Maintaining the marginally stable three-dimensional architecture of proteins often requires a complicated network of weak interactions, including hydrogen bonds. The study of individual hydrogen bonds in the context of such complex protein structures is often a formidable task. A kind of “molecular surgery” (Thi To et al., 2022) with “atomic mutations” (Minks et al., 1999) is required to remove or replace individual atoms or groups of atoms in the context of complex proteins containing thousands of atoms with a complicated network of, for example, covalent, non-covalent or hydrogen bonding and van der Waals interactions. It has been suggested that the function of the Orange Carotenoid Protein (OCP), a blue light-driven photoswitch involved in photoprotection of cyanobacteria (Holt and Krogmann, 1981), strongly depends on two critical H-bonds between the ketocarotenoid cofactor and a tryptophan (Trp) and a tyrosine (Tyr) residue in the surrounding protein matrix (Figure 1). However, classical protein mutagenesis does not allow the precise parameterization of the contributions of these H-bonds to a protein’s functional integrity.

Structurally, the 35 kDa OCP consists of an N-terminal domain (NTD) composed entirely of α-helices and a mixed α/β-type C-terminal domain (CTD) of about equal size, both encapsulating a single ketocarotenoid molecule in a common central cavity (Kerfeld et al., 2003). Upon absorption of a photon, OCP converts from the dark-adapted, orange OCP^O state to the red, active OCP^R state. The embedded ketocarotenoid is at the heart of OCP photoswitching activity and function. In native cyanobacteria this cofactor is 3'-hydroxyechinenone (Muzzopappa and Kirilovsky, 2020); whereas the chemical identity in the equally active recombinant form is a 4-ketolated xanthophyll. In the compact OCP^O state, for which X-ray crystal structures are available (e.g., PDB codes 3MG1, 4XB5, 5UI2), the 4-keto oxygen atom of the xanthophyll is involved in an H-bonding network to the two highly conserved residues within the C-terminal domain (Tyr201 and Trp288 according to amino acid numbering of Synechocystis OCP). This local network fixes the cofactor in a rigid, slightly bent and twisted conformation (Figure 1) (Kerfeld et al., 2003; Leverenz et al., 2015). The H-bonds in this network break upon photoexcitation, releasing the terminal β-ring of the xanthophyll, leading to a number of detectable spectroscopic and conformational changes, e.g., xanthophyll isomerization and translocation, separation of domains, and a red-shift in absorbance (Leverenz et al., 2015).

The structure of the OCP^R state has not yet been elucidated because it is very elongated and flexible (Gupta et al., 2015; Sluchanko et al., 2017a). In OCP^R, the protein is able to interact closely with the water-soluble antenna complexes of cyanobacteria, the phycobilisomes (PBSs). By exposing the chromophore to the terminally emitting pigments of the PBSs, OCP^R acts as an efficient quencher of PBS fluorescence, protecting the underlying photosystems from the deleterious effects of overexcitation and dissipating the electronic excitation energy into heat.

Mutations of Tyr201 and Trp288 often destabilize the compact OCP^O form and generate permanently red-shifted OCP forms with separated domains that are - despite not being photoactive - constitutively capable of PBS fluorescence quenching (Maksimov et al., 2016). As an example, the OCP-W288A mutant protein upon heterologous expression yields a mixture of species with a dominant fraction of photo-inactive red-purple protein, which is capable of PBS fluorescence quenching. At the same time, the absorbance spectra of many dark-adapted OCP species show spectral heterogeneity with admixtures of red states in the OCP^O form (Yaroshevich et al., 2021). However, mutational analyses occasionally revealed OCP variants with markedly reduced contamination by red states in the compact, dark-adapted OCP^O form. Such properties were observed for the OCP-Y201W mutant, whose spectrally pure, “super-orange” OCP^O state prove advantageous for femtosecond pump-probe UV-Vis spectroscopy to investigate the photocycle (Yaroshevich et al., 2021). For OCP-Y201W, the analysis of femtosecond transient UV-Vis absorption experiments combined with quantum chemical calculations led to the formulation of a reaction mechanism for the photochemistry in the early events that occur upon light-triggered H-bond breaking and involve the formation of an oxocarbenium ion on the ketocarotenoid (Yaroshevich et al., 2021).

