Introduction

Microbial rhodopsins, a superfamily of heptahelical photoreceptive membrane proteins using a retinal chromophore, are crucial to understanding how microorganisms adapt to and harness light energy. These proteins exhibit a remarkable structural and functional diversity1,2,3, which enables their application in optogenetics and synthetic biology. Among these, some members of the proteorhodopsin (PR), xanthorhodopsin (XR), and related families stand out due to their unique dual chromophore system, which includes a retinal moiety and a non-covalent antenna in the form of a xanthophyll molecule (carotenoid with oxygen atoms). So far, since the discovery of SrXR, the first antenna-binding XR from the extreme halophilic bacterium Salinibacter ruber, only a few prokaryotic rhodopsins have been demonstrated to form such complexes4,5,6,7,8. One of the best studied XRs with a carotenoid antenna is GR from the thylakoid-less terrestrial cyanobacterium Gloeobacter violaceus6, which, similarly to SrXR, recruits a 4-ketocarotenoid antenna (echinenone) to transfer light energy to the retinal chromophore9. Nevertheless, increasing evidence suggests that, in contrast to the classical 4-ketocarotenoid-binding rhodopsins GR and SrXR, rhodopsin proton pumps binding 3-hydroxycarotenoids might be more common7,8,10,11. As there are no substantial differences between hydroxylated and ketolated xanthophylls in the efficiency of energy transfer to the retinal molecule within rhodopsins (30–40%5,6,8,11,12), the apparent preference for the hydroxylated species is probably due to their higher abundance in aquatic environments8.

The appearance of ion-pumping rhodopsins in organisms with chlorophyll-based photosystems is well documented, although their physiological roles and the interplay between the two types of photosystems are known only in isolated cases7,13,14,15,16,17,18. Recently, a possible connection between anoxygenic photosynthesis and rhodopsin-based phototrophy was suggested by Kopejtka and coworkers7, who described a freshwater alphaproteobacterium capable of using two different systems for light-harvesting: a proton-pumping rhodopsin and an anoxygenic photosynthetic apparatus with bacteriochlorophyll in the reaction center. Interestingly, both systems were shown to use the same carotenoid (nostoxanthin) as an antenna. This overlap in antenna systems suggests a functional convergence where organisms leverage similar molecular components across distinct photosystems, potentially facilitating efficient energy transfer and flexibility in response to varying light conditions.

Cyanobacteria play a crucial role in aquatic and terrestrial ecosystems through oxygenic photosynthesis, contributing significantly to primary production and biogeochemical cycles19. Rhodopsin ion pumps from several families are known from cyanobacteria20,21 but the understanding of their relationship to the photosynthetic apparatus is very limited. A complex interplay between photosynthesis and XRs in particular is hinted at by the contribution of GR to proton gradient and ATP production in the atypical cyanobacterium G. violaceus13. While the presence of rare XR genes in thylakoid-possessing cyanobacteria has been recently mentioned21, the extent of this phenomenon and the potential functions of these XRs remain entirely unknown. The current study aims to elucidate the distribution of XRs and their potential for antenna binding among cyanobacteria beyond GR. We report the discovery of an XR in an Antarctic cyanobacterium (psXR, standing for “psychrophilic Pseudanabaena XR”) that binds a hydroxylated carotenoid. Through bioinformatic analysis, spectroscopic techniques, and functional assays, we have characterized psXR and its potential role in enhancing cyanobacterial adaptation to diverse light environments. Our findings suggest that this XR might complement oxygenic photosynthesis (separately or in parallel) in Antarctic cyanobacteria, potentially contributing to the ecological success of cyanobacteria.

Results

Phylogenetic analysis of XRs in cyanobacteria

Our search for rhodopsin genes in cyanobacterial genomes revealed that 11.7% and 14.7% of representative genomes possess them in GTDB and JGI GEM, respectively, with XRs present in 14.1%/21.5% of the rhodopsin-carrying genomes, surpassed by bacterial halorhodopsins (anion pumps) and members of the xenorhodopsin family (inward proton pumps and sensory rhodopsins) (Supplementary Fig. S1). Despite these modest numbers, XR genes were detected in diverse and often unrelated cyanobacterial families. To our surprise, the majority of the cyanobacterial XRs came from families of filamentous cyanobacteria, not only MAGs but also isolates of Phormidesmis priestleyi (previously mentioned in literature21), Nodosilinea antarctica (non-axenic cultures), Halomicronema spp., Leptothoe sp. (Leptolyngbya sp. Heron Island J), Leptolyngbyaceae sp. JSC-12 and Rivularia radiosa (Plectonema cf. radiosum LEGE 06105). Interestingly, all of the XRs from cyanobacterial genomes appeared to cluster together in a single clade (Fig. 1A). This clade was nested among XR sequences originating from Deinococcota with good support, thus indicating that cyanobacteria might have acquired the first XR genes from this extremophile group.

Fig. 1: Phylogenetic relationships between xanthorhodopsins found in cyanobacteria.
Fig. 1: Phylogenetic relationships between xanthorhodopsins found in cyanobacteria.The alternative text for this image may have been generated using AI.
Full size image

A Phylogenetic tree of XR proteins from diverse genomes classified to phylum. Proteins characterized previously or otherwise mentioned in literature are shown for reference. Numbers in parentheses next to the names of the phyla indicate the number of the representative sequences from the corresponding phylum used for the tree reconstruction. Only phyla with at least two sequences are shown. The position of the XRs from cyanobacteria (phylum Cyanobacteriota, class “Cyanobacteriia”) is indicated. Other unrelated Cyanobacteriota XRs belong to the non-photosynthetic “Ca. Sericytochromatia”. The tree is outgroup-rooted, and the outgroups are not shown. Gray dots indicate branches with ultra-fast bootstrap support values ≥95. B Phylogenetic relationships among the cyanobacterial XRs. Symbol shapes and colors indicate the habitats from which the corresponding isolates or MAGs originate. Labeled are sequences coming from genome assemblies: MAGs are referred to by their accession numbers, and isolates are referred to by the names of the corresponding organisms. Genera assigned to the corresponding genomes are specified in parentheses (if different from the original classification in the case of the cultured isolates). The genome labels are colored by the inferred taxonomy based on GTDBtk. Note that the inference about the morphotypes for MAGs is based on the knowledge available for isolates from the corresponding families. The tree is outgroup-rooted, and the outgroups are not shown. Rapid bootstrap support values above 60 are shown.

