Structural determinants for red‐shifted absorption in higher‐plants Photosystem I

Summary

• Higher plants Photosystem I absorbs far‐red light, enriched under vegetation canopies, through long‐wavelength Chls to enhance photon capture. Far‐red absorption originates from Chl pairs within the Lhca3 and Lhca4 subunits of the LHCI antenna, known as the ‘red cluster’, including Chls a603 and a609.

• We used reverse genetics to produce an Arabidopsis mutant devoid of red‐shifted absorption, and we obtained high‐resolution cryogenic electron microscopy structures of PSI‐LHCI complexes from both wild‐type and mutant plants.

• Computed excitonic coupling values suggested contributions from additional nearby pigment molecules, namely Chl a615 and violaxanthin in the L2 site, to far‐red absorption. We investigated the structural determinants of far‐red absorption by producing further Arabidopsis transgenic lines and analyzed the spectroscopic effects of mutations targeting these chromophores. The two structures solved were used for quantum mechanics calculations, revealing that excitonic interactions alone cannot explain far‐red absorption, while charge transfer states were needed for accurate spectral simulations.

• Our findings demonstrate that the molecular mechanisms of light‐harvesting under shaded conditions rely on very precise tuning of chromophore interactions, whose understanding is crucial for designing light‐harvesting complexes with engineered absorption spectra.

Introduction

Photosynthetic organisms use solar radiation as their primary energy source to convert carbon dioxide and water into oxygen and sugars. The light reactions of oxygenic photosynthesis involve two multimeric pigment‐binding protein complexes: photosystems (PS) I and II, which work in series. The initial step of light harvesting is catalyzed by PSII, which is responsible for water splitting and oxygen evolution. Meanwhile, PSI mediates the reduction of NADP⁺ to NADPH, serving as a temporary storage for reducing power (Croce & van Amerongen, 2020). PSs share a common structure that includes a core complex housing the reaction centers (RC) where charge separation reactions occur and an antenna system (Light Harvesting Complexes (LHC) enhancing the light‐absorbing cross‐section of the complexes (Pan et al., 2020)). Despite these similarities, there are striking differences in their spectral properties, with PSI exhibiting a red‐shifted absorption profile (Rivadossi et al., 1999). The red‐most PSI absorption arises from special Chl pairs that absorb at lower energies than RC P700, referred to as ‘red Chl forms’ or simply ‘red forms’ (RF; Croce & Van Amerongen, 2013).

These pigments extend the absorption capacity into the far‐red spectrum (λ ≥ 700 nm), providing advantages under canopy or in dense culture conditions where most visible wavelengths are absorbed by the upper leaf layers (Martínez‐García et al., 2010) while far‐red radiation is strongly enriched, resulting in a far‐red to red ratio higher than 5 (Park & Runkle, 2017). It is worth noting that, although RF contribute only a small percentage of the total absorption (Gobets & Van Grondelle, 2001), they are preferentially excited under far‐red enriched light conditions (Croce et al., 1998; Rivadossi et al., 1999). Moreover, RF are highly effective in energy transfer and trapping, as c. 80% of the PSI excitation transits through these pigments to reach P700 (Croce et al., 1998). Since the excited states associated with these low‐energy Chls are highly populated, they may play crucial roles in photoprotection and/or the concentration of excitation energy (Rivadossi et al., 1999; Jennings et al., 2003; Carbonera et al., 2005). However, their exact physiological role(s) remain to be fully understood (Jennings et al., 2013).

RF are present in nearly all types of PSI complexes, and their occurrence is closely related to the availability of far‐red radiation. In cyanobacteria (e.g. Arthrospira platensis), these pigments are localized within the core complex and allow for the absorption of far‐red photons, resulting in PSI low temperature fluorescence emission peaks between 730 and 760 nm (Rakhimberdieva et al., 2001). In green algae (e.g. Chlamydomonas reinhardtii) and mosses (e.g. Physcomitrium patens), red Chls are associated with the LHCI antenna system, resulting in a PSI‐LHCI fluorescence emission peak wavelength c. 715 and 722 nm, respectively (Mozzo et al., 2010; Gorski et al., 2022; Fig. 1). Upon land colonization, environmental niches enriched in far‐red radiation became widespread under canopy, resulting in a progressively increased red‐edge Q_Y absorption associated with LHCI (Croce et al., 1998). Indeed, the PSI‐LHCI fluorescence emission from flowering plants leaves shifts toward the red, peaking at 730 nm or even beyond (Akhtar & Lambrev, 2020; Bos et al., 2023; Li et al., 2024).

Fig. 1.

Evolution of low‐absorbing spectral features in Viridiplantae. (a) Time tree of representative Viridiplantae species and their divergence times (in million years, Myr). The most red‐shifted fluorescence emission peak of each species' experimental spectra is highlighted in bold. Note that the PSI core complex emits at c. 720 nm. Thus, red forms associated with LHCI proteins can be readily detected only when emitting at >720 nm. The emergence of red forms (RF) can be traced back to c. 489–403 Myr co‐eval with the appearance of tall trees and canopies. (b) Low temperature (77 K) fluorescence emission spectra of representative species from green algae (Chlamydomonas reinhardtii), mosses (Physcomitrium patens), and full sun land plant angiosperms: Zea mays, Oryza sativa (Poaceae) and Arabidopsis thaliana (Brassicaceae); shade plants: Fittonia albivenis (Acanthaceae); and seagrasses (Posidonia oceanica and Cymodocea nodosa). Shade plants thrive in far‐red enriched light, while seagrasses thrive in the absence of far‐red radiation. (c) Multiple alignments of the Lhca4 (corresponding to Lhca8 for C. reinhardtii and Lhca2b for P. patens and Lhca3 protein sequences from C. reinhardtii (Cr), P. patens (Pp)), P. oceanica (Po), C. nodosa (Cn), A. thaliana (At), Z. mays (Zm), Ananas comosus (Ac), and F. albivenis (Fa). Residues binding red form Chls (Chl a603 and Chl a609) are colored in red and shown in red rectangles. Coordinating residues for Chl a615 are colored green for species with reported PDB structures. Details about the acquisition of the time tree in panel (a) and the spectra in panel (b) are reported in the Supporting Information. PS, photosystem; LHC, light harvesting complex.

In higher plants, LHCI binds the PSI core complex as functional heterodimers: Lhca1‐Lhca4 and Lhca2‐Lhca3, both of which exhibit similar spectral properties and emit far‐red fluorescence at c. 730 nm (Wientjes & Croce, 2011). In vitro reconstitution of recombinant Lhca complexes (rLhca) showed that the far‐red shift is primarily caused by the Lhca4 and Lhca3 components of the dimers (Croce et al., 2002; Castelletti et al., 2003), and is linked to a specific sequence substitution unique to these two LHCs, where an Asn serves as the binding residue for Chl a603 instead of a His residue, as seen in all other LHC proteins (Jansson, 1999; Morosinotto et al., 2003; Fig. 1c). Indeed, the substitution of Asn (N) with His (H) (a603‐NH substitution) resulted in the complete loss of the far‐red absorption and emission forms in both rLhca3 and rLhca4 complexes (Morosinotto et al., 2003). It was hypothesized that the presence of Asn promotes strong excitonic interactions between the Chl a603–a609 pair (A5 and B5 according to Kühlbrandt's nomenclature; Kühlbrandt et al., 1994) leading to a charge transfer (CT) character (Romero et al., 2009) and ultimately resulting in the formation of the RF (Morosinotto et al., 2003). Complementing koLhca4 Arabidopsis lines with mutant Lhca4, carrying either His or nonpigment‐binding residues at the Chl a603 ligand position (Li et al., 2023) resulted in a blue shift of c. 3 nm at 77 K.

