Cryo-EM structure of a photosystem I supercomplex from Euglena gracilis containing pentameric LHCE and dimeric LHCII

Abstract

Euglenophyta is a representative phylum of green-lineage secondary endosymbiotic eukaryotes. These organisms have evolved far-red light-harvesting complexes (LHCs) composed of diadinoxanthin and chlorophyll a/b, which are now classified as the LHCE family. Here we report a 2.35-Å cryo-electron microscopy structure of photosystem I (PSI) supercomplex from Euglena gracilis, with all subunits and their paralogs assigned. This structure reveals a minimal PSI core associated with twelve LHCE and four LHCII subunits. Most LHCE subunits are organized into dimers through a helix C–to–helix C interaction. Two dimers, together with a monomeric LhcE8, assemble into a (2 + 2 + 1)-type LHCE pentamer. The red-shifted pairs in the two LhcE6 subunits likely contribute to far-red absorption. The LHCII subunits form a distinct heterodimer and associate with the PsaF side. Overall, these results provide a structural basis for understanding energy transfer and dissipation, antenna subunit assembly, and far-red light-harvesting strategies in green-lineage secondary endosymbiotic organisms.

Euglena gracilis is a representative green-lineage secondary endosymbiont. Here, the authors present a 2.35-Å cryo-EM structure of its photosystem I supercomplex. The structure reveals the organization of diadinoxanthin-rich LHCE antenna subunits.

Introduction

Eukaryotic photosystem I (PSI) is a membrane pigment–protein supercomplex composed of a reaction center surrounded by light-harvesting antenna complexes (LHCs). It functions as a highly efficient energy conversion apparatus, generating reducing power to drive the reduction of ferredoxin¹. Cyanobacterial PSI is considered the evolutionary ancestor of all eukaryotic PSI supercomplexes². Eukaryotic photosynthetic organisms, shaped by multiple endosymbiotic events, are broadly classified into the red and green lineages^3,4.

In red algae and red-lineage secondary endosymbiotic organisms, PSI–LHCI supercomplexes exhibit substantial structural diversity in their LHC-binding sites^5,6. In contrast, the LHC-binding sites in plants^7–9 and green algae^10–12, where four to ten LHCI subunits associate with the PSI core, are relatively conserved. The minimal PSI complex¹³, which lacks PsaG, PsaH, PsaK, PsaL, and PsaO, has been proposed as a structural “stepping stone” for the assembly of green PSI–LHCI supercomplexes¹⁴. Specifically, PsaG and PsaK are critical for binding the conserved LHCI belt¹⁵, whereas the PsaO, PsaL, and PsaH enable the association of phosphorylated LHCII with PSI during state transitions^16–18. Nevertheless, high-resolution structures of PSI supercomplexes from green-lineage secondary endosymbiotic organisms remain unclear.

Euglena gracilis belongs to the Euglenophyta, a lineage derived from secondary endosymbiosis involving the engulfment of a green-lineage photosynthetic organism (a green alga)^3,4, and utilizes conserved chlorophyll a/b for efficient energy transfer¹⁹. However, unlike typical green-lineage organisms, which employ the violaxanthin–antheraxanthin–zeaxanthin (VAZ) cycle for photoprotection^20,21, E. gracilis has adopted the diadinoxanthin–diatoxanthin (Ddx–Dtx) cycle²²; this cycle is commonly found in planktonic red-lineage chlorophyll a/c-containing marine algae^21,23–25.

In addition, Euglena has developed a chlorophyll a far-red–absorbing antenna complex in E. gracilis–LHCE²⁶, which is distinct from the Lhca and Lhcb proteins found in green plants, as well as Lhcf and Lhcr proteins in red-lineage species¹⁹. LHCE subunits have been shown to possess far-red light absorption capabilities. Five LHCE protein species (E1–E4 and E12) from the peripheral antenna pool are proposed to form a pentameric assembly, as suggested by low-resolution negative-stain EM data²⁶. LhcE9 was suggested to be associated with PSII, and most LHCE subunits (E5–8, 10-13) are mainly distributed in PSI, along with 2–3 LhcbM proteins²⁶.

In this work, to elucidate the high-resolution structure of PSI from a secondary endosymbiotic photosynthetic organism of green algal origin, and to understand the structural basis of its association with far-red–absorbing LHCE antenna supercomplexes, as well as their pigment network. Here, we purify the PSI supercomplex from Euglena gracilis (Eg) and determine its structure using single-particle cryo-electron microscopy (cryo-EM) at an overall resolution of 2.35 Å, with improved local resolution in the peripheral antenna regions achieved through seven rounds of local refinement. As a result, we model the Eg-PSI–LHCE–LHCII supercomplex, which comprises two LHCE pentamers, one LHCE dimer, one LHCII dimer, and two LHCII monomers. This structure provides a structural basis for understanding excitation energy transfer and dissipation, antenna subunit assembly, and far-red light-harvesting strategies in the PSI supercomplex of E. gracilis.

Results

Overall structure of the PSI–LHCE–LHCII supercomplex

We purified the PSI-LHCE-LHCII supercomplex from E. gracilis cells (GY-D32, Guangyu Biotech Co., Ltd., China) (Methods, Supplementary Fig. 1a) and confirmed the presence of LHCE and LHCII subunits through SDS-PAGE and mass spectrometry (Source data, Supplementary Fig. 1b). Under room temperature, the absorption spectrum of this supercomplex shows a Qy band peaking at 676 nm (Supplementary Fig. 1c), while the 77 K fluorescence emission spectrum peaks at 733 nm (Supplementary Fig. 1d)—approximately 20 nm red-shifted compared to typical green algal PSI¹⁸ but similar to plant PSI which contains red chlorophyll pairs^7,9,27. This red shift corresponds to the red-shifted spectral properties of the abundant LHCE subunits incorporated within the PSI supercomplex²⁶, and the emission spectrum shoulder at 686 nm corresponds to the LHCII subunits associated (Supplementary Fig. 1d). High-performance liquid chromatography (HPLC) analysis revealed the Eg-PSI-LHCE-LHCII supercomplex contains five major pigment species: chlorophyll a (Chl a), chlorophyll b (Chl b), diadinoxanthin (Ddx), neoxanthin (Neo) and β-carotene (Bcr) (Supplementary Fig. 1e).

The isolated Eg-PSI–LHCE–LHCII supercomplex was imaged by single-particle cryo-EM at 300 kV, and the data were processed using cryoSPARC²⁸. Following Ab-initio reconstruction, 1,059,170 PSI particles were selected. A second round of 2D classification was then performed based on the heterogeneous refinement, and 51,871 particles were selected, which were subsequently refined to a resolution of 2.35 Å. To improve the local density in the peripheral antenna regions, seven rounds of masked focused refinement were carried out on distinct peripheral regions, resulting in reconstructions from 2.60 Å to 3.20 Å resolution (Supplementary Fig. 2 and 3). A composite cryo-EM map was generated from these refinements and used for model building and refinement (Supplementary Fig. 4 and 5; Supplementary Table 1; Methods). We built the initial model using ModelAngelo²⁹ and CryoAtom³⁰, as well as the recently reported full-length LHC sequences of E. gracilis^26,31,32. We identified all subunits and their paralogs and performed manual model fitting, guided by sequence differences^26,31 and the local cryo-EM density.

The model of Eg-PSI–LHCE–LHCII supercomplex accommodates twelve LHCE subunits, comprising two LHCE pentamers and one LHCE dimer, four LHCII subunits, comprising one LhcbM dimer and two LhcbM monomers, and a minimal PSI core. LHCE pentamer 1 consists of LhcE5.4, LhcE5.7, LhcE6.2, LhcE7.3, and LhcE8.2; Homologous LHCE pentamer 2 consists of LhcE5.7, LhcE5.4, LhcE6.4, LhcE7.2, and LhcE8.2. The LhcE8 paralog in Pentamer 2 was tentatively assigned as LhcE8.2 based on local density fit and sequence homology to the corresponding subunit in Pentamer 1, as the resolution was insufficient to unambiguously distinguish between LhcE8 paralogs. Pentamer 1 and pentamer 2 are located on the PsaA and PsaB sides of the PSI core, respectively. The LHCE dimer is composed of LhcE10.2 and LhcE11.1 and is positioned at the interface between the PsaA and PsaJ subunits. The LHCII dimer comprises LhcbM2.1 and LhcbM8.2. The two LhcbM4 monomers are distributed separately, LhcbM4.6 associates directly with the PsaB side, facilitated by LhcbM2.1, while LhcbM4.10 is inserted between the two LHCE pentamers (Fig. 1a).