Site-directed mutagenesis studies to probe the role of Trp288 and Tyr201 in OCP suffer from the limited ability to substitute aromatic amino acids within the standard amino acid repertoire. As a result, inaccurate conclusions have been drawn about which amino acid residues or their chemical properties are necessary for particular functional aspects, leaving unanswered the question of which chemical functionalities are sufficient for function. In this work, we performed a kind of “molecular surgery” in which a canonical amino acid residue was site-specifically replaced with a non-canonical analogue to remove the critical H-bond donor function of Trp288 of OCP with atomic precision. To this end, we used a recently evolved orthogonal pair (Tseng et al., 2020) consisting of a custom-tailored Methanosarcina mazei pyrrolysine tRNA synthetase (MmPylRS) and its cognate tRNA to replace Trp288 in OCP with 3-benzothienyl-L-alanine (BTA), an isostructural analogue of tryptophan in which the imino nitrogen is substituted by a sulphur atom. The resulting OCP-W288_BTA protein is photoactive and, according to the X-ray crystal structure solved at 1.8 Å resolution, fully isomorphous to the wild-type protein, allowing us to parametrize the consequences of the absence of a single H-bond by spectroscopic, thermodynamic and computational methods.

We used an Escherichia coli expression host for the biosynthesis of the BTA-containing OCP holoprotein variant (OCP-W288_BTA). Our expression system is based on three plasmids (Figure 2). Affinity chromatography using His-tag yielded an intensely stained protein solution (blue spectrum in Figure 3A) showing characteristic features of dark-adapted OCP with a small amount of a red-shifted species, which shows up as a shoulder at 570 nm (Sluchanko et al., 2017a). After an additional purification step with preparative size-exclusion chromatography (SEC), an orange and a pink protein species could be separated based on their different elution profiles (Figure 3B) (Sluchanko et al., 2017b). While the pink species [OCP-W288_BTA(P)] eluted earlier, the orange fraction [OCP-W288_BTA(O)] corresponded to the retention time of an OCP monomer. Figure 3A shows the representative spectra of a protein preparation that contained approximately 35% pink and 65% orange species. The purified proteins of the orange OCP-W288_BTA(O) variant and wild-type OCP were subjected to electrospray LC-MS analysis (Figure 3C), which revealed 100% incorporation of BTA in the variant protein.

The carotenoid content of the protein samples was determined by HPLC (Table 1). While wild-type OCP almost exclusively contained ECN, the orange OCP-W288_BTA(O) species had two peaks of approximately the same amplitude (54%/46%) corresponding to ECN and CAN. Different amounts of ECN/CAN in OCP samples obtained from xanthophyll-producing E. coli strains is a common phenomenon (de Carbon et al., 2015) and may depend on the employed β-carotene ketolase gene or on the expression level of OCP in the cells. However, the photoswitching ability is not affected by the presence of ECN or CAN (de Carbon et al., 2015) (see below). In contrast, the pink OCP-W288_BTA(P) species contained more than 96% CAN.

When both the orange and pink OCP-W288_BTA protein fractions were illuminated with blue light (3-W LED, 455 nm) at 15°C, the OCP-W288_BTA(O) species was photoswitchable (Figure 4A) like wild-type OCP (Figure 4B), whereas no spectral changes were observed for the OCP-W288_BTA(P) species (Figure 4C). Spectral decomposition showed that the spectrum of the dark-adapted OCP-W288_BTA(O) species contained a much smaller contribution of a red non-photoswitchable species compared with the wild-type OCP (Figure 4A) than the dark-adapted wild-type OCP (5% vs. 12%, respectively) (Figure 4B). The absorption spectrum of the orange OCP-W288_BTA(O) species exhibits a distinct vibrational fine structure, with even more pronounced vibrational bands than the wild-type OCP^O (Figure 4A, inset). These features of the OCP-W288_BTA(O) protein, a spectrally more homogenous dark-adapted state and a deeper vibrational substructure are typical attributes of so-called “super-orange” OCP species, as first described for the OCP-Y201W mutant (Yaroshevich et al., 2021).

