Color-Tuning Mechanism of the Lit Form of Orange Carotenoid Protein

Abstract

Orange carotenoid protein (OCP) of photosynthetic cyanobacteria binds to ketocarotenoids noncovalently and absorbs excess light to protect the host organism from light-induced oxidative damage. Herein, we found that mutating valine 40 in the α3 helix of Gloeocapsa sp. PCC 7513 (GlOCP1) resulted in blue- or red-shifts of 6-20 nm in the absorption maxima of the lit forms. We analyzed the origins of absorption maxima shifts by integrating X-ray crystallography, homology modeling, molecular dynamics simulations, and hybrid quantum mechanics/molecular mechanics calculations. Our analysis suggested that the single residue mutations alter the polar environment surrounding the bound canthaxanthin, thereby modulating the degree of charge transfer in the photoexcited state of the chromophore. Our integrated investigations reveal the mechanism of color adaptation specific to OCPs and suggest a design principle for color-specific photoswitches.

INTRODUCTION

Cyanobacterial carotenoid-binding proteins (CBPs) bind carotenoids such as zeaxanthin, echinenone, and canthaxanthin (Cax). CBPs absorb and dissipate excess light energy and thereby protect the photosynthetic apparatus of host organisms from photo-oxidative damage under intense illumination (Liang et al., 2018). Soluble orange carotenoid proteins (OCPs) are CBPs that bind ketocarotenoids in a non-covalent manner. Structural studies demonstrated that water-soluble CBPs comprise two structural domains, an all-helical N-terminal domain (NTD) and a C-terminal domain (CTD) (Kerfeld et al., 2003), which are connected by a flexible linker domain (Bao et al., 2017b; Melnicki et al., 2016). Asymmetric positioning of ketocarotenoids in OCPs stabilizes the globular form of the protein via intramolecular salt bridges and hydrogen bonds (Kirilovsky and Kerfeld, 2016; Sluchanko et al., 2017).

An ever-increasing number of cyanobacterial genome sequences and phylogenetic analyses of the resulting data have revealed three types of OCPs: OCP1, OCP2, and OCPx (Muzzopappa and Kirilovsky, 2020). A further two closely-related families of OCPs, helical carotenoid proteins (HCPs) homologous to the NTD and C-terminal domain homologs (CTDHs), are present in many cyanobacterial genomes and bind carotenoids in a non-covalent manner (Bao et al., 2017a; Melnicki et al., 2016). Among the identified OCP families, the canonical OCP1 is the most well-studied in terms of structural and spectroscopic properties. OCPs are thought to be derived from a fusion event between HCPs and CTDHs, and OCP2 is evolutionarily intermediate between OCP1 and HCPs and CTDHs (Bao et al., 2017b; Melnicki et al., 2016). Unlike OCP1, OCP2 is not regulated by fluorescence recovery protein (FRP) (Bao et al., 2017b). Furthermore, OCP2 exhibits faster photoconversion but is less structurally flexible than the canonical OCP (Bao et al., 2017b; Kuznetsova et al., 2020).

Photophysical measurements revealed the detailed photoprotection mechanism of OCPs (Fig. 1). Blue-green light activates OCPs into the active (lit) form. In the absence of light, OCPs exist as a stable inactive form (OCP^O) that is orange in color. Once activated, the protein transitions to a metastable active form (OCP^R) that is red in color and binds to cyanobacterial light-harvesting proteins known as phycobilisomes (PBSs). Binding to PBSs causes dissipation of excess energy as heat via the non-photochemical quenching mechanism (Gwizdala et al., 2011). In addition, OCPs regulate reactive oxygen species that accumulate as by-products of photosynthesis (Kerfeld et al., 2017) or following abiotic stresses such as salt stress (Yang et al., 2020).

Fig. 1. Schematic view of photoconversion of orange carotenoid proteins (OCPs) and its role in cyanobacteria.

High irradiance induces photoconversion of OCPs from inactive (OCP^O) form to active (OCP^R) forms followed by binding to phycobilisomes. High energy entering through phycoerythrin (red rods) and phycocyanin (blue rods) of phycobilisomes is dissipated as heat. Carotenoid translocates through the N-terminal domain (NTD), and the C-terminal domain (CTD) detaches from the NTD following illumination. Conversion of OCP^R to OCP^O is triggered via binding of fluorescence recovery protein (FRP) to the CTD. The N-terminal extension (NTE) of NTD attaches to the CTD of the OCP^O state. Modified from the article of Bao et al. (2017a) (Curr. Opin. Plant Biol. 37, 1-9) with original copyright holder’s permission.

Activation of OCP^R involves interdomain interactions and translocation of the bound carotenoid (Liu et al., 2016). Light activation disrupts hydrogen bonds between the β1-ring of the carotenoid and Trp-288 and Tyr-201, leading to unwinding of α-helix of the N-terminal extension (NTE) and a 12 Å carotenoid translocation from CTD to NTD. Recovery to the OCP^O state requires relocation of the carotenoid from NTD to CTD via π-stacking interactions between the carotenoid β-ring and Trp-41 and Trp-101, and formation of hydrogen bonds with Trp-288 and Tyr-201 (Sluchanko et al., 2017). Translocation of the bound carotenoid is responsible for the observed color changes. Atomistic modeling and quantum mechanical calculations revealed the influence of charge polarization on the color-tuning of β-crustacyanin bound to two astaxanthin carotenoids (Spezia et al., 2017). More recently, it was demonstrated that the bound carotenoid experiences different polar environments in the active (OCP^R) and inactive (OCP^O) forms, which influences the excited state electronic structure of the pigment upon photoexcitation, and ultimately the color profile of the protein (Bondanza et al., 2020).