Fig. 1. Overall structure of the PSI–LHCE–LHCII supercomplex from E.gracilis.

a Structure of the PSI–LHCE–LHCII supercomplex viewed from the stromal side along the membrane. Antenna subunits are labeled on the left; PSI core subunits and the overall antenna assembly are labeled on the right. b Side view of the supercomplex, highlighting the stromal components PsaC, PsaD, and PsaE, and the antenna subunits are colored as in (a).

The minimal PSI core consists of only eight subunits: PsaA, PsaB, PsaC, PsaD, PsaE, PsaF, PsaJ, and PsaM (Fig. 1a, b; Supplementary Fig. 4a). Compared with the “stepping stone” minimal PSI¹³, the PSI core in E. gracilis employs an additional PsaM subunit and an extended PsaD (Supplementary Fig. 6). Notably, the extended PsaD and the additional PsaM mediate the association of LHCE pentamer 1 and LHCE pentamer 2 with the PSI core at the PsaA and PsaB sides, respectively (Fig. 1). The cryo-EM map also reveals 289 chlorophylls (Chls), 56 carotenoids (Cars) and 5 cofactors and 51 lipids, including 281 Chls a, 8 Chls b, 12 Bcrs, 42 Ddxs, and 2 Neos, 2 phylloquinones, 3 Fe₄S₄ clusters, 3 digalactosyldiacylglycerol (DGDG) molecules; 21 monogalactosyl-diacylglycerol (MGDG) molecules; 23 phosphatidylglycerol (PG) molecules and 4 sulfoquinovosyldiacylglycerol (SQDG) molecules (Supplementary Table 2).

Structure and assembly of the peripheral LHCE subunits

Compared with typical green-lineage Lhca and Lhcb proteins^7,33,34 and red-lineage FCPs^23,35,36, all LHCE subunits identified in this study form a distinct evolutionary clade of antenna proteins (Supplementary Fig. 7). In detail, most LHCE subunits exhibit an extended AC-Loop (except LhcE8) and a shorter BC-Loop (except LhcE10) (Fig. 2a, Supplementary Fig. 8a).

Fig. 2. Structures and pigment organization of LHCE subunits and their assembly within the Eg-PSI–LHCE–LHCII supercomplex.

Chls a and Ddxs are shown as green and blue sticks, respectively; only rings of Chls a are displayed, with phytol chains omitted. Lipids are shown in orange as ball and sticks. a Structures and pigment arrangements in each LHCE subunit. b LHCE dimers (LhcE5–E6, LhcE5–E7, and LhcE10–E11) assemble via a helix C–to–helix C interaction. c The LHCE pentamer adopts the (2 + 2 + 1) manner, comprising two LHCE dimers and a monomeric LhcE8.2. Associated lipid molecules are labeled, and interactions are indicated by black lines.

All the LHCE subunits bind only two types of pigments: Chl a and Ddx. Each LHCE subunit combines two conserved Ddx-binding sites (Ddx616 and Ddx617) and eleven conserved Chl a-binding sites (Chl a601–604 and 608–614) (Supplementary Table 3). Notably, LhcE6 and LhcE7 coordinate an additional Chl a615 via an asparagine residue (Fig. 2a, Supplementary Fig. 8a, b). LhcE6–LhcE8 further coordinate an additional Chl a622 through a glutamate residue (Fig. 2a, Supplementary Fig. 8a, c). LhcE10.2 utilizes its extended BC-Loop to bind an extra Chl a621 (Fig. 2a, Supplementary Fig. 8a, d). As shown in Supplementary Fig. 7, phylogenetic analysis reveals that the LHCE subunits cluster into a distinct clade and share higher sequence similarity with adjacent Lhca proteins from green-lineage C. reinhardtii and plants than with red-lineage Lhcr proteins. This observation supports the evolutionary homology between LHCE and the light-harvesting complex family of green-lineage photosynthetic organisms. In addition, the chlorophyll-coordinating residues and carotenoid-binding sites of LHCE subunits also show homology to those in C. reinhardtii LHCI (Supplementary Fig. 9).

Most LHCE subunits employ the extended AC-Loop and utilize the Aspartic acid–Arginine–Serine (DRS) polar domain and tryptophan to form dimeric assemblies (Supplementary Fig. 8a, e). Three LHCE dimers: LhcE5–LhcE6, LhcE5–LhcE7, and LhcE10–LhcE11, were identified in this study (Fig. 2b). The LHCE dimer adopts a helix C–to–helix C interaction (Fig. 2b, Supplementary Fig. 8e), which differs from the helix A–to–helix C arrangement observed in green algae and plants^7,9,10,33 but resembles the dimeric organization of red-lineage diatom FCPs²³ (Supplementary Fig. 10a–c). In addition, the two conserved Ddx sites in LHCE subunits are similar to those of Ddx303 and Ddx305 in the dimeric FCP encoded by Lhcf7³⁵. In contrast, LhcE8.2 lacks the DRS polar domain and tryptophan therefore cannot form a stable dimer (Supplementary Fig. 8a, f).

Furthermore, two dimers: LhcE5–LhcE6 and LhcE5–LhcE7, along with a monomeric LhcE8, assemble into a (2 + 2 + 1)-type LHCE pentamer (Fig. 2c). This pentameric organization is mediated by the N-terminus of LhcE6 and LhcE8, which insert into the dimeric interfaces of the LhcE5–LhcE7 and LhcE5–LhcE6 dimers (Fig. 2c). The extended C-terminus of LhcE6 and LhcE8 are critical for pentamer stability (Fig. 2a, c; Supplementary Fig. 8a). Four MGDG and two PG lipid molecules are embedded at the subunit interfaces and strengthen the inter-subunit interactions (Fig. 2c). The LHCE pentamer resolved in this study differs from the previously reported (4 + 1)-type FCP pentamer in diatoms (FCP tetramer and one monomer)³⁷ (Supplementary Fig. 10d), and is also distinct from the hypothetical LHCE pentamer proposed under low-resolution conditions from negative stain²⁶. In contrast, the LhcE10–LhcE11 dimer, due to its extended BC-Loop, sterically conflicts with adjacent subunits and therefore cannot participate in pentamer formation (Supplementary Fig. 8a, g).

The two homologous LHCE pentamers (P1 and P2) are located on the PsaA and PsaB sides of the PSI core, respectively (Fig. 1), with only minor structural differences arising from shifts in LhcE5.4 and LhcE5.7 (Supplementary Fig. 11). Notably, an extra Ddx618 is clearly resolved in LhcE8.2 of pentamer 1 (Supplementary Fig. 5a), where it is tightly bound at the interface between LhcbM4.10 and the LHCE pentamer. The density for Ddx618 is absent in pentamer 2, suggesting that it may be loosely bound in the peripheral region of the supercomplex.

Structure and assembly of the peripheral LHCII subunits

In plants and green algae, the PSI–LHCI supercomplex can associate with phosphorylated LHCII trimers during state transitions to form the PSI–LHCI–LHCII supercomplex^{16–18,38,39}, thereby enabling the qT-type photoprotective mechanism^40,41. In mosses, the PSI–LHCI–LHCII supercomplex has also been proposed to enhance overall light harvesting under low-light conditions^42,43.

