Active-site environmental factors customize the photophysics of photoenzymatic old yellow enzymes

Abstract

The development of non-natural photoenzymatic systems has reinvigorated the study of photoinduced electron transfer (ET) within protein active sites, providing new and unique platforms for understanding how biological environments affect photochemical processes. In this work, we use ultrafast spectroscopy to compare the photoinduced electron transfer in known photoenzymes. 12-oxophytodienoate reductase 1 (OPR1) is compared to to Old Yellow Enzyme 1 (OYE1) and morphinone reductase (MR). The latter enzymes are structurally homologous to OPR1. We find that slight differences in amino acid composition in the active sites of these proteins determine their distinct electron transfer dynamics. Our work suggests that the inside of a protein active site is a complex/heterogeneous dielectric network where genetically programmed heterogeneity near the site of biological ET can significantly affect the presence and lifetime of various intermediate states. Our work demonstrates additional tunability in Old Yellow Enzyme active site reorganization energy that could be leveraged for photoenzymatic redox approaches.

TOC GRAPHICS

Introduction

Photoinduced electron transfer (ET) is a key step in biological metabolic processes ranging from photosynthesis to DNA repair and bird magnetoreception^1–3. In systems that undergo photoinduced ET to perform their natural function, there have been numerous reports which reinforce the proposal that photochemically-active proteins have structural properties which serve to optimize those functions^4–11. In this context, it is imperative to understand the methods by which biology controls the efficiency of photochemical processes towards the ultimate goal of improving artificial photocatalysis or commandeering proteins for non-natural photochemistry in both efficiency and scope^12–18.

Specifically, one of our groups has made great strides in developing novel photoinduced chemical transformations utilizing the photoredox and redox properties of the flavin mononucleotide (FMN) cofactor in some ‘ene’-reductases. Two of these novel transformations are highlighted in Fig. 1. The reactions make use of either an excited FMN hydroquinone charge-transfer state or the anionic FMN semiquinone state to reduce the substrates and induce formation of an α-acyl radical, subsequently cyclizing to form specific enantiomers of the desired product^17,18. While careful spectroscopic mechanistic studies support the claimed mechanisms, we questioned what the photophysical behavior is of the natural, oxidized forms of the FMN cofactor in this particular family of ‘ene’-reductases, termed the Old Yellow Enzyme family, in order to get a more holistic understanding of the cofactor’s photoinduced dynamics within the enzyme active site. While these studies are of fundamental interest and not of biological relevance, reconciling the possibility of intraprotein photoinduced ET using Marcus theory will provide us with a better understanding of how this protein family engages the substrates in ET in reaction conditions and may hold the key to unlocking new photochemical substrate activation mechanisms for further reaction discovery.

Figure 1.

(A) Previous work in non-natural reaction discovery by combining optical irradiation with ‘ene’-reductases. (B) This work addressing fundamental questions regarding the photophysical properties of a select group of ‘ene’-reductases in the Old Yellow Enzyme family.

Much work has been done in applying principles of Marcus theory in order to predict biological electron transfer rates and efficiencies^19,20. In cases like the bacterial photosynthetic reaction center, where the photoinduced electron transfer time constant is 2–3 ps, the competition between hopping and superexchange needed to be considered, but the basis of Marcus theory was adequate^21,22. Otherwise, researchers have been largely successful in fitting electron transfer dynamics within the framework of both Marcus and other electron transfer theories^23–27. In many biological systems where ET is observed, one can in principle use steady-state spectroscopic and electrochemical methods to estimate the driving force for the ET, and time-resolved spectroscopic methods to acquire the photoinduced ET rate. It is straightforward to then find the resulting electronic coupling and reorganization energy which gives rise to the observed rate, allowing one to combine biological information about the system with spectroscopic and electrochemical experiments to better understand biological electron transfer²⁸.

One amino acid that occupies a privileged space in biological electron transfer is the tyrosine residue, as is evidenced by their pervasiveness in both natural and non-natural enzyme function²⁹. Largely due to its reversible and relatively accessible proton-coupled oxidation, many have found in various biological contexts that tyrosine oxidation plays an integral role in a range of intraprotein electron transfer processes, perhaps most notably in ribonucleotide reductases which employ a chain of tyrosine residues to shuttle electrons over large distances^6,30,31. While tyrosine residues acting in a direct redox capacity is not uncommon especially under conditions of cofactor or auxiliary photocatalyst excitation, exploration and discovery of the role of this and similar residues’ effects on environmental properties that affect enzyme activity remains somewhat elusive^32–34. Boxer and coworkers, for example, have elucidated unique circumstances in which directional electric fields generated by the permanent dipole of tyrosine residues in the active site explicitly enhance catalytic activity of ketosteroid isomerase^35,36. Similar modulations of catalytic processes of oxidoreductases or other enzymatic transformations due to the existence and positioning of tyrosine residues near, but not directly participating in the electron transfer process would demonstrate a new dimension of biocatalytic control of chemistry.