However, accumulating evidence suggests that the substitution of Asn vs His as the ligand of Chl a603 alone cannot account for the variability in far‐red spectral properties found in nature (Li et al., 2023; Elias et al., 2024). For instance, in the Lhca2/a4/a9 subunits of C. reinhardtii, where the Chl a603 ligand is Asn, emission peaks range between 690 and 717 nm, with significantly lower absorption beyond 700 nm compared with higher plants Lhca4 (Mozzo et al., 2010). Moreover, the a603‐HN (H111N) mutation, although introducing a significant red‐shift in the PSII antenna complex Lhcb4 (from 680 to 682 nm), did not achieve a shift as large as the 40 nm one observed in the Chl a603‐NH mutant of Lhca3 or Lhca4 (from c. 730 to c. 690 nm; Morosinotto et al., 2003; Guardini et al., 2020; Sardar et al., 2024), despite the high structural similarity between the antenna complexes.

In addition to the arrangement of relevant Chls, the surrounding microenvironment can be important for optimizing low‐energy spectral forms. Structural models of PSI‐LHCI reveal a chromophore cluster comprising three Chls – Chl a603, Chl a609, and Chl a615 – and two xanthophylls (Xan) – violaxanthin (Vio) in site L2 and a lutein (Lut; Qin et al., 2015). Notably, Chl a615 is present only in the Lhca3 and Lhca4 subunits – the red‐shifted subunits – while it is absent in all other LHC proteins, regardless of whether they serve PSI or PSII. This observation raised the question of its potential role in forming excitonic interactions (Melkozernov & Blankenship, 2003; Wientjes et al., 2012).

In this study, we investigated the structural and biophysical determinants of the red Chl forms in Arabidopsis thaliana PSI‐LHCI by combining in vivo site‐directed mutagenesis and single‐particle cryogenic electron microscopy (Cryo‐EM). We obtained two high‐resolution structures of the PSI‐LHCI supercomplexes from A. thaliana wild‐type (WT) and the a603‐NH mutant genotype, which lacks RF. We then computed the excitonic interactions within PSI‐LHCI pigments, focusing on the interaction between the L2 Xan and the Chl a603–a609 pair in forming low‐lying energy spectral states.

Materials and Methods

Plant materials

The Arabidopsis thaliana (L.) Heynh koLhca3 koLhca4 mutant was generated by crossbreeding koLhca3 and koLhca4 NASC insertional lines, following the methodology outlined by Bressan et al. (2016). Complementation of the koLhca3 koLhca4 mutant was achieved through Agrobacterium tumefaciens‐mediated transformation, as described by Zhang et al. (2006), resulting in a603‐NH, a615‐HA, a615‐HI, A3WT‐A4WT, A3NH‐A4WT, and A3WT‐A4NH lines.

Purification and characterization of PSI‐LHCI samples

To purify PSI‐LHCI supercomplexes, A. thaliana plants were grown for c. 6 wk in a phytotron (150 μmol photons m⁻² s⁻¹, 23°C, 70% relative humidity, 8 h : 16 h, day : night). Before the isolation procedure, Arabidopsis leaves were dark‐adapted for 60 min at 4°C. Unstacked thylakoid membranes were prepared as described in Bassi et al. (1985). Thylakoid membranes were resuspended to a Chl concentration of 1 mg ml^‐1 in 10 mM HEPES pH 7.5 and solubilized by adding an equal volume of 2% dodecyl‐β‐d‐maltoside (β‐DM). The samples were vortexed for 30 s and incubated on ice for 10 min, and the insoluble material was removed by centrifugation at 20 000  g for 10 min at 4°C. The supernatants were fractionated by sucrose gradient ultracentrifugation at 284 000  g at 4°C for 18 h (Beckman SW40 Ti rotor, Beckman Coulter, Brea, CA, USA) or at 141 000  g at 4°C for 30 h (Beckman SW28 Ti; Supporting Information Fig. S1). Bands containing PSI‐LHCI were harvested using a Hamilton syringe.

For Cryo‐EM preparations, PSI‐LHCI samples were concentrated to a final volume of c. 800 μl, and sucrose was removed by dialysis overnight against a solution containing 10 mM HEPES pH 7.5 and 0.05% β‐DM. Finally, the samples were concentrated to a Chl concentration of c. 1.5–2 mg ml⁻¹.

Purification of PSI core and LHCI

PSI core complex and LHCI were purified from WT and mutant lines as described by Croce et al. (1998) and Wientjes & Croce (2011) with some modifications. Briefly, PSI‐LHCI from sucrose gradient ultracentrifugation was diluted in 10 volumes of 5 mM tricine (pH 7.8) and centrifuged for 3 h at 411 000  g , using a T‐865.1 rotor, Sorvall. Pellets were resuspended in a buffer containing 5 mM Tricine (pH 7.8) and 0.05% β‐DM. The Chl concentration was adjusted to 0.3 mg ml⁻¹, and the samples were solubilized by adding 1% β‐DM and 0.5% Zwittergent 3–16. The samples were kept on ice with gentle agitation for 25 min, then rapidly frozen in liquid nitrogen, and slowly thawed to enhance the yield of detached LHCI. Solubilized complexes were fractionated by sucrose gradient ultracentrifugation at 485 000  g at 4°C for 6 h, using a Beckman SW60 Ti rotor.

Spectroscopy and pigment analysis

Absorption spectra were recorded at room temperature (RT, 22°C) using an SLM‐Aminco DW‐2000 spectrophotometer in a buffer containing 10 mM HEPES pH 7.5 and 0.05% β‐DM for native complexes or 80% acetone buffered with Na₂CO₃ for pigment extracts, in standard 1 cm path quartz cuvettes. Low temperature (77 K) absorption spectra were recorded by diluting samples at c. 0.6 OD at the Q_Y peak in 70% glycerol, 20 mM HEPES, pH 7.5, following the method outlined in Yang et al. (2003) in 1‐ml methacrylate disposable cuvettes.

Emission and excitation fluorescence spectra were recorded at cryogenic temperatures (77 K) using a Jobin–Yvon Fluoromax‐3 spectrofluorometer. Samples were diluted at c. 0.2 OD at the Q_Y peak in a buffer containing 50% w/v glycerol, 10 mM HEPES, pH 7.5, 0.05% β‐DM. One milliliter of samples were frozen into Aldrich ColorSpec NMR tubes and excited at 440 nm.