A fine-grained phylogenetic analysis of cyanobacterial XRs together with related sequences from unbinned environmental contigs, showed that XRs coming from genomes assigned to the same cyanobacterial family often do not form a holophyletic grouping as is the case with the genes from the Chamaesiphonaceae, Elainellaceae, Leptolyngbyaceae, Nostocaceae, Phormidesmiaceae and Pseudanabaenaceae which further indicates that horizontal gene transfer between cyanobacterial groups played a major in spreading the XR genes among them (Fig. 1B). In this analysis, the cyanobacterial clade is further subdivided into four well-supported subclades (A–D), each with its environmental and taxonomic trends. The largest subclade, A, is composed of XR genes from unrelated cyanobacterial families with dissimilar morphotypes ranging from the unicellular thylakoid-less G. violaceus (the well-characterized GR) to the family of the epiphytic Chamaesiphonaceae and filamentous families (including psXR characterized here). This subclade covers sequences originating from diverse terrestrial, freshwater, and saline habitats. Most conspicuously, all XR genes from cold environments: glaciers, meltwater, cryoconite, and polar deserts, belong to this subclade. Further, considering that sequences from temperate locations, such as GR, constitute a minority of this subclade, this suggests that subclade A might have undergone adaptation to low temperatures. The smaller subclades B, C, and D were found in isolates and environmental datasets originating strictly from aquatic habitats, from temperate waters or hot springs and lakes. The cyanobacteria possessing these genes, when known or predictable, are filamentous, except for Halothece spp., which possess subclade-B XR genes.

The genomic context of the XR genes in different cyanobacteria was found to be highly diverse, with no indication that any of them belong to an operon. Thus, no genes for retinal synthesis: blh, typically encountered in the same operon as proteorhodopsin and xanthorhodopsin genes in diverse bacteria22,23 or genes for cyanobacterial retinal-producing apo-carotenoid dioxygenases24,25, or any other gene for carotenoid biosynthesis were found in the vicinity of the cyanobacterial XR genes (Supplementary Fig. S2). As exemplified by the Pseudanabaena genome possessing psXR, genes for cyclic carotenoid biosynthesis are present in other genomic locations, indicating that both the retinal and potential carotenoid antennas can be supplied endogenously (Supplementary Fig. S3), as demonstrated experimentally for G. violaceus9,13. The most frequent gene neighbors of XR genes instead were hli genes coding for high-light inducible proteins (e.g., upstream of the GR gene) and nblA coding for the proteolysis adapter NblA, in the vicinity of XR genes from subclades A and D. Light-dependent up-regulation of hli26 and XR genes7,27, might explain their genomic proximity. Noteworthily, NblAs participate in controlled degradation of phycobilisomes under nutrient starvation28,29 and possibly under high light30, suggesting that XR expression in cyanobacteria might be co-regulated under similar conditions as well.

Given the recently reported gene regulatory activity of GR27, we also assessed the appearance of genes coding for ArsR-like helix-turn-helix transcription regulators similar to GvTcR, GR’s interaction partner, in XR-possessing cyanobacteria. Only genomes from the Elainellaceae genus g__JAAUTN01 were found to have related genes (≥50% identity to GvTcR) indicating that the interaction between GR and GvTcR might be specific to G. violaceus.

Xanthorhodopsins are well known from a wide range of strictly unicellular prokaryotic and eukaryotic microbes (see Fig. 1A). Their appearance in multicellular (filamentous) cyanobacteria is thus noteworthy, which motivated us to choose psXR, a representative of an XR lineage from Antarctic Pseudanabaenacea (see Fig. 1B) for characterization. psXR adds to the short list of ion-pumping rhodopsins characterized from filamentous cyanobacteria, together with the bacterial halorhodopsin MrHR (= MastR) from Mastigocladopsis repens PCC 1091420,31,32, and the cyanorhodopsin-II P7104R from Nodosilinea nodulosa PCC 710421.

Functional characterization of psXR

To investigate the ion transport function of psXR, the protein was expressed in E. coli cells. The transformed psXR-expressing Escherichia coli exhibited a purple color, indicating the formation of a functional rhodopsin (Fig. 2A). The proton transport activity of psXR in the E. coli cells was assayed by monitoring the light-induced changes in pH of the external solvent, which was used to characterize the functions of other microbial rhodopsins, like outward and inward proton, inward Cl−, and outward Na+ pumps33,34,35. The cell suspension of E. coli expressing rhodopsin proton pumps shows acidification (pH decrease) or alkalinization (pH increase) of the external medium upon illumination by outward or inward proton transport, respectively, and the signals are eliminated in the presence of a protonophore, CCCP. We observed acidification of the external solvent, and it was largely eliminated by the addition of 10 μM CCCP, suggesting that psXR functions as a light-driven outward proton pump (Fig. 2A). The large pH decrease observed for psXR following the addition of CCCP is consistent with the fact that it poses a DTE motif (Fig. 2B and Supplementary Fig. S4), which is the most common among the outward proton-pumping rhodopsins and is known for its effectiveness in achieving higher transport activity1,34,36,37.

Fig. 2: Amino acid sequence and light-driven active proton transport of psXR.
Fig. 2: Amino acid sequence and light-driven active proton transport of psXR.The alternative text for this image may have been generated using AI.
Full size image

A Proton transport activity assay of psXR, derived from an Antarctic cyanobacterium, in E. coli cells without (blue line) and with (green line) 10 μM CCCP. The cells were illuminated with light (λ > 500 nm) for 150 s (yellow line). The picture of the pellet of E. coli cells expressing psXR, highlighting the strong purple color of the circle, is shown next to the result. B Amino acid sequence of psXR in TM3 compared with other XRs and PRs. Black triangles indicate the motif residues. C Proton transport activity assay of GR (violet) and psXR (blue) at different temperatures. D Initial H+ transport rates of GR and psXR at different temperatures; E Difference in absorption spectra before and after HA bleaching reaction of psXR in solubilized E. coli membranes at pH 7.0. The λmax value was determined based on the position of the maximum absorption peak indicated, and the absorption of retinal oxime produced by the hydrolysis reaction of RSB and HA was observed as a peak at 360–370 nm.

Our experiments comparing the proton-pumping activities of GR and psXR revealed striking differences in temperature dependence. When expressed in E. coli cells, these two phylogenetically related rhodopsins displayed different responses to temperature reduction (Fig. 2C, D). GR exhibited the expected enhancement of apparent proton-pumping activity as temperature decreased from 20 °C to 5 °C, consistent with previous observations using heterologous E. coli expression systems. This temperature-dependent increase in apparent activity (approximately 3-fold stronger signal at 5 °C compared to 20 °C) is primarily attributed to reduced activity of endogenous E. coli proton transporters at lower temperatures, which decreases the rate of passive proton backflow across the membrane. Remarkably, and contrary to expectations, psXR demonstrated temperature-independent initial proton-pumping rates across the tested temperature range (5 °C, 12.5 °C, and 20 °C). As shown in Fig. 2D, the initial slopes of light-induced pH changes for psXR remained nearly constant across all temperatures, while GR showed a clear negative correlation between temperature and initial pumping rate. This unexpected behavior suggests that psXR’s intrinsic pumping activity might actually decrease at lower temperatures, likely due to its high activation enthalpy barrier that renders the photocycle significantly slower under low-temperature conditions than that in GR. The reduction in the proton-pumping turnover rate of psXR counterbalances the reduced activity of endogenous E. coli proton transporters that would otherwise lead to enhanced signals at lower temperatures. The contrasting temperature responses between GR and psXR indicate potential adaptations in the molecular mechanism of psXR. This finding is particularly intriguing given that some cold-adapted organisms maintain membrane fluidity through increased proportions of unsaturated fatty acids. The native lipid environment of psXR in cyanobacteria likely differs significantly from that in the E. coli membrane, which may influence its temperature-dependent functionality.

