Femtosecond Dynamics of Short-Range Protein Electron Transfer in Flavodoxin

Abstract

Intraprotein electron transfer (ET) in flavoproteins is important to understanding the correlation of their redox, configuration, and reactivity at the active site. Here, we used oxidized flavodoxin as a model system and report our complete characterization of a photoinduced redox cycle from the initial charge separation in 135–340 femtoseconds, to subsequent charge recombination in 0.95–1.6 picoseconds, and to following cooling relaxation of the product(s) in 2.5–4.3 picoseconds. With 11 mutations at the active site, we observed that these ultrafast ET dynamics, much faster than the active-site relaxation, mainly depend on the reduction potentials of the electron donors with minor changes by mutations, reflecting a highly localized ET reaction between the stacked donor and acceptor at a van der Waals distance and leading to a gas-phase type of bimolecular ET reactions confined in the active-site nanospace. Significantly, these ultrafast ET reactions assure our direct observation of vibrationally excited reaction product(s), suggesting that the back ET barrier is effectively reduced due to the decrease in the total free energy in the Marcus inverted region, leading to the accelerated charge recombination. Such vibrationally coupled charge recombination should be general to flavoproteins with the similar configurations and interactions between the cofactor flavin and neighboring aromatic residues.

Electron transfer (ET) is ubiquitous in biology and essential to a variety of biological activities such as converting chemical energy,^1,2 catalyzing enzymatic reactions,^3–5 and triggering biological signaling.^6,7 To understand ET dynamical behaviors in proteins, for the past decades various experiments of long-range (>10 Å) ET have been designed and characterized,^8–11 and theoretical models^12–15 were developed to elucidate the relationship of ET rates with donor-acceptor distance (r) or electronic coupling constant (J), driving force (ΔG^∘), and reorganization energy (λ). Typically, these ET dynamics occur on nanoseconds or longer and the systems are in equilibrium with local environments. These experimental observations can basically be understood based on the original Marcus theory in the normal (−ΔG^∘ ≤ λ) or inverted region (−ΔG^∘ >λ). However, when the donor-acceptor separation becomes less than 10 Å, the ET dynamics can occur ultrafast in the range of femtoseconds (fs) to hundreds of picoseconds (ps), which are on the similar time scales of local environment relaxation of solvent and the protein.^2–4 Thus, the ET dynamics are in nonequilibrium and the processes follow non-exponential dynamical behaviors.^3,4,16 Usually, the extended Sumi-Marcus two-dimensional model¹⁷ with quantum correction¹⁸ can be used to simulate experimental results, providing the molecular insights of short-range protein ET. The vibrational modes could be observed to greatly enhance ET dynamics in chemical systems,^18,19 effectively reducing the ET barrier, but such vibrational quantum effect has not been directly observed in protein ET. A theoretical study was recently proposed to explain the direct coupling of two low-frequency vibrational modes to charge separation in bacterial photosynthetic reaction centers.²⁰

Furthermore, when the donor and acceptor is in close proximity such as at van der Waals contact, the short-range ET dynamics in proteins could be even faster than the local relaxation, leading to a ‘frozen’ protein environment with a heterogeneous distribution of electrostatics. In this case, although the energetics is related to the local electrostatics, the ET reaction is relatively “isolated” and would be a confined bimolecular reaction. In this report, as our enduring efforts^16,21 to elucidate the molecular mechanism of ultrafast nonequilibrium protein ET at the short range, we report our careful characterization of the ET dynamics in a model system of D. vulgaris flavodoxin.

Flavodoxin is one class of flavoproteins and functions as an electron shuttle through the cofactor of flavin mononucleotide (FMN).²² Structurally, flavodoxin consists of five-parallel β-strands surrounded by several α-helices as shown in Figure 1 (left). The prosthetic FMN group is noncovalently but tightly bound by a series of interactions with protein residues.²³ One of the key interactions is the π-π stacking of the isoalloxazine ring with Y98 and W60 at van der Waals contacts (Fig. 1, right) to form a sandwich configuration. Such recognition of FMN excludes water out of the binding pocket and only the partial o-xylene ring is exposed to surface hydration water molecules. We have recently characterized the dynamics of the active-site solvation in three redox states (FMN, FMNH^•, and FMNH⁻) with site-directed mutagenesis to replace the electron donors of Y98 and W60 with the inert aromatic residue F (a double mutant W60F/Y98F).²⁴ The local relaxation was observed to occur on the wide time scales from a few, to tens and to hundreds of picoseconds. With well-characterized active-site relaxation dynamics, we can quantitatively examine the short-range nonequilibrium ET dynamics in flavodoxin.

