Ultrafast Dynamics of Fatty Acid Photodecarboxylase in Anionic Semiquinone State

Abstract

Fatty acid photodecarboxylase is a newly identified blue-light driven photoenzyme that catalyzes decarboxylation of fatty acids. The catalytic reaction involves a transient anionic semiquinone of flavin cofactor (FAD^•−) as an intermediate, but photochemical properties of this anionic radical are largely unknown. Here, we have anaerobically produced the wild-type FAP in FAD^•− state and conducted femtosecond-resolved fluorescence and absorption measurements. We have observed the multiphasic deactivation dynamics of excited states on multiple timescales from a few picoseconds even to a few nanoseconds through conical intersections between various electronic states. Interestingly, the nanosecond component can only be observed from higher electronic excited states. Our results show the complexity of the energy landscapes of various excited states and rule out the occurrence of electron or proton transfer with nearby residue(s) in the active site.

TOC GRAPHICS

Fatty acid photodecarboxylase (FAP) is a blue-light driven photoenzyme that catalyzes decarboxylation of fatty acids.¹ Discovered in microalgae Chlorella variabilis, the enzyme converts cheap fatty biomass into hydrocarbon biofuel, which has aroused tremendous industrial and academic interests. Since the discovery of FAP in 2017, numerous studies have been reported to increase enzyme sustainability, improve production efficiency, broaden substrate scope, and obtain stereospecificity.^2–8 In recent mechanistic studies, spectroscopic observations show that the catalysis begins with a single electron transfer from the anionic substrate RCOO⁻ (R denotes the hydrocarbon chain) to the photoexcited flavin adenine dinucleotide (FAD) cofactor at the active site. Upon receiving the electron, FAD is converted to the anionic semiquinone FAD^•− state, which later returns an electron to decarboxylated R^• radical in hundreds of nanoseconds.^1,9,10,11 As the proposed FAD^•− radical