To examine the λmax and the expression level of psXR, a hydroxylamine (HA) bleaching assay was conducted after solubilization of E. coli membranes with detergent, n-dodecyl-β-D-maltopyranoside (DDM). HA hydrolyzes the protein-bound retinal, resulting in free retinal-oxime, so that the λmax of the original protein can be estimated by calculating the difference absorption spectra between before and after the HA treatment. Figure 2E shows the result representing the photobleaching of psXR in the presence of 500 mM HA. The positive peak at ~542 nm corresponds to the absorption of the rhodopsin with the protonated retinal Schiff base (RSB) present prior to the bleaching. The small positive peak at ~425 nm was attributed to the absorption of rhodopsin with the deprotonated RSB, while the negative peak at 361 nm was derived from the absorption of the reaction product, retinal-oxime. This result indicates that the expressed psXR has the absorption spectrum mainly peaked in the visible region.

Molecular properties and photocycle

To further study the molecular properties, we purified and spectroscopically investigated psXR. The purified psXR showed a λmax at 545 nm in detergent (DDM) (Fig. 3A). The spectra of purified psXR at different pH values are shown in Supplementary Fig. S5. The absorption peak in the visible wavelength region corresponding to the state with the deprotonated RSB, denoted by λmax = 546 nm, decreased and slightly redshifted when the pH was raised, and simultaneously, another peak appeared in the ultraviolet (UV) region (λmax = 366 nm) (Supplementary Fig. S5A, top). The difference absorbance values at these peak wavelengths were plotted against pH and fitted with the Henderson-Hasselbalch equation (Supplementary Fig. S5A, top), which determined that the RSB in psXR exhibits two pKa values of 10.8 ± 0.4 and 11.19 ± 0.02. These values are in line with the RSB pKa of typical microbial rhodopsins (e.g., the RSB pKa of bacteriorhodopsin is 13.3 ± 0.3)38. In contrast, a 13 nm redshift in the λmax was observed on the acidic side (Supplementary Fig. S5B, top). This is caused by the protonation of the counterion in the third transmembrane helix (TM3) (aspartic acid (D115), Fig. 2B), as known for many microbial rhodopsins1,39,40,41.

Fig. 3: Absorption spectra and photocycle of psXR.
Fig. 3: Absorption spectra and photocycle of psXR.The alternative text for this image may have been generated using AI.
Full size image

A UV–vis absorption spectrum for psXR. The picture of the purified protein is also shown. B Transient absorption spectra at different time points. C The two-dimensional plot of transient absorption change. D Absorption spectra of photointermediates. E Time course of the transient absorption change; F Photocycle model of psXR.

The retinal configuration was also investigated using high-performance liquid chromatography (HPLC) on retinal oximes produced by the hydrolysis reaction of the RSB in psXR with HA (Supplementary Fig. S6). Most retinal chromophores derived from dark-adapted (DA) proteins showed an all-trans configuration with trace amounts of 13-cis and 11-cis ones (Supplementary Fig. S6A). During illumination at λ = 550 ± 10 nm, the 13-cis-retinal form accumulated substantially, while the amount of 11-cis-retinal form slightly increased (yellow bars in Supplementary Fig. S6B and Supplementary Table S1). The all-trans form represents the functional state of psXR, similar to typical microbial rhodopsins2,42.

Next, the photocycle reaction of psXR was studied by the laser-flash photolysis method, in which the psXR protein was solubilized in DDM. Figure 3B shows the transient change in absorption of psXR upon excitation at λmax = 550 nm, representing the accumulation of blue-shifted (M) and redshifted (K and O) photointermediates (Fig. 3C) at 413 nm and 618 nm, respectively, in addition to L, N, and psXR′ intermediates with λmax close to the initial state (Fig. 3D). Simultaneously, the maximum bleaching signal at 543 nm was observed, indicating the M intermediate accumulation at t~155 μs. To monitor the transient absorption change representing the short-lived photointermediates prior to M, we conducted measurement at specific probe wavelengths using a photomultiplier tube at higher time resolution (Fig. 3E). The photocycle and the absorption spectra of the initial state and five photointermediates (K2/L1/M1, K3/L2/M2, M3/N1, N2/O1, and psXR′) were determined by global fitting analysis of the transient absorption changes (Fig. 3F).

Xanthophyll binding to psXR

The ability to bind a carotenoid antenna in microbial rhodopsins requires the presence of a fenestration, an opening in the protein moiety that exposes the retinal-binding pocket to the external environment. This structural feature facilitates energy transfer between the two chromophores and is determined by a conserved glycine residue at a specific position in the amino acid sequence (see Supplementary Fig. S3, black circle)6. XRs (as well as PRs) exhibit variability at this position, with some representatives possessing a bulky tryptophan or phenylalanine that precludes carotenoid antenna binding, while others maintain the glycine residue, allowing for such binding6,8,43,44 (if no other side chain blocks the fenestration secondarily).

As exemplified by psXR, structures predicted for all collected cyanobacterial XRs featured a fenestration exposing the retinal ring (Fig. 4A) similar to SrXR5, thanks to the glycine residue at the SrXR position 156 in TM5 (G172 in psXR, Supplementary Fig. S4). This hints at the possibility that besides GR, many, if not all, cyanobacterial XR might be able to bind carotenoid antennas as well. We hypothesized that psXR could potentially utilize a ketolated carotenoid as a light-harvesting antenna, as observed in previous energy transfer studies involving light-harvesting ketocarotenoids and light-driven proton pump XRs in bacterial systems such as SrXR with salinixanthin5 and GR with echinenone9 (Fig. 4B). To investigate the impact of carotenoid antenna binding on the properties of the retinal-binding pocket, we attempted to purify psXR with canthaxanthin, a ketolated xanthophyll known to be produced by several Pseudanabaenaceae isolates45,46 (Supplementary Fig. S7A). Our findings revealed that canthaxanthin did not bind to psXR (Supplementary Fig. S7B).

Fig. 4: XRs fenestrations and carotenoid structures.
Fig. 4: XRs fenestrations and carotenoid structures.The alternative text for this image may have been generated using AI.
Full size image

A Structural comparison of fenestrated XRs: psXR (AlphaFold 3 model)73, SrXR (PDB: 3DDL)5, and Kin4B8-XR (PDB: 8I2Z)8. Low-quality N- and C-terminal extensions are not shown (see “Methods” section for details). The structures are aligned based on the retinal β-ionone ring position, and the fenestration zone is highlighted. For clarity, salinixanthin and zeaxanthin are not shown in the enlarged fenestration zone of SrXR and Kin4B8-XR structures, respectively. B The chemical structures of all-trans-retinal and various ketocarotenoids (salinixanthin and echinenone) and hydroxycarotenoids (zeaxanthin and lutein). The type of the ionone ring (β and ε) of each carotenoid is also shown.