We can connect this concept of tyrosine residues modifying local electron transfer processes back to the Marcus formulation of ET by realizing that the outer sphere reorganization energy is partially derived from dielectric environment surrounding the site of ET, which one might imagine is a complicated function of the surrounding protein structure^37–39, and also from non-dielectric factors, like changes in explicit solvent or amino acid orientation⁴⁰. Many researchers have reported large reorganization energies (greater than 1 eV) for electron transfer within protein active sites by fitting biological electron transfer rates with Marcus parameters in conjunction with electrochemical analyses^4,41,42. For the most part, these the large reorganization energies found in protein active sites have been justified by the presence of water molecules within the active site because of the large dielectric constant of bulk water (80.2) in comparison of that to dry protein (2-4). Less attention has been focused on the contribution to the external reorganization energy by the heterogeneous dielectric environment comprising charged, hydrogen bonding, and nonpolar amino acids within the active site that usually differ between even closely structurally homologous proteins, that give rise to both dielectric and non-dielectric reorganization energy⁴³. Further, only recently has there been theoretical and experimental investigation as to how amino acid-water interactions might further complicate the dielectric properties of the active site, and thereby influence the external reorganization energy of ET in their vicinity^23,44,45.

In this work, we use ultrafast spectroscopy to compare the photoinduced electron transfer in 12-oxophytodienoate reductase 1 (OPR1), to Old Yellow Enzyme 1 (OYE1) and morphinone reductase (MR), which were previously used as photoenzymes in one of our group’s previous work^17,18. The latter enzymes are structurally homologous to OPR1. We find that slight differences in amino acid composition in the active sites of these proteins determine their distinct electron transfer dynamics. We rationalize the differences in their photoinduced dynamics within a classical Marcus framework by primarily invoking a large reorganization energy which is highly sensitive to the active site composition. Then, we study mutants of OPR1 with ultrafast spectroscopy. We find that using site-directed mutagenesis we can modify the ET dynamics by the replacement of select tyrosine residues in the active site which do not constitute the ET donor or acceptor to phenylalanine. Our work demonstrates that the inside of a protein active site is a complex/heterogeneous dielectric network where genetically programmed heterogeneity near the site of biological ET can significantly affect the presence and lifetime of various intermediate states.

Results and Discussion

The crystal structures for OPR1, OYE1, and MR are shown in Figure 2. Their overlaid crystal structures show clear structural homology between the three proteins within the large family of Old Yellow Enzymes, yet their specific composition in the active site in some cases differ^46–48. The flavin mononucleotide cofactor (FMN) in each case is noncovalently bound to the protein via multiple hydrogen bonding interactions, although the specific amino acids around the FMN are somewhat variable in both identity and in closest distance to the FMN. Certain amino acids are known to be directly involved in flavin protein photophysics and function as electron donors or acceptors. Here we focus attention to other nearby amino acids to explore whether and how they might tune the ET. These “redox-affecting” amino acids and their proximity to the FMN cofactor are noted in Table 1.

Figure 2.

(A) Overlaid crystal structures of OPR1 (green), OYE1 (blue), and MR (purple), displaying their overall structural homology (PDB: 1ICS, 1OYA, and 1GWJ, respectively). (B) Zoomed-in picture of the active site of OPR1, OYE1, and MR, also overlaid. Amino-acids B-K denote hypothesized redox-affecting residues, and their names and distances are noted in Table 1 relative to their respective FMN cofactors, denoted A. (C) Absorption spectra of the aforementioned proteins in comparison to that of the free FMN cofactor in tricene buffer.

Table 1.

Redox-affecting residues in OPR1, OYE1, and MR

Residue group	OPR1	OYE1	MR	Distance to FMN (Å)a
B	T37	T37	T31	2.7, 2.7, 2.7
C	Y358	Y376	Y356	3.3, 4.1, 3.3
D	W112	W116	W106	6.4, 5.4, 5.9
E	Y78	Y82	Y72	8.8, 10.0, 8.4
F	Y192	Y196	b	7.1, 6.5
G	b	b	F246	9.2
H	H187	H191	H186	3.5, 3.6, 3.7
I	Y246	F250	b	7.3, 7.0
J	H190	N194	N189	3.4, 3.3, 3.2
K	b	F296	b	7.0

^aThis distance reflects the closest distance of the amino acid to that of the isoalloxazine ring of the FMN, considering only the non-hydrogen atoms.

^bWhile amino acids are present in the locations which correspond to the respective residue groups, they are either non-redox active or point in directions away from the FMN.

The steady-state absorption of OPR1, OYE1, and MR are quite comparable despite their various active site compositions, Figure 2c. The increase in oscillator strength and redshift in the linear absorption spectrum of the studied flavoproteins as compared to FMN in solution has been previously discussed, and is not directly reflective of any charge-transfer type character of the lowest energy transition, but is instead due to increased structural rigidity of the isoalloxazine backbone⁴⁹. Further, hydrogen bonding interactions with the isoalloxazine core of the FMN cofactor further redshift the lowest energy absorption feature of the studied proteins, as can be seen in the absorption spectrum of the T37A OPR1 mutant (S7).