Emission spectra of intact leaves were recorded at 77 K using an Ocean Insight SR‐6NVN500‐50 spectrofluorometer. Leaves were frozen into a nitrogen bath and directly measured with an optical fiber.

Circular Dichroism (CD) spectra were recorded at 4°C on a Jasco J1500 spectropolarimeter.

The pigment composition of LHCI complexes was assessed from the deconvolution of acetonic spectra as described in Chazaux et al. (2022). High‐performance liquid chromatography (HPLC) pigment separation and quantification were performed according to Gilmore & Yamamoto (1991) by using an Agilent 1260 Infinity II HPLC system.

Cryo‐EM sample preparation and data acquisition

The PSI supercomplex at a Chl concentration of c. 1.5–2 mg ml⁻¹ was vitrified with a Mark IV Vitrobot (Thermo Fisher Scientific, Waltham, MA, USA). Three microliters of the sample was applied to a Quantifoil R 0.6/1Cu 300‐mesh grid previously glow‐discharged at 30 mA for 30 s in a GloQube (Quorum Technologies, Laughton, East Sussex, UK). Immediately after sample application, the grids were blotted in a chamber at 4°C and 100% humidity and then plunge‐frozen into liquid ethane.

Vitrified grids were transferred to a Talos Arctica (Thermo Fisher Scientific) operated at 200 kV and equipped with a Falcon 3 direct electron detector (Thermo Fisher Scientific). 2716 and 2601 movies were acquired for WT and a603‐NH mutant, respectively, at a nominal magnification of 120 000×, corresponding to a pixel size of 0.889 Å/pixel, in electron counting mode, with a nominal defocus range of −0.8 to −2.4 μm and with a total dose of 40 e⁻/Ų equally distributed on 40 frames. The Cryo‐EM experiments were conducted at the NoLimits Center of the University of Milan.

Cryo‐EM data processing and image reconstruction

All image processing and reconstruction steps were performed using CryoSparc v.4.4 (Punjani et al., 2017). The experimental workflows are outlined in Figs S2 and S3. After patch motion correction and Contrast Transfer Function (CTF) estimation, 2313 and 2257 manually curated micrographs were used for initial particle picking for WT and a603‐NH mutant, respectively.

For AtPSI‐WT, after a first round of reference‐free autopicking and 2D classification, the best classes were used for template‐based autopicking, resulting in an initial number of 575 271 particles. Particles were extracted with a box size of 448 px, binned 2 × 2, and subjected to several rounds of 2D classification. The best 2D classes (282 376 particles) were used for ab initio reconstruction (three classes) and heterogeneous refinement, resulting in an initial map at 3.74 Å resolution. After 3D classification, local and global CTF refinements were used to determine per‐particle defocus and to correct for higher order aberrations (beam tilt and trefoil) and anisotropic magnification, respectively. The best particles were re‐extracted at full size and used for a final round of nonuniform refinement (Punjani et al., 2020), obtaining a final density map at 3.13 Å resolution.

For AtPSI‐a603‐NH, 531976 particles generated from a first round of reference‐free autopicking were subjected to 2D classification, and the best classes were used to train the neural network in Topaz (Bepler et al., 2019) on a subset of 157 micrographs. The trained model was then used to pick the entire dataset, resulting in a total number of 115 589 particles. These particles were extracted with a box size of 448 px, binned 2 × 2, and after several rounds of 2D classification, the best particles were selected for ab initio reconstruction (3 classes) and heterogeneous refinement, resulting in an initial 4.19 Å resolution map. After two rounds of 3D classification, particles corresponding to the best 3D class were re‐extracted without downscaling and, following local and global CTF refinement as described above, further subjected to homogeneous and nonuniform refinement, yielding a final map with a resolution of 3.29 Å.

Before model building, the maps were sharpened with a global B factor of 83.7 Ų and 91.8 Ų for WT and a603‐NH mutant, respectively. The resolution was estimated by the ‘gold‐standard’ Fourier shell correlation (FSC = 0.143) criterion. Local resolution estimation was performed as implemented in CryoSPARC on the unsharpened map. The final EM maps (colored according to the local resolution) and the FSC curves are shown in Figs S2 and S3 (panels D and E).

Model building and refinement

The high‐resolution structure of the P. sativum PSI‐LHCI complex (Protein Data Bank (PDB) code: 7DKZ) was used as the starting model. After initial docking of the model in the map with UCSF ChimeraX (Goddard et al., 2018), each protein chain was independently rigid‐body fitted and refined with Phenix‐refine (Adams et al., 2010). The amino acid sequences of the different subunits were manually mutated to match the sequences of A. thaliana using COOT (Emsley & Cowtan, 2004), and the resulting model underwent several rounds of real‐space refinement in Phenix and manual rebuilding in COOT. Ligands and water molecules were modelled when unambiguously identified in the density map and refined to a reasonable B factor. Ligand restraints for refinement were generated with eLBOW (Moriarty et al., 2009). The stereochemical quality of the final model was assessed with MolProbity (Chen et al., 2010). Data collection and model refinement statistics are summarized in Table S3. High‐resolution figures were prepared with ChimeraX and PyMol (The PyMOL Molecular Graphics System, v.2.0 Schrödinger, LLC).

Energy transfer and excitonic coupling calculations

We used a point‐dipole approximation (PDA) to compute the excitonic couplings (EC), as described by van Amerongen & van Grondelle (2001), Müh et al. (2010), Liguori et al. (2015), and Sen et al. (2021). The Chl point dipoles were positioned at the geometric centers of the four nitrogen atoms of the chlorin ring (Friedl et al., 2022) $\mathbf{R}=\frac{1}{4}\left({\mathbf{R}}^{\mathrm{NA}}+{\mathbf{R}}^{\mathrm{NB}}+{\mathbf{R}}^{\mathrm{NC}}+{\mathbf{R}}^{\mathrm{ND}}\right)$.

The Chl Q_Y and Q_X transition dipole moments were aligned parallel to the nitrogen ND‐NB and NC‐NA axis, respectively, as in van Amerongen & van Grondelle (2001); Georgakopoulou et al. (2007); Liguori et al. (2015). The carotenoids (Car) point dipole was placed on the C15 atom, with transition dipole moments (transition S2 ← S0) oriented parallel to the central part of the polyene chain (atoms C11–C33), as modelled in previous studies (Georgakopoulou et al., 2007; Liguori et al., 2015). The Q_Y transition dipole moments were used for couplings between Chls (Chl a–Chl a, Chl a–Chl b, and Chl b–Chl b), while couplings between Chls and Cars were computed with the Q_X transition dipole moments for Chls (Croce et al., 2001; Polívka & Frank, 2010).