Figure 1.

(Left) X-ray crystallographic structure of oxidized D. vulgaris flavodoxin (PDB code 2FX2). (Right) A close-up view of the local configuration at the FMN-binding site. The FMN cofactor (in yellow) is sandwiched between two neighboring aromatic residues W60 (in green) and Y98 (in orange) at the van der Waals contact. Also shown are the acidic residue D95 and the backbone carbonyl group of G61 in close proximity of FMN.

We report here the characterization of both forward and backward ET in the complete photoinduced redox cycle of oxidized FMN. Although the ET dynamics have been reported in several flavoproteins,^25–28 including flavodoxin,²⁸ the majority of studies only examined the forward ET dynamics. These studies showed that the intraprotein ET can occur on the wide time scales from hundreds of femtoseconds to tens of picoseconds. Here, with site-specific mutation and femtosecond resolution, we systematically studied 11 mutants to evaluate the short-range ET dynamics with different redox energies, including two mutations of neighboring residues of G61 and D95 which have been identified to highly affect the redox states of FMN.^29,30 These 11 mutants and the wild type can be classified into four cases in Table 1 (also see Fig. 1, right): (i) Y98 as the only ET donor with mutations of W60F and W60A; (ii) W60 as the only ET donor with Y98F, Y98A, Y98H and Y98R; (iii) two similar ET donors of Y60/Y98 and W60/W98 with mutations of W60Y and Y98W, respectively; and (iv) both W60 and Y98 as the ET donors with the wild type and G61A, G61V, D95N mutants. By the extensive characterization of these ET dynamics for 12 proteins, we observed ultrafast nonequilibrium ET dynamics for both forward (FET) and backward (BET), modulated by different redox potentials. More significantly, we directly observed the vibrationally involved BET process and subsequent vibrational cooling dynamics of the product(s) at the active site of the protein.

Table 1.

Time scales of ET and subsequent cooling dynamics and related energies.^a

Case	Mutants	Donor	τFETb	τFETc	τBETd	τcooling	Em,7(FMN/FMNH)e	ΔG∘FETf	ΔG∘BETf
(i)	W60Ag	Y98	0.34	0.49	0.95	3.7–4.0	−0.16	−0.66	−1.84
	W60F	Y98	0.30	0.49	0.95	3.7–4.0
(ii)	Y98F	W60	0.26	0.37	1.61	3.3–3.8	−0.15	−0.99	−1.50
	Y98A	W60	0.25	0.37	1.00	2.5–3.5	−0.19	−0.95	−1.54
	Y98H	W60	0.20	0.32	1.43	3.2–4.2	−0.19	−0.96	−1.54
	Y98R	W60	0.19	0.32	1.10	2.6–3.6	−0.17	−0.98	−1.51
(iii)	W60Y	Y60, Y98	0.26	0.46	0.95	3.7–4.0
	Y98W	W60, W98			1.10	3.0–4.0	−0.15
(iv)	WT	W60, Y98	0.16	0.18	1.56, 0.95	3.0–4.3	−0.15
	G61V	W60, Y98	0.15	0.16	1.23, 0.95	3.0–3.5	−0.33
	G61A	W60, Y98	0.15	0.16	1.23, 0.95	3.0–3.5	−0.26
	D95N	W60, Y98	0.14	0.18	1.51, 0.95	3.0–4.0	−0.15

^aAll times in picoseconds and the energies in eV.

^bFrom fluorescence detection.

^cFrom transient absorption detection.

^dFor the dual donors of W and Y, the first column for W and the second for Y.