Motivated by this result, we explored the binding potential of psXR with a hydroxylated carotenoid, as observed in other bacterial XRs, such as Kin4B8-XR8 and AAP5 XR7 (see Fig. 4B). Incubation of purified psXR with lutein, a non-symmetric isomer of zeaxanthin (Fig. 4B and Supplementary Fig. S7C), resulted in binding (Fig. 5A). Energy transfer from lutein to the retinal molecule of psXR was investigated by laser-flash photolysis (Fig. 5B). It revealed an enhancement in the transient absorption (TA) signal for psXR bound to lutein when excited at the 450–490 nm region, supporting the energy transfer from the carotenoid to the retinal, enhancing the retinal isomerization. The excitation energy dependence of the TA signal intensity of psXR (Supplementary Fig. S8A) showed that the linearity of the transient absorption signal intensity was maintained up to an energy value (0.55 mJ cm−2), which is as high as in other carotenoid-binding rhodopsins, suggesting that psXR’s retinal isomerizes as efficiently as in these rhodopsins8,10,11. psXR photocycle was enhanced by a factor of 1.3 upon excitation at 430 nm when bound to lutein, which is similar compared to the value of Kin4B8-XR and a little more than half compared to the value of HeimdallR1 (1.6- and 3.2-fold enhancement, respectively)8,10 bound to lutein. The excitation energy transfer quantum yield (ϕEET) between the retinal and lutein chromophores was estimated to be more than 20% and nearly constant (see table of percentage values on Fig. 5B) over all excitation wavelengths. Subsequently, the retinal chromophore will be isomerized using the energy from the carotenoid. However, its isomerization quantum yield is not clear, and it will be determined in the future by further experiments

Fig. 5: Absorption spectra and photocycle of psXR with lutein.
Fig. 5: Absorption spectra and photocycle of psXR with lutein.The alternative text for this image may have been generated using AI.
Full size image

A UV–vis absorption spectrum of purified psXR with lutein. The picture of the purified protein is also shown. B Ratios of transient absorption changes in psXR with and without lutein at different excitation wavelengths. Bars are colored according to the color of the excitation light. Raw data of transient absorption signals at each excitation wavelength are included in Supplementary Fig. S8B. The absorption spectra of psXR without (purple line) and with (orange line) lutein were overlaid. The red dashed line indicates no difference between with and without lutein. The table with the ϕEET percentages and the pictures of the purified proteins is shown next to the corresponding results. C Transient absorption spectra at different time points. D The two-dimensional plot of transient absorption change. E Time course of the transient absorption change. F Absorption spectra of photointermediates. Asterisks represent the absorption change of lutein. G Photocycle model of psXR with lutein.

To investigate how the retinal-binding pocket is affected by lutein binding, we compared the absorption of purified psXR with and without lutein (Figs. 3A and 5A). Binding lutein did not display a shift in the observed λmax when compared with psXR alone, which was also observed in bacterial rhodopsins investigated previously6,8. The absorption peak of the retinal was decreased and slightly redshifted when the pH was raised, corresponding to the structural change affecting the deprotonated RSB in psXR, reproducing the behavior observed in psXR alone (Supplementary Fig. S5A, bottom). In the case of the acidic side, a redshift in the λmax representing the protonation state of the counterion was observed as psXR alone (Supplementary Fig. S5B, bottom). Hence, no difference in the λmax was observed between psXR with and without lutein, indicating the impact of the lutein binding on the property of the retinal-binding site is minor. The influence of the lutein binding on the retinal isomer composition in psXR was also investigated using HPLC of retinal oximes produced by hydrolyzing the RSB with HA (Supplementary Fig. S6B). The binding of lutein increased the fraction of the all-trans form in the dark from 92.6% to 95.5%, which represents the functional form to initiate proton-pumping photocycle (Supplementary Fig. S6B, bottom and Supplementary Table S1). The transient absorption changes observed for psXR with lutein are roughly similar to those observed for psXR without lutein (Fig. 5C–F). Multi-exponential analysis identified 5 photointermediates having different absorption spectra (K2/L1, K3/L2/M1, M2/N1, N2/O1, and psXR’) (Fig. 5F, G). Two additional sharp peaks around 465 nm and 485 nm (see asterisks in Fig. 5C, D, F), which were absent for psXR alone, were observed when lutein was bound to the protein, indicating that a conformational change of the protein occurs during the photocycle that alters the structure and the absorption of lutein. The observed biphasic spectral feature upon carotenoid antenna binding to the protein, characterized by two distinct sharp peaks, contrasts with the monophasic spectral profile, exhibiting a single sharp peak, previously reported for Kin4B8-XR and HeimdallR18,10.

Structure of psXR with lutein

To investigate the xanthophyll-binding properties of psXR, we determined the cryo-electron microscopy structure of the lutein-bound psXR at a nominal resolution of 2.64 Å (Supplementary Fig. S9A–C; Supplementary Table S2). psXR assembles into a pentameric configuration (Fig. 6A), similar to Kin4B8-XR (Protein Data Bank ID: 8I2Z)8. The monomeric units of psXR exhibit highly conserved features with other XRs (Kin4B8-XR and SrXR44), including similar carotenoid-binding orientations (Supplementary Fig. S9D).

Fig. 6: Structure of psXR bound to lutein.
Fig. 6: Structure of psXR bound to lutein.The alternative text for this image may have been generated using AI.
Full size image

A Electron microscopy map and pentameric structure of the lutein-bound psXR, viewed from the extracellular side. Lutein and all-trans retinal (ATR) are shown as stick models. B Overall structure of the monomeric unit, with lutein and the retinal chromophores. C Lutein-binding site with the cryo-EM density. The extended carotenoid is tightly bound to the transmembrane surface of psXR, traversing nearly the entire bilayer, with an inclination of about 60° to the membrane normal. Fenestrations in psXR (D) and Kin4B8-XR (E). F Positional relationship between the hydroxyl ring of lutein, retinal, and surrounding residues. G Conservation of the residues surrounding the fenestrations in XRs. Glycine is colored orange. Polar and hydrophobic residues are colored green and blue. Tyrosine is colored cyan.

In psXR, lutein is positioned transversely along the outer surface of TM6 (Fig. 6C), with its 3-hydroxy-β-ionone ring fitting into the fenestration in the protein enabled by the residue G172. This binding mode closely resembles that of zeaxanthin in Kin4B8-XR (Fig. 6D, E), suggesting a conserved mechanism of energy transfer. However, unlike Kin4B8-XR, the fenestration at the base of psXR is less hydrophilic, although the backbone carbonyl group of I169 may still form a hydrogen bond with lutein. Furthermore, the residue corresponding to A205 in Kin4B8-XR is L224 in psXR, whose bulkier side chain imposes a steric constraint. This constraint slightly shifts the orientation of lutein’s ring to zeaxanthin in Kin4B8-XR (Fig. 6F, G; Supplementary Fig. S9E). Such structural features of the fenestration may contribute to maintaining a stable energy transfer mechanism by reducing the distance between the lutein ring and the β-ionone ring of the retinal, while also preventing the intrusion of water molecules into the fenestration.

Discussion

Our study presents the discovery and characterization of a xanthorhodopsin (psXR) from an uncultured Antarctic Pseudanabaena sp., expanding our understanding of microbial rhodopsins in these photosynthetic organisms.