The ECs between two pigments i and j, in the PDA were calculated using the following formula (cm⁻¹):

$$V_{\mathit{ij}}=\frac{f_1^2\left|{\mathrm{\mu }}_i\right|\left|{\mathrm{\mu }}_j\right|}{{\mathrm{\epsilon }}_{\mathrm{r}}}\cdot 5.04\frac{{\kappa }_{\mathit{ij}}}{R_{\mathit{ij}}^3}$$

where f ₁ is the local field correction factor, $\left|{\mathrm{\mu }}_i\right|$ and $\left|{\mathrm{\mu }}_j\right|$ are the module of the transition dipole moment of pigments i and j, ε_r is the relative dielectric constant, here equal to 2.4 (van Amerongen & van Grondelle, 2001; Liguori et al., 2015) $R_{\mathit{ij}}$ is the module of the distance between the center of the dipole moment vector, and ${\kappa }_{\mathit{ij}}$ is the orientation factor between Chl i and j

$$k_{\mathit{ij}}=\left({\widehat{\mathrm{\mu }}}_i\cdot {\widehat{\mathrm{\mu }}}_j\right)-3\cdot \left({\widehat{\mathrm{\mu }}}_i\cdot {\widehat{\mathbf{r}}}_{\mathit{ij}}\right)\left({\widehat{\mathrm{\mu }}}_j\cdot {\widehat{\mathbf{r}}}_{\mathit{ij}}\right)$$

with ${\widehat{\mathrm{\mu }}}_i$ and ${\widehat{\mathrm{\mu }}}_j$ are the normalized transition dipole moment vector and ${\widehat{\mathbf{r}}}_{\mathit{ij}}$ is the normalized distance vector. Dipole moment values were taken as 4 D, 3.4 D, and 4.5 D for Chl a, Chl b, and Cars, respectively (van Amerongen & van Grondelle, 2001; Georgakopoulou et al., 2007; Liguori et al., 2015). In the case of ε_r = 2.4, (f ₁ ²μ²)/ε_r was calculated to be 17.6 D² for Chl a (Gobets & Van Grondelle, 2001). The calculations were run via homebuilt codes using Python 3.8.

Excited state calculations

We performed quantum mechanics/molecular mechanic (QM/MM) optimizations and polarizable QM/MM (Bondanza et al., 2020) excited‐state calculations of the Chls in the Lhca4 structure from the present work. Before QM/MM calculations, the structure was refined in a pure MM protocol as detailed in the Supporting Information. All Chls were optimized independently, except for a603–a609, which were optimized together. Excited‐state calculations were performed for all Chls in Lhca4 and for the a603–a609 dimer; control calculations were performed by also including L2 Vio with Chls a603 and a609. We employed a diabatization procedure to extract CT energies and couplings from the dimer calculations (Nottoli et al., 2018). Finally, we built an exciton model considering all Q_Y states of Chls and the CT states within the a603–a609 dimer, which we used to simulate absorption spectra of Lhca4 (Sláma et al., 2023). Detailed computational methods are reported in the SI (Methods S1).

Statistics

Statistical analyses were performed in OriginPro using one‐way analysis of variance (ANOVA); means were separated with Tukey's post hoc test at a significant level of P < 0.05 (see figure legends for details). Error bars represent the SD.

Results

The koLhca3 koLhca4 A. thaliana genotype was complemented with WT or mutant isoforms of Lhca3 and Lhca4, in which the Asn responsible for Chl a603 binding was replaced with a His (a603‐NH mutant) in both antenna subunits (Asn99 in AtLhca4 and Asn103 in AtLhca3). In the leaves of plants transformed with the a603‐NH mutation, we observed a shift in the 77 K fluorescence emission spectrum from 736 to 724 nm (Fig. S4).

PSI‐LHCI supercomplexes from both WT and a603‐NH mutant lines were purified by sucrose gradient ultracentrifugation (Fig. S1). RT absorption spectra confirmed the loss of far‐red spectral forms in the a603‐NH mutant complex, which peaked at 704 nm and extended to 750 nm. This was accompanied by an increased absorption in the 650–700 nm region (the negative peak at 679 nm visible in the difference spectrum; Fig. 2a). Low temperature (77 K) emission spectra revealed a 13‐nm blue shift in the a603‐NH mutant PSI supercomplexes compared with WT, its emission peaking at 721 nm (Fig. 2b). Emissions c. 720 nm can be attributed to the PSI core complex (Bassi & Simpson, 1987), leading us to conclude that the a603‐NH mutation effectively abolished Lhca‐associated RF in vivo, consistent with previous reports (Morosinotto et al., 2003; Li et al., 2023). No change in emission peak was observed in the complex purified from ko plants complemented with WT Lhca3 and Lhca4 (A3WT‐A4WT; Fig. S5a,b). The introduction of the a603‐NH mutation in either Lhca3 (A3NH‐A4WT) or Lhca4 (A3WT‐A4NH) induced spectral properties intermediate between WT and a603‐NH genotypes (Fig. S5c–e). Although the absorption of the PSI‐LHCI complex in the far‐red region (700–750 nm) was the same in the two single mutants (Fig. S5d), the low‐temperature emission of A3WT‐A4NH was slightly shifted toward blue (731 vs 733 nm; Fig. S5e), thus confirming that Lhca4 is the most red‐shifted LHCI, as previously reported (Morosinotto et al., 2003; Li et al., 2023).

Fig. 2.

Spectral analysis of PSI‐LHCI and LHCI from Arabidopsis thaliana wild‐type (WT) and a603‐NH (a, c) Room temperature (RT) absorption and (b, d) 77 K fluorescence emission spectra of PSI‐LHCI (exc. λ = 440 ± 5 nm) (a, b) and LHCI (c, d) from A. thaliana WT and a603‐NH. The WT minus NH difference spectra are shown as black lines in the corresponding plots. The amplitude of the absorption difference spectra was magnified by a factor of 3, while the fluorescence difference spectra were multiplied by a factor of 0.25 in order to plot them on the same axis. Key wavelengths are indicated in nm above the respective peaks. Experiments were repeated twice independently, with similar results. PS, photosystem; LHC, light harvesting complex.

To assess the impact of the a603‐NH mutation on the spectral properties of LHCI complexes compared with PSI‐LHCI supercomplexes, we purified LHCI heterodimers (Lhca1–Lhca4 and Lhca2–Lhca3) from WT and a603‐NH mutant. RT absorption spectra of LHCI complexes purified from the a603‐NH mutant showed the loss of a broad absorption tail, covering the range from 680 to 730 nm and peaking at 698 nm (Fig. 2c). The 77 K emission maximum of LHCI WT dimers was recorded at 728 nm, consistent with (Wientjes & Croce, 2011). By contrast, LHCI dimers from the a603‐NH line showed an emission peak at 689 nm, highlighting a significant blue shift of c. 39 nm (Fig. 2d), which aligns with observations made in vitro (Morosinotto et al., 2003). Conversely, the PSI core purified from both WT and a603‐NH lines exhibited unchanged optical properties, maintaining a peak at c. 720 nm (Fig. S6).

Structures of PSI‐LHCI from WT and a603‐NH plants

To investigate the structural determinants underlying the blue‐shifted absorption/emission caused by the a603‐NH mutation, we determined the Cryo‐EM structures of the A. thaliana PSI‐LHCI WT and a603‐NH supercomplexes. The final reconstructions, at 3.13 Å and 3.29 Å resolution, respectively, revealed well‐defined density maps for both the PSI core and the LHC subunits (Fig. S7), allowing for the construction of accurate models for the two complexes.