^eThese data from refs. 29, 31, 41, and 42. E(FMN/FMN⁻) could be calculated from ${\mathrm{E}}_{\mathrm{m},7}(\text{FMN}/\text{FMNH})+0.06\times log[\frac{1+[H^{+}]/K_{ox}}{1+[H^{+}]/K_{\mathit{red}}}]$ with the known values of [H⁺]=10⁻⁷, K_ox=1 and K_red=10^−10.5.

^fThe free energies are not calculated for the dual donors and could be separately estimated for W and Y as for the single donor with E_m,7.

^gThere is another minor component with 1.82 ps and 4% amplitude in the FET by the fluorescence detection.

MATERIALS AND METHODS

Protein Preparation

The purification procedures for flavodoxin and the mutants have been well established.^31,32 For fs-resolved experiments, the flavodoxin was prepared at the concentration of 60–250 μM in 50 mM phosphate buffer solution at pH 7. To remove the unbound FMN molecule, the sample was passed through a mini-desalting column (Sephadex G-25 media) right before use. During laser experiments, the sample was in aerobic condition at room temperature (22 °C).

Femtosecond Methods

Fs-resolved measurements were carried out using the fluorescence upconversion and transient absorption methods. The experimental layout has been described elsewhere.³³ Briefly, the pump pulse was set at 400 nm with the energy attenuated to ≈100 nJ before being focused into the sample cell. For the fluorescence upconversion experiments, the fluorescence emission at 538 nm was gated by another 800-nm laser beam in a 0.5-mm thick β-barium borate crystal (BBO, type). For the transient absorption experiments, the probe pulse at the desired wavelengths between 410 and 800 nm were generated from two optical parametric amplifiers (OPA-800C and TOPAS, Spectra-Physics). The instrument response is 250–300 fs and 160–300 fs for the fluorescence and transient-absorption detection, respectively. All experiments were taken at the magic angle (54.7°). To avoid heating and photobleaching, the sample was kept in stirring quartz cells with 1 or 5 mm thickness during laser irradiation.

RESULTS AND DISCUSSION

Reaction Schemes and Probing Strategies

For 12 proteins with 4 different groups of ET reactions, we can summarize the kinetics as follows:

[1] [2]

For cases (i) and (ii), the proteins only contain one ET donor and thus the ET dynamics correspond to the reaction [1] and [2], respectively. For cases (iii) and (iv), both reactions [1] and [2] occur, but for the former the two donors are either the same Y or W in the reactions. In Figure 2, we show all the absorption spectra of various species involved in the reactions in the visible region. Figure 2B shows the special absorption spectrum of the mutant Y98W and the usual absorption profile of the wild-type (WT) protein and other 10 mutants with the peaks of S₁←S₀ (450 nm) and S₂←S₀ (380 nm) transitions of FMN in the binding pocket. For Y98W, the absorption extends to more than 650 nm and this tail suggests the charge-transfer character due to electronic delocalization of FMN through stacking with the two aromatic tryptophan residues of W60 and W98.³¹ If the high-frequency vibrational modes are involved in the BET reactions and thus enhance the BET rates, the formed hot FMN^† product could be detected at the red edge of the ground-state absorption (insert in Fig. 2B). To directly observe such the vibrational effect, i.e., the formation of FMN^† product, the ET reaction rates, k₁ and k₃ in [1] or k₂ and k₄ in [2], must be faster than the cooling relaxation (k₅). Otherwise, the FMN^† could not be accumulated due to the rate-determining slow ET dynamics. Thus, we need to resolve all elementary steps by monitoring the dynamics of each intermediate involved in the reactions. Figure 2A shows the normalized absorption spectra of three intermediate radicals and excited FMN.^34–36 We can tune different wavelengths to follow the whole dynamics of reactants (FMN^*), intermediates (FMN^•−, W⁺ and Y⁺) and products (FMN^† and FMN) and thus map out the entire redox cycle and hot vibrational cooling, if any, in the protein.

Figure 2.