In this study, the Eg-PSI core employs a minimal structure and lacks several key subunits, including PsaO, PsaH, PsaL and PsaK, that are essential for LHCII binding in previously reported PSI–LHCI–LHCII complexes^16–18,38. Despite this loss, we identify a previously unobserved LHCII heterodimer composed of LhcbM2.1 and LhcbM8.2. The monomer–monomer interaction within this heterodimer resembles that of the conserved LHCII trimer (Fig. 3a, b), and the heterodimer binds to the PsaF side of the PSI core (Fig. 1). PsaF has recently been suggested to function as a regulatory subunit that promotes the assembly of peripheral antenna complexes in PSI⁸.

Fig. 3. Structures and pigment organization of LHCII subunits in the Eg-PSI–LHCE–LHCII supercomplex.

Chls a, Chls b, Ddxs, and Neos are shown as green, cyan, blue, and pink sticks, respectively. Only the rings of Chl a and Chl b are displayed; phytol chains are omitted. a, b Structures and pigment arrangements in the two dimeric LHCII subunits, whose monomer–monomer interfaces resemble those of the conserved LHCII trimer (PDB: 1RWT). c Structural comparison and pigment organization of the homologous subunits LhcbM4.6 and LhcbM4.10.

In the heterodimer, both LhcbM2.1 and LhcbM8.2 coordinate an additional Chl a620 in their C-terminal regions (Fig. 3a, b). Compared with the LHCE subunits in this study, all Eg-LHCII subunits bind two conserved Car sites, 616, 617 and an extra 619 site. In the plant, mosses and green algal LHCII heterodimer, the conserved car619 site mostly modeled as neoxanthin^34,42,44. In contrast, in this study, only two car619 sites in two LhcbM4 subunits could be modeled as Neo619 (Fig. 3c), based on the characteristic C29 chiral center (Supplementary Fig. 5a), two car619 sites in LHCII heterodimer are tentatively modeled as Ddx619 (Fig. 3a, b). Compared with plant and green algal LHCII subunits, which typically bind eight Chls a and six Chls b molecules^34,44, only two coupled Chl b sites, Chl b606 and Chl b607, in the Eg-LHCII subunits could be tentatively identified (Fig. 3 and Supplementary Fig. 5b).

Interactions between antenna subunits and the PSI core

Interactions between antenna subunits and PSI core subunits were analyzed using van der Waals (VDW) contact analysis³⁵ (Methods, Supplementary Fig. 12a). In addition, lipid molecules play crucial roles in stabilizing the oligomerization interfaces within the photosynthetic supercomplex⁴⁵. We also characterized the distribution of lipid-binding sites in the PSI–LHCE–LHCII supercomplex, revealing their essential contributions to structural stability (Supplementary Fig. 12b).

In the complete PSI–LHCE–LHCII supercomplex, the two homologous LHCE pentamers (P1 and P2) are located on the PsaA and PsaB sides of the PSI core, respectively (Figs. 1 and 4a). LHCE pentamer 1 associates with the PSI core primarily through interactions between LhcE6.2 and LhcE7.3 of P1 and PsaA, facilitated by the extended N-terminus of PsaD (Supplementary Fig. 6). Six lipid molecules, inserted at the interface, strengthen these interactions (Fig. 4b), including SQDG 861, MGDG 624, 862, 863, 873, and PG 625.

Fig. 4. Interactions between antenna subunits and the PSI core.

a The Eg-PSI–LHCE–LHCII supercomplex viewed from the stromal side. The regions enclosed by black dotted squares are magnified in (b–f). PSI and antenna subunits involved in protein–protein interactions are colored as in Fig. 1. Amino acid residues and lipid molecules mediating the interactions are labeled in (b–f). Lipids are shown in orange as ball and sticks. b Interactions between LHCE pentamer 1 and the PSI core, viewed from both stromal and lumenal side. c Interactions between LHCE pentamer 2 and the PSI core, viewed from both stromal and lumenal side. d Interactions between LhcbM4.10 with LhcE8.2 (pentamer 1), and LhcE7.2 (pentamer 2), viewed from the stromal side. e Interactions between the LHCII dimer and the PSI core, viewed from the stromal side. f Interactions among LhcbM4.6, adjacent LhcbM2.1, and the PSI core, viewed from the stromal side. g Interactions between the LHCE dimer and the PSI core, viewed from the lumenal side.

For the LHCE pentamer 2 on the PsaB side, interactions between LhcE7.2 of P2 and PsaB on the stromal side are relatively weak. Instead, stable binding is mainly mediated by LhcE6.4 and LhcE7.2 of P2 interacting with PsaB and PsaM on the lumenal side. Two lipid molecules (SQDG856 and MGDG625) embedded at the lumenal interface strengthen the interactions (Fig. 4c). Between the two homologous LHCE pentamers, LhcbM4.10 is inserted into the interface between LHCE pentamer 1 and pentamer 2, strengthening their interactions with the assistance of three lipids: PG620, PG627, and SQDG623 (Fig. 4d).

The LhcE10–LhcE11 dimer associates directly with the PSI core through the extended BC-Loop of LhcE10.2 (Fig. 4e, Supplementary Fig. 8a), LhcE11.1 interacts with PsaA, facilitated by the lipid MGDG624 on the lumenal side (Fig. 4e). The LHCII heterodimer utilizes three lipid molecules (MGDG406, MGDG409, and SQDG407) to associate with PsaF and PsaJ of the PSI core (Fig. 4f). In addition, LhcbM4.6 forms interactions with the adjacent LhcbM2.1 subunit and binds to PsaB, facilitated by two lipid molecules, MGDG410 and MGDG852 (Fig. 4g).

Excitation energy transfer pathways within the LHCE pentamer and from peripheral antenna subunits to the PSI core

In this study, four Chl b606–Chl b607 pairs were assigned to the four LHCII subunits (Supplementary Table 3), as the additional carbonyl oxygen of each Chl b participates in a stable hydrogen-bonding network between the two pigments, producing a discernible density feature in the map (Supplementary Fig. 5b). Other potential Chl b sites in LhcbM subunits were tentatively modeled as Chl a, guided primarily by the higher overall abundance of Chl a indicated by HPLC results (Supplementary Fig. S1e). All chlorophyll-binding sites in the LHCE subunits are modeled as Chl a. The minimal PSI core, together with two LHCE pentamers, one LHCE dimer, one LHCII dimer, and two LHCII monomers, forms efficient excitation energy transfer (EET) pathways within the complete Eg-PSI–LHCE–LHCII supercomplex. We analyzed these EET pathways using Förster resonance energy transfer (FRET) analysis^46,47 (Methods, Supplementary Fig. 13, Supplementary Table 4 and 5). The FRET parameters used in this study were derived from experimentally validated FRET constants obtained from previously reported spectroscopic models of Chls a and b in plant light-harvesting systems^46,47. These parameters have been widely adopted in the field for modeling energy transfer in organisms containing Chls a and b, such as Chlamydomonas³⁴ and moss⁴³. The point–dipole approximation becomes unreliable for Mg–Mg distances below ~10 Å. Chlorophyll pairs within this range were excluded from quantitative FRET analysis due to potential inaccuracies in the model and their likely departure from the Förster regime. At such short distances, excitonic coupling and sub-picosecond energy transfer may occur, processes that are not adequately described by incoherent Förster theory^48,49. To ensure the validity of our FRET analysis, our analysis is restricted to inter-subunit energy transfer pathways with Mg–Mg distances greater than 10 Å.

In the (2 + 2 + 1)-type LHCE pentamer of the complete supercomplex (Fig. 5a, b), EET pathways within the LHCE dimer are mediated by Chl a609–Chl a609 pairs, homologous with the Chl a406–Chl a406 pairs in diatom FCP²³. EET pathways from dimer to dimer and from dimer to monomer are mediated by the Chl a615 in LhcE7 to Chl a613 in LhcE6, and from Chl a615 in LhcE6 to Chl a613 in LhcE8, respectively (Fig. 5b).