The structure of the individual subunits and the positioning of the chromophores within the AtPSI‐LHCI WT supercomplex were very similar to those observed in other land plants (Qin et al., 2015; Mazor et al., 2017; Iwai et al., 2024; Nelson, 2024; Figs S8–S13, Table S1). In both AtPSI‐LHCI WT and a603‐NH structures, the primary difference in pigment composition between the different Lhcas was found in the Chla : Chlb ratio (Fig. S14). Specifically, Lhca1/a2/a3 each bound 14 chl, while Lhca4 bound 15 Chls.

Chl a615 was coordinated by His168 and His151 from helix C in Lhca3 and Lhca4, respectively. These pigments showed clear density maps in both structures, allowing for accurate modelling of the Chls (including the chlorin rings and parts of the phytol tails) and the Vio molecule (Fig. S15). An additional Xan molecule (Lut) was located close to Chl a615 in Lhca4 only (Fig. S14), as previously observed also in the pea PSI‐LHCI structure (Qin et al., 2015). Since this Lut was absent in the red‐emitting subunit Lhca3, it was not considered a potential component of the red‐emitting cluster.

In the a603‐NH mutant structure, the bulkier His side chain at position 103 (Lhca3) and 99 (Lhca4) caused a c. 0.7 Å displacement of the Chl a603 chlorin ring (Figs 3, S15–S18), clearly visible in the corresponding density maps (Fig. S15). The chlorin ring and part of the phytol tail of Chl a603 were well defined in both the WT and a603‐NH density maps, allowing the reliable assignment of the position of Chl a603 in both structures (Fig. S15b,c). Surprisingly, the distance between the chlorin ring centers of Chl a603 and Chl a609 remained essentially unchanged across both WT and mutant proteins (9.3 Å in WT vs 9.1 Å in the a603‐NH for Lhca3, and 9.3 Å vs 9.2 Å for Lhca4), as well as the distance between the centers of the closest pyrroles of the two Chl (ring C), which changed slightly from 4.2 to 4.4 Å in Lhca3 and from 4.3 to 4.6 Å in Lhca4 (Fig. S16).

Fig. 3.

Structural superposition and excitonic coupling analysis of Lhca4 and Lhca3 wild‐type (WT) and a603‐NH. (upper part) Structures of WT and a603‐NH red clusters of Lhca4 and Lhca3 from Arabidopsis thaliana PSI‐LHCI. WT structures are colored in red and orange, while a603‐NH mutant structures are colored in blue and cyan. Lines are drawn between the pigments to highlight the excitonic couplings (EC) reported in the table (lower part, values in cm⁻¹). The thickness of the line is proportional to the EC value.

At the same time, the distance between Chl a603 and Vio in L2 increased from 5.2 Å in WT to 5.9 Å in a603‐NH, and from 5.3 Å to 6.0 Å in both Lhca3 and Lhca4, with a change of c. 12% in both cases.

We then computed EC values, which account for the relative spatial orientation between each pair of pigments in the cluster. The EC between Chl a603 and Vio decreased from 534 cm⁻¹ and 524 cm⁻¹ for Lhca3 and Lhca4 in the WT to 375 cm⁻¹ and 384 cm⁻¹ in the a603‐NH mutant, representing a change of 27–29% (Figs 3, S19). By contrast, the introduction of the a603‐NH mutation resulted in quite small changes in the EC values between the Chl a603 and Chl a609 pair (6–8 cm⁻¹). Furthermore, while the EC values between the Chl pairs were in the limited range of 87 cm⁻¹ and 101 cm⁻¹ for all Lhca, the EC values for the WT Chl a603–Vio L2 coupling were significantly higher for Lhca3 and Lhca4 than for Lhca1 and Lhca2, which do not show red‐shifted absorption. Notably, the a603‐NH mutation decreased the ECs of Lhca3 and Lhca4 to values comparable to those of the ‘non‐red’ subunits. Together, this evidence suggests that the Vio ligand might play a role in tuning the absorption properties of Lhca3 and Lhca4 toward low energy levels.

To investigate the role of the Vio at the L2 site of Lhca3 and Lhca4 in modulating RF‐inducing interactions, we compared the 77 K fluorescence excitation spectra of WT and a603‐NH PSI‐LHCI, highlighting the contribution of different wavelengths and associated chemical species to their far‐red fluorescence emissions. In the 480–510 nm region, where the largest contribution in absorption comes from Cars (Ashenafi et al., 2023), the a603‐NH mutant showed only slightly lower signal than WT (Fig. 4a), suggesting a small reduction in energy transfer efficiency from Car to Chls in the mutant (Fig. 4a). The difference between the 77 K absorption and excitation (Abs‐Exc) spectra in Fig. S20(a) provides insight on the efficiency by which the absorbed energy is transferred to the lowest energy emitter at each wavelength. In the 480–510 nm region, the Abs minus Exc values for the a603‐NH mutant were somewhat larger than those of the WT, suggesting a stronger energy dissipation and a reduced energy transfer toward Chl a, consistent with the lower calculated EC value (Fig. 3). We proceed to further investigate the role of Vio in promoting RF. Since the removal of Vio in site L2 is not feasible due to its essential role in protein folding and stability (Dall'Osto et al., 2013), we analyzed the effect of altering L2 occupancy on the PSI‐LHCI far‐red absorption. In the npq2 genotype, zeaxanthin (Zea) replaces Vio in all sites due to zeaxanthin epoxidase inactivation (Niyogi et al., 1998; Ballottari et al., 2014). 77 K fluorescence emission spectra showed a slight, c. 2 nm, blue shift in npq2 PSI‐LHCI vs WT (Fig. S21a). Notably, the Vio → Zea exchange also led to an apparent decrease in Chl a/b ratio in the PSI‐LHCI supercomplex, reflected in enhanced absorption between 462 and 492 nm, consistent with previous findings (Ballottari et al., 2014), which, however, was not conserved upon purification of the LHCI dimers (Fig. S21c,d). Pigment composition analysis of LHCI also revealed an altered Chl/Car ratio in the npq2 line, with a significant decrease in Lut content with respect to both WT and a603‐NH. This was partially compensated by the accumulation of Zea. In this view, the small spectral differences observed in npq2 compared with WT could also stem from secondary factors, such as altered carotenoid profiles (Fig. S21).

Fig. 4.

Spectral analysis of PSI‐LHCI and LHCI from Arabidopsis thaliana wild‐type (WT) and a603‐NH. (a) Low‐temperature (77 K) fluorescence excitation spectra of PSI‐LHCI from A. thaliana WT and a603‐NH. Far‐red fluorescence emission (734 ± 5 nm for WT and 721 ± 5 nm for the a603‐NH mutant) was followed by exciting samples from 400 to 550 nm. Spectra were normalized to the integrated area under the curve. (Solid lines represent the average of n = 3 independent biological replicates, broken lines represent the SD) The difference spectrum was magnified by a factor of 3 for better visualization. (b) circular dichroism (CD) (4°C) spectra of purified LHCI complexes from WT and a603‐NH mutants. Spectra are normalized to the same absorption in the Q_Y region. The difference spectra are shown as black lines in the corresponding plots. The experiments were independently repeated twice with similar results. PS, photosystem; LHC, light harvesting complex.