Cyanobacteria with rhodopsin proton pumps represent the only case of photoautotrophic prokaryotes in which two mechanisms of converting light energy into electrochemical proton gradients co-exist: the chlorophyll-based photosystems and a lower-efficiency and lower-maintenance retinal-based rhodopsin “photosystem”. As illustrated by psXR and GR, cyanobacterial XRs are well-suited for capturing light in the green-yellow part of the spectrum, potentially complementing the chlorophyll-based photosynthetic apparatus (blue and red absorption). Coincidentally, the giant antenna complexes of cyanobacteria, the phycobilisomes, absorb in the green-yellow light range as well. As phycobilisome biosynthesis and maintenance incur a substantial nitrogen cost, both physiological mechanisms of dealing with nitrogen starvation: controlled degradation of phycobilisomes enabled by the “non-bleaching” genes28,29, and partial loss of phycobilisome genes in nitrogen-limited environments, such as loss of genes for the peripheral green-absorbing pigment phycoerythrin in polar environments47, are known. It is thus possible that the acquisition of green-yellow absorbing rhodopsins by cyanobacteria may compensate for phycobilisome deficiency, physiological or evolutionary. More broadly, rhodopsin proton pumps might be of benefit to cyanobacteria under conditions where photosynthesis is inhibited, as has been suggested for diatoms expressing XRs under iron limitation14,15,17. Indeed, the associations between XR genes and nblA and genes for high-light inducible proteins (see Supplementary Fig. S2) hint at the existence of a regulatory mechanism by which up-regulation of XR genes might be coupled with degradation of the phycobilisomes and photoinhibition of the photosynthetic apparatus, akin to the switching mechanism between the XR and the bacteriochlorophyll-based photosystem in Sphingomonas glacialis AAP57. In Gloeobacter, both kinds of light-absorbing systems are active under non-limiting laboratory conditions13 with GR perhaps serving as a compensation for the lack of thylakoids48. It stands to reason that for cyanobacteria with a higher degree of compartmentalization, the role of rhodopsin proton pumps is linked to physiological and genetic responses to environmental stress. Further, the abundance of XRs in polar cyanobacteria, of which psXR is an example (see Fig. 1A), suggests a selective advantage for XRs specifically in this harsh environment.

The complexity of the XR “photosystem” in cyanobacteria is further increased by the potential for recruiting carotenoid antennas. The detailed photocycle analysis of psXR, both with and without xanthophyll, provides insights into the mechanism of this interaction. The observed enhancement of the photocycle by lutein binding, although modest, suggests a potential role in optimizing light capture under variable light conditions of Antarctic ecosystems. The unique spectral changes observed during the photocycle of lutein-bound psXR, particularly the two sharp peaks at 465 nm and 485 nm, indicate changes in protein-carotenoid interactions, as previously observed for SrXR. In that case, they were interpreted as reflecting a transient loosening of the carotenoid-binding site caused by a conformational change in the protein5. Interestingly, the temperature dependence of the proton-pumping activity of psXR differs substantially from that of GR (Fig. 2C). The molecular mechanism underlying this difference will be studied by investigating the temperature dependence of the photocycle kinetics in the near future.

It is intriguing that psXR binds a 3-hydroxycarotenoid, as this stands in contrast to GR, which was found to bind 4-ketocarotenoids instead. This demonstrates that despite a relatively recent common origin (see Fig. 1A), cyanobacterial XRs can bind diverse carotenoid antennas. The energy transfer from lutein to the retinal chromophore in psXR showed an efficiency similar to other XRs, suggesting that this diversity might be incidental and reflects adaptation to the pool of endogenous xanthophylls. Indeed, in G. violaceus, unusually for cyanobacteria, xanthophylls are represented exclusively by echinenone49, while Pseudanabaena isolates have mainly 3-hydroxy- and less abundantly 4-ketocarotenoids45,46. It is important to stress that in our experiments, zeaxanthin’s isomer, lutein, was used as a generic 3-hydroxycarotenoid due to its commercial availability. Similarly to the ability of GR to bind a range of non-native carotenoids sharing the 4-keto-β-ionone ring6,9,50, psXR likely binds several native carotenoids with the 3-hydroxy-β-ionone ring (cf. Supplementary Fig. S3).

The cryo-electron microscopy structure of lutein-bound psXR provides detailed insights into the xanthophyll-binding mechanism of this XR variant. The xanthophyll-binding fenestration in psXR induces a rotation of the carotenoid’s β-ionone ring: residue L224 protrudes into the pocket and pushes the ring upward (Fig. 6D, Supplementary Fig. S9E). This rotation shortens the effective conjugation length and accounts for the uniform ~3 nm hypsochromic shift of the three lutein absorption peaks (431, 454, 484 nm) relative to those reported for Kin4B8-XR8. As snow and ice preferentially remove longer wavelengths, the ambient light spectrum is enriched in shorter wavelengths, making this blueshift a plausible spectral adaptation. The same steric effect also narrows the distance between the lutein and retinal ionone rings, a geometry expected to enhance energy transfer. Thus, a single pocket residue can simultaneously tune the antenna spectrum and improve chromophore coupling, underscoring the adaptive flexibility of xanthorhodopsins.

Further research into the diversity and function of XRs in cyanobacteria might shed light on the evolutionary strategies that enable these organisms to dominate a wide range of aquatic environments.

Methods

Search for rhodopsin genes in cyanobacteria

The incidence of different rhodopsin genes in cyanobacteria was estimated based on genomes in JGI’s Genomic catalog of Earth’s Microbiomes (GEMs)51 operational taxonomic unit [OTU] representatives, and assemblies of GTDB r. 21452 species representatives. Assemblies assigned to class “Cyanobacteriia” were translated using getorf from EMBOSS v. 6.6.053 (minimum ORF size 150 nt), and rhodopsins were identified using Pfam profile PF01036.24 with hmmsearch from HMMER v. 3.454 with the default gathering cutoff. The matching proteins were classified into families using a curated database of regular prokaryotic microbial rhodopsins with blastp from NCBI BLAST v. 2.16.055. Best matches with an E-value of <1e-10 and ≥50% identity to the reference were used for family assignment. The results were visualized using ComplexUpset v. 1.3.356.

Phylogenetic analyses

For the phylogeny of xanthorhodopsins across bacterial taxa, the following strategy was utilized. SrXR was used to search JGI GEM genomes and OTU representatives, and assemblies of GTDB species representatives with tblastn from NCBI BLAST. Genes on matching scaffolds (E-value threshold of 1e-10 and at least 200 residues alignment length) were predicted with prodigal v. 2.6.357 in metagenomic mode. To extract XR sequences, the resulting proteins were searched against the curated rhodopsin database (see above) with blastp, and best matches to reference XRs with an E-value threshold of 1e-30 and ≥40% identity were retained. The XRs were combined with outgroup sequences (NQ pumps and clade P1 representatives QsActR, KrActR, and DSE009) and reference XRs, clustered at a 95% identity level with CD-HIT v. 4.8.158, aligned with MAFFT v. 7.52559 in automatic mode, trimmed with trimAl v. 1.4.160 allowing up to 10% gaps per position, and a maximum likelihood phylogeny was reconstructed with IQ-TREE v. 2.2.2.361 with automatic model selection (LG + F + R10 selected as the best-fit model) with 1000 replicates for ultra-fast bootstrap62. In the resulting phylogeny, the ingroup appeared monophyletic (confirming the classification of the sequences), and the outgroups were used for rooting. For the fine-grained phylogenetic analysis, sequences belonging to the cyanobacterial XR clade were extracted from a database containing rhodopsins from JGI GEMs, GTDB species representatives, proteins from JGI’s Integrated Microbial Genomes & Microbiomes (IMG/M)63 assigned to the Pfam family PF01036 (downloaded in November 2023), and the UniRef100 r. 2023_0564 database based on similarity to the initially collected cyanobacterial XRs and preliminary phylogenetic analyses. Additional Pseudanabaenales MAGs with psXR-related genes were recruited from NCBI WGS based on blast searches. Sequences at least 240 residues in length were combined with outgroup XRs, clustered at 100% identity level with CD-HIT, aligned with MAFFT in automatic mode, trimmed with trimAI in atomated1 mode, and a maximum likelihood phylogeny was reconstructed using RAxML v. 8.2.1265 under the PROTGAMMAAUTO model and with 1000 rapid bootstrap replicates. The corresponding cyanobacterial assemblies were taxonomically classified using GTDBtk v. 2.3.266 with GTDB r. 214 as a reference database.