(A) Normalized (transient) absorption spectra of protein-bound FMN^* (red) and anionic flavin semiquinone (dark yellow), and free cationic radicals of indole (blue) and phenol (dark purple) in organic solvents. The profile of phenol radical cation was extracted by direct subtraction of long-lived phenoxyl radical absorption from the mixed spectrum (34). The various probe wavelengths for the transient absorption experiments are marked on the top. (B) Steady-state absorption spectra of wild-type (blue) and mutant Y98W (red) flavodoxin. The pump wavelength at 400 nm and multiple probe wavelengths (inset) for probing the vibrationally hot FMN molecules are marked by a series of arrows.

Femtosecond Charge Separation, Frozen Active-Site Configuration and Critical Free Energies

Figure 3 shows the typical fluorescence transients gated at the emission peak (538 nm) for the wild type and several mutants. Except for the mutant of Y98W, for all other 11 proteins the fluorescence transients show the dynamics in 135–340 fs. We did not observe any fluorescence dynamics for Y98W due to the nature of charge-transfer excitation. Except for W60A, all other 10 fluorescence transients show a single-exponential decay behavior. Specifically, for case (i), we obtained 302 fs for W60F; for case (ii), we got 258, 247, 204, and 193 fs for Y98F, Y98A, Y98H and Y98R, respectively; for case (iii), we had 258 fs for W60Y; and for case (iv), we found 157, 154, 149 and 135 fs for the wild type, G61V, G61A, and D95N, respectively. For W60A, we can fit a double-exponential decay in 340 fs with the amplitude of 96% and 1.82 ps (4%) or a stretched exponential decay with 345 fs and a stretched parameter of 0.9.^3–5 These results are listed in Table 1. We noticed that the W60A mutant is less stable than W60F because the aromatic residue W60 behaves as a critical gate for entering of FMN into the binding pocket.³⁷ Thus, the dominant 340-fs component represents the major ET dynamics and the minor 1.82-ps component probably reflects a loose, unstable configuration. The observation of the single-exponential decay is in contrast to the reported double-exponential decay dynamics for mutants studied by the Mataga group,²⁸ possibly a question of the freshness of samples used in their reports. Nevertheless, our single exponential decay dynamics of the mutants are consistent with the major components reported by the Mataga group. For example, our results of 157 fs in the WT, 258 fs in the Y98F and 302 fs in the W60F are similar to the reported values of 158 fs (92%), 245 fs (85%), and 322 fs (83%) for the same mutants,²⁸ respectively.

Figure 3.

Normalized femtosecond-resolved fluorescence transients of FMN^* in flavodoxin mutants gated at 538 nm. Note that in some transients a small long-lifetime component (1–2%) due to the presence of free FMN molecule was removed for clarity.

The obtained dynamics of charge separation are ultrafast in 135–340 fs and our previously reported active-site relaxation takes longer than 1 ps.²³ Thus, upon initial excitation the electron transfer is much faster than the local environment relaxation. The ET reaction occurs in a nonequilibrium configuration. On the time scale of electron transfer, the active site appears nearly frozen. The concept of reorganization energy (λ) is not be applicable here and the forward ET reaction mainly depends on the reduction potentials (mainly enthalpy difference) of the donor and acceptor, the driving force of the reaction, similar to the ET reaction in gas phase that is determined by the donor ionization energy and acceptor electron affinity. The reduction potential of W^•+/W is about +1.15 V vs. NHE³⁸ and Y^•+/Y about +1.47 V vs. NHE^39–42 at pH 7. The reduction potentials of FMN/FMN^•− can be obtained from those of FMN/FMNH^•39 for the different mutants^{29,31,43–45} in a range of −(0.35–0.54) V vs. NHE. Thus, the corresponding FET and BET free energies are also listed in Table 1. Clearly, for case (i) of Y donor only and case (ii) of W donor only, the FET is faster for case (ii), 0.2–0.25 ps (1/k₂), than for case (i), 0.3–0.34 ps (1/k₁), mainly due to the difference of the reduction potential energies of two Y and W donors. For case (iii), the dual Y donors (0.258 ps) are faster than the single Y donor in case (i). For case (iv), the dual donors of Y and W clearly accelerate the FET reaction, resulting in even faster dynamics in about 0.15 ps. We did not observe a measurable signal for the dual W donors (W60/W98) with our time resolution and probably the FET dynamics is super ultrafast (<50 fs) or the system upon excitation is already in the charge-separated state due to the charge-transfer character in the ground- or excited-state.³¹