In order to understand their photoinduced dynamics, we performed ultrafast pump-probe spectroscopy by photoexciting the FMN cofactor using a 370 nm pump pulse and resolving the subsequent photoinduced dynamics in time and frequency using a white-light supercontinuum probe pulse generated from a CaF₂ crystal. More details can be found in the experimental section in the Supporting Information. Global analysis was performed on the photoinduced dynamics using the Glotaran software package⁵⁰, generating evolution-associated spectra (EAS) of OPR1, OYE1, and MR shown in Figure 3. We found that a sequential model with four states was adequate to describe the ultrafast dynamics in each system studied and we began each fit starting at 500 fs. The last two states in the model were almost always dominated by free FMN signatures, and hence their time constants were fixed to 3.2 ns and 100 ns, the latter of which is great enough to be considered an infinite time constant and reflects the long-lived nature of the free FMN triplet state. Goodness of fit is addressed in the Supporting Information (S1–3). For MR, analysis is included in the Supporting Information using a parallel sequential decay pathway model to account for a relatively large contribution from free flavin in the transient spectrum, which instead generates species-associated spectra (SAS). We further found that photoexcitation to the S₂ transition at 370 nm of the FMN cofactor afforded an unobscured view of many of the transient features required for photophysical analysis due to the absence of pump scattering in the ground-state bleach (GSB) region, and there was no difference observed on the picosecond timescales between excitation of the S₂ or S₁ transitions (S4).

Figure 3.

(A) Global analysis EAS spectra of OPR1. (B) Global analysis EAS spectra of MR. (C) Global analysis EAS spectra of OYE1. (D) Raw kinetics of the aforementioned proteins in both the GSB region (E) and in the SE region. MR and OYE1 EAS spectra do not include the 3^rd and 4^th EAS spectra, which are components reflecting a minor composition of free FMN dynamics that dominate the pump probe spectrum after 100 ps in some samples, as seen in the uncorrected kinetics of the MR trace in (E).

The main features in the transient absorption for an initially photoexcited free FMN S₁ state are also apparent in the excited-state dynamics of OPR1, mainly a GSB at 460 nm, a clear stimulated emission (SE) signature at 580 nm, and a sharp excited-state absorption (ESA) between them, centered at 530 nm. Note that although the SE signature is only apparent at regions lower in energy than 550 nm, the emission spectrum of free FMN has its maximum at 530 nm, meaning that the ESA between the GSB and the SE is likely overlapping with a strong SE signal in that wavelength region.

The spectral signatures of OPR1 are the only likeness that it shares with free FMN, considering that it appears that the excited state decays rapidly without any indication of charge transfer dynamics. This is a surprising result because, typically, flavoproteins have short-lived excited states due to charge transfer interactions with nearby amino acids with accessible oxidation potentials—this photophysical phenomenon is easily distinguishable as it should only result in changes to the ESA and SE regions of the spectrum^33,51–54. In OPR1, we observe a global decrease in the intensity of all spectral signatures which is well-fit with a biexponential kinetics on the picosecond timescale. We find that EAS 1 and EAS 2, the first and second representative spectra, are almost identical, aside from their amplitudes, further complicating the interpretation of the ultrafast dynamics.

In order to gain some context into the possible photophysical mechanisms for excited-state deactivation in OPR1, i.e. ET vs internal conversion, we can look at the same ultrafast experiments performed on MR and OYE1, also shown in Figure 3. While OYE1 has an almost identical EAS 1 to that of OPR1, EAS 2 displays noticeably different spectral features in the 500-700 nm region. Namely, the SE peak present in EAS 1 evolves on the picosecond timescale to form a broad, relatively unstructured ESA with two peaks at 530 and 625 nm. This evolution is concomitant with a decrease in intensity in the GSB region as well, and the overall evolution has a time constant which is quite similar to that of EAS 1 for OPR1. Overall, the shape of EAS 2 resembles greatly that of the neutral semiquinone FMNH•, the reduced and protonated form of the cofactor FMN, and is commonly seen as one of the photoinduced intermediates in the BLUF domains for cellular signaling⁵⁵. This EAS decays with a timescale of almost double that of EAS 2 in OPR1, reflecting a unique deactivation pathway from this photoproduct compared to the possible transient intermediate in OPR1. Moreover, it decays approximately 6 orders of magnitude faster than FMNH• in signaling proteins.

On the other hand, MR displays some completely unique transient signals as compared to both OPR1 and OYE1, including ultrafast dynamics which we could not fully capture due to the temporal resolution of our narrowband pump-probe setup. We begin the global analysis of MR at 1 ps to isolate the subsequent dynamics, meaning that EAS 1 is likely a mixture of the initial photoexcited state and the first transient intermediate. As such, EAS 1 in MR has a few key features: the relative intensity of the ESA at 530 nm compared to the GSB has increased, indicating that the new transient species has a greater absorption in this region than the initial FMN S₁ state, and there exists a broad ESA which is superimposed with the SE signal in the 560 nm region, indicating that there is likely a mixture of species in solution. Subpicosecond dynamics are shown in the Supporting Information and confirm the growth of ESA in the range from 500 to 640 nm and the reduction of negative signal in the SE region (S5). An assignment of at least one of the contributing species to the EAS 1 spectrum is straightforward, since the anionic semiquinone, FMN⁻, has increased absorption compared to FMN in the 500-550 spectral range. Its formation should remove the overlaying SE in that region, leading to a large growth in the positive signal in that area⁵⁴. EAS 2 for MR looks almost identical to that for OYE1, leading to a straightforward assignment of that transient intermediate to the neutral semiquinone once more. The fact that their respective EAS 2 have almost identical lifetimes further supports this claim. In the Supporting Information, we do a complementary analysis of MR using parallel sequential decay pathways with a total of five states to rigorously account for free flavin that is spectroscopically visible at nanosecond pump-probe delay times (S3).