XR gene neighbors

Regions of at most 20,000 nt surrounding the XR genes were extracted and annotated using PGAP v. 2024-04-27.build742667. Immediate neighbors chosen for further annotation included genes and operons within ca. 1800 nt of the XR gene. Representative genomic fragments were chosen per cluster of similar genomic contexts, and homologous genes were identified with Proteinortho v. 6.0.2568 and based on functional annotations.

Pathway analysis

Potential for biosynthesis of cyclic carotenoids in the Pseudanabaena MAG 3300012044_11 was analyzed with KAAS69 and additional manual blastp searches with queries representing characterized cyanobacterial enzymes involved in carotenoid biosynthesis.

Construction of the expression plasmid

The gene chosen for expression originated from JGI GEM genome 3300012044_11, an Antarctic MAG from Dry Valleys (JGI Genomes Online Database analysis Ga0136636) classified in the genus Pseudanabaena (Pseudanabaenales: Pseudanabaenaceae). psXR was codon-optimized for expression in Escherichia coli, synthesized (GenScript, China), and cloned into pET21a(+) vector (Novagen, Merck KGaA, Germany) with a C-terminal 6×His-tag, using NdeI and XhoI restriction sites.

Protein expression and purification

E. coli cells harboring the psXR-cloned plasmid were cultured in 2 × YT medium containing 50 μg/mL ampicillin. The expression of C-terminal 6× His-tagged protein was induced by 0.1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) in the presence of 10 μM all-trans-retinal (Toronto Research Chemicals, Canada) at 37 °C for 4 h. The harvested cells were sonicated (Ultrasonic Homogenizer VP-300N; TAITEC, Japan) for disruption in a buffer containing 50 mM Tris–HCl (pH 8.0) and 5 mM MgCl2. The membrane fraction was collected through ultracentrifugation (CP80NX; Eppendorf Himac Technologies, Japan) at 142,000 × g for 1 h. The protein was solubilized in a buffer containing 50 mM MES–NaOH (pH 6.5), 300 mM NaCl, 5 mM imidazole, 5 mM MgCl2, and 3% n-dodecyl-β-D-maltopyranoside (DDM) (ULTROL Grade; Calbiochem, Sigma-Aldrich, MO). Solubilized protein was separated from an insoluble fraction through ultracentrifugation at 142,000 × g for 1 h. Protein was purified using a Co-NTA affinity column (HiTrap TALON crude; Cytiva, MA). The column was washed with a 15-column volume buffer containing 50 mM MES–NaOH (pH 6.5), 300 mM NaCl, 50 mM imidazole, 5 mM MgCl2, and 0.1% DDM. Protein was eluted in a buffer containing 50 mM Tris–HCl (pH 7.0), 300 mM NaCl, 300 mM imidazole, 5 mM MgCl2, and 0.1% DDM. Eluted protein was immediately concentrated using a 50-mL centrifugal ultrafiltration filter with a 30-kDa molecular weight cutoff (Amicon Ultra-4, Millipore, Merck KGaA, Germany), and the sample was dialyzed against a buffer containing 50 mM HEPES–NaOH (pH 7.0), 150 mM NaCl, 10% glycerol, and 0.1% DDM.

Proton transport activity assay

Rhodopsin-expressing E. coli cells were collected through centrifugation at 4800 × g at 20 °C for 2 min (CF15RF; Eppendorf Himac Technologies, Japan) and washed with 100 mM NaCl. The cells were equilibrated thrice with rotational mixing in 100 mM NaCl at room temperature for 10 min. Finally, the cells were suspended in 7.5 mL of unbuffered 100 mM NaCl, and the optical density (OD) at 600 nm was adjusted to 2. The cell suspension was placed and stirred in the dark in a glass cell at 5, 12.5, and 20 °C and illuminated at λ = 550 ± 5 nm from the output of a 300 W xenon light source (MAX-303, Asahi Spectra, Japan) through a bandpass filter (HMX0550, Asahi Spectra, Japan). Light-induced pH changes were measured using a pH electrode (9618S-10D, HORIBA, Japan). The measurements were repeated under the same conditions after the addition of carbonyl cyanide m-chlorophenylhydrazone (CCCP, final concentration = 10 μM).

To quantitatively compare ion transport activity, the amount of protein was determined by measuring near-UV absorption of retinal oxime produced by the hydrolysis reaction between the retinal Schiff base (RSB) in the protein and hydroxylamine (HA). Briefly, rhodopsin-expressing E. coli cells were washed with a buffer containing 133 mM NaCl and 66.5 mM Na2HPO4 (pH 7.0). The washed cells were treated with 1 mM lysozyme and a small amount of DNase I for 1 h and disrupted through sonication. To solubilize rhodopsins, 3% DDM was added, and the samples were stirred overnight at 4 °C. The rhodopsins were bleached with 500 mM HA and illuminated with visible light (λ > 500 nm) from the output of a 300 W xenon lamp (MAX-303, Asahi Spectra, Japan) through long-pass (Y-52, AGC Techno Glass, Japan) and heat-absorbing (HAF-50S-50H, SIGMAKOKI, Japan) filters. The absorption changes due to the bleaching of rhodopsin by the hydrolysis reaction between retinal and HA and the formation of retinal oxime were measured using a UV-visible spectrometer (V-750, JASCO, Japan). The amount of rhodopsin expressed in E. coli cells was estimated by the absorbance of produced retinal oxime and its differential molar extinction coefficient (ε) (33,600 M−1 cm−1)70 relative to the original absorption of unhydrolyzed retinal (Fig. 2C).