Overall, the observed FET reaction times follow the trend of their driving-force changes, but for the fixed donor(s) the FET dynamics have minor variations with mutations in case (iv). For the mutants of G61A and G61V, the large changes of the reduction potentials for FMN/FMNH^• (E_m,7 in Table 1) are due to the second-step protonation process of one-electron reduced FMN^•− to FMNH^•, the key effect of the G61 mutation,²⁹ and thus the charge separation for the first-step reduction of FMN/FMN^•− would not be significantly affected. Thus, the change of the FET dynamics is mainly from the difference (320 mV) of the reduction potentials of W and Y. The reduction of FMN^* to FMN^•− depends on the total available aromatic residues (W and/or Y) in proximity. The Mataga group has recently discussed in detail the ultrafast FET mechanism in flavoproteins using the average distance instead of the edge-to-edge distance between the donor and acceptor.^46,47 However, the usage of solvent reorganization energy (λ) in their discussion is probably not appropriate as we pointed out early, due to the much faster ET dynamics than the active-site relaxation, i.e., the nearly frozen active site on the femtosecond time scale.

Ultrafast Charge Recombination, Hot Product Formation and Subsequent Vibrational Cooling

Figures 4–6 show the typical absorption transients for four ET cases with a wide range of detection wavelengths from 800 to 400 nm to monitor the dynamics from the reactants (FMN^*), to charge-separated intermediates (FMN^•−, Y^•+/W^•+), and to final products (FMN^† and FMN). Specifically, Fig. 4 shows the complete ET dynamics of Y98 (W60F mutant) in case (i) and Fig. 4A shows the entire evolution of the transients. At 800 and 740 nm, both transients are the same and we only detected the excited reactant FMN^*. The dynamics follow a single exponential decay in 0.49 ps, slower than that from the fluorescence detection (0.3 ps), which seems a general phenomenon and was observed in many other systems.⁴⁸ The absorption method may sense more donor-acceptor configurations than the fluorescence detection. When the probe wavelength was tuned to the blue side, the dynamics clearly become slower and the time scales get longer (Fig. 4A). In particular, when the dynamics were probed at the red side of the ground-state absorption from 540 to 500 nm, the transients become slower and slower. Clearly, if we only detected the excited FMN^* and intermediates FMN^•−/Y^•+, the dynamics should not change dramatically even though the percentage of each species could change. Thus, the only possibility is that we observed the dynamics of hot ground state FMN^† after charge recombination (Fig. 4B); such vibrationally excited states have absorption move to the red side of the original ground-state absorption (Fig. 2B). Therefore, we observed a series of hot ground-state relaxation and the dynamics change from faster to slower as the vibrational cooling from higher to lower energy states. At 480 nm, we observed the ground-state recovery and the overall dynamics becomes faster again due to the cancellation of slow cooling contributions by ground-state formation signals (Fig. 4C). At 410 nm, we observed the positive signal again from the contributions of the excited FMN^* and the intermediates of FMN^•− and Y^•+. Thus, the global fitting of all these transients gave 0.95 ps (k₃⁻¹) for charge recombination and 3.7–4.0 ps (k₅⁻¹) for vibrational relaxation of hot ground states.

Figure 4.

Normalized femtosecond-resolved absorption transients of mutant W60F flavodoxin probed from 800 to 410 nm for the ET reaction [1] with only donor Y98. Inset A shows the entire evolution of the dynamics with the probe wavelengths. Insets B and C show the deconvolution of the transients into various species probed at 510 and 480 nm, respectively. Note that at 480 nm the transient becomes faster again as a result of cancellation of ground-state FMN recovery (light-orange line) by the positive signal of FMN^† (green line).

Figure 6.