Recent studies demonstrated the potential utility of OCPs as tunable photoswitches. The two structural domains of OCP form a heterodimer in the presence of carotenoid, and following illumination, the heterodimer dissociates and adopts the lit form (OCP^R) (Andreoni et al., 2017). OCP can be integrated into an artificial light-harvesting system that responds to light of a particular wavelength (Andreoni et al., 2017). These findings suggest that OCPs can be incorporated into a photoresponsive system that responds to a specific wavelength of light (Dominguez-Martin and Kerfeld, 2019; Kirilovsky and Kerfeld, 2013).

In the present work, in the search for rational design principles to control the light response properties of OCPs, we investigated the influence of point mutations in the ligand-binding pocket of NTD on the light absorption properties of the lit form (OCP^R) of OCP1 originating from the unicellular cyanobacterium Gloeocapsa sp. PCC 7513 (GlOCP1) (Leverenz et al., 2015; Wilson et al., 2010; 2011; 2012). Initially, we found that the variants of GlOCP1 with single mutations at Val-40 exhibited a shift in the absorption maximum in the lit form relative to that of wild-type (WT) GlOCP1. Intrigued by this observation, we explored the origin of the shift in absorption maximum by combining experimental and computational methods. Firstly, we obtained an X-ray crystallography structure of GlOCP1. We then prepared homology models of the site-directed variants at Val-40, based on the obtained high-resolution structure. Conformational ensembles of WT GlOCP1 and its point mutants were obtained through molecular dynamics (MD) simulations. Hybrid quantum mechanics/molecular mechanics (QM/MM) simulations were conducted to refine the sampled conformations of the bound Cax. Our integrated experimental and computational studies revealed the detailed influence of point mutations at Val-40 on the light absorption properties of bound carotenoid, and suggested a novel design principle for utilizing OCPs as tunable photoswitches.

MATERIALS AND METHODS

In silico analysis of GlOCPs

Amino acid sequences of full-length OCPs were acquired from GenBank. A maximum likelihood (ML) phylogenetic tree of OCPs was constructed as described previously (Bao et al., 2017a) with some modifications. Sequences were aligned in Jalview (Waterhouse et al., 2009) and trimmed with the TrimAL suite of Phylemon 2 (Sánchez et al., 2011). The ML tree was constructed via the online version of PhyML (Guindon et al., 2010) with shape and color edited using iTOL (Fig. 2) (Letunic and Bork, 2019). To analyze variation in the amino acid sequences of OCPs from Microcoleus sp. B353 and Gloeocapsa sp. PCC 7513, full-length amino acid sequences were aligned using MUSCLE (Kumar et al., 2018) together with canonical Synechocystis sp. PCC 6803 OCP1 (SyOCP1), Tolypothrix sp. PCC 7601 OCP1 and OCP2, and Cyanothece sp. PCC 7424 OCP2. Amino acid sequence identity was obtained using PFAAT (Caffrey et al., 2007). Using the OCP alignment, carotenoid protein configuration (CPC) residues (Bao et al., 2017b; Leverenz et al., 2015) were identified, and the secondary structure was mapped using the crystal structure of canonical SyOCP1 (PDB ID: 4XB5), as shown in Fig. 3.

Fig. 2. Maximum likelihood phylogenetic tree of orange carotenoid proteins (OCPs).

OCPs from Synechocystis sp. PCC 6803 (Slr1963), Gloeocapsa sp. PCC 7513 (Gpl0480), and Microcoleus sp. B353 (Mbr0566) are grouped in the OCP1 clade, while those of Tolypothrix sp. PCC 7601 (Fdi7374) and Gloeocapsa Gpr1727 are in the OCP2 clade.

Fig. 3. Multiple sequence alignment of OCP1s.

. Orange carotenoid proteins (OCPs) from Synechocystis sp. PCC 6803 (SyOCP1), Tolypothrix sp. PCC 7601 (ToOCP1), Gloeocapsa sp. PCC 7513 (GlOCP1), Calothrix elsteri CCALA 953 (CaOCP1, WP_095720570.1), and Microcoleus sp. B353 (MiOCP1) were aligned using MUSCLE (MEGAX). Val-40 is a specific (non-conserved) residue in GlOCP1 and CaOCP1. Secondary structures are based on SyOCP1 (PDB ID: 4XB5). Circles and squares show residues involved in orange and red form configurations, respectively. Red, variations in CPC between GlOCP1 and MiOCP1 proteins. Blue, residues involved in OCP dimerization. Green, additional residues involved in photoactivity of OCP proteins.