Fig. 5. Pigment arrangement and potential EET pathways in the Eg-PSI–LHCE–LHCII supercomplex.

a Pigment arrangement in the PSI–LHCE–LHCII supercomplex, viewed from the stromal side. Antenna subunits are colored as in Fig. 1; pigments are colored as in Figs. 2 and 3. Only the rings of Chls a are shown; phytol chains are omitted. Black dotted squares indicate regions magnified in (b–h). EET pathways are denoted by dashed lines. The pigment distances and coupling values were labeled in blue. b EET pathways between LHCE dimers, and EET pathways within the LHCE pentamer. Inter-dimer pathways are indicated by black lines. c EET pathways from LHCE pentamer 1 to the PSI core. d EET pathways from LHCE pentamer 2 to the PSI core. e EET pathways between LhcbM4.10 with LhcE8.2 (pentamer 1) and LhcE7.2 (pentamer 2). f EET pathways from the LHCII dimer to the PSI core. g EET pathways from LhcbM4.6 to the PSI core. h EET pathways from the LHCE dimer to the PSI core.

EET pathways from LHCE pentamer 1 to the PSI core are mainly mediated by Chl a614_E8.2→Chl a834_PsaA, Chl a612_E6.2→Chl a833_PsaA, Chl a611_E6.2→Chl a845_PsaA, and Chl a612_E7.3→Chl a824_PsaA (Fig. 5c). EET pathways from LHCE pentamer 2 to the PSI core are mainly mediated by coupled Chl a611-a612 pair_E6.4→Chl a811-a812 pair_PsaB (Fig. 5d). EET pathways between the two LHCE pentamers are mediated by Chl a608 in LhcE8.2 (P1) to Chl a601 in LhcE7.2 (P2), and EET pathways from LhcbM4.10 to the two pentamers are mainly mediated by Chl a609_E6.4→Chl a608_E8.2(P1) and Chl a601_E7.2(P2), respectively (Fig. 5e).

EET pathways from the LHCE dimer to the PSI core are mainly mediated by the Chl a621_E10.2, which is coordinated by its extended BC-Loop (Supplementary Fig. 8d), to Chl a815_PsaA, and by Chl a604_E11.1 to Chl a816_PsaA (Fig. 5f). EET pathways from the LHCII dimer to the PSI core are mainly mediated by the Chl a620 (Fig. 3a, b), with the pathway proceeding as Chl a620_bM2.1→Chl a620_bM8.2→Chl a403_PsaF (Fig. 5g). EET pathways from LhcbM4.6 to the PSI core are mainly mediated by Chl a603_bM4.6→Chl a841_PsaB and Chl a609_bM4.6→Chl a822_PsaB (Fig. 5h).

We acknowledge the methodological limitations inherent in cryo-EM and Förster theory as previously described^46,47. The cryo-EM data in this study represent a frozen snapshot ( ~ 90 K). Thermal flexibility at 300 K could alter pigment orientations and coupling. In addition, our reconstruction includes particles from different illumination or exposure times, and structural dynamics may blur subtle pigment arrangements. Even at 2.35 Å resolution, uncertainties of ±0.5 Å in Mg–Mg distances and ±10–20° in dipole orientation remain. Because EET rates scale with R⁻⁶ and κ², these uncertainties can alter computed rates by up to 50%. In this study, we do not address several known limitations of cryo-EM and related modeling methods—such as beam-induced motion, local resolution variability, radiation damage, and preferred particle orientation. In addition, the potential role of interfacial water molecules and surrounding electrolytes may also contribute to pigment stability and excitation energy transfer^50–52. Because FRET rates depend sensitively on R⁻⁶, these experimental factors can introduce positional uncertainties of approximately 0.5 Å in the pigment coordinates^53,54, potentially affecting the accuracy of energy transfer calculations. We acknowledged the methodological limitation. Previous studies about time-resolved fluorescence on PSI–LHCI revealed multiple kinetic components corresponding to distinct chlorophyll domains^55,56, the current Förster model assumes a single homogeneous rate, which likely oversimplifies the dynamics. Limited quantitative validity of classical Förster modeling. The Förster-based kinetic model provides qualitative trends but not absolute rates. Without including site-energy disorder, reorganization energies, and coherence effects, the reported transfer constants should be interpreted as semi-quantitative.

Structural basis for red-shifted chlorophyll a absorption in Eg-PSI–LHCE–LHCII supercomplex

In green-lineage PSI–LHCI supercomplexes, red-shifted absorption is typically contributed by red chlorophylls^7,9. In land plants, the LHCI subunits (Lhca1-Lhca4-Lhca2-Lhca3) bind conserved red Chl a603–a609 pairs⁵⁷, and the far-red shift is primarily caused by the Lhca3 and Lhca4^58–60. The identification of red chlorophylls should be based on multiple factors, including the axial ligands of the chlorophyll molecules and their local protein environment²⁷. The variations in site energies of 100–200 cm⁻¹, induced by protein electrostatics, can strongly influence the formation of red-shifted states in PSI–LHCI complexes^58,61,62.

Previous studies have shown that the substitution of His with Asn to Chl a603 can enhance far-red absorption^27,63, the replacement of Asn by His at this position (a603-NH substitution) results in the complete loss of far-red absorption and emission in both reconstituted Lhca3 (rLhca3) and rLhca4 complexes⁶³. The 77 K fluorescence emission of canonical green-lineage PSI–LHCI supercomplexes varies across species: 734 nm in Arabidopsis thaliana²⁷, 722 nm in Physcomitrium patens⁶¹, and 715 nm in Chlamydomonas reinhardtii³³. These differences correlate with the axial ligand of Chl a603. In P. patens, Lhca3 uses Asn, but its Lhca4-homologous subunit Lhca2b retains His⁶¹, contributing to reduced red-shifting. In C. reinhardtii, both Lhca3 and Lhca8 use His for coordination^27,33, explaining its weak far-red emission (Supplementary Fig. 14). In addition, the “large canopy”, defined as a cluster of bulky hydrophobic residues near the Chl a603–a609 pair, contributes to strongly red-shifted emission^9,64,65. In Fittonia albivenis Lhca3, the sequence Phe-Trp-Phe-Gln forms an extended hydrophobic canopy that may lower the excited-state energy and supports 753 nm emission⁹. In contrast, A. thaliana, P. patens and C. reinhardtii Lhca3 contain the smaller Gly-Phe-Ile-Glu motif³³ (Supplementary Fig. 15).

In this study, we found that the Chl a603–a609 pairs in the two LhcE6 subunits exhibit two key structural determinants of red-shifted chlorophylls: Asn-mediated ligation of Chl a603 and a distinct LhcE-specific canopy-like environment. Specifically, we performed simplified excitonic Hamiltonian calculations, which tentatively show that the Chl a603–a609 pairs may play an important role in the far-red spectral band splitting observed in LHCE subunits, as these pairs exhibit the largest excitonic energy splitting among all pigment pairs within each monomer (Supplementary Table 6). In addition, we found that the Chl a603–a609 pairs in LhcE6.2 and LhcE6.4 of pentamer 1 and pentamer 2 utilize Asn to coordinate with the Chl a603, in contrast to other LhcE subunits in this PSI supercomplex (Fig. 6a). On the other hand, each Chl a603–a609 pair in the LhcE dimer is surrounded by the LhcE-specific Arg-Met-Arg (RMR) domain (Fig. 6a) and the adjacent helix C (Fig. 6b), a feature arising from the helix C–to–helix C dimeric assembly of LHCE subunits (Fig. 2b), which creates a distinctive “canopy” effect (Fig. 6b). Furthermore, LhcE6.2 and LhcE6.4 in both pentamer 1 and pentamer 2 are located at the center of the LHCE pentamers (Fig. 6b), such that their Chl a603–a609 pairs are fully surrounded by adjacent LHCE subunits. This arrangement further enhances the canopy effect in LhcE6.2 and LhcE6.4. These structural features, which are associated with site energy and protein electrostatics, including Asn ligation, extended shielding by the RMR domain and adjacent helix C regions, and central positioning within the pentamer, provide potential insight into how a green alga-derived photosynthetic apparatus, following secondary endosymbiosis, evolved plant-like far-red light-harvesting capabilities, a feature likely attributable to the two LhcE6 subunits.