Purification of carotenoid-binding psXR

Carotenoid-binding psXR was obtained by mixing purified protein with carotenoids (lutein (Sigma-Aldrich, MO, USA) or canthaxanthin (Sigma-Aldrich, MO, USA)). First, carotenoids were dissolved in dimethyl sulfoxide (DMSO) at twice the molar amount of psXR to be bound. The molar amounts of each were estimated by measuring the absorption spectrum, using a molar extinction coefficient of ε = ~50,000 M−1 cm−1 at the absorption maximum of rhodopsin and ε = 145,000 M−1 cm−1 at 445 nm for carotenoids. The volume of DMSO was adjusted so that it did not exceed 2% of the rhodopsin solution volume. After preparing the carotenoid solution as described above, it was added to psXR and left on ice and in the dark for 2 h, gently mixing the solution every 15 min. To remove the excess of carotenoids not bound to psXR, psXR complexed with carotenoids was purified again using the same method as the purification of the original psXR protein. The purified protein was immediately concentrated using a 50-mL centrifugal ultrafiltration filter with a 30-kDa molecular weight cutoff (Amicon Ultra-4, Millipore, Merck KGaA, Germany), and the sample was dialyzed against a buffer containing 20 mM HEPES–NaOH (pH 7.0), 100 mM NaCl, and 0.05% DDM.

High-performance liquid chromatography (HPLC) analysis of retinal isomers

Retinal configuration in psXR was analyzed by HPLC using purified proteins in a buffer containing 20 mM Tris–HCl (pH 8.0), 100 mM NaCl, and 0.05% DDM. Before the measurements, the OD of the samples at their maximum absorption wavelength (λmax) was adjusted to ~0.2 (protein concentration of approximately 0.1 mg mL−1), and the proteins were stored at 4 °C overnight for dark adaptation. The HPLC system was equipped with a silica column particle size of 3 μm (150 × 6.0 mm, Pack SIL, YMC, Japan), a pump (PU-4580, JASCO, Japan), and a UV–vis detector (UV-4570, JASCO, Japan). The solvent was composed of 15% (v/v) ethyl acetate and 0.15% (v/v) ethanol in n-hexane and with a flow rate of 1.0 mL min−1. To denature the protein, 280 μL of 90% methanol solution was added to the 75 μL sample. Retinal oxime formed by the hydrolysis reaction with 25 μL of 2 M HA solution was extracted with 800 μL of n-hexane, and 200 μL of the solution was injected into the HPLC system. For measurements during light illumination, the sample solutions were illuminated at λ = 550 ± 10 nm using a bandpass filter (AGC Techno Glass, Japan) for 1 min, followed by denaturation and hydrolysis of the retinal chromophore under illumination. For measurements of light-adapted samples, the sample solutions were illuminated at λ = 550 ± 10 nm using a bandpass filter (AGC Techno Glass, Japan) for 1 min, and after waiting for 1 min, denaturation and hydrolysis reactions of the retinal chromophore were performed. The molar compositions of the retinal isomers were calculated from the areas of the corresponding peaks in the HPLC patterns using the molar extinction coefficients at 360 nm for each isomer (all-trans-15-syn: 54,900 M−1 cm−1; all-trans-15-anti: 51,600 M−1 cm−1; 13-cis-15-syn: 49,000 M−1 cm−1; 13-cis-15-anti: 52,100 M−1 cm−1; 11-cis-15-syn: 35,000 M−1 cm−1; 11-cis-15-anti: 29,600 M−1 cm−1). Three independent measurements were performed to estimate experimental errors. The compositions of the retinal isomers are listed in Supplementary Table S1. Peaks were assigned by comparing the elution time with those of authentic retinal oxime isomers.

pH titration

To investigate the pH dependence of the absorption spectra, the concentration of proteins was adjusted to OD≈0.5 (protein concentration of approximately 0.25 mg mL−1) at λmax and solubilized in a 6-mix buffer (trisodium citrate, MES, HEPES, MOPS, CHES, CAPS (10 mM each, pH 7.0), 100 mM NaCl, and 0.05% DDM). The pH was adjusted to the desired values by adding small aliquots of HCl and NaOH. Absorption spectra were recorded using a UV–vis spectrometer (V-750, JASCO, Japan). The measurements were performed at every 0.3–0.6 pH value.

Laser-flash photolysis

For the laser-flash photolysis spectroscopy, the protein was solubilized in 20 mM HEPES–NaCl (pH 7.0), 100 mM NaCl, 0.05% DDM. The OD of the rhodopsin was adjusted to ~0.4–0.5 (protein concentration of approximately 0.2–0.25 mg mL−1 at the λmax), and then the rhodopsin sample solution was placed in a 10 mm path-length quartz cell. The laser-flash photolysis measurement was conducted as previously described71,72. Nanosecond laser pulses from an optical parametric oscillator (excitation wavelength (λexc) = 550 nm, 10 mm diameter, 4.5 mJ pulse−1 cm−2, 3.3 Hz, basiScan, Spectra-Physics, CA) pumped by the third harmonics of Nd-YAG laser (λ = 355 nm, INDI40, Spectra-Physics, CA) were used for the excitation of psXR w/wo lutein. The transient absorption (TA) spectra were obtained by monitoring the intensity change of white light from a Xe-arc lamp (L9289-01, Hamamatsu Photonics, Japan) passed through the sample using an ICCD linear array detector (C8808-01, Hamamatsu Photonics, Japan). To increase the signal-to-noise ratio, 40–60 spectra were averaged, and the singular-value decomposition analysis was applied. To measure the time evolution of transient absorption change at specific wavelengths, the output of a Xe-arc lamp (L9289-01, Hamamatsu Photonics, Japan) was monochromated using monochromators (S-10, Soma Optics, Japan) and focused onto the sample solution with a spot size of 3 × 5 mm. The change in the probe beam intensity after the photoexcitation was monitored using a photomultiplier tube (R10699, Hamamatsu Photonics, Japan). To increase the signal-to-noise ratio, 200–400 signals were averaged. Global fitting of the signals using a multi-exponential function was performed to determine the lifetimes and absorption spectra of each photointermediate.

To measure the fold change in transient absorption change between psXR with and without lutein, nanosecond laser pulses from an optical parametric oscillator (basiScan, Spectra-Physics, CA) pumped by the third harmonics of Nd-YAG laser (λ = 355 nm, INDI40, Spectra-Physics, CA) were used for the excitation of psXR at different wavelengths (λexc = 430, 455, 470, 485, 520, 545, and 585 nm). The pulse energy was adjusted to 0.55 mJ cm−2, which falls within the range that maintains the linearity between the number of absorbed photons and the transient absorption change. To calculate the quantum efficiency of excitation energy transfer (ϕEET) (see table percentages on Fig. 5B), sample solutions of rhodopsins, one without a bound lutein and the other with a bound lutein, and their absorbances at the excitation wavelengths, \({{Abs}}^{0}\) and \({Abs}\), respectively, were considered. The number of photons absorbed by the retinal in the rhodopsin without a bound lutein, \({N}_{R}^{0}\) is expressed as:

$${N}_{R}^{0}=(1-{10}^{{-{Abs}}^{0}}){I}_{0}$$

where \({I}_{0}\) is the intensity of the incident light.