Cloning and expression of cyanobacterial β-carotene ketolase and OCP genes

Genes encoding OCP of Synechocystis sp. PCC 6803 (SyOCP1, Slr1963, GenBank BAA18188), Gloeocapsa sp. PCC 7513 (GlOCP1, Glp0480, GenBank MT268542), Microcoleus sp. B353 (MiOCP1, Mbr0566, GenBank MT268544), and OCP variants used in this study were amplified by PCR using specific primers listed in Supplementary Table S1 and cloned into trcp+pAC-BETA, pBAD HisB, and MycHisC (Invitrogen, USA), respectively. Escherichia coli LMG194 (Invitrogen) was transformed with the trcp+pAC-BETA+ScyCrtW plasmid, and a Cax high accumulation colony was used to prepare competent cells that were subsequently transformed with the pBAD-OCP plasmids. The content of ketocarotenoids in trcp+pAC-BETA+ScyCrtW was analyzed by high-performance liquid chromatography (Yang et al., 2020). The carotenoid content of holo-OCP was calculated (Bao et al., 2017a; de Carbon et al., 2015; Moldenhauer et al., 2017). For expression of OCP holoproteins, E. coli LMG194 cells carrying trcp+pAC-BETA+ScyCrtW and pBAD+OCP plasmids were grown in rich medium (RM) media in the presence of chloramphenicol (35 µg/ml) and ampicillin (200 µg/ml) antibiotics at 37°C for 5 h to accumulate ketocarotenoids. Cultures were diluted 2-fold with fresh RM media containing antibiotics (35 µg/ml of chloramphenicol and 200 µg/ml of ampicillin) and 0.02% arabinose, and grown for 20 h at 25°C to induce expression of OCP holoproteins. All induction processes were performed in the dark.

Purification of OCP holoproteins

E. coli pellets were dissolved in lysis buffer (50 mM Tris-HCl pH 8, 200 mM NaCl, 10% glycerol, 0.1% octylphenoxypolyethoxyethanol [IGEPAL], 2 mM dithiothreitol [DTT], 1 × ethylenediaminetetraacetic acid [EDTA]-free Protease Inhibitor Cocktail, 20 mM imidazole), and sonicated for 1 h or disrupted using an LM20 microfluidizer (Microfluidics, USA) at 20,000 psi (Yang et al., 2020). The supernatant was incubated with Ni-NTA Agarose (Qiagen, Korea) matrix for 2 h at 4°C. Nickel resin was transferred to a chromatography column (Bio-Rad, USA) and washed (50 mM Tris-HCl pH 8, 300 mM NaCl, 0.1% IGEPAL, 10% glycerol, 20 mM imidazole). OCP proteins were eluted (50 mM Tris-HCl pH 8, 300 mM NaCl, 250 mM imidazole) and purified by gel filtration chromatography using a Hiprep TM 26/60 Sephacryl TM S-200 HR column (GE Healthcare, USA) equilibrated with 50 mM Tris-HCl pH 8 and 300 mM NaCl. Finally, purified proteins were washed with 50 mM Tris-HCl (pH 8) and stored on ice or at –70°C until needed. Protein parameters were determined using Compute pI/MW (https://web.expasy.org/compute_pi/). To assess the tertiary structures of OCP1 and OCP2, purified proteins were separated by native polyacrylamide gel electrophoresis (PAGE) as described previously (Bao et al., 2017a). All procedures were performed in the dark.

GlOCP1 crystallization and structure determination

Crystals of WT GlOCP1 bound to Cax were grown in mother liquor containing 0.2 M sodium formate and 22% PEG3350 at 18°C via the hanging-drop vapor diffusion method. Crystals were frozen in mother liquor supplemented with 20% glycerol as a cryoprotectant. X-ray diffraction data were collected at the 5C beamline of Pohang Accelerator Laboratory (PAL, Korea). Data were processed, indexed, integrated, and scaled using HKL2000 (Table 1) (Otwinowski and Minor, 1997). The structure was determined by molecular replacement using the structure of Fremyella diplosiphon OCP1 (PDB ID: 6PQ1) and beta,beta-carotene-4,4′-dione (PDB ID: 45D) as a search model. Refinement and model building were performed using Phenix (Adams et al., 2010; Park et al., 2021) and COOT (Emsley and Cowtan, 2004; Hyun et el., 2021). Structure factors and atomic coordinates have been deposited in the PDB database (PDB ID: 7EKR).

Table 1.

Crystallographic data collection and refinement statistics

GlOCP1	Value
Data collection
Wavelength (Å)	1.00
Space group	P41212
Unit cell dimensions
a, b, c (Å)	82.94, 82.94, 197.64
α, β, γ (°)	90, 90, 90
Resolution (Å)	49.41-2.17 (2.25-2.17)a
Rsym	0.25 (0.75)a
I/σ (I)	38.91 (9.82)a
Completeness (%)	99.30 (93.20)a
Redundancy	16.30 (13.50)a
Refinement
Resolution (Å)	49.41-2.17
No. of reflections	37,051
Rwork/Rfree	0.18/0.23
No. of atoms
Protein	4,779
Ligand	84
Water	365
RMSD deviation
Bond lengths (Å)	0.90
Bond angles (°)	0.007
Average B factor
Protein	22.32
Ligand	18.17
Water	26.43
Ramachandran plot (%)
Favored/allowed/outliers	98.85/1.15/0

OCP1 absorbance spectra and dark reversion kinetics

UV-visible spectra were measured using a UV1900 spectrophotometer (Shimadzu, Japan) equipped with a temperature-controlled cuvette holder (TCC-100-230VCE) in dark conditions. OCP1^O to OCP1^R conversion was achieved by exposing purified proteins to blue light (460 nm light emitting diode [LED], ~500 µmol photons m^-1 s^-1) for varying periods. In the dark state (OCP1^O), the vibrational band resolution of S₀→S₂ transitions was observed, denoted as Peak II and Peak III in Table 2 (Christensen, 1999).