Fig. 6. Sequence and structural comparison of the Chl a603–a609 pair in LHCE subunits, with analysis of site energies and protein electrostatic effects.

a All LHCE subunits coordinate the Chl a603–a609 pairs. Notably, LhcE6 utilizes an Asn residue to bind Chl a603, whereas other LHCE subunits use His. All LHCE subunits contain the Arg-Met-Arg (RMR) domain. b In E. gracilis PSI supercomplex, LhcE6.2 and LhcE6.4 in both pentamer 1 and pentamer 2 are located at the center of the LHCE pentamers. Each Chl a603–a609 pair in the LhcE dimer is surrounded by the LhcE-specific RMR domain and the adjacent helix C (a feature enabled by the helix C–to–helix C assembly mode of LHCE subunits), providing a distinctive canopy that may contribute to red-state stabilization.

Insights into the evolution and antenna assembly mechanism of Eg-PSI supercomplex

The overall architecture of the Eg-PSI–LHCE–LHCII supercomplex reveals an assembly mechanism distinct from those of previously characterized green-lineage PSI supercomplexes^5,6. Compared to the canonical PSI–10LHCI supercomplex of C. reinhardtii³³, the Euglena PSI–16LHCI complex binds more antenna subunits (Fig. 7). Notably, red-lineage secondary endosymbiotic algae extend their light-harvesting capacity by adding second- and third-layer LHCI antennas on the PsaA side^5,25,66, a strategy that is derived from conserved antenna-binding sites in primary red-lineage phototrophs such as Porphyridium purpureum⁶⁷, Galdieria sulphuraria⁶⁸, and Cyanidioschyzon merolae⁶⁹ (Fig. 7). In contrast, as a secondary green-lineage alga, Euglena employs an arrangement: the species-specific LHCE pentamer 1 and pentamer 2 form a second peripheral antenna layer that associates with both the PsaA and PsaB sides of the PSI core (Fig. 1). While antenna expansion on the PsaA side is common in other systems⁵, the PsaB-side expansion observed here is unprecedented among primary green-lineage plants and algae, as well as other photosynthetic eukaryotes^5,6 (Fig. 7).

Fig. 7. Possible evolutionary development of PSI–LHCI supercomplexes.

The Euglena PSI–16LHCI supercomplex features a more extensive antenna system than the canonical PSI–10LHCI of C. reinhardtii. While PsaA-side expansion is common, PsaB-side antenna addition is unprecedented in green-lineage and other eukaryotic phototrophs. Structures used: C. reinhardtii PSI-LHCI (PDB: 6IJO); P. patens PSI-LHCI (PDB: 7KSQ); P. sativum PSI-LHCI (PDB: 7DKZ); T. elongatus PSI (PDB: 1JB0); C. paradoxa PSI (PDB: 7DR2); C. merolae PSI-LHCI (PDB: 5ZGB); G. sulphuraria PSI-LHCI (PDB: 9KC5); P. purpureum PSI-LHCI (PDB: 7Y5E); C. placoidea PSI-LHCI (PDB: 7Y7B); C. gracilis PSI-LHCI (PDB: 6LY5); A. carterae PSI-LHCI (PDB: 8JW0); E. huxleyi PSI-LHCI (PDB: 9JJ8);.

Notably, similar to PSI–LHCI supercomplexes in red-lineage secondary endosymbionts, no phosphorylation sites were identified at the LHCII–PSI core interface in this study, suggesting that the LHCII subunits are constitutively associated with the PSI core, thereby expanding the light-harvesting cross-sectional area. In red-lineage secondary endosymbiotic organisms, Lhcf and Lhcr family proteins are primarily associated with the PSII and PSI cores, respectively^21,70. However, minor Lhcf subunits are constitutively bound to the PSI core and have been proposed to expand the light-harvesting capacity^25,36,66,71, and minor Lhcr subunits associate constitutively with the PSII core and are suggested to stabilize antenna assembly^35,72,73. This indicates that the association of PSII-related antenna subunits with the PSI core is not uncommon in secondary endosymbiotic organisms. In this context, the observation in Euglena, where LhcE9 is also predicted to associate with the PSII supercomplex²⁶, fits a broader evolutionary pattern of flexible antenna allocation to optimize light-harvesting strategies during secondary endosymbiosis.

Discussion

E. gracilis originated from a secondary endosymbiosis involving Chloroplastida⁴, and belongs to the Euglenophyta, a lineage in which a phagotrophic protist acquired a green algal chloroplast⁷⁴. At the 2.35 Å resolution solved in this study, and with the aid of ModelAngelo²⁹ and CryoAtom³⁰, as well as the available complete LHC sequences²⁶, we were able to build a complete structural model of the PSI–12LHCE–4LHCII supercomplex. An obvious difference from other green-lineage PSI–LHCI supercomplexes is the minimal PSI core (Fig. 1). Most green algal PSI–LHCI supercomplexes utilize conserved multi-subunit PSI core and bind 6 to 10 Lhca subunits^5,6, such as those in Ostreococcus³⁸, Chlamydomonas^10,33, Bryopsis¹¹ and Chlorella¹². In contrast, the E. gracilis genome appears to have lost several PSI core subunit genes²⁶. Based on the structure of the PSI–LHCE–LHCII supercomplex, there is no available space within the complex to accommodate additional core subunits, such as PsaG, PsaH, PsaK, PsaL, and PsaO. The loss of core subunits and the resulting minimal PSI core likely drove the evolution of a distinct antenna organization in Eg-PSI.

On the other hand, Euglena has adopted the Ddx–Dtx xanthophyll cycle²⁰, which is typically found in planktonic red-lineage secondary endosymbiotic algae^23–25, as its primary mechanism for quenching excess excitation energy. This cycle enables a rapid photoprotective response^20,75, which is advantageous for a motile organism like Euglena that experiences frequent and rapid light fluctuations due to its fast-spinning flagellar movement⁷⁶. In contrast, sessile or symbiotic red-lineage secondary endosymbiotic algae^75,77, retain the Vio–Zea cycle. Thus, the use of the Ddx–Dtx cycle in Euglena likely represents an evolutionary adaptation to its dynamic lifestyle⁷⁴, allowing efficient photoprotection in heterogeneous light environments. In this study, we report a pigment network rich in diadinoxanthin within the Euglena PSI supercomplex (Fig. 5a). Each LHCE monomer binds two conserved Ddx616 and Ddx617 molecules (Fig. 2a), which are located at the core of the antenna protein and embedded in the grooves formed by the crossing helices B and C, and by helices A and B, respectively (Fig. 2a). The location of these Ddx molecules suggests a potential role in quenching excess excitation energy from adjacent Chls a. Notably, LhcE8.2 in pentamer 1 binds an additional Ddx618, which may enable it to dissipate excess energy received from the adjacent peripheral LhcbM4.10 subunit (Fig. 5e). These structural features provide a molecular basis for the potential energy dissipation mechanism in Euglena.

The helix C–to–helix C dimeric assembly was previously observed in diatoms^23,35,73 and haptophytes⁷⁸, organisms derived from red-lineage secondary endosymbiosis⁴. In contrast, green-lineage LHCI typically adopt a helix A–to–helix C arrangement^7–9,27. In this study, the helix C–to–helix C dimerization of LHCE subunits is mediated by a conserved DRS domain within the LHCE sequence (Supplementary Fig. 8). This dimeric organization exhibits three distinct functional advantages. First, the closely coupled Chl a609 molecules at the monomer–monomer interface enable ultrafast excitation energy transfer on a timescale of 0.48–0.49 ± 0.1 ps (Fig. 5b, Supplementary Table 4 and 5), facilitating efficient energy exchange between the two subunits and enhancing the overall ability in light-harvesting. A previous study reported ultrafast energy transfer between symmetric Chl a–Chl a pairs in a helix C–to–helix C LHC dimer within the diatom PSII supercomplex, with a Mg–Mg distance of approximately 11.8 Å and an experimentally determined transfer time of ~0.38 ps⁷⁹. In contrast, the energy transfer time of 0.48–0.49  ± 0.1 ps, calculated in this study based on a simplified FRET model using a distance of 14.4–14.5 Å, is semi-quantitative and does not account for advanced effects such as quantum coherence, site-energy disorder, or excitonic coupling. Given the longer inter-pigment distance, this value is consistent with the experimental trend and falls within a reasonable range predicted by the R⁻⁶ dependence of Förster theory. Although coherent effects are not quantified in this study, coherent coupling within tightly coupled chlorophyll pairs may contribute to rapid energy transfer in the far-red absorbing domains^80,81, as suggested in other photosynthetic systems. The second advantage is that the helix C–to–helix C interactions provide a canopy effect for the adjacent Chl a603-a609 pairs (Supplementary Fig. 15b), thus may stabilize the excited state for the red-shifted chlorophylls. Third, as the minimal oligomeric unit of the antenna system²³, the dimeric organization enabling rapid reorganization and regulatory flexibility under fluctuating light conditions. Notably, Euglena could further assemble these LHCE dimers into larger (2 + 2 + 1)-type pentamers, thereby achieving substantial light-harvesting ability. The integration of dynamic LHCE dimers into stable pentamers support both efficient light capture and regulatory flexibility, a feature not previously described in any green or red-lineage algae.