With assigned EAS for both MR and OYE1, we can now assign the ultrafast dynamics themselves as being mixtures of photoinduced charge transfer and subsequent protonation of the reduced flavin, shown in Fig. 4a. MR shows the most straightforward case, where we can temporally resolve parts of the evolution directly from the FMN S₁ state to FMN⁻, and then from there protonation of the flavin to form FMNH•; importantly, this sequence occurs with sequentially slower kinetics to be able to observe the build-up of each transient intermediate. Comparing this to OYE1 begs the question of whether or not to interpret the decay of EAS 1 into EAS 2 as a proton-coupled electron transfer from a nearby residue, although it is also possible that OYE1 follows the same decay pathway as MR but with k_PT > k_ET, thereby preventing appreciable formation of FMN⁻. While these two possibilities are spectroscopically indistinguishable, we continue with the simplest case for comparison to MR and OPR1, though, which is to consider the ET and PT steps separately, arriving at the sequential relative rates for the photoinduced dynamics of all three proteins shown in Fig. 4a.

Figure 4.

(A) Kinetic scheme of the ultrafast decay pathways within the ‘ene’-reductases studied, where the relative rates in their respective proteins reflect the transient populations observed in the ultrafast spectroscopy. (B) Free energy surfaces which generally reflect the feasibility of inverted electron transfer kinetics in OPR1 considering fixed driving forces (−0.4 and −2.0 eV for ET and bET, respectively) and excited-state energies (2.4 eV) as compared to a small reorganization energy (C) or a large reorganization energy. Changes in the reorganization energy are implemented by both a change in the curvature of the parabolas as well as by shifting their minima relative to one another. (D) Free energy surfaces which show that inverted ET kinetics are also possible by changing the energy of the radical ion pair (RIP) state with the same reorganization energy as in Fig. 4b.

While we can rationalize the photophysics of OYE1 and MR, those of OPR1 remain unclear, despite the active-site similarity to OYE1 and MR. This leads us to the hypothesis that OPR1 exhibits inverted charge transfer kinetics, where FMN⁻ forms with the lifetime of EAS 1, but the back-electron transfer (bET) to reform the ground state FMN is much faster than the formation rate of the FMN⁻. We reject the hypothesis that this is internal conversion directly from the FMN S₁ state to the ground state based on the low likelihood of the access of a distorted geometry to relax through a conical intersection and the lack of report of any oxidized FMN which exhibits a similar picosecond timescale internal conversion process. Further, we found that by mutating out the two nearest potential electron donors to the FMN excited state (Y358 and W112), we could achieve a lifetime more than 20 times longer than the wild-type, supporting our hypothesis that the deactivation we are observing is rooted in charge transfer (S6). This phenomenon can be examined by considering the various system properties which affect the ET rate from the perspective of the Marcus formulation, which incorporates parameters like the electronic coupling between initial and final states V, the driving force for charge transfer ΔG, and the reorganization energy λ, shown in equation 1.

$$k_{ET}=\frac{4\pi V^2}{h\sqrt{4\lambda k_{B}T}}\exp \left(\frac{{-\left(∆G_{ET}+\lambda \right)}^2}{4\lambda k_{B}T}\right)$$ (1)

It is worth stating that one must be careful when using the Marcus formulation to analyze such ultrafast reactions—electron transfer on the timescale of picoseconds is appropriate to interpret in the framework only when the solvent relaxation time is on a faster timescale, which is satisfied in this case because the relaxation time of water is partially on the subpicosecond timescale⁵⁶. Within this framework, there are a series of situations which would explain the observed inverted charge transfer kinetics in OPR1 and why it is not observed in OYE1 and MR:

1. In OPR1, the coupling between the electronic states for the bET process is much greater than the coupling between the electronic states for ET, such that the rate of bET is much faster than ET despite rate constant considerations from ΔG and λ for the ET and bET processes. In MR and OYE1, the coupling for ET is greater than that for bET, reversing the situation in OPR1, again with rate constant considerations from ΔG and λ amounting to a negligible contribution to the magnitude of rates for ET and bET.

2. In OPR1, rate considerations from ΔG and λ for bET and ET dominate their differences in rates, either because the electronic coupling V for each process are similar in magnitude or because their difference in magnitude does not compensate for the rate considerations from differences in ΔG and λ for the ET and bET processes. Thus, the values for ΔG and λ are more similar in magnitude for bET than ΔG and λ for the ET process, such that the rate for bET is greater than the rate of ET. In OYE1 and MR, the situation is reversed for the values of ΔG and λ for ET and bET, such that the rate of ET is greater than that for bET. This possibility is shown graphically in Fig. 4b–d.

The first possibility, that the sufficiently different electron couplings for photoinduced ET and bET are responsible for the inverted kinetics, is challenging to experimentally to verify without prior knowledge of the other thermodynamics properties of the charge transfer processes as well as their reorganization energies. In one report, photoinduced ET and bET (in their report denoted “electron return”) originally between flavin adenine dinucleotide in fully reduced form (FADH⁻) with a thymine dimer (T<>T) and then between FADH• and the thymine anion (T⁻) following splitting of the cyclobutene ring in DNA photolyase are found have coupling values of 3.0 meV and 2.6 meV, respectively⁵⁷. In many Marcus type analyses of photoexcited flavin cofactors which undergo photoinduced ET with nearby amino acids, the electronic coupling (in the references reported as tunneling parameters) for ET and bET are assumed to either be the identical or to vary about some small range^4,58,59.