In the OCP-Y201W mutant protein, the lack of one H-bond donor [according to the X-ray crystal structure solved in (Yaroshevich et al., 2021), the introduced Trp residue is too far away from the C=O group of the ketocarotenoid] leads to an increase in the importance of the remaining H-bond (from the C=O group of the carotenoid to the N-H group of Trp288), making the ground state less susceptible to spontaneous carotenoid fluctuations. In the OCP-W288_BTA variant, the same effect could also occur in the opposite direction: Removal of the imino nitrogen by a sulphur atom also strengthens the importance of the remaining H-bond to Tyr201.

The spectra of the OCP-W288_BTA(P) species and the light-adapted OCP-W288_BTA(O) species are very similar (Supplementary Material S1, Supplementary Figure S2). This suggests that the spectroscopic properties of the OCP-W288_BTA(P) species are determined by features attributable to domain separation and complete translocation of the chromophore into the NTD, as present in the photoactivated form of OCP. The aforementioned spectra are also very similar to the absorbance spectrum of the Red Carotenoid Protein when it coordinates CAN (i.e., the OCP-NTD holoprotein produced by heterologous expression in E. coli) (Moldenhauer et al., 2017), but slightly blue-shifted and distinct from the absorbance spectrum of the pink fraction of the OCP-W288A mutant protein previously characterized (Maksimov et al., 2016) (Supplementary Material S1, Supplementary Figure S2).

We were able to crystallize the orange species of OCP-W288_BTA containing a mixture of ECN and CAN and determined its structure (PDB: 7ZXV) with a resolution of 1.8 Å (see Table 2 for crystallographic information) by molecular substitution with the structure of wild-type OCP as a template (PDB: 5TUW) containing an ECN cofactor. As expected, the overall structure is very similar to the wild-type OCP, with RMSD values of 0.180 Å and 0.203 Å for monomer and dimer alignments at the Cα-positions, respectively. A close inspection of Figure 5A clearly shows that the BTA mutation has no significant effect on the global OCP structure, and the different geometry of the BTA side chain (Figure 5B; C-S-C angle of 92.5°) compared to Trp288 (C-N-C angle of 109°) is nicely reflected by the electron density (Figure 5B and Supplementary Material S1, Supplementary Figure S3). The introduced BTA residue matches well with its native Trp288 counterpart, resulting in an almost identical spatial arrangement (i.e., isomorphic replacement). Interestingly, the distance to the keto oxygen atoms of the ECN/CAN cofactors increases only slightly from 2.9 Å in the wild-type protein to 3.2 Å and 3.3 Å, respectively (Figures 5C, D), despite the potential repulsion of the lone electron pairs of sulphur, which, however, do not appear as a significant electron density in the 2mFo-DFc map (Figure 5B, and Supplementary Material S1, Supplementary Figure S3). In addition, the distance between Tyr201 and the keto oxygen of ECN and is almost identical at 2.5 Å compared to 2.6 Å in wild-type OCP. However, the distance to CAN is slightly increased to 2.8 Å (Figures 5C, D). Here, the carbonyl carbon atom of CAN has an sp ² hybridization state, whereas the corresponding carbon atom of ECN is sp ³ -hybridized, which affects the overall ring geometry in the NTD (Figure 5E) and is probably responsible for the small shift at the distal end of the xanthophyll in the CTD. Nevertheless, the protein environment of the xanthophylls seems to be unaffected by the presence of an additional keto function in the NTD, and the CAN configuration is almost identical to the one of the CAN cofactor in another Synechocystis wild-type OCP structure (PDB: 4XB5; Supplementary Material S1, Supplementary Figure S4). The only observable difference is that Cδ of Ile40 seems to be rotated 120° compared to the wild-type (Figure 5F), yet this does not result in a subsequent conformational change. Overall, the differences due to the BTA mutation and the different cofactor notably result in very minor structural changes.