To analyze pigment‐pigment interaction in WT and a603‐NH LHCI complexes, we recorded CD spectra in the visible region (350–750 nm). In the Q_Y region, the spectra of both WT and a603‐NH line displayed signature peaks typical of LHC (i.e. −/+/−), indicating a similar and conserved structural conformation and pigment organization (Mozzo et al., 2008). The CD spectra of the WT and a603‐NH complexes revealed the largest difference in the far‐red region (λ > 700 nm), where the mutant showed a markedly reduced (−) signal associated with the excitonic interaction responsible for the RF (Fig. 4b). The difference spectrum (black line) evidenced the disappearance of a (−) low‐energy band in a603‐NH LHCI (690–730 nm), which was compensated by a high‐energy (+) band appearing at 683 nm. This conservative signal aligns with the loss of excitonic interactions between Chl a603 and a609, as previously reported (Morosinotto et al., 2003; Wientjes et al., 2012). In the Soret region, the interpretation of the CD spectra is complicated by the superimposition of signals from Chls and Cars. Minor differences are also present between 465 and 530 nm when comparing the a603‐NH with the WT. These differences, however, can be attributed to actual changes in pigment‐pigment interactions as well as to small losses of Cars during purification (Fig. S21d).

Overall, our results suggest that the Vio in L2, despite strong EC with Chl a603 in Lhca3 and Lhca4, has only a modest influence on the spectral properties and red‐light absorption of the a603–a609 pair.

Excited state calculations

To model the effect of the a603‐NH mutation on the red‐shifted absorption at the molecular level, we performed structure‐based polarizable QM/MM calculations of excited states on the Lhca4 WT and a603‐NH mutant models, including charge‐transfer excitations (Fig. 5a). In the spectra simulated with the standard exciton model (noCT in Figs 5c, S22), neither the WT nor the a603‐NH exhibited a signature of the RF observed in the experiments, notably the broad band peaking at > 700 nm. Conversely, by adding the CT contribution for the a603–a609 dimer in the calculations, we observed a low energy band (c. 710 nm) in the simulated spectrum of the WT (Fig. 5b), resembling the experimental observations in the isolated rLhca4 (Wientjes et al., 2012). By contrast, virtually no changes in the spectrum were observed for the mutant (Fig. 5c). This indicates that the major contribution to the WT/a603‐NH absorption shift resides in the different extent of the coupling of CT states to local excitations between Chl a603–a609, supporting a previous proposal (Wientjes et al., 2012; Sláma et al., 2023). CT has a similar effect on the energy of the exciton state regardless of the CT direction; therefore, we included both CT states (a603 + a609− and a603–a609+) in our calculations. It turned out that both CT states are coupled to the Q_Y states as found previously (Sláma et al., 2023), but the a603 + a609− is significantly lower in energy than the opposite state. We can therefore conclude that the lowest exciton state of WT Lhca4, responsible for the red forms, is a mixture of exciton and CT states, with a larger component of the a603 → a609 CT (Table S4).

Fig. 5.

Role of charge‐transfers (CTs) in enhancing far‐red absorption. (a) Schematic depiction of local excitations (left) and charge transfer excitations (right). In local excitations, electron transitions occur within orbitals of the same molecule; in charge‐transfer excitations, electrons are promoted from orbitals localized on one molecule to orbitals localized on another molecule. (b) Comparison of wild‐type (WT) and a603‐NH absorption spectra in the simulations (left) and experiments for isolated Lhca4 (right). (c) Effect of adding CT excitations to the exciton model in the a603‐NH mutant (left) and WT (right). In (b, c) the simulated spectra were rigidly shifted by −1800 cm⁻¹ to account for time‐dependent density functional theory (TD‐DFT) systematic error. (d) Energy of the lowest excited state in the a603–a609 dimer and in the a603–a609–Vio trimer. The experimental spectra of WT and a603‐NH are adapted with permission from Wientjes et al. (2012). Copyright © 2012 Elsevier B.V. All rights reserved.

A secondary effect can be noticed in the WT spectrum simulated without CT states (Fig. 5c), namely the appearance of a red‐shifted shoulder, corresponding to a603–a609 absorption. While this shoulder is clearly more blue‐shifted than the red band at c. 710 nm, it indicates red‐shifted site energies for a603–a609, in contrast with the a603‐NH mutant. Nonetheless, the model predicts the significantly red‐shifted band observed in the experiment only when including CT states.

We also assessed the electronic interactions between the Chl dimer and L2 Vio. To this end, we computed the excited states for two supermolecules, one formed by a603 and a609 (dimer), and the other formed by the same two Chls and the Vio (trimer). We investigated the energy of the lowest excited state responsible for fluorescence emission, to explore the effect on the electronic structure of including or excluding Vio (Fig. 5d). A change by c. 15 and 70 cm⁻¹ was observed when comparing the dimer to the trimer supermolecule, for both a603‐NH and WT, respectively, suggesting a marginal involvement of the Vio in L2 in tuning the low‐energy absorption of the cluster, consistent with our experimental results (Fig. 4). However, part of this difference is already explained by the varying treatment of the Vio (as point charges in the dimer vs an active molecule in the trimer). Therefore, such a difference is not large enough to conclude that Vio in L2 effectively mixes its orbitals with the two Chls or contributes to the CT excitation within the a603–a609 pair (Fig. S27). We conclude that the difference between the WT and the a603‐NH species (Δ) was not significantly changed by the inclusion of Vio.

The role of the extra chromophore Chl a615

Lhca3 and Lhca4 bound an additional Chl molecule, a615 (referred to as Chl a617 in Qin et al., 2015), which was coordinated by a His residue located on either the third or fourth turn of the C helix of Lhca3 and Lhca4, respectively (Fig. 3). This pigment was absent in the ‘blue’ subunits Lhca1 and Lhca2, leading to speculation about its involvement in RF formation (Melkozernov & Blankenship, 2003; Fig. 3), since the positioning of Chl a615 allowed favorable dipolar coupling with Chl a609. Notably, the EC value between Chl a615 and Chl a609 was 79 cm⁻¹ for Lhca3, whereas it dropped to 19 cm⁻¹ in Lhca4 due to the differing orientation of the chlorin ring in relation to Chl a609. Although His residues were also present in the second turn of the C helix of Lhca1 and Lhca2, the lack of electronic density from Cryo‐EM suggested that Chl a615 was absent in these subunits.

An additional Lut molecule was located near the chlorin ring of Chl a615 in Lhca4 (Fig. S23a), positioned to allow strong dipolar coupling (492 cm⁻¹) with this Chl. This Lut had previously been observed only in the Cryo‐EM structure of Zea mays PSI‐LHCI (PDB 5ZJI; Pan et al., 2018) and in the X‐ray structure of P. sativum PSI‐LHCI (PDB 4XK8; Qin et al., 2015). The Lut was situated at the interface between Lhca1 and Lhca4, in contact with both subunits, and one of its hydroxyl groups formed a hydrogen bond with the carbonyl oxygen of Ser210 in Lhca1 (Fig. S23b). Consequently, it might get lost during the purification of monomeric Lhca4.