By comparing the active sites of OPR1, OYE1, and MR, it is clear that the residues likely playing the role of the FMN reductants are conserved, namely residue groups C, D, and H, all with almost equivalent distances to the FMN cofactor, as can be seen from Table 1. The electronic coupling for photoinduced ET is a quantum mechanical interaction which depends on various parameters of the wavefunction in the initial and final diabatic ET state, such as the shape and symmetry of the relevant orbitals on the donor and acceptor, as well as intermolecular parameters such as relative orientation and an exponential distance-dependence at donor-acceptor distances greater than the sizes of the donor and acceptor themselves^60,61. The absence or presence of inverted kinetics does not seem to obviously depend on the FMN-residue distance or orientation for these proteins, for example in the case of OPR1 and MR in which both Y358 and Y356 are equidistant and similarly oriented with respect to the FMN cofactor in the crystal structure. We note that the electronic couplings themselves for ET and bET may vary depending on the protein because it is crucially dependent on small changes in intermolecular distance. However, this coupling variation is distinct from the discussion of the relative magnitude of electronic couplings for ET and bET within the same protein, which would have to reverse in relative magnitude when comparing different proteins for this to be an adequate explanation of inverted kinetics in OPR1 but normal kinetics in OYE1 and MR. Based on this logic and the cited literature which finds these electronic coupling values to be similar or assumes them to be the same, we thus disfavor the explanation afforded by invoking highly different electronic couplings for photoinduced ET and bET for the observed inverted kinetics^57,59.

Remaining is the possibility that the relative magnitudes of the expression $\exp \left(\frac{{-\left(\mathrm{\Delta }G+\lambda \right)}^2}{4\lambda k_{B}T}\right)$ for photoinduced ET and bET are not only such that in OPR1 the rate of bET is greater than the rate of photoinduced ET, but that also in OYE1 and MR this situation is reversed giving non-inverted charge transfer kinetics. The driving force parameter can be estimated with the knowledge of the redox potentials of the electron donor and acceptor as well as the energy of the excited state as described by Weller⁶². In the case of OPR1, OYE1, and MR, their excited state energies are likely similar as indicated by their nearly identical absorption spectra, and it has been reported that the one electron reduction potential of the FMN can varies only by 20 meV between MR and OYE1 via electrochemical or chemical methods^63,64. More drastic variability in the reduction potentials of the proposed electron donors, namely tyrosine and tryptophan residues, have similarly been reported, complicating straightforward calculation of the driving force^65,66.

Despite the complication of not exactly knowing the driving force for the photoinduced ET and bET processes for the proteins studied, these values in and of themselves do not determine the rates of ET—only in the context of the reorganization energy of the electron transfer of interest does the driving force affect the rate of ET, giving rise to the well-known Marcus normal and inverted regimes. In DNA photolyase, it has been deduced through ultrafast experiments that electron transfer from FAD excited state to nearby tryptophan residues is in the Marcus normal region whereas the back ET is in the inverted region due mostly to the large derived reduction potential for the proximal tryptophan (1.48 V vs NHE) resulting in a −0.7 eV for photoinduced ET. The derived reorganization energy for the forward step the authors narrowed to a range of between 1.0 and 0.84 eV, and with a nearly equal reorganization energy for the back ET their location in the normal and inverted regions are clear to see⁵⁹. We note that while comparable values for the reduction potential of tyrosine residues have been reported, their reduction potentials vary by as much as 0.8 V depending on the measurement conditions, especially between conditions in the absence of a protein-like environment^30,31.

Taking a range of tyrosine reduction potentials (1.0 to 1.8 V vs SHE) as well as the known one electron reduction potential from OYE1 (−0.24 V vs SHE) gives an approximate driving force range of −1.26 to −0.46 eV for the photoinduced ET and −1.24 to −2.04 eV for bET. In most simple electron donor-acceptor systems, the reorganization energy for photoinduced ET is much smaller than 1 eV, which typically results in the bET to the ground being placed far in the inverted region (assuming that the reorganization energies for the two ET processes are comparable)⁶⁷. Even before considering more elaborate models for electron transfer that include, e.g. the participation of high-frequency modes and solvent frictional effects^68–70, we find that the inverted kinetics observed in OPR1 but not in OYE1 or MR could be adequately explained by assuming large reorganization energies for the photoinduced and bET steps, which are more similar in magnitude to the bET driving force than that of the photoinduced ET for the proteins where we observe inverted kinetics, sketched in Figure 4c. Along similar lines, an alternative, but ultimately similar explanation could also be that the various proteins contain FMN cofactors and electron-donating amino acids with highly variable redox potentials, modulating the ET driving forces for photoinduced and bET such that the driving force becomes more similar to reorganization energy for bET than for photoinduced ET in OPR1, which exhibits inverted kinetics.