During refinement, the electron density of a Cl⁻ ion was delineated that was also present in another OCP mutant (Y201W, PDB: 6T6K) at the same position; which, however, is not present in the 3MG1 structure. The only potential OCP interaction partner in the proximity is Thr275. Interpretation of the xanthophyll cofactor electron density based on the wild-type structure with only one ECN molecule resulted in a difference density in the NTD (Figure 5G). Subsequently, the electron density was interpreted with an additional CAN molecule and occupancy refinement (Figure 5H), which led to the conclusion that both xanthophylls are present in the OCP-W288_BTA mutant: The ratio of ECN/CAN ratio (45:55) corresponds to the ratio of xanthophylls extracted from the purified protein (Table 1).

To gain insight into the character of the interaction in the ECN-Tyr201-BTA288 complex, we performed a 10 ps quantum dynamics (QD) calculation. A comparison of the distances between the key atoms in the ECN-Tyr201-Trp288/BTA complexes (Supplementary Material S1, Supplementary Figure S5) clearly shows the absence of the energetically favourable interaction between the sulphur atom in BTA and the keto oxygen in ECN. The distance of 3.3 Å between the keto oxygen and the sulphur atom indicates the dispersion nature of the interaction in the ECN-BTA complex. Despite the high polarizability of the bulky sulphur atom, such an interaction does not significantly affect the OCP-W288_BTA absorption spectrum in the dark-adapted state. Moreover, a slight trend toward shortening of the ECN-Tyr201 H-bond is observed in the presence of BTA. This effect is consistent with the hypothesis that the absence of a competing H-bond donor for the keto oxygen enhances the interaction with the remaining H-bond donor (Yaroshevich et al., 2021). In the case of OCP-W288_BTA, the remaining H-bond is formed with Tyr201, making it similar to the previously reported OCP-Y201W mutant (Yaroshevich et al., 2021).

With the introduction of BTA into OCP, several new effects could be expected from the appearance of a bivalent sulphur atom in close proximity to the keto group of the bound xanthophyll. For example, the electrostatic interaction between the sulphur atom of a methionine and the carboxylate oxygen atom was suggested to play a major role in enzymatic catalysis (Taylor and Markham, 1999). Further analysis of structural data revealed that the bivalent sulphur atom forms specific non-bonded (sometimes called hypervalent) interactions with non-hydrogen atoms (O, N, C) within the protein backbone (Iwaoka and Isozumi, 2012). These relatively weak interactions can be classified into several types: Two-center-three-electron (Mishra et al., 2009), bifurcated (Zhang and Wang, 2009) and sulphur-carbene (Zhao et al., 2011) interactions. In addition, a theoretical analysis has shown that electron deficient, bivalent sulphur atoms can interact with electron donors by accepting their lone pair into low-lying σ* orbitals (Beno et al., 2015). All these types of specific interactions are sensitive to the mutual orientations of the interacting counterparts and expected to be energetically significant at S···O distances less than 3 Å (Iwaoka and Isozumi, 2012; Beno et al., 2015). However, with respect to the interaction of low-lying σ* orbitals of the C–S in the BTA bond with electron donors, we note that, first, the electron deficiency at the sulphur atom in BTA should be low [see Figure 8 in (Beno et al., 2015)], and, second, the sulphur (acceptor), oxygen (donor), and carbon (covalently bound to the sulphur in BTA) should be collinear (180° C-S-O angle) for a favourable interaction to occur. Yet, the corresponding C-S-O angle from the 7ZXV crystal structure deviates substantially (160°), arguing against a significant contribution of the aforementioned non-bonded interactions.