To investigate the potential role of Chl a615 in far‐red light absorption, we generated mutants lacking this chromophore in both Lhca3 and Lhca4 by substituting the His‐binding residue with nonbinding Ala (a615‐H → A) or Ile (a615‐H → I; Fig. S24; Remelli et al., 1999; Guardini et al., 2022). Immunoblotting analysis of thylakoid membranes from the a615‐HA and a615‐HI mutant lines revealed a slight reduction in Lhca3 and Lhca4 protein levels (Fig. S26a,b). While this reduction was not statistically significant for Lhca3, it likely reflects a general decrease in the stability of Lhca complexes lacking Chl a615, rather than a transcriptional effect. This interpretation is supported by the comparable accumulation of Lhca proteins in A3WT–A4WT lines and WT Arabidopsis, and is consistent with previous in vivo observations of other Chl‐deficient LHC complexes (Guardini et al., 2020).

When recording 77 K emission spectra from intact leaves of genotypes with and without Chl a615, we observed a small blue‐shift from 736 nm to 733 ± 1 nm (Fig. 6a,b). However, in the isolated PSI‐LHCI complex, both genotypes showed the same emission at 734 nm (Fig. 6c). The c. 2–3 nm red‐shift in leaf samples compared with isolated supercomplexes can be attributed to self‐absorption effects (Weis, 1985). The a603‐NH mutant and koLhca3 koLhca4 exhibited a blue‐shifted λ_max to c. 724 nm, similar to the emission of the PSI core complex (Croce et al., 1998). We conclude that Chl a615 does not play a role in forming RF. To explain the 3 nm shift observed in leaf emission with and without Chl a615, we analyzed the pigment‐protein organization of thylakoids by sucrose gradient ultracentrifugation. Fig. S25 compares the fractionation patterns from solubilized thylakoids from the WT and Chl a615‐less mutants. In addition to the lowest (higher MW) band containing the fully assembled PSI‐LHCI supercomplex, a prominent green band containing the PSI core complex was present in both a615 mutant lines, while it was faint in WT (Wientjes et al., 2009). Consistently, the upper band (with the lowest MW) was enriched in the mutants compared with the WT. We interpreted these results as indicating a de‐stabilization of the dimeric Lhca1‐Lhca4 and Lhca2‐Lhca3 dimers consequent to the missing Chl a615. This was further confirmed by a second sucrose gradient fractionation upon treating the isolated PSI‐LHCI with Zwittergent 3–16, a procedure that allows the isolation of LHCI dimers from the PSI core (Figs 6d, S26). The pattern from the WT yielded both monomers and dimers of Lhcas, while only monomers were observed in both Chl a615 mutant lines. It is worth noting that Chl a615 was localized at the interface between Lhca subunits in both Lhca2‐Lhca3 and Lhca1‐Lhca4 subunits (Fig. S24), consistent with the hypothesis of its role in the stabilization of LHCI dimers and the binding to PSI core complex.

Fig. 6.

Spectral characteristic of different Chl‐binding mutants of Arabidopsis thaliana. (a) Fluorescence emission spectra measured on frozen intact leaves for different genotypes of A. thaliana: wild‐type (WT), koLhca3 koLhca4 lines complemented with WT sequences of Lhca3 and Lhca4 (A3WT‐A4WT), mutants lacking Chl a615 (a615‐HA and a615‐HI), koLhca3 koLhca4 (koA3A4), and a603‐NH mutant (a603‐NH), and normalized to the λ_max. (b) Barplot of the peak emission wavelength (λ_max), measured on the same genotypes. The values on the individual bars represent the mean λ_max in nm; the error bars correspond to the SD (n ≥ 5 independent biological samples). Values that are significantly different (ANOVA followed by Tukey's post hoc test at a significance level of P < 0.05) are marked with different letters. (c) PSI‐LHCI 77 K fluorescence emission spectra measured on the WT, a615‐HA, and a615‐HI genotypes (exc λ = 440 ± 5 nm). (d) Sucrose gradient fractionation of solubilized PSI‐LHCI from WT, a615‐HA, and a615‐HI plants. Five pigment‐containing bands were resolved and identified as: free‐pigments, monomeric LHCI, dimeric LHCI, PSI core complex, and PSI‐LHCI supercomplex. Experiments in panels (c) and (d) were repeated twice independently, with similar results. PS, photosystem; LHC, light harvesting complex.

Discussion

A key evolutionary trend in the green lineage is the enhancement of far‐red absorption in PSI‐LHCI, associated with the appearance of RFs. These adaptations likely emerged c. 489–403 m in response to light filtering under dense vegetation canopies (Fig. 1). Despite clear differences in spectral traits across taxa, the structural similarity of LHCI subunits (Iwai et al., 2024; Fig. 7) raises questions about the molecular determinants of far‐red absorption. The presence of an Asn as a ligand for the Chl a603 has been reported as required for red‐shifted states (Morosinotto et al., 2003; Wientjes et al., 2012; Li et al., 2023). However, this correlation proves overly simplistic. Indeed, in higher plants which exhibit the strongest far‐red emission (i.e. A. thaliana, Z. mays, and F. albivenis), Chl a603 is coordinated by Asn residues in both Lhca3 and Lhca4 (Pan et al., 2018; Li et al., 2024) while in C. reinhardtii, Chl a603 is coordinated by a His residue in Lhca3 and Lhca8 (the Lhca4 homolog; Naschberger et al., 2022). Furthermore, Asn appears in loosely bound Lhcas (Stauber et al., 2009; Su et al., 2019), which are located in more distal positions relative to the PSI core complex (Huang et al., 2021). In P. patens, which occupies an intermediate position both evolutionarily and in terms of absorption properties, an Asn residue coordinates Chl a603 only in Lhca3 (Gorski et al., 2022). Additionally, seagrasses have retained the Asn ligand while losing RF, suggesting that a603 coordination alone is insufficient to explain RF (Fig. 1b,c).

Fig. 7.

Comparative structural and excitonic coupling analysis of red‐shifted pigments across species. Structural superposition and excitonic coupling analysis (values in cm⁻¹) of the red cluster pigments from Fittonia albivenis (PDB 8WGH), Zea mays (PDB 5ZJI), Arabidopsis thaliana, Physcomitrium patens (PDB 7XQP), and Chlamydomonas reinhardtii (PDB 7ZQC). The analysis focused on (a) Lhca4 or the corresponding subunit, that is Lhca8 for C. reinhardtii and Lhca2b for P. patens (Yan et al., 2021; Gorski et al., 2022) and (b) Lhca3.