In a recent study²⁶, an LHCE pentamer in the peripheral antenna pool was reported through negative stain EM analysis. Due to limited resolution, only the overall shape of the pentamer was resolved, and no detailed information on the subunit arrangement was obtained²⁶. In this study, the LHCE pentamers were resolved at 2.35 Å, revealing that LhcE5–LhcE8 form a (2 + 2 + 1)-type pentamer—composed of two dimers (LhcE5–LhcE6 and LhcE5–LhcE7) and one monomeric LhcE8 (Fig. 2c). This arrangement differs from both the (4 + 1)-type pentamer observed in diatom FCP pentamer³⁷ and the previously suggested model²⁶ (Supplementary Fig. 10d). Phylogenetic analysis of all LHCE subunits further supports structural modularity in pentamer formation (Supplementary Fig. 7). Previous studies have proposed that a distinct peripheral LHCE pentamer is composed of LhcE1–4 and LhcE12²⁶. In our phylogenetic tree, LhcE1, LhcE2/12, LhcE3, and LhcE4 cluster closely with LhcE7, LhcE8, LhcE5, and LhcE6, respectively (Supplementary Fig. 7). Despite differential localization, LHCE subunits can form functionally similar pentameric antenna complexes by recruiting paralogous isoforms. This study reveals that the PsaA- and PsaB-side LHCE pentamers share overall structural similarity (Fig. 1) but differ in subunit composition. While LhcE1–4 + LhcE12 typically form free complexes, LhcE6.2 and LhcE6.4 are specialized for PsaA- and PsaB-side association, respectively. The reciprocal exchange of LhcE5.4 and LhcE5.7 between the two pentamers introduces structural asymmetry, potentially enabling position-dependent tuning of energy transfer or assembly.

The structural and functional integration of red-shifted chlorophyll pairs in LHCE subunits support the far-red light harvesting capability of the Eg-PSI–LHCE supercomplex. Previous studies proved that the isolated LHCE pentamer, a component of the peripheral antenna pool, has been shown to exhibit far-red absorption^26,82. Difference spectroscopy between LHCE and LHCII indicates that LHCE contains minimal Chls b, which is consistent with its red-shifted absorption profile²⁶. Room-temperature fluorescence measurements further reveal that LHCE emits at 698 nm, significantly red-shifted compared to LHCII, which peaks at 684 nm. This property is attributed to the presence of red-shifted Chls a⁸³. Notably, the Eg-PSI core associated with LHCE also displays a red-shifted emission relative to the Eg-PSI core. This demonstrates that far-red light harvesting is not limited to the free LHCE pentamer but is functionally integrated into the PSI–LHCE supercomplex²⁶. The LHCE subunits are enriched under low-light conditions, which aligns with the physiological role of red-shifted antenna systems in enhancing light capture under low-light conditions²⁶. In this study, detailed structural and biophysical analysis of the Chl a603–a609 pairs across all LHCE subunits, taking into account site energies and protein electrostatics, reveals that the Chl a603–a609 pair in LhcE6 possesses potential key features associated with the far-red emission, including Asn coordinated with Chl a603, extended canopy effect by the LhcE-specific RMR domain and adjacent helix C, and central positioning within the pentamer (Fig. 6). Based on phylogenetic relatedness and structural similarity, we proposed that LhcE4²⁶, a close homolog of LhcE6 (Supplementary Fig. 7), is also likely to harbor similar red-shifted properties within the LhcE1–4/12 pentamer. This functional conservation among LhcE subunits may contribute to the red shift observed in the LHCE subunits.

In this study, it should be noted that several key physical effects were not explicitly incorporated into our Förster-based energy transfer model. These include site-energy disorder, spectral broadening and dynamic averaging. Site-energy disorder, arising from conformational and dynamic heterogeneity, was not modeled beyond average site energies and could create transient low-energy traps that influence energy transfer direction and efficiency. The FRET constants were adopted from plant Chl a/b systems, but inhomogeneous broadening due to local microenvironment variations was not parameterized, potentially affecting spectral overlap and absolute rate accuracy. Pigment orientations were fixed from the cryo-EM structure, ignoring fast-timescale dynamics that could average transition dipoles and reduce the sensitivity of FRET to precise geometry. Therefore, these simplifications mean that our model provides a semi-quantitative estimate of energy transfer pathways rather than a fully predictive kinetic simulation.

The identity of the two LhcbM4 subunits in this study has been a subject of debate previously (LhcE13), as they occupy a phylogenetic position between the LhcbM and LHCE clades in the phylogenetic tree²⁶ (Supplementary Fig. 7). In this study, we found that these two subunits differ from canonical LHCE subunits, which bind only Chl a and Ddx. Instead, each subunit binds two Chls b and one neoxanthin, a pigment composition characteristic of the LhcbM family⁴⁴. This structural evidence supports their classification within the LhcbM family, thus named LhcbM4 in this study.

In conclusion, the cryo-EM structure of the PSI–LHCE–LHCII supercomplex from E. gracilis, determined at 2.35 Å resolution in this study, reveals detailed structural features of the minimal PSI core and its association with one LHCE dimer, two LHCE pentamers, one LHCII dimer, and two LHCII monomers. The LHCE subunits exhibit structures and pigment compositions. Two LHCE dimers, together with a monomeric LhcE8, assemble into a distinct (2 + 2 + 1)-type pentamer. LhcE6.2 and LhcE6.4 potentially bind the red-shifted Chl a603–a609 pairs, enabling the entire PSI supercomplex to exhibit far-red light absorption. The overall supercomplex is assembled in an organization relative to canonical green algal PSI–LHCI supercomplexes, which may optimize light-harvesting efficiency on the minimal core. These findings provide insights into the photosynthetic strategies of green plastid-harboring secondary endosymbiotic organisms.

Methods

Purification of PSI-LHCE-LHCII from E. gracilis

PSI–LHCE–LHCII was purified from E. gracilis (GY-D32, Guangyu Biotech Co., Ltd., China). Cells were cultured in TAP medium¹⁸ under continuous illumination of ~100 μmol photons m^-2 s^-1, with 3% CO₂ bubbling, at 25 °C, on a 16-h light/8 h dark cycle under stirring at 120 rpm. All subsequent procedures were performed under green light and at 4 °C or on ice.

Cells were harvested by centrifugation at 4000 g and resuspended in buffer containing 50 mM N-2-hydroxyethylpiperazine-N’-2-ethanesulfonic acid (HEPES)-KOH (pH 7.5), 1 M betaine, 10 mM MgCl₂, and 5 mM CaCl₂. The cell suspension was disrupted using a French press (3 cycles at 30 MPa). Unbroken cells were removed by centrifugation at 1000 g for 10 min. The supernatant was then centrifuged at 30,000 g for 30 min, and the resulting thylakoid pellet was resuspended in buffer containing 50 mM HEPES-KOH (pH 7.5), 1 M betaine, 10 mM NaCl, and 5 mM CaCl₂ to a chlorophyll (a + b) concentration of 1 mg/mL.