Resonance Raman spectra of the photoswitchable OCP-W288_BTA protein were recorded for dark- and light-adapted samples and compared with those of wild-type OCP (Figure 6), and those in the literature (Leverenz et al., 2014; Kish et al., 2015; Maksimov et al., 2016). In agreement with the established band assignment for carotenoids (Schlücker et al., 2003), the ν₁ band position around 1,525 cm⁻¹ is mainly due to C=C double bond stretching vibrations of the conjugated polyene chain, while the ν₂ band around 1,160–1,180 cm⁻¹ consists of contributions from stretching vibrations of C–C single bonds coupled with C–H in-plane bending modes. The ν₃ band at 1,000 cm⁻¹ arises from in-plane rocking vibrations of methyl groups on the conjugated polyene chain, coupled with in-plane bending modes of the neighboring C–H groups, and collectively considered as a fingerprint of the conjugated end-cycle configuration. Finally, the ν₄ band around 960 cm⁻¹ stems from C–H out-of-plane wagging motions coupled with C=C torsional modes (out-of-plane twists of the carbon backbone). When the conjugated system of the carotenoid is planar, these out-of-plane modes are not coupled with the electronic transition, and these bands are not resonance-enhanced. However, distortions around C–C single bonds increase the coupling of these modes with the electronic transition, leading to an increase in their intensity (Kish et al., 2015). The ν₁, ν₂, ν₃ and ν₄ bands of dark-adapted wild-type OCP and the OCP-W288_BTA variant were at the same positions and had very similar amplitudes (Figure 6).

After photoconversion, there was a clear red shift of the ν₁ band from 1,525 cm⁻¹ to 1,517 cm⁻¹ for wild-type OCP, and even to 1,515 cm⁻¹ for OCP-W288_BTA, which is consistent with the correlation between the position of the ν₁ band and the position of the maximum of the chromophore’s absorbance spectrum (Kish et al., 2015). Theoretical studies have shown that this correlation is due to a combination of two effects: (i) the rotation of the β-ionone rings with respect to the conjugated system (Mori, 2016) and (ii) the local electric field in the chromophore-binding pocket (Gamiz-Hernandez et al., 2015). The red form of OCP-W288_BTA even appears to be slightly more red-shifted than wild-type OCP, which can be attributed to the effect of CAN embedment, which also causes a slight red-shift in the absorbance spectrum compared to ECN embedment (de Carbon et al., 2015). After photoconversion, both proteins show a decrease in the intensity of the ν₄ band since the carotenoid assumes a more planar configuration in the photoactivated state. The dark-minus-light difference spectra for the wild-type OCP and the BTA-bearing variant are virtually identical, leaving only the slight red shift in the ν₁ band of the photoactivated OCP-W288_BTA by 2 cm⁻¹ as a difference (Figure 6B).

Next, we investigated the kinetics of transitions between the dark-adapted OCP^O state and the light-adapted OCP^R state and monitored the OCP^O→OCP^R photoconversion and subsequent recovery as the change in absorbance at 550 nm, always at the same illumination intensity (fluence F) at different temperatures (Figure 7A). Fitting the rising and decaying phase of the curves with a monoexponential function yielded the rate constants for photoconversion and recovery [rate constants F∙σ and k _r , respectively, see the procedure described by (Shirshin et al., 2017)]. The resulting Arrhenius diagrams for the photoconversion and recovery rate constants for OCP-W288_BTA variant and wild-type OCP are shown in Figures 7B, C, respectively. The data show that the OCP^O→OCP^R photoconversion of OCP-W288_BTA is significantly slowed compared with wild-type OCP. Moreover, the activation energy (E_A) for OCP-W288_BTA determined from the Arrhenius plots is substantially reduced by ∼40 kJ/mol compared with wild-type OCP. The same trend is observed for the rate constant for recovery to the ground state: The rate constant of OCP-W288_BTA is slowed down compared to wild-type OCP, and E_A is lower by ∼33 kJ/mol. Assuming that the differences in E_A are due to the deletion of a single H-bond, the determined energy differences classify the bond between Trp288 and ECN as a moderately strong intermolecular H-bond, since the dissociation energies of H-bonds in the condensed phase range from 0.8 kJ/mol to 162 kJ/mol [the latter being an exceptionally large value unique for HF₂ ⁻ (Steiner, 2002)], which is much larger than the value of 9 kJ/mol for H-bonds in water at 298 K (van der Spoel et al., 2006).