As the number of photons absorbed by the retinal and the lutein in the rhodopsin with a bound lutein, \({N}_{R}\) and \({N}_{L}\), respectively, are proportional to their absorbances, \({N}_{R}\) and \({N}_{L}\) can be expressed as:

$${N}_{R}=(1-{10}^{-{Abs}})\frac{{{Abs}}^{0}}{{Abs}}{I}_{0}$$

and

$${N}_{L}=(1-{10}^{-{Abs}})\frac{{Abs}-{{Abs}}^{0}}{{Abs}}{I}_{0}$$

The ratio between the transient absorption signals of the rhodopsin samples without and with lutein, \(\Delta {{Abs}}^{0}\) and \(\Delta {Abs}\), respectively, can be expressed as:

$$\frac{\Delta {Abs}}{{\Delta {Abs}}^{0}}=\frac{{N}_{R}\,+\,{\phi }_{{EET}}{N}_{L}}{{N}_{R}^{0}}$$

The \({\phi }_{{EET}}\) can be expressed as follows using \({N}_{R}^{0}\), \({N}_{R}\), \({N}_{L}\), \(\Delta {{Abs}}^{0}\), and \(\Delta {Abs}\), which can be experimentally determined:

$${\phi }_{{EET}}=\left(\frac{\Delta {Abs}}{{\Delta {Abs}}^{0}}{N}_{R}^{0}-{N}_{R}\right)\frac{1}{{N}_{L}}$$

Protein expression and purification for structural analysis

pBAD-psXR was transfected into E. coli C41 (Rosetta). The transformant was grown in LB supplemented with 50 µg ml−1 ampicillin and 10 µg ml−1 at 220 rpm. at 37 °C. When the OD600 reached 0.6, expression was induced using a 0.1% final concentration of ʟ-arabinose (A3256, Sigma-Aldrich). The induced culture was grown at 120 rpm. overnight (more than 16 h) at 25 °C. Then, the pooled plate content was incubated with 20 µM all-trans retinal for more than 4 h in the dark. The collected cells were disrupted by sonication in a buffer containing 20 mM Tris–HCl, pH 8.0, 150 mM NaCl, and 10% glycerol. The crude membrane fraction was collected by ultracentrifugation at 180,000 × g for 1 h. The membrane fraction was solubilized in buffer containing 20 mM Tris–HCl, pH 8.0, 150 mM NaCl, 1% DDM, and 10% glycerol for 2 h at 4 °C. The supernatant was separated from the insoluble material by ultracentrifugation at 180,000 × g for 20 min and incubated with Ni-NTA resin (Qiagen) for 30 min. The resin was washed with 10 column volumes of wash buffer containing 20 mM Tris–HCl, pH 8.0, 500 mM NaCl, 0.03% DDM, 10% glycerol, and 20 mM imidazole. The protein was eluted in buffer containing 20 mM Tris–HCl, pH 8.0, 150 mM NaCl, 0.03% DDM, 10% glycerol, and 300 mM imidazole. Eluted psXR was concentrated and purified through size-exclusion chromatography on a Superose 6 Increase 10/300 GL column (GE Healthcare) with buffer containing 20 mM Tris-NaCl, pH 8.0, 150 mM NaCl, and 0.03% DDM. The protein was incubated with a buffer containing 20 mM Tris-NaCl, pH 8.0, 150 mM NaCl, 0.03% DDM, and 1.75 mM lutein.

XR structure predictions

For the assessment of the presence of fenestrations in XRs without experimental structures, the structures were predicted with AlphaFold 373 with retinal as a covalent modification and 20 random seeds. PyMOL v. 3.1.3.174 application programming interface and graphical user interface, with the default solvent settings, were used to screen the structures. To visualize the structural prediction of psXR, the low-quality N- and C-terminal extensions were trimmed to exclude terminal residues with the per-residue average pLDDT (predicted local distance difference test) score of backbone atoms below 80.

Cryo-EM single-particle analysis of the psXR-lutein complex

For the cryogenic electron microscopy (cryo-EM) grid preparation of the lutein-bound psXR, the protein was reconstituted in nanodiscs. psXR, MSP2N2, and SoyPC were mixed at a molar ratio of 1:4:200, respectively, and rotated at 4 °C for 1 h. Detergents were removed by adding Bio-Beads SM2 (Bio-Rad) to 40 mg ml−1, followed by gentle agitation. The Bio-Beads (equal amount) were incubated at 4 °C overnight. The Bio-Beads were then removed, and the solution was ultracentrifuged before size-exclusion chromatography.

The ultracentrifuged sample was concentrated and purified through size-exclusion chromatography on a Superose 6 Increase 10/300 GL column (GE Healthcare) with buffer containing 20 mM Tris-NaCl, pH 8.0, and 150 mM NaCl. The peak fractions of the protein were collected and concentrated to an absorbance (A280) of 15, using a centrifugal filter unit (50 kDa molecular weight cutoff; Merck Millipore). Of protein, 3 µl was loaded onto glow-discharged holey carbon grids (Quantifoil Au 300 mesh R1.2/1.3), after which, these were plunge-frozen in liquid ethane, using a Vitrobot Mark IV (Thermo Fisher Scientific).

Cryo-EM imaging data were collected on a Titan Krios at 300 kV, equipped with a Gatan K3 Summit detector and controlled using the EPU software (Thermo Fisher’s single-particle data collection software). Images were obtained at a dose rate of about 15.5940 e− Å−2 s−1, with a defocus ranging from −0.6 to −1.6 μm. Total exposure time was 2.3 s, with 48 frames recorded per micrograph. A total of 12,126 videos were collected. All acquired movies were dose-fractionated and subjected to beam-induced motion correction implemented in cryoSPARC75. The contrast transfer function (CTF) parameters were estimated using cryoSPARC76. Particles were initially picked from a small fraction with Gaussian blob picking and subjected to 2D classification. Class averages showing reasonable features of the psXR pentamer in various orientations were selected as templates for template-based particle picking. Particles from these class averages generated an ab initio model in cryoSPARC. For each full dataset, extracted particles were downsampled to 3.32 Å, followed by two rounds of 2D classification to remove ‘junk’ particles. 3D classification into two classes with the ab initio model as a reference was performed using C5 symmetry. After multiple rounds of 3D classification, particles were re-extracted with a pixel size of 1.10 Å and a box size of 288 pixels. Multiple rounds of local CTF refinement and non-uniform refinement77 were performed using cryoSPARC. Finally, the 327,293 particles in the best class were reconstructed using non-uniform refinement, resulting in a 2.64 Å resolution reconstruction, with the gold-standard Fourier shell correlation (FSC = 0.143) criteria78. The quality of the map was sufficient to build a model manually in Coot79,80.

Model building was performed based on the initial model generated by ModelAngelo81 from the amino acid sequence of psXR. We manually readjusted the model into the density map using Coot and refined it using phenix.real_space_refine v. 1.1982,83. The asymmetric lutein molecule was tested in two orientations due to the lack of clear density to determine its precise orientation. The final placement was selected based on the orientation that best fit the density map.

Statistics and reproducibility

No sample-size calculations were performed. After initial optimization trials, the biophysical experiments (flash photolysis, pH titration) were generally performed once, given the high reproducibility of such experiments and their mutual support. The HPLC experiment was performed on three technical replicates. For the proton-pumping activity experiments, the spheroplast preparations were done in batches and thus batch was included as a random effect in the mixed models; data exclusion criteria are outlined in the corresponding section in the “Methods” section. The investigators were not blinded to allocation during experiments and outcome assessment.

Reporting summary

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