Table 2.

Spectroscopic values of orange carotenoid proteins (OCPs)

	Maximum (nm)
	OCPO			OCPR
	Peak II	Peak III
Wild type
GlOCP1	477	502		520
SyOCP1	476	501		526
MiOCP1	476	502		526
GlOCP1 variants
V40I	477	501		529
V40L	476	502		530
V40T	480	502		516
V40G	495	NA		514
V40W	482	NA		535
V40Y	478	502		532
V40D	490	NA		522
V40H	477	502		522
V40R	495	NA		507

OCPs from Gloeoecapsa sp. PCC 7513 (GlOCP1), Synechocystis sp. PCC 6803 (SyOCP1), and Microcoleus sp. B353 (MiOCP1).

NA, not available.

Homology modeling and MD simulations of GlOCP1^R

For molecular simulations of GlOCP1^R, we constructed a homology model of GlOCP1^R. We employed the NTD of the crystal structure determined in this study as the template for homology modeling. Using PyMol v2.3.3 (Schrödinger, 2016), we mutated the CPC residue (Val-40) to five other amino acids (V40W, V40I, V40L, V40T, and V40R). Notably, the crystal structure obtained represents the dark form, as judged by the location of the bound Cax. Previous structural analyses of SyOCP1 demonstrated that the structure of NTD of OCP did not change significantly between dark and lit forms; the root-mean-squared deviation (RMSD) value of the NTDs of the two forms was 1.6 Å (PDB IDs: 4XB4 and 4XB5). Therefore, we assumed that the NTD of GlOCP1 in its lit form will be comparable with that of the dark form. The configuration of Cax was deduced from that of SyOCP1^R (PDB ID: 4XB4) and manually inserted into the binding pocket of the constructed model of GlOCP1^R. The structural models were protonated at pH 7.0 and solvated with ~20,000 TIP3P water molecules (Jorgensen et al., 1983). The AmberFF14SB force field (Maier et al., 2015) was used to describe the physical properties of GlOCP1^R-NTD. We obtained the covalent bond parameters of the ligand using General Amber Force Field 2 (GAFF2) (Wang et al., 2006). Restrained electrostatic potential (RESP) (Bayly et al., 1993) was used for assigning atomic partial charges of the ligand. Initially, the geometry of the ligand was optimized at the B3LYP/6-31G(d) theory level (Becke, 1993). Using the Using the lowest energy structure, a single point calculation was performed to obtain the electrostatic field around the molecule at the HF/6-31(d) theory level (Supplementary Table S2). We utilized AmberTools16 and Gaussian09 software suites for parameterization of the protein and the ligand (Case et al., 2005; Frisch et al., 2009).

Each solvated system was (i) energy-minimized in 4,000 steps with restraints imposed on Cα atoms, (ii) heated from 0 to 100 K over 2 ns of dynamics, (iii) equilibrated in the constant pressure-temperature ensemble (NPT) with temperature gradual increasing from 100 to 300 K over 2 ns, (iv) equilibrated again for 2 ns, and (v) propagated in the NPT ensemble until the model converged. Temperature was regulated by employing a Langevin thermostat with a collision frequency of 5.0 ps^-1 using an integration time step of 2 fs. Long-range electrostatic interactions were modeled using the Particle Mesh Ewald (PME) method and Lennard-Jones interactions were cut-off at 8 Å. We used AMBER16 molecular simulation software for energy minimization and subsequent MD simulation steps (Case et al., 2005). Calculations were conducted on an in-house cluster equipped with graphical processing units (GPUs).

Structural clustering analysis of MD trajectories

We monitored the convergence of MD simulations using structural cluster analysis. MD trajectories were saved every 10 ps from 1 to 3 µs for detailed analysis. The root-mean-square-fluctuation (RMSF) of each residue was calculated using the bio3D package implemented in the R programming language (Grant et al., 2006; R Core Team, 2015). Residues that fluctuated less than 2 Å relative to the X-ray crystal structure were identified as the invariant region of the protein. Next, we aligned the sampled coordinates by comparing the RMSD of the invariant region. We used the k-mean algorithm implemented in VMD software (Humphrey et al., 1996) for structural cluster analysis.Surface electrostatic potentials of representative structures were computed using Adaptive Poisson-Boltzmann Solver (APBS) implemented in PyMol v2.3.3 (Schrödinger, 2016).