The membranes were solubilized with 1.6% (w/v) n-dodecyl-α-D-maltoside (α-DDM) (Anatrace, Maumee, OH, USA) and incubated on ice for 20 min. Insoluble material was removed by centrifugation at 12,500 g for 10 min. The supernatant was loaded onto a continuous sucrose density gradient (0.1–1.3 M in buffer with 0.01% α-DDM) and centrifuged at 36,000 rpm for 18 h in a SW40 rotor (Beckman Coulter, USA). The PSI–LHCE–LHCII fraction was collected using a syringe (Supplementary Fig. 1a). Sucrose was removed by dilution with buffer (50 mM HEPES-KOH, pH 7.5, 110 mM NaCl, 5 mM CaCl₂, 0.03% α-DDM), and the sample was concentrated to 2.2 mg/mL total chlorophyll (a + b) using an ultrafiltration centrifuge tube (100 kDa molecular weight cutoff; Amicon, Merck Millipore, Germany).

Characterization of the PSI–LHCE–LHCII supercomplex

The subunit composition of the PSI–LHCE–LHCII supercomplex was analyzed by SDS-PAGE on a 12% polyacrylamide gel. The gel was stained with Coomassie Brilliant Blue R-250 (Sigma-Aldrich, Germany). Protein identification was performed by mass spectrometry. Briefly, Coomassie-stained bands were excised and subjected to in-gel digestion using sequencing-grade modified trypsin. The resulting peptides were extracted and analyzed by mass spectrometry (Source data).

Room-temperature absorption spectrum and fluorescence emission spectra at 77 K were recorded using a UV-Vis spectrophotometer (Shimadzu, Japan) and a fluorescence spectrophotometer (F-7000, Hitachi, Japan), respectively. Measurements were performed on the PSI–LHCE–LHCII supercomplex at a chlorophyll concentration of 50 μg Chl (a + b)/mL, with excitation at 436 nm.

Pigment composition was determined by HPLC⁴². Samples collected from the sucrose density gradient were mixed with 90% (v/v) acetone and incubated for 30 min at 4 °C to extract pigments. After centrifugation at 12,500 g for 15 min, the supernatant was loaded onto a C-18 column (5 μm, 100 Å, 250 mm × 4.6 mm; Grace, USA). Pigments were eluted at room temperature with a flow rate of 1 mL/min using a linear gradient from buffer A (90% acetonitrile in water) to buffer B (100% ethyl acetate). Pigment identities were assigned based on absorption spectra and retention times monitored at 445 nm. HPLC analysis revealed the presence of Chl a, Chl b, Neo, Ddx, and Bcr in the Eg-PSI–LHCE–LHCII supercomplex (Supplementary Fig. 1e). The Chl a/b ratio is about 10.6.

Phylogenetic analysis

To investigate the gene families of each antenna subunit, the sequences of the thirteen antenna proteins resolved in this study were compared with representative light-harvesting complex (LHC) proteins from green-lineage organisms, Arabidopsis thaliana²⁷ and C. reinhardtii¹⁰, and the red-lineage diatom Thalassiosira pseudonana^35,36. Multiple sequence alignment was performed using TBtools⁸⁴, with protein sequences aligned using MAFFT⁸⁵ under default settings. The resulting alignment was used to construct a phylogenetic tree in IQ-TREE2⁸⁶. The best-fitting substitution model was selected using the Bayesian Information Criterion implemented in IQ-TREE2, and the tree was visualized using iTOL v6⁸⁷.

Cryo-EM data acquisition and processing

The PSI–LHCE–LHCII sample, at a concentration of 2.2 Chl (a + b) mg/mL, was applied to a Quantifoil Cu R1.2/1.3, 300-mesh grid and vitrified using a Vitrobot (Thermo Fisher Scientific, Mark IV, USA) at 4 °C and 100% humidity. Blotting was performed for 2 s with a blot force of 1.

Data were collected on a 300 kV Titan Krios microscope (Thermo Fisher Scientific) equipped with a Gatan Quantum energy filter (slit width: 20 eV) and a K3 Summit camera (Gatan, USA). Movies were acquired using EPU 1.9 software in super-resolution mode at a nominal magnification of ×130,000, yielding a calibrated pixel size of 0.92 Å. A total of 15,238 movies were recorded with a total dose of ~50.0 e⁻/Ų, a dose rate of 20 e⁻/pixel/s (30 frames), and a defocus range of −1.0 to −2.0 μm. All micrographs were binned during processing, resulting in a final pixel size of 0.92 Å for downstream image processing.

Movies were imported into cryoSPARC²⁸, and 3,935,064 particles were picked using the Blob/Template Picker. Particles were extracted and subjected to multiple rounds of two-dimensional (2D) classification to remove junk particles. A subset of 1,059,170 particles was selected for Ab-Initio reconstruction. Following two rounds of heterogeneous refinement, 51,871 particles were used to reconstruct the PSI–LHCE–LHCII supercomplex at a global resolution of 2.35 Å, as determined by the gold-standard Fourier shell correlation (FSC) at a cutoff of 0.143.

To improve the resolution of the peripheral antennae, local masks were made from 2.60 Å to 3.20 Å resolution, respectively. A composite map was generated by combining the global and local reconstructions and was used for model building and refinement (Supplementary Fig. 2).

Model building and refinement

For model building of the Eg-PSI–LHCE–LHCII supercomplex, the initial model building was assisted by ModelAngelo²⁹ and CryoAtom³⁰, followed by manual model fitting, guided by sequence differences^26,31 and local cryo-EM density features (Supplementary Fig. 2 and 3). The assignments of the paralogs LhcE5, LhcE6, LhcE7, and LhcbM4 were based on distinct structural features and clear differences in the cryo-EM density maps, including side-chain configurations and loop conformations (Supplementary Fig. 16 to 19). For LhcE8 in Pentamer 2, the local resolution is insufficient to unambiguously distinguish between closely related paralogs. Based on the best-fit density and sequence homology, this subunit was tentatively modeled as LhcE8.2.

Final model building and real-space refinement were performed using Coot 0.8.9⁸⁸. All protein residues and pigments were fitted using locally optimized map weights. Pigment assignment was according to the HPLC analysis (Supplementary Data Fig. 1e). Discrimination between Chl a and Chl b was based on the presence of the additional carbonyl oxygen atom of Chl b⁸⁹. However, due to the limited resolution of the cryo-EM density, unambiguous identification of Chls b is tentative. During revision of our work, a 2.82 Å resolution structure of smaller E. gracilis PSI–13 LHCI was reported⁹⁰. Due to low resolution in the local LHCI region, no Chl b molecules could be modeled in the structure. In contrast, in our study, four Chl b606-Chl b607 pairs in the four LHCII subunits (Supplementary Table 3) could be assigned, as the extra oxygen atom of each Chl b participates in a stable hydrogen-bonding network between the two pigments—providing a discernible density feature in the map (Supplementary Fig. 5b), other potential Chl b sites in LhcbM subunits are tentatively modeled as Chl a, guided primarily by the higher overall abundance of Chl a indicated by HPLC (Supplementary Fig. S1e).

Discrimination between Ddx and Bcr was based on the presence of bulged densities corresponding to the terminal oxygen atom (Supplementary Fig. 5a). Neo molecules could be distinguished based by its characteristic C29 chiral center (Supplementary Fig. 5a). In this study, only the density at the Car619 site in LhcbM4 clearly exhibits this bent conformation (Supplementary Fig. 5a), allowing us to assign it as Neo619. For other carotenoid sites in the antenna subunits, the densities were tentatively modeled as Ddxs, guided primarily by the higher overall abundance of Ddx indicated by HPLC (Supplementary Fig. 1e).

The Eg-PSI–LHCE–LHCII model was finally refined using real-space refinement in Phenix 1.20⁹¹. Model validation, rebuilding, and refinement were carried out iteratively using MolProbity 4.2⁹², Coot 0.8.9⁸⁸, and Phenix 1.20⁹¹. Statistics for cryo-EM data collection, model building, and refinement are summarized in Supplementary Table 1. Figures were prepared using UCSF ChimeraX 1.10⁹³.