Normalized femtosecond-resolved absorption transients of wild-type flavodoxin (Upper) for the ET reactions [1] and [2] with the dual donors W60 and Y98, and Y98W mutant (Lower) with the dual donors of W60 and W98 (reaction [2]) probed with a series of wavelengths from 740 to 410 nm, showing the striking evolution when the probe wavelength was tuned to the blue side. The black dashed line in the lower panel shows a simulated 100-fs decay for comparison with 1.1 ps-decay in 740-nm transient. The negative signals probed at 500 nm are not completely shown in this display.

The observation of ground-state vibrational excitation after BET in protein flavodoxin is significant. The vibrationally coupled ET through high-frequency modes was observed in BET of several chemical systems.^18,19,49 For flavodoxin studied here, the BET dynamics (<1 ps) is even faster than the active-site relaxation (≥ 1 ps) and thus the active-site motions are only partially coupled with the BET dynamics. With the van der Waals distance of 3.34 Å between the donor tyrosine and acceptor isoalloxazine ring,²² the formed ion pair FMN^•−···Y^•+ would proceed to vibrational motions as observed in gas-phase benzene···iodine charge-transfer complex.⁵⁰ The ultrafast BET, faster than the active-site relaxation, must efficiently channel ionic energy into high-frequency vibrational energy of ground-state FMN (and/or Y), leading to the formation of hot FMN^†, a molecular picture similar to hot ground-state benzene formation in BET of benzene···iodine charge-transfer complex.⁵⁰ Such ultrafast ET reactions in flavodoxin bear some similarity with gas-phase bimolecular charge-transfer reactions.^50,51 If we consider the absorption shift from 500 to 540 nm due to hot ground-state formation, a total vibrational energy of ~1500 cm⁻¹ is obtained for hot FMN^† with a few vibrational quantum numbers (ν=1–3), given some high-frequency vibrational modes of FMN in 500–1500 cm⁻¹.⁵² The cooling dynamics occur in a few picoseconds, slower than all ET dynamics, and thus enable us to catch the hot FMN^† formation and subsequent cooling dynamics. The time scale in 3–4 picoseconds is similar to that of the FMN^† cooling in polar solvents,^53,54 implying one possibility that the vibrational energy may flow into the water molecules around the entrance of the active site. Such cooling dynamics is also mixed with the active-site relaxation induced by fast charge recombination. The mutant W60A exhibits the similar results as for W60F; see in Table 1.

Figure 5 shows the absorption transients of mutant W60 (Y98F) and a similar dynamic pattern was observed. At longer probing wavelengths of 800-580 nm (insets of Figs. 5A and 5B), we detected the excited FMN^* dynamics in 0.37 ps (k₂⁻¹) as well as the formation and decay dynamics of intermediate W^•+ in 0.37 and 1.4 ps, respectively. The transient at 580 nm shows the dominant contribution of the intermediate state due to the strong absorption of tryptophan cation W^•+.⁴⁰ Similarly, we observed the vibrational cooling dynamics at shorter wavelengths than 540 nm in 3.3–3.8 ps, similar to that observed in mutant Y98 (W60F). Also, the ground-state recovery transient of the product FMN at 480 nm becomes faster, again due to the cancellation of contributions of the cooling dynamics and relaxed ground-state formation dynamics. For the different other mutants in Table 1, we obtained the similar time scales with minor changes. Typically, the charge recombination of photoinduced charge separation is in the Marcus inverted region because the free energy of charge recombination is usually larger than the reorganization energy. As shown in Table 1, the BET reaction for W60 has a free energy, −(1.50–1.54) eV, smaller than that of Y98 (−1.84 eV). But, we observed the slower BET dynamics for W60 and faster for Y98, indicating an unusual behavior given that the reorganization energy for the ET reaction of anionic flavin radical in flavoproteins is usually smaller than 1.5 eV.^55,56 This phenomenon must result from the faster BET and slower active-site relaxation.

Figure 5.

Normalized femtosecond-resolved absorption transients of mutant Y98F flavodoxin probed from 800 to 410 nm for the ET reaction [2] with only donor W60. Insets A–D show the deconvolution of the transients into various species probed at 800, 580, 510 and 480 nm, respectively, showing the step-by-step detection of the initial reactant (FMN^*), intermediates (FMN^•− and W^•+) and final products (FMN^† and FMN). Note that at 480 nm the transient becomes faster again as in Fig. 4.