Hybrid QM/MM and time-dependent density functional theory (TDDFT) calculations

Hybrid QM/MM (Warshel and Levitt, 1976) computations were conducted to quantify the absorption spectra of the simulated GlOCP1^R proteins. Initially, the sampled conformations from MD simulations were optimized at the CAM-B3LYP/6-31G(d,p)/AmberFF14SB/GAFF theory level. Optimizations were necessary as the bond lengths and angles of the ligand were not optimal for calculating the absorbance wavelength (Bondanza et al., 2020). The QM region spans the entire ligand and amino acid 40, while all atoms within 8 Å of the ligand were treated as point charges polarizing the QM region based on the electrostatic embedding (EE) method (Senn and Thiel, 2007). After optimizing the geometries up to 2,000 steps, singlet excited state electronic structures with transition dipole moments were computed using TDDFT at the TD-CAM-B3LYP/6-311G(d,p)/AmberFF14SB/GAFF theory level (Binkley et al., 1980; Grimme et al., 2010; Yanai et al., 2004). In TDDFT calculations, the electronic energy difference (ΔE[S₁-S₀]) between the electronic ground state (S₀) and the excited singlet state (S₁) as well as the oscillator strength of the excited state was computed. Notably, ΔE(S₁-S₀) could be compared across different CPC mutants since the quantum mechanical Hamiltonian of each mutant was different. The QChem v5.0 interface (Q-Chem, USA) integrated with AMBER16 was used for QM/MM calculations (Shao et al., 2015). Natural transition orbital analyses were then carried out to characterize the electronic nature of computed excited state electronic structures (Martin, 2003).

Statistical analysis of the computed absorption energy profile

For WT and mutant proteins, absorption energy spectra were computed for 100 snapshots sampled from the MD trajectories. We computed the histogram of absorption energy weighted by the oscillator strength for multiple TDDFT results using the following formula:

$$f(\nu ,\nu +\Delta \nu )={\Sigma }_i{f}_i({\nu }_i)\left\{{\nu }_i|\nu \le {\nu }_i<\nu +\Delta \nu \right\}$$

where v is the wave number, f(v,v + Δv) is the oscillator strength in the interval from v to v + Δv (Δv is 2 nm), and I is the index of configuration that runs from 1 to 100. As the oscillator strength is proportional to the relative absorption strength in the spectrum measurements, the weighted average was assumed to account for the most significant contributions of the competent molecular structures of Cax to the computed absorption spectrum. Notably, the computed first excited singlet state (S₁) displayed the maximum oscillation strength, contributing the most to the absorption energy peak in the histogram. Previous photophysical studies of Cax showed that the S₁ state is a dark state that does not absorb light, whereas the second excited singlet state (S₂) is light-absorbing (Fiedor et al., 2016). However, previous benchmark studies of QM methods suggested that the S₁ state computed using the TDDFT method corresponds to the light-absorbing state (S₂) observed experimentally (Spezia et al., 2017). Accordingly, we analyzed the energetics and the electronic structures of the first computed singlet excited states of Cax in detail to rationalize the influence of CPC mutation to the photophysical properties of OCP^R. Standard errors of the computed absorption maxima were estimated based on the bootstrap method (Efron and Tibshirani, 1986). In brief, we selected 100 computed excited state energies with oscillator strengths and evaluated the absorption maximum from this sample. The random resampling process was repeated over 10,000 times, thereby providing the standard error of the computed maximum.

RESULTS

The phylogenetic tree of OCP proteins suggested that natural variations in CPC residues located within 4 Å of the bound carotenoid are rare events. The ML tree of OCPs showed that GlOCP1 belongs to clade 1 (Fig. 2) alongside the canonical Synechocystis OCP (SyOCP1), Tolypothrix OCP (ToOCP1), and Calothrix elsteri OCP (CaOCP1). As they are in the Synechocystis genome, FRPs are tandemly located with OCPs in Gloeocapsa genomes ( Fig. S1). Despite amino acid sequence variations in the range of 20%-25% among OCP1 paralogs, CPC residues that are crucial for photoactivity (SyOCP1 numbering, N104, E174, R185, N204, D209, and W277) as well as dimerization are strongly conserved (Fig. 3) (Leverenz et al., 2015; Muzzopappa et al., 2019; Wilson et al., 2011). Minor variations were observed including I40 in GlOCP1. Among 200 OCP1s, Ile-to-Val substitution at residue 40 was observed only in OCP1 of C. elsteri CCALA 953 (WP_095720570.1) and Gloeocapsa sp. PCC 7513.

We investigated the role of the conserved residue 40 in the absorption maxima of OCPs. Although CPC residues are known to be in contact with the bound carotenoid, the consequences of CPC mutations to the spectroscopic properties of OCPs have not been reported. Specifically, we envisioned that Val-40, strongly conserved among different species, may have a critical role in determining the specific functions of OCPs. Site-directed mutagenesis studies showed that mutations of the conserved Val-40 caused red and blue shifts in absorption maxima. First, we substituted Val-40 of GlOCP1 with Ile, yielding V40I GlOCP1. As shown in Table 2, in the lit form (OCP^R), the absorption peak of V40I GlOCP1 was red-shifted from 520 to 529 nm, which is comparable with those of SyOCP1 and MiOCP1 (Table 2, Supplementary Fig. S2). On the other hand, in the dark form (OCP^O), mutations of Val40 did not affect the dual absorption peaks at 477 nm and 502 nm.