We thus examine this driving force-reorganization energy interplay and how it could lead to inverted kinetics via the Marcus formulation by calculating how the relative rate constants for photoinduced and back ET change over a range of realistic driving forces and reorganization energies for the both pathways, shown in Fig. 5. The reorganization energies of the forward and back ET are treated as independent variables where the relative rate of the photoinduced and back ET rates can be evaluated for three driving force conditions, ranging from very similar (−1.2 for photoinduced ET and −1.3 for back ET) driving forces to more different (−0.4 for photoinduced ET and −2.1 for back ET). The relative driving forces between photoinduced ET and bET are restricted such that the sum of the driving forces must be equal in magnitude to the energy provided by the initial excited state, estimated to be 2.5 eV. Relative rates as plotted in the z-axis in Fig. 5 that are above 1 indicate a prediction of inverted kinetics within the standard Marcus picture shown evidenced by Equation 1.

Fig. 5.

Relative photoinduced and back ET rates using realistic driving forces and reorganization energies for each ET step. We assume the electronic couplings are equal.

Fig. 5 clearly demonstrates not only that the occurrence of inverted kinetics is controlled by both driving force and reorganization energy considerations, but that also the extent of kinetic inversion, how large the value of k_bET/k_ET is, is similarly sensitive to particular regions in energetic parameter space. The results at each point reflect an interplay between the driving forces of each ET step and their reorganization energies, thus spanning a range of Marcus regimes for each ET process. The global trend within this wide range of parameters is that with larger reorganization energies for both steps, the relative rates for photoinduced and back ET trend towards the appearance of inverted kinetics. The explanation is that in all driving force cases investigated the bET driving force was larger in magnitude than that of the photoinduced ET, thus upon increasing the reorganization energy of both ET steps when the photoinduced ET and back ET are in the Marcus normal and inverted regimes, respectively, will give rise to an increasing rate of bET with a decreasing rate of photoinduced ET. Interestingly, the main change between the relative rates as a function of increasing the energy of the RIP state, which decreases the driving force for photoinduced ET and increases the driving force for back ET, is to change the sensitivity of the relative ET rates to the reorganization energies. For the results in purple, reflecting the highest energy RIP state, the relative rates range from 0.025 to 2.83, whereas the results in black for the lowest energy RIP state investigated range from 0.44 to 1.81. Based on this model and its results, it is evident that both changes in the reorganization energy and driving forces for the ET processes in OPR1, OYE1, and MR can cause the photoinduced dynamics to switch between normal and inverted kinetics and increase the relative magnitudes of photinduced ET and bET.

In order to experimentally verify the role of the reorganization energy and driving forces on the existence of inverted kinetics in OPR1 in the context of the results obtained from the Marcus formulation of ET, we used site-directed mutagenesis to replace the redox-affecting amino acids in the active site with nonpolar or redox-innocent residues. Our hypothesis is that since OPR1, MR, and OYE1 all differ in amino acid composition in the active site, that this difference in composition influences the reorganization energy and driving forces to be modulated between the different proteins, thus giving rise to their differing ET dynamics. With the exception of T37A, which had a blueshifted absorption spectrum by almost 10 nm, the steady-state properties were quite comparable to OPR1 (shown in the Supporting Information, S7), although many of the mutants had minor contributions of free FMN which appeared as background signals in the ultrafast spectroscopy, and whose contributions to the global analysis EAS of the various mutants can be addressed using similar methods used for MR giving Figure 6.

Figure 6.

Global analysis spectra of various mutants of OPR1, namely (A) H187N (B) T37A (C) Y192F (D) Y358F (E) Y78F, Y246F. (F) Transient absorption kinetics of selected mutants at λ_probe = 610 nm, resonant with the lowest energy absorption of FMNH• and SE of the FMN S₁ state.

The lifetimes of EAS 1 and EAS 2 for the above mutants, as well as W112A, are shown in Table 2. W112A showed identical spectral evolution to the OPR1 wild-type (WT). T37A also shows essentially identical spectra as compared to the WT with a slightly slower overall decrease in the transient signals. While the EAS lifetimes for T37A and Y358F don’t differ much themselves as compared to the WT, Figure 6 clearly shows that the ratio of the amplitudes between EAS 1 and 2 are not equivalent, with the longer-lived EAS 2 showing a greater overall contribution to the signal than the WT. This observation supports the argument for conformational flexibility of FMN in OPR1, with which the mutation of Y358, which is almost in van der Waals contact, likely increases. It is interesting to note that despite the close proximity of Y358 and W112 to the FMN cofactor in OPR1, their individual mutations do not lead to an overall change in the deactivation pathway of the FMN cofactor excited state. This likely reflects a cooperative effect, that when Y358 does not function as the electron source for the photoinduced charge transfer, W112 replaces it while still affording inverted kinetics with respect to the back-electron transfer (bET). This hypothesis was confirmed by showing that the double mutant of Y358 and W112 indeed showed an even longer excited-state lifetime (S6). The observation of similar EAS spectra as the wild-type but with slower kinetics is in support for our Marcus formulation rate analysis, since in analogy to OYE2, W112 serves as a secondary electron donor in the absence of Y358. Since the appearance of inverted kinetics seams mostly sensitive to large values of the reorganization energy, even when switching from Y358 to W112 as the electron donors in the photoinduced ET reaction there is likely to still be inverted kinetics despite the different electronic coupling values and driving forces for ET.