Since the first high-resolution crystal structures of OCP were published (Kerfeld et al., 2003; Wilson et al., 2010), several studies on OCP have investigated the role of the absolutely conserved Tyr201 and Trp288 residues in the functional activity of this protein (Wilson et al., 2010; Wilson et al., 2011; Maksimov et al., 2016; Kerfeld et al., 2017; Yaroshevich et al., 2021). However, the limited possibilities to substitute aromatic amino acids within the standard amino acid repertoire led to ambiguous results, and overwhelmingly to the loss of the most important property of OCP—its photoactivity. To understand the role of H-bonds, one must be able to control their presence or absence, the most difficult task being neither to disrupt the OCP^O form of the protein, nor the ability to photoswitch. These requirements are met by the approach adopted in this work, in which the isosteric exchange of a nitrogen by a sulphur atom in the OCP-W288_BTA variant precisely removes one of the two H-bond donors to the ketocarotenoid.

The X-ray crystal structure clearly reflects the minimal-invasive, hence isomorphous nature of this exchange, and the structural and spectroscopic properties of the OCP^O state as well as photoactivity were preserved. Thus, we clearly demonstrate that Trp288 is not required for OCP photoactivity. The same is true for Tyr201, as shown by our previous work on the OCP-Y201W variant generated by classical site-directed mutagenesis (Yaroshevich et al., 2021). Thus, it is obvious that the presence of at least one H-bond donor at either of the two sites is sufficient to ensure proper formation of the OCP^O state and photoactivity. Interestingly, removing one of the two H-bond donors competing for the keto oxygen leads to the intriguing consequence that the carotenoid forms a more stable interaction with the remaining H-bond donor. As a result, the protein lacks the spectral heterogeneity characteristic of the dark-adapted state of the wild-type OCP protein.

From the point of view of a protein engineer or a spectroscopist, the spectral homogeneity of the dark-adapted OCP^O state and, consequently, the higher yield of the S* photoproduct, which was proposed to be crucial for the productive photocycle (Yaroshevich et al., 2021), could be considered as advantageous evolutionary selection criteria. Therefore, one may speculate why nature solved the problem of designing the OCP photoswitch with two H-bond donors, because obviously the selection criterion was neither the spectral homogeneity of the OCP^O ground state nor the quantum efficiency of the photoconversion. Rather, the incorporation of two H-bond donors competing for the ketocarotenoid was a compromise to ensure that the formation of the dark-adapted OCP^O state from the highly flexible structural arrangement of the domain-separated OCP^R state occurs with high reliability. Apparently, the two competing H-bond donors provide the necessary flexibility or functional redundancy for the compact dark-adapted OCP^O state to be formed.

Replacement of Trp288 with BTA reduced the activation energies of photoconversion and recovery by approximately 40 kJ/mol and 33 kJ/mol, respectively, compared with wild-type OCP. Attributing these enthalpy differences to the deletion of a single H-bond, the determined values classify the N-H⋯O = C bond between Trp288 and ECN as a moderately strong intermolecular H-bond. Since all H-bonds can be considered as precursors to proton transfer reactions, which may be well advanced for strong H-bonds (Steiner, 2002), the determined enthalpy differences due to the deletion of an H-bond in OCP are important, especially in view of the recently proposed photoactivation pathway of OCP involving an oxocarbenium ion at the ECN cofactor (Yaroshevich et al., 2021).

Undoubtedly, we now have a much better understanding of the nature of the interactions between the carotenoid ligands and protein matrix of the OCP. Moreover, how the presence of a “surrogate” or “redundant” hydrogen bond may affect the photoactivation process and its biological role has not been studied before. The low quantum yield in the formation of the primary and all subsequent photoproducts in the OCP photocycle (Shirshin et al., 2017) is obviously important for maintaining the balance between photosynthetic activity and photoprotective processes in cyanobacterial cells. We have demonstrated that the use of chemically controlled molecular operations based on the incorporation of non-canonical amino acids offers great potential for better elucidation of protein design by nature, but also for the engineering of such complex and biologically relevant chemical photoswitches in the future.