We further investigated the influence of Val-40 by substituting the residue with conserved (Leu and Thr), hydrophobic (Trp and Gly), hydrophilic (Tyr), acidic (Asp), and basic (His and Arg) amino acids (Table 2, Supplementary Fig. S2). Among the substituted residues, red-shifted spectra in OCP^R were observed in V40I, V40L, V40W, V40Y, V40D, and V40H variants of GlOCP1. By contrast, blue shifts in GlOCP1^R spectra were observed for V40T, V40G, and V40R variants. In summary, site-directed mutagenesis studies suggested that mutation of Val-40 with an amino acid larger than Val, including Ile, Leu, Trp, and Tyr, resulted in a red-shift in the absorption spectra of OCP^R, except for Arg, which carries a positive charge at neutral pH. Conversely, mutation with a smaller sidechain, including Gly and Thr, caused a blue-shift in the absorption spectra.

To investigate the contribution of Val-40 to the absorption maximum, we determined the crystal structure of full-length GlOCP1 in complex with Cax using molecular replacement with F. diplosiphon OCP1 (PDB ID: 6PQ1) as a search model. The structure was fully refined to 2.35 Å resolution with a crystallographic Rwork value of 18% and an Rfree value of 23% (Table 1). The obtained crystals contained two copies of full-length GlOCP1-Cax per asymmetric unit. The structure of GlOCP1 is similar to other OCP1 structures (Leverenz et al., 2015). The structure of GlOCP1 consists of an NTD with 10 α-helices and a CTD with five α-helices and six β-sheets. NTD and CTD are connected via a flexible linker. The NTE functions to stabilize the domains by interacting with the surface of the β-sheets of the CTD (Fig. 4A). OCPs are highly conserved among different species, with the exception of GlOCP1, which has a single isoleucine to valine substitution at position 40. In order to gain insight into the environment surrounding Cax in GlOCP1, we conducted a detailed structural analysis. The Cax molecule in GlOCP1 is positioned in an NTD to CTD direction. The first cyclohexene ring of Cax is oriented towards the NTD and is surrounded by bulky amino acids such as L37, Val40, Y44, W110, and V158, while the second cyclohexene ring of Cax is surrounded by L207, Y203, V275, and W290 in the CTD (Fig. 4B). In particular, valine at position 40 in the NTD is located 4 Å away from the oxygen atom of Cax, and leucine at position 37 is located 3.3 Å away. The ketone group of the second cyclohexene ring of Cax is located 2.8 and 2.5 Å away from the aromatic amino acids Y203 and W290 in GlOCP1, respectively (Fig. 4). This suggests that the bulky and aromatic amino acids in GlOCP1 play an important role in binding to Cax.

Fig. 4. OCP1 from Gloeocapsa sp. PCC 7513 (GlOCP1) with bound canthaxanthin (Cax).

(A) GlOCP1 is composed of N-terminal domain (NTD) and C-terminal domain (CTD) regions connected via a linker. Cax is bound within GlOCP1 bridging NTD and CTD. (B) Cax mainly interacts with hydrophobic residues. Val-40 is located above the first cyclohexene ring of Cax. NTE, N-terminal extension.

We prepared homology models of GlOCP1^R and conducted MD simulations to investigate in detail the influence of mutating residue 40 of GlOCP1^R on the photoexcited state properties of protein-ligand complexes. We constructed atomistic models of NTDs of the WT protein and V40I, V40T, V40R, V40L, and V40W variants based on the crystallographic structure of NTD of GlOCP1^O as described in the Materials and Methods section. All five mutants exhibited significant shifts in the absorption maximum (Table 2). For each atomistic model of NTD of GlOCP1, 3 µs MD simulations were carried out at 300 K and 1 atm. Upon completion of the MD simulations, the backbone RMSD with the crystal structure of GlOCP1^R was measured (Supplementary Fig. S3). Deviations in RMSD were <2.5 Å throughout the simulations, suggesting the mutations do not alter the overall arrangement of NTD. We also measured the RMSD of amino acids from 1 to 3 µs (Supplementary Fig. S4). The RMSD of residues in the ligand-binding pocket were <2 Å, confirming that the pocket structures remain intact following mutation. Our MD simulations of WT and CPC mutants implied that the conformations of the NTDs and the ligand-binding pockets were consistent across all simulations.

We performed QM/MM and TDDFT calculations to quantify the absorption maximum of WT GlOCP and its V40I, V40T, V40R, V40L, and V40W variants. As shown in Table 3, the experimental absorption maxima of the five variants spanned 40 nm, which led us to compare and analyze the shifts in the absorption spectrum based on QM calculations. For each simulation, a representative ensemble of structures was determined using structural clustering analysis. The geometries of bound Cax were optimized at the CAM-B3LYP-D3/6-31G(d,p)/AmberFF14SB/GAFF theory level. QM/MM optimization of the bound ligand was necessary since the bond lengths of the ligand from MD simulations did not reproduce the results of higher-level QM calculations (Bondanza et al., 2020). Both the bound ligand (Cax) and the mutated residues were optimized at a higher theory level while restraining the atomic coordinates of the rest of the protein, which were subjected to the MM theory level. Upon completion of QM/MM optimization, TDDFT computation was performed. The calculated absorption energies, namely, the energy difference between the excited singlet state (S₁) and the electronic ground state (S₀), were averaged over 100 sampled snapshots.

Table 3.