Table 2.

Global analysis lifetimes of the studied ‘ene’-reductases

	EAS 1 (ps)	EAS 2 (ps)	EAS 3 (ps)
OPR1 (WT)	2.6	14
OYE1	3.7	24
MR	3.2	24
H187N*	1.8	11	220
T37A	2.1	18
Y192F	3.9	159
Y358F	3.2	27
Y78F, Y246F	3.2	70
W112A	2.0	11

^*This global fitting was done using a scheme that included parallel and sequential decay pathway, and thus should formally be denoted SAS.

Since Y192 exists in OPR1 but there is no tyrosine in that physical position in MR, and similarly since Y246 exists in OPR1 but not OYE1, we leveraged site-directed mutagenesis to express OPR1 proteins that replaced those amino acids with phenylalanine. EAS 1 in both the Y192F and Y78F, Y246F cases gave rise to a spectrum which looks qualitatively similar to the WT, but with important differences, namely the ratio of the GSB signal to the ESA signal at 530 nm; the relative equal amplitude of these positive and negative signatures is reminiscent of the case for MR, where we assigned the transient species to FMN⁻. As clear stimulated emission signals are also present, EAS 1 likely represents linear combinations of cofactor conformations which undergo either the inverted kinetics like in the WT or have a decreased rate of bET such that the FMN⁻ is observable. The evolution from EAS 1 to EAS 2 in both cases exhibits to a large degree repopulation of the ground state, likely in the conformers similar to the WT, but also reveals a small degree of subsequent PT, affording the relatively long-lived FMNH•. Overall, these site-directed mutagenesis experiments show that we can essentially tune the OPR1 photophysical properties so that they become similar to OYE1 and MR. Somewhat surprisingly, this was achieved by substituting amino acids further than 7 Å from the FMN cofactor.

Interestingly, we also found that H187N affords a completely unique spectral evolution with respect to OYE1, MR, or any of the mutants. Like the aforementioned tyrosine mutants, the first global analysis spectrum clearly shows signs of FMN⁻ for the same reasons as discussed above, namely the ratio of the ESA and SE signals as compared to the GSB. However, instead of undergoing subsequent proton transfer, the reduced FMN instead directly undergoes bET, revealing a population of proteins composing which presumably undergoes bET with a slower rate, likely again due to morphological variation, arriving at the neutral semiquinone to a similar degree as the tyrosine mutants. For H187N, a more complicated, parallel sequential decay pathway model with 5 states was used due to the large amount of free flavin in solution. The goodness of fit for the Global Analysis on each OPR1 mutant is demonstrated in the Supporting Information (S8–13).

The observation that the OPR1 photophysical dynamics are most sensitive to mutations of tyrosine residues far from the ET site, but within the active site itself is significant. The presence of tyrosines Y76, 192, and 246 in OPR1 therefore play a defining role in the reorganization energy of ET within the active site, reflected clearly in its ultrafast dynamics. As compared to residues like phenylalanine, tyrosine residues are much more polar, and as a result might serve to increase the dielectric constant and thus the reorganization energy associated to the active site as compared to nonpolar residues. Changes in the driving forces for photoinduced and back ET are also possible, as the Rehm-Weller equation includes terms that account for the Coulombic interaction between the newly generated RIP state. The redox potentials of the donor and acceptor also depend on the dielectric constant of their environment, usually accounted for using the Born correction term to the Rehm-Weller equation; this type of correction can be large, and explains well the range of driving forces reported before for tryptophan and tyrosine residues^65,71. The observation of single mutations within a protein active site affecting the reorganization energy of an intra-protein ET has been reported in a few careful studies on DNA photolyase^4,59,72. However, this study remains distinct from those, as they clearly only observe normal kinetics with photoinduced ET always outpacing back ET, likely due to smaller reorganization energies than necessary to observe inverted kinetics, possibly due to its own active site environment. However, changes to the dielectric constant alone likely do not explain such drastic changes to the reorganization energy. It has been discussed that corrections to the ET driving force are of similar magnitude to changes in the reorganization energy as a result of changing dielectric properties of the environment, or more specifically upon changes to the polaron coupling constant ($\frac{1}{{\epsilon }_o}-\frac{1}{{\epsilon }_s})$, where ε_o and ε_s are the optical and static dielectric permittivities of the surrounding medium⁴⁰. Importantly, this “cancelling out” of changes to the driving force and reorganization energies with changes to the medium dielectric properties is reversed when thinking of bET, since the lower energy RIP state results in a decreased driving force for bET on top of an increasing reorganization energy. If bET is in the Marcus inverted region, then this combination of effects increases its rate while the ET rate is less affected.