VDW force analyses between PSI core and antenna subunits

VDW contact analysis was performed using a previously described method³⁵. Parameters were set using the ‘Contact’ tool in UCSF ChimeraX 1.10⁹³, and VDW interactions within individual antenna subunits and within the PSI core subunits were excluded. Only intermolecular VDW contacts between antenna subunits and the PSI core, and those mediated by lipid molecules are displayed (Supplementary Fig. 12).

Computational analysis of FRET

The equation of FRET is described as follows:

$$k_{\mathrm{FRET}}=\frac{C{\mathit{\kappa }}^2}{n^4{R}^6}$$ 1

where C is the spectral overlap between the electron donor’s fluorescence spectrum and the acceptor’s absorption spectrum. The applied C values for Chl a→Chl a, Chl b→Chl a, Chl a→Chl b, and Chl b→Chl b were 32.26, 1.11, 9.61, and 14.45, respectively, which were estimated from plant PSI-LHCI and LHCII previously^46,47.

The dipole orientation factor is defined as:

$${\mathit{\kappa }}^2={\left[\widehat{{\mathbf{u}}_D}\cdot \widehat{{\mathbf{u}}_A}-3\left(\widehat{{\mathbf{u}}_D}\cdot \widehat{{\mathbf{R}}_{DA}}\right)\left(\widehat{{\mathbf{u}}_A}\cdot \widehat{{\mathbf{R}}_{DA}}\right)\right]}^2$$ 2

The$\widehat{{\mathbf{u}}_D}$ and $\widehat{{\mathbf{u}}_A}$ are the unit vectors of displacement vectors from NB to ND atom in donor and acceptor Chl, representing the direction of dipole moment in the dipole approximation of the Chl. $\widehat{{\mathbf{R}}_{DA}}$ is the unit vector of the displacement vector from the magnesium atom in donor Chl to the magnesium atom in acceptor Chl. n stands for the refractive index, and an estimated value of n = 1.55 was taken from previous studies. $R_{DA}$ is the magnesium-to-magnesium distance between donor and acceptor Chls. The lifetime and half life is defined as:

$$t_{\mathrm{FRET}}=\frac{1}{k_{\mathrm{FRET}}},t_{1/2}=\frac{\ln 2}{k_{\mathrm{FRET}}}$$ 3

Error propagation in FRET

The FRET calculation is based on structural parameters extracted from the cryo-EM resolved pigment geometry. To incorporate finite accuracy of structure determination, we explicitly include two independent uncertainty sources:

Distance uncertainty: we treat the donor–acceptor separation R as having a fixed error of ${\sigma }_R=0.5\mathrm{\AA }$ for all pairs.

Dipole-orientation uncertainty: we allow the donor and acceptor transition dipoles to deviate by ±20° from their nominal directions. Dipole uncertainty is projected onto an uncertainty in the orientation factor ${\mathit{\kappa }}^2$ through a statistical (Monte Carlo) procedure. Specifically, we model this ±20° angular uncertainty by uniformly sampling perturbed dipole directions within a cone of half-angle $\delta ={20}^{\circ }$ around each dipole. For each Monte Carlo draw i, we compute κ² from the standard Förster orientation factor. Repeating this procedure for N samples yields a distribution ${\mathit{\kappa }}_i^2$(N = 50000). We define:

$$⟨{\mathit{\kappa }}^2⟩=\frac{1}{N}{\sum }_{i=1}^N{\mathit{\kappa }}_i^2,{\sigma }_{{\kappa }^2}=\sqrt{\frac{1}{N}{\sum }_{i=1}^N{\left({\mathit{\kappa }}_i^2-⟨{\mathit{\kappa }}^2⟩\right)}^2}$$ 4

$⟨{\mathit{\kappa }}^2⟩$ is used as the effective orientation factor in FRET calculation, and ${\sigma }_{{\mathit{\kappa }}^2}$ is carried as the uncertainty originating from the ±20° dipole-angle uncertainty.

Since FRET rate constant $k_{DA}\propto {\mathit{\kappa }}^2{R}^{-6}$, the relative uncertainty is:

$${\left(\frac{{\sigma }_{\mathit{k}}}{\mathit{k}}\right)}^2={\left(\frac{{\sigma }_{{\mathit{\kappa }}^2}}{⟨{\mathit{\kappa }}^2⟩}\right)}^2+{\left(6\frac{{\sigma }_R}{R}\right)}^2$$ 5

The absolute uncertainty of $k_{DA}$ is then:

$${\sigma }_{\mathit{k}}=k_{DA}\sqrt{{\left(\frac{{\sigma }_{{\mathit{\kappa }}^2}}{⟨{\mathit{\kappa }}^2⟩}\right)}^2+{\left(6\frac{{\sigma }_R}{R}\right)}^2}$$ 6

And the absolute uncertainties of τ and $t_{1/2}$ are given by:

$${\sigma }_{\tau }=\tau \frac{{\sigma }_{\mathit{k}}}{{\mathit{k}}_{DA}}$$ 7

$${\sigma }_{t_{1/2}}=t_{1/2}\frac{{\sigma }_{\mathit{k}}}{k_{DA}}$$ 8

Calculation of excitonic Hamiltonian modeling of two Chls

For two coupled chlorophylls (donor D and acceptor A), the excitonic Hamiltonian is written as:

$$\mathit{H}=\left(\begin{array}{cc}\hfill E_D\hfill & \hfill J\hfill \\ \hfill J\hfill & \hfill E_A\hfill \end{array}\right)$$ 9

where E_D and E_A are the site energies (excitation energies) of the two Chls, and J is the electronic coupling⁹⁴. We approximate J by a point-dipole interaction:

$$J=\frac{1}{4\pi {\epsilon }_0{\epsilon }_r{R}_{DA}^3}\left[{\mathit{\mu }}_{\mathit{D}}\cdot {\mathit{\mu }}_{\mathit{A}}-3\left({\mathit{\mu }}_{\mathit{D}}\cdot \widehat{{\mathit{R}}_{DA}}\right)\left({\mathit{\mu }}_{\mathit{A}}\cdot \widehat{{\mathit{R}}_{DA}}\right)\right]$$ 10

We use the same NB to ND vector to approximate the dipole orientation and fixed transition dipole magnitudes taken from literature:

$$\mu \left(\mathrm{Chl}a\right)=4.49,\mu \left(\mathrm{Chl}b\right)=3.13$$ 11

The eigenvalues of the Hamiltonian are:

$$E_{\pm }=\frac{E_D+E_A}{2}\pm \sqrt{{\left(\frac{E_D-E_A}{2}\right)}^2+J^2}$$ 12

If both donor and acceptor are Chl A (E_D = E_A):

$$E_{\pm }=E_0\pm ∣J∣$$ 13

And the energy splitting under this condition is:

$$\mathrm{\Delta }E=2∣J∣$$ 14

Author contributions

Y.F. conceived the project. Y.F. purified the proteins, performed data collection and structure determination, carried out model building and biochemical analysis, and wrote the manuscript. T.C., B.S., X.B., K.G. and P.Y. assisted with protein sequence, pigment and model analysis. X.Z. calculated the EET pathways. All authors contributed to the analysis and the final version of the paper.

Peer review

Peer review information

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

Data availability

The composite cryo-EM map and atomic coordinates for the PSI–LHCE–LHCII supercomplex have been deposited in the Electron Microscopy Data Bank under accession code EMD-63994 and the Protein Data Bank under accession code 9UAS. The atomic coordinates data used in this study are available in the Protein Data Bank database under the accession codes 1JB0, 5ZGB, 6IJO, 6LY5, 7DKZ, 7DR2, 7DZ7, 7KSQ, 7XQP, 7Y5E, 7Y7B, 8JW0, 9JJ8, 9KC5. Source data for Supplementary Fig. 1 are provided in the Source data file.  Source data are provided with this paper.

Code availability

The Python script used for the calculation of FRET rates is available at [10.5281/zenodo.3250649]^46,47.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-71601-x.