Absorption maxima of wild-type GlOCP1 and its variants

GlOCP1 variants	λmax (Calc.)a	ΔEavg (Calc.)b	λmax (Exp.)
Wild type	450 ± 11	2.72	520
V40W	468 ± 7.4	2.64	535
V40I	454 ± 4.3	2.72	529
V40L	458 ± 8.4	2.72	530
V40T	452 ± 12.8	2.82	516
V40R	434 ± 26.2	2.75	507

GlOCP1, OCP1 from Gloeocapsa sp. PCC 7513; Calc., computed; Exp., experimentally obtained.

^aAverage absorption maxima and standard deviation are shown. Values for absorption maxima are in nanometers (nm).

^bThe average energy difference between the ground state and the 1st excitation state is shown. Values for energy are in electron volts (eV).

Analysis of the computed excited state electronic structures showed that the bound Cax exhibited the characteristics of a charge-transfer (CT) state. We conducted natural transition orbital (NTO) analysis of Cax to probe the vacant (hole) and occupied (electron) portions of the molecule in its photoexcited state (Martin, 2003). Figure 5 shows the results of NTO analysis of the computed first photoexcited state (S₁) of Cax. When bound to GlOCP1^R, Cax exhibited the characteristics of a CT state in which a photoexcited electron migrates toward the end of the Cax molecule, resulting in a spatially separate electron and hole charges. When comparing the frontier molecular orbitals of the ground state (Supplementary Figs. S5 and S6), we discovered that the one-electron excitation from the HOMO (highest occupied molecular orbital) to the LUMO (lowest unoccupied molecular orbital) constitutes the CT state. We also found a spatial overlap between the electron and the hole charges, suggesting a partial CT character in the photoexcited state (Kim et al., 2020). By contrast, in a vacuum, the excited state of Cax bears the characteristics of π-π* transition and does not display any CT character. The computed absorption peak of the isolated Cax was 430 nm, blue-shifted by ~30 nm relative to those computed for Cax bound to GlOCP1^R, suggesting that the CT character may explain the observed shifts in absorption maxima. Subsequently, we demonstrated that the CT character in photoexcited Cax explains the observed shifts in absorption maxima (vide infra).

Fig. 5. Illustration of charge separation in the photoexcited state of canthaxanthin (Cax) bound to GlOCP1R (V40W).

The NTOs (natural transition orbitals) of photoexcited state (S₁) species are shown. NTD, N-terminal domain; OCP, orange carotenoid protein.

Figure 6 compares the computed and experimented absorption energy profiles of WT GlOCP1 and its five variants, and Table 3 summarizes the computed and experimental absorption maxima of the five variants. The computed absorption maxima of WT, V40W, V40I, V40L, V40T, and V40R were 450 nm, 468 nm, 454 nm, 458 nm, 452 nm, and 434 nm, respectively. The values of computed absorption maxima spanned 34 nm (434 to 468 nm), in qualitative agreement with the experimental observations that spanned 28 nm (507 to 535 nm). The tryptophan mutation (V40W) showed the most significant red-shift and the arginine mutation (V40R) exhibited the most significant blue-shift in both experimental and computational analyses. The computed absorption maxima of the other three variants (V40I, V40L, and V40T) were comparable with that of the WT protein. Notably, the computed maxima were slightly blue-shifted from the experimental values, consistent with previous reports showing that computed absorption spectra of biological chromophores based on TDDFT-QM/MM approaches suffer from overestimation of the absorption energy (Rocha-Rinza et al., 2009). The degree of charge separation, measured as the distance between the hole and electron in NTO analysis, correlates with the red-shifts in the absorption maxima. For instance, the average hole-electron distance of V40W was 3.30 Å, significantly longer than that of V40R (2.97 Å). This finding is in accord with the quantum mechanical nature of electrons in excited states since charge separation effectively lowers Coulombic repulsion between opposite spin electrons in a singlet excited state (Griffiths and Schroeter, 2018). We also found that the degree of charge separation in the photoexcited state was positively correlated with the red-shift of the absorption energy peak (Fig. 7). This finding is in line with previous experimental observations showing that the electrostatic field along the direction of the major axis of polyenes can increase the oscillator strength in the UV-Vis absorption region (Zheng et al., 2020). Indeed, in the presence of external electric fields applied in the direction of the major axis of Cax, our computational analysis of an isolated Cax molecule demonstrated a positive correlation between the red-shift in the absorption maxima and the degree of charge separation (Supplementary Fig. S7). By contrast, in a vacuum, the computed hole-electron distance of Cax was 0.4 Å, 10-fold smaller than the average charge separation of Cax bound to the GlOCP1 variants. Our computational analyses of the absorption maximum and the degree of charge separation in photoexcited Cax suggests that the ligand-binding pocket induces charge separation in the photoexcited state of Cax, resulting in shifts in absorption maxima.

Fig. 6. Comparison of computed (A) and experimental (B) absorption energy spectra of OCP1 from Gloeocapsa sp. PCC 7513 (GlOCP1) wild-type (WT) and variants (V40I, V40T, V40R, V40L, and V40W).

The absorption maxima are normalized.

Fig. 7. Correlation between the computed absorption wavelength and the hole-electron distances.