Figure 6 show the ET reactions with dual donors of Y and W and overall a similar pattern was observed with the striking vibrational cooling dynamics again. As seen in reactions [1] and [2], for the wild type, we observed the ultrafast FET dynamics (k₁+k₂) mainly at 740 and 700 nm. At 580 nm, we observed the dominant W^•+ contribution and thus obtained the decay dynamics of BET in 1.56 ps (k₄⁻¹) by the reaction [2], similar to the mutant W60 (Y98F) in Fig. 5. The overall fitting gave the BET dynamics of reaction [1] in about 0.95 ps for FMN^•−+Y^•+ charge recombination, similar to the single Y-donor BET dynamics. Although we observed a total FET rate from both W and Y, we resolved the individual charge-recombination dynamics of Y and W (k₃ and k₄). This observation is important and we can compare the obtained BET dynamics from two donors with the results from the single donors. Here, we observed the similar BET dynamics (k₃ or k₄) for the same donor even for the different initial FET reactions. The cooling dynamics occur in 3.0–4.3 ps. For the mutant Y98W, because the absorption shows the charge-transfer character (Fig. 2) we did not observe any FET dynamics as shown in Fig. 6 (bottom panel). At 800 nm, the signal is negligible and is 1–2 orders of magnitude weaker than the mutant W60 (Y98F) in Fig. 5. At 740 nm, the signal is also very weak and represents the only W^•+ dynamics of charge recombination in 1.1 ps (k₄⁻¹); see the dashed line of a simulated 100-fs decay for comparison. From 630 nm, we started to detect the cooling dynamics (Fig. 2) after charge recombination. The transient at 580 nm appears faster than that at 630 nm due to the dominant W^•+ contribution in the former and actually the cooling dynamics is similar around 3 ps (k₅⁻¹) in both transients. The transients at the shorter wavelengths follow the similar pattern as other mutants and the cooling relaxation takes 3–4 ps. All other mutants with the dual donors also have the similar dynamics given in Table 1.

Photoinduced Redox Cycle, Reaction Time Scales, and Vibrational Coupling Generality

The photoinduced redox cycle at the active site of flavodoxin is given in Figure 7A. Four fundamental processes are involved in the redox cycle: Initial forward ET (τ_FET), subsequent back ET (τ_BET), following vibrational product cooling (τ_C), continuous active-site solvation (τ_S)²⁴ that was determined in our previous studies. The active-site relaxation evolves with the entire dynamic process from the initial excitation, to charge separation, to charge recombination and finally to vibrational cooling. The complete solvation at the active site occurs on the multiple time scales in 1.0 ps (53%), 25 ps (26%) and 670 ps (21%), involving various motions at the active site including neighboring hydration water molecules.²⁴ The 1.0-ps dynamics mainly comes from the initial local relaxation of hydration water networks around the active site at the protein surface.^24,57 Thus, the four time scales involved are critical to our observation and the understanding of the ET molecular mechanism.

Figure 7.

(A) The complete photoinduced redox cycle in flavodoxin with all resolved elementary dynamics and their time scales. The active-site relaxation (not shown) is involved in all these processes on the picosecond time scales. (B) Schematic presentation of the three potential surfaces involved in the photoinduced redox cycle along two coordinates, intramolecular distortion and solvent reorganization. The three dots represent the equilibrium configurations on the three surfaces. The high-frequency ground-state vibrational excitation is not shown during the transition from the ionic state to the ground state.