Another interpretation of the mutation-sensitive ultrafast dynamics which decreases the reliance on specific amino acid identities is that these tyrosine residues support a network of hydrogen bonding interactions between one another and with water molecules within the active site. While hydrogen-bonding networks have been invoked in explaining anomalous photoinduced ET behavior within proteins, they have not yet been described within a part of the active site dielectric environment except theoretically^44,45. Thus, when some of those tyrosine residues are absent, like in MR, OYE1, or the relevant OPR1 mutants, the hydrogen-bonding network might break down and reduce the reorganization energy associated with ET, or those replacement amino acids may not have large changes in orientation upon ET. This is likely related to literature discussions of “non-dielectric reorganization energy,” similar to how one might think about internal reorganization energy for ET processes, but for the molecules in the surrounding environment instead of the ET donor and acceptor^40,42,73. These non-dielectric contributions to the reorganization energy are thought to come from changes to the position or orientation of nearby charged residues or water molecules within the active site, but changes to hydrogen bonding networks has so far not been implicated for non-dielectric reorganization energy contributions to biological ET events. Clearly more detailed experimental investigation into the contributions of the solvent medium and explicit protein medium are necessary for a complete understanding of the role that these distant tyrosine residues have on the reorganization energy for ET in protein active sites.

Naturally arising from our work is a new distinction between the functions of amino acids in the active site, illustrated in Figure 7. Within the residues affecting the redox properties and identified for the case of these Old Yellow Enzymes, the ones 5 Å or less from the FMN cofactor likely serve as redox editors, changing the driving force of the ET by modifying the free energy driving force via direct hydrogen bonding, or redox partners, donating the electron to the FMN upon photoexcitation. Good examples of the above assignments from our study are the T37A and Y358F mutants of OPR1, respectively. The T37A mutant is shown in the literature to affect the reduction potential of the FMN cofactor in OYE1⁷⁴. This explains the decreases in ET rates in our time-resolved studies. Similarly, Y358F slows down photoinduced ET, although in this case the proximity to the FMN cofactor and the tyrosine’s accessible redox potential makes it more likely to be a redox partner. Building on this framework, we introduce the concept of a reorganization energy-editor, which is a residue that affects the ET pathways or rates due to its changes to the electrostatic properties and reorganization energies for ET within the active site. In the case of OPR1 it seems that modification of reorganization energy-editing Y76, Y192, and Y246 which are all more than 7 Å from the FMN cofactor have the greatest effect on the observed photoinduced ET dynamics. The mutation of these residues serves to change thermodynamics parameters critical ET, namely the driving force and reorganization energy, within the active site, which we observe as a reversal from inverted to normal kinetics in their mutants. While rational design of enzyme reorganization energies for biocatalysis has is still an active area of research, our proposal of at least partial reorganization energy control of ET processes in photoenzymes is unique in an important dimension⁴³. For ground state biological ET reactions, modulation of the reorganization energy to match the driving force can be a powerful tool to lower the activation free energy of catalysis. In the case of photoinduced ET, reorganization energies instead drastically affect the lifetime of transiently generated redox intermediates due to the competition between the photoinduced and back ET steps, and thus requires more careful management for the rational design of photoenzymes.

Figure 7.

The roles of various redox-affecting amino acids in the OPR1 active site in determining the rates of intraprotein ET processes.

Conclusions

We reported ultrafast dynamics on three Old Yellow Enzymes. Despite their overall large degree of structural homology, they display distinct ultrafast ET dynamics. While the analysis of ET in these Old Yellow Enzymes does not have direct implications for their native chemical functionalities, in the context of non-natural photoenzymatic catalysis these studies provide insight into the basics of their photochemical functionalities, and thus their utility as photocatalysts. Using a Marcus ET framework, we explained that the differences observed between their photoinduced ET pathways depend on an interplay between the driving forces for photoinduced and back ET with their reorganization energies. In the case of OPR1, the reorganization energy is large enough to induce inverted ET kinetics. By using site-directed mutagenesis, we found that by replacing hydrogen-bonding residues far from Van der Waals contact with the FMN cofactor, the photoinduced ET pathway in OPR1 can be modified so that it resembles that of OYE1 and MR. We conclude that the removal of hydrogen-bonding amino acids in the active site could change the reorganization energy and ET driving forces by amounts large enough to change the photochemical response of OPR1, both presumably resulting from perturbation of residue-water interactions or the dielectric properties within the active site.

Our work demonstrates another biological handle for optimizing protein function which in this case does not only depend on the augmentation of the thermodynamic driving force of the transformation. From an evolutionary standpoint, this reflects an optimization pathway of the protein which may allow redox-active enzymes like OPR1 to specialize for a natural substrate in addition to tuning the explicit substrate binding interactions. Further, in photoenzymatic catalysis, mutations targeting the peripheral environment surrounding ET could yield longer-lived reactive intermediates that, in turn, would facilitate non-natural diffusion-based photocatalysis, which has yet to be realized in practice. In a similar vein, achieving photooxidation chemistry using this family of enzymes may require careful adjustment of the active site amino acid composition, since the reorganization energy of the active site may deactivate the oxidation via fast back ET, akin to the inverted kinetics phenomenon discovered in the absence of a substrate. One unpredictable solution to this conundrum may arise from the substrate binding event itself modifying the reorganization energy within the active site, likely due to the expulsion of active site water molecules decreasing the dielectric constant or changing the hydrogen bonding structure in the active site. This work considering the role of fluctuations of the electrostatics within the protein active site serves to complement the pioneering work by Boxer and coworkers in the context of electric field catalysis. Supplementary transient infrared experiments suggests the role of explicit electric fields generated by tyrosine residues within the Old Yellow Enzyme active sites to photoenzymatic catalysis^35,36,75. Finally, our work has shown the sensitivity of intra-protein electron transfer dynamics within the active site through and the influence of amino acids on the surrounding dielectric environment.