APE1 has to occur just before MST experiments because it strongly impairs APE1 stability. MST experiments were performed on a NanoTemper ® Monolith NT.115 with red filter (NanoTemper Technologies GmbH). Samples were loaded into standard treated capillaries. Measurements were performed at 25 °C using 100% excitation on medium MST power position. Data analysis was performed using the NanoTemper® analysis software, where K_D constants were calculated using the saturation binding curve at equilibrium.

Superoxide detoxification assays

Superoxide detoxification activity was evaluated by a specific assay kit (CS0009, Sigma-Aldrich), according to the manufacturer’s instructions or diluted 4 times. Superoxide detoxification activity was proportional to the decrease in the color signal of WST-1 (Water Soluble Tetrazolium dye) due to O₂^•− production by xanthine/xanthine oxidase, determined by reading the absorbance at 450 nm wavelength after a given reaction time at 25 °C (Infinite M200 spectrophotometer, Tecan Ltd) after subtracting the background (buffer control). Molar extinction coefficient of WST-1 at 450 nm is ε_450nm = 35.2 mM⁻¹.cm^{−1 44}. In the 96-well plate with 200 µl assay volume, the optical path is 0.5 cm. In the experimental conditions of Fig. 4C, absorbance increased by 0.1 (O.D.) every 20 min, corresponding to 5.7 µM formazan, each formazan molecule quenching two O₂^•− molecules. The rate of generation of superoxide radical anions was therefore 0.57 µM.min⁻¹. In Fig. 4D, the rate of generation was calculated as 1.14 µM.min⁻¹

EPR spectroscopy

X-band continuous wave EPR spectra were recorded with a Bruker Elexsys 500 X-band spectrometer equipped with a standard ER 4102 (Bruker) X-band resonator, a Bruker teslameter, an Oxford Instruments cryostat (ESR 900) and an Oxford ITC504 temperature controller. Spectra were collected at 30 K; modulation amplitude, 10 G; microwave power, 0.02 mW; microwave frequency, 9.5 GHz; modulation frequency, 100 kHz. 50 µM protein samples were suspended in 100 mM HEPES pH 6.5 with equimolar concentrations of CuCl₂. Spectra were fitted using EasySpin, version 6⁴⁵.

Metal determination and quantification

Metal content of the proteins (Cu, Zn) was quantified by MP-AES (microwave-plasma-atomic-emission-spectroscopy). Samples were treated with 50% nitric acid and heated to 96 °C for 2 h, then diluted to 10% nitric acid prior to measurement. After ten-fold dilution in trace-metal-free water, the metal content of the samples was determined by atomic emission spectroscopy using an MP AES 1200 spectrometer (Agilent, USA).

Statistics

The statistical relevance was calculated using GraphPad Prism 10 by one-way Analysis of variance (ANOVA), a comparison between the means of two or more groups by analyzing variance, using Tukey’s post-hoc multiple comparisons, and shown as pairwise comparisons (p ≤ 0.001*** or p ≤ 0.0001 ****) or compact letter display (CLD).

Experimental set up for measuring plastocyanin accumulation in vivo

The protocol was adapted based on⁴⁶. Cells were spun down and resuspended in Ficoll and dark-adapted. Cells were then poisoned with DCMU and DBMIB to restrict electron flow to only twosecondary electron donors to photo-oxidized P₇₀₀⁺ Rieske protein, cytochrome f and the primary electron donor PC. Flash-induced absorbance changes were measured using the Joliot Type Spectrophotometer (JTS-10) using two detection wavelengths (705 nm for P₇₀₀⁺, and 740 nm to subtract PC⁺ contribution). Saturating single turnovers of PSI were induced by a < 10 ns laser flash; three biological replicates were measured.

Analysis of genes co-expressed with APE1

Co-expression analyses were performed to identify genes coordinately expressed with the three APE1 orthologs chosen in this work. For Chlamydomonas APE1 (Cre16.g665250), the PhytoMine expression compendium tool of JGI Phytozome v14 (https://phytozome-next.jgi.doe.gov/) was queried; genes exhibiting a Pearson correlation coefficient (PCC) ≥ 0.80 with CrAPE1 across 518 public RNA-seq datasets were retained. For Synechocystis APE1 (slr0575), the AlcoDBcyano portal (https://alcodb.jp/cyano/) was used to retrieve the top-ranking co-expression partners; only genes with a mutual-rank (MR) ≤ 50 were kept. For Arabidopsis APE1 (At5g38660), co-expression data were obtained from ATTED-II v12.0 (https://atted.jp/), and genes with Logit Score > 3.0 (≈ PCC ≥ 0.70) were selected. Genes were manually assigned to functional categories according to their primary annotation; when no annotation was available, assignments were based on additional information obtained from the KEGG database (https://www.genome.jp/kegg/) or from targeted literature searches.

Reporting summary

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

Author contributions

Conceptualization: S.M., A.V.M., M.C., C.B., J.L., S.V., P.A., F.C., S.C., C.C.C., J.A., and X.J.; Methodology: S.M., A.V.M., E.G., M.C., A.K.L., C.B., J.L., S.V., L.T., P.A., P.D., C.M., C.C.C., M.S., and X.J.; Investigation: S.M., A.V.M., E.G., M.C., A.K.L., C.B., J.L., S.V., L.T., P.A., P.D., C.M., C.C.C., M.S., J.A., and X.J.; Funding acquisition: X.J., J.A., A.K.L., C.C.C., and S.C.; Project administration: X.J.; Supervision: X.J., S.M., J.A., C.C.C., S.C.; Writing – original draft: X.J., J.A., A.V.M., S.M.; Writing – review & editing: S.V., P.A., C.C.C., F.C., S.C., A.K.L., L.T., J.A., and X.J.

Data availability

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/s41467-026-71893-z.