For flavodoxin, τ_FET is ultrafast and the solvation appears frozen, a gas-phase type of bimolecular charge-transfer reactions. τ_BET is also very fast on the similar time scale of initial solvation. Thus, the BET is a nonequilibrium ET dynamics and theoretically could be treated by the extended Sumi-Marcus two-dimensional model (solvent and nuclear coordinates), leading to a nonexponential behavior typically in a stretched exponential decay. However, we observed the dominant single-exponential decay for BET (such as W⁺ decay at 580 nm), indicating that the coupling of BET with solvent relaxation is not severe yet. Due to the stacking nature of donor and acceptor and the hydrating water molecules not intercalating between the donor and acceptor, the charge-separated complex (FMN^•−··· Y^•+/W^•+) could have significant intermolecular vibrations and thus lead to high-frequency vibrational excitation of the products FMN^†···Y^†/W^†, resulting in very fast BET dynamics. However, if τ_FET and τ_BET are much longer than τ_S, we would also observe a single-exponential decay for both equilibrated ET dynamics with a relaxed actives-site configuration. If τ_FET or τ_BET is longer than τ_C, we would not observe the vibrational-cooling process because the rate-determining step is the longer ET dynamics and the hot products cannot be accumulated. Thus, the protein flavodoxin is an ideal model system for direct observation of the vibrationally accelerated BET dynamics with all resolved elementary processes in a redox cycle in the active-site environment of the protein.

With the 11 mutants and wild-type flavodoxin, we resolved all three elementary dynamics of FET, BET and vibrational cooling as well as active-site solvation dynamics. With the reported redox potentials of FMN in these mutants (Table 1), the ET dynamics mainly follow the redox trend of the donor W or Y but slightly change with the reduction potentials of FMN by mutations. For the stacked, compact FMN···Y/W pair, the ET reaction mainly occurs between the donor and acceptor. The environment has a main static effect with a minor dynamic influence. Figure 7B shows the schematic potential surfaces of three states along two coordinates of intramolecular distortion (q) and solvent reorganization (X). The initial FET mainly evolves along the intramolecular nuclear coordinate due to the relatively slow solvation. Thus, the observed slowdown of the FET in the crystal of a similar FMN-binding protein⁵⁶ is probably not due to the slow solvation in crystal⁵⁸ because the FET dynamics is much faster than the solvation process, even more in the crystal. After charge separation, the ionic complex continuously moves along the nuclear coordinate and meanwhile also evolves along the solvation coordinate to reach the minimum crossing seam to the ground-state surface with high-frequency vibrational excitation (not shown in Figure 7B) to conserve the total energy. The molecular picture revealed here should be applicable to many other flavoproteins due to the common structural configurations and interactions between aromatic residues and cofactor flavin. Thus, the vibrational excitation after charge recombination and subsequent cooling dynamics should be general to flavoproteins even though the cooling dynamics could be often buried in the rate-determining slow ET process.

CONCLUSION

We reported here our complete characterization of a photoinduced short-range cyclic electron transfer of oxidized flavin (FMN) in the protein flavodoxin. With femtosecond resolution, we can follow the entire redox reactions, forward and backward electron transfer, from the reactants, to all intermediates, and finally to products (Fig. 7A). The forward electron transfer occurs ultrafast with stacked aromatic residue(s) (tryptophan or tyrosine) in 130–350 fs, much faster than the active-site relaxation which is on the picosecond time scales. With 11 mutations at the active site, the electron-transfer dynamics have minor changes, a reaction highly localized between the donor and acceptor at a van der Waals distance, leading to a gas-phase type of bimolecular reactions confined in the active-site nanospace. The environment exerts an electrostatic effect on the reactions with a minor dynamic influence.

The charge-separated complex evolves along nuclear and solvent coordinates (Fig. 7B). Along the nuclear coordinate, the ionic donor-acceptor vibrations strongly couple with the donor/acceptor high-frequency modes. Along the solvent coordinate the ionic complex couples with the active-site motions of the protein and neighboring hydration water molecules (around the entrance of the active site). As a result, the hot ground-state product(s) was formed with accelerated charge recombination in 1–2 ps. The vibrationally excited product cools down in 3–4 ps at the active site, presumably channeling the energy to hydrating water molecules. The faster electron-transfer dynamics and slower cooling relaxations enable to directly observe such vibrational effect on speeding up the back electron-transfer reaction. The observed vibrational effect should be general to flavoproteins based on their similar configurations and interactions between the cofactor flavin and neighboring aromatic residues at the active sites and should be considered in charge-recombination reactions in flavoproteins.

ABBREVIATIONS

ET

electron transfer

FMN

flavin mononucleotide