The emergence of a CT state for photoexcited Cax is accredited to the polar electrostatic nature of the ligand-binding pocket. Figure 8 shows the surface electronic potential of the Cax-binding pocket of WT GlOCP1^R, revealing a polar electrostatic environment. Here, the two charged residues situated at the ends of the ligand-binding pocket, namely, Glu-34 and Arg-155, create an electronic dipole moment that polarizes the bound Cax. These observations concomitantly support the idea that the polarity of the ligand-binding pocket induces the CT character of the ligand in its photoexcited state.

Fig. 8. Map of surface electrostatic potential of the N-terminal domain of Gloeocapsa sp. PCC 7513.

The unit of electric field is kT/e at 300 K.

Comparison of the surface electrostatic potential of the ligand-binding pocket suggest that the electrostatic polarity of the binding pocket and the relative position of Cax may determine the absorption maximum of GlOCP1^R. Figure 9 depicts the electrostatic environments of the Cax-binding pockets of WT GlOCP1^R and the simulated variants. For V40R, a positively polarized region of the binding pocket extends toward the center of Cax, which decreases the polarity of the binding pocket. This mutation places a cationic residue in the middle of the binding pocket, which effectively lowers the net polarity of the binding pocket (Fig. 9D). This finding is in line with the observed blue-shift in the absorption maximum of V40R (Table 2). For the four non-polar and polar-neutral mutants (V40W, V40L, V40T, and V40I), the surface electrostatic potential appears to be comparable with that of WT GlOCP1^R. Interestingly, for V40W, which showed the most significant red-shift in the absorption maximum, the distance between the center of the ligand and the Cα of CPC residues is markedly different from that in the WT protein; the distances are 5.5 Å and 5.1 Å for WT and V40W, respectively, when averaged over the last 2 µs of MD trajectories (Supplementary Fig. S8). These findings suggest that as the fraction of Cax under the influence of the polar environment increases, the CT character increases, and the absorption energy of the photoexcited state decreases. Our analysis of the electrostatic field of the binding pocket suggests that the CPC mutations modulate the microscopic environment that surrounds the bound Cax, which controls the degree of CT and the absorption maximum of the ligand.

Fig. 9. Electrostatic potential in the ligand-binding pocket of canthaxanthin (Cax).

(A) Wild-type (WT). (B) V40W. (C) V40L. (D) V40R. (E) V40T. (F) V40I. C12 of each Cax molecule is highlighted with a triangle (▲). The unit of electrostatic potential is kT/e at 300 K.

DISCUSSION

Non-conserved CPC residues are likely to perform specific functions in the photodynamic properties of OCP1 proteins (Supplementary Fig. S3). Among natural variations in CPC residues examined in the present study, Val-40 in helix α3 of GlOCP1, equivalent to Ile-40 of SyOCP1 (Supplementary Fig. S4), is a critical residue determining the light absorption properties of OCP1R. Previously, CPC residues responsible for carotenoid attachment have been characterized, such as Trp-288, Tyr-201, Tyr-44, Trp-110, and Arg-155 (Slonimskiy et al., 2019; Wilson et al., 2011). Herein, we identified a CPC residue that is involved in determining the light absorption peak in the active lit form, without significantly affecting ligand chromophorylation.

Our integrated experimental and computational analysis suggested that mutation of CPC residues alters the electrostatic environment surrounding Cax, thereby modulating the absorption maxima of GlOCP1^R. Based on the map of the protein surface electrostatic field (Figs. 8 and 9), we showed that the ligand-binding pocket has an electrostatic dipole moment, exerting an external electric field onto Cax in the ligand-binding pocket. The dipole moment mainly originates from the asymmetric placement of charged side chains of Glu34 and Arg155. Mutation of the CPC residue resulted in two different outcomes: first, mutation to large non-polar residue (V40W) shifts the center of Cax closer to the CPC residue, which increases the electrostatic polarity surrounding the ligand; second, the V40R mutation reduces the net polarity of the binding pocket, reducing the electrostatic polarity near the ligand. Notably, mutation of the CPC residue did not significantly alter the geometries of Cax relative to the WT protein.

A relatively large red-shift of 15 nm in the absorption peak was observed for the V40W variant, in which the indole ring of tryptophan and the conjugated plane of Cax are aligned in a near-parallel configuration. NTO analysis of the first light-absorbing excited state precluded a significant electron transfer from Cax to the side chain of tryptophan (Fig. 5). However, we found that Trp40 and the methyl groups of Cax are within 3 to 4 Å, close enough for weak CH-π interactions. Such interactions likely pull the ligand toward the center of the ligand-binding pocket, increasing the distance between the hole and the electron. As a consequence, the mutation causes a significant 15 nm red-shift in the absorption maximum.

In summary, we identified mutations in the Cax-binding pocket that modulate the light absorption maximum of OCP1 in its light-activated form (OCP1^R). Our combined analysis demonstrated that changes in the electric field surrounding the chromophore affect the light absorption properties of the bound Cax in the NTD of OCP1. These observations suggest novel rational design principles for the Cax-binding pocket of the lit form (OCP1^R), allowing fine-tuning of the light-harvesting properties as well as the color profile of cyanobacteria. Given the possibility of utilizing OCPs and associated photoprotection mechanisms to engineer renewable feedstock processing units and optogenetic sensors, our findings are relevant to rational engineering of OCPs.

Supplemental Materials

Note: Supplementary information is available on the Molecules and Cells website (www.molcells.org).