Our PSI‐LHCI a603‐NH mutant showed a c. 13 nm fluorescence blue shift at 77 K. Upon isolation of LHCI, we observed a c. 39 nm emission shift, confirming the direct effect of the Asn → His substitution on the antenna itself (Fig. 2). Cryo‐EM structures revealed a c. 0.7 Å repositioning of the a603 chlorin ring in the a603‐NH mutant (Fig. S16), weakening coupling with neighboring pigments, particularly Vio in site L2 (Fig. 3). While EC calculations suggested a possible role for Vio L2 in the tuning of the excitonic interaction originating RFs, the QM/MM excited‐state simulations revealed that the low‐energy absorption is primarily due to CT states between a603 and a609 (Wientjes et al., 2012; Sláma et al., 2023; Fig. 5). Although L2 Vio does not participate directly (quantum‐mechanically) in the lowest exciton state (Figs 3, 5), it seems to be important in influencing local pigment geometry, and its EC value with Chl a603 could still be an indicator to predict the presence of RFs (Fig. 7).

Additionally, Chl a615, close to the red cluster (Melkozernov & Blankenship, 2003; Qin et al., 2015), showed strong coupling with a609 (Fig. 3) but its removal did not affect far‐red emission (Fig. 6c); rather, it appears to play a structural role in LHCI dimers stabilization and in their association with PSI core (Figs 6d, S23b).

In conclusion, our integrated structural, spectroscopic, and computational analyses support a model where CT interactions between a603 and a609 are the primary drivers of far‐red spectral features (Sláma et al., 2023). Other pigments and protein environments modulate, but do not determine, the presence of RF.

Further research will be necessary to better clarify the contribution of the protein environment surrounding the ‘red cluster’, for example, by comparing high‐resolution structures of red‐shifted and blue‐shifted Lhcas, such as from F. albivenis (Li et al., 2024) and seagrasses.

Understanding these structure–function relationships is essential for future efforts aimed at fine‐tuning Chl absorption toward underutilized spectral regions, an approach with strong potential to enhance photosynthetic efficiency under real field conditions (Ort et al., 2015; Cutolo et al., 2023).

Competing interests

None declared.

Author contributions

RB, SC and LD'O conceived the work and designed the experiments. ZG, VFP and AA carried out the construction of mutants and performed their biochemical and spectroscopical characterization. ZG, SC and AA carried out the preparation of the samples for Cryo‐EM. DMVB and AC‐S conducted the Cryo‐EM sample preparation and data collection. SC and DM analyzed the Cryo‐EM data and reconstructed the PSI‐LHCI structures. DM performed the bioinformatics analysis and EC calculations. EB, CJ, LP‐G, LC and BM carried out the quantum chemical calculations and analyses. DM, SC and ZG wrote the original text draft. SC, ZG, DM and VFP contributed equally to this work. All authors discussed the results and contributed to drafting the manuscript.

Disclaimer

The New Phytologist Foundation remains neutral with regard to jurisdictional claims in maps and in any institutional affiliations.

Supporting information

Fig. S1 Sample preparation and characterization.

Fig. S2 Cryo‐EM workflow, images and map quality for AtPSI‐LHCI WT.

Fig. S3 Cryo‐EM workflow, images and map quality for AtPSI‐LHCI a603‐NH.

Fig. S4 Emission spectra of leaves of Arabidopsis thaliana WT and a603‐NH mutant.

Fig. S5 Characterization of PSI‐LHCI purified from WT, A3WT‐A4WT, A3NH‐A4WT and A3WT‐A4NH mutant lines.

Fig. S6 Characterization of solubilized PSI‐LHCI from WT and a603‐NH.

Fig. S7 Atomic models of PSI‐LHCI subunits and selected ligands superimposed on Cryo‐EM maps.

Fig. S8 Overall architecture of the PSI‐LHCI WT supercomplex.

Fig. S9 Positions of ligands in the PSI‐LHCI WT of Arabidopsis thaliana (PDB 9GBI).

Fig. S10 Lipid arrangement in PSI‐LHCI supercomplexes.

Fig. S11 Superposition of the PSI WT from Arabidopsis thaliana (PDB 9GBI, white) and from P. sativum (PDB 7DKZ).

Fig. S12 Global superposition RMSD of the AtPSI‐WT (PDB 9GBI) structure with PSI‐WT structures from PDB 8J7B, 8JZA (both from Cryo‐EM data), and 7DKZ.

Fig. S13 Chl, xanthophyll and lipid arrangement in PSI‐LHCI WT.

Fig. S14 Pigment content of the LHCI antenna subunits of Arabidopsis thaliana WT (PDB 9GBI).

Fig. S15 Atomic models of the ‘red cluster’ pigments.

Fig. S16 Distances between Chl a603, Chl a609, and Violaxanthin L2 in Lhca3 and Lhca4 subunits of Arabidopsis thaliana.

Fig. S17 Superposition of selected Chls and xanthophylls of the Lhca3/Lhca4 WT and a603‐NH.

Fig. S18 Superposition of selected Chls and xanthophylls of the Lhca1‐4 WT and a603‐NH.

Fig. S19 Excitonic coupling absolute values between pigments in LHCI WT and a603‐NH.

Fig. S20 Difference spectra (absorption – excitation spectra, measured at 77 K in the 400–550 nm region) of PSI‐LHCI from Arabidopsis thaliana WT and a603‐NH.

Fig. S21 Spectral analysis and pigment composition of PSI‐LHCI from WT and npq2.

Fig. S22 Site energies and simulated absorption spectra.

Fig. S23 Structural analysis of the Chl a615‐Lutein cluster.

Fig. S24 Structural diagram of PSI‐LHCI supercomplex, Lhca3 and Lhca4 from WT and the a615‐HI line.

Fig. S25 Sucrose gradient fractionation of thylakoid membranes of WT and Chl a615 mutant lines.

Fig. S26 Characterization of Chl a615 mutant lines.

Fig. S27 Visual representation of the overlap between LUMO orbitals of Chls a603 and a609 as computed for the WT and a603‐NH structures.

Fig. S28 Alignment of the sequences (exons and introns) of the synthetic genes.

Fig. S29 Superposition of Cryo‐EM structures of Lhca4 WT or a603‐NH mutant with the optimized models used for QM/MM calculations.

Methods S1 Additional methods.

Table S1 AtPSI‐a603‐NH structural model.

Table S2 List of the primers used to obtain and characterize the a603‐NH mutant lines.

Table S3 Cryo‐EM data collection, refinement, and validation statistics.

Table S4 Energies and couplings of the Q_Y and CT states of WT and a603‐NH.

Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Data availability

Sequence data from this article can be found in the Arabidopsis Genome Initiative under accession nos.: At1g61520 (LHCA3), At3g47470 (LHCA4), and At5g67030 (ZE). The KO lines were obtained in the NASC under stock nos.: N876497 (koLhca3) and N679009 (koLhca4). Mutant line npq2 was a kind gift of Prof. K.K. Niyogi (University of California at Berkeley). The Cryo‐EM maps and coordinates have been deposited in the EMDB and wwPDB, respectively: PSI–LHCI WT (Cryo‐EM map, EMD‐51219; consensus refinement map; PDB: 9GBI) and PSI–LHCI a603‐NH mutant (Cryo‐EM map, EMD‐51227; PDB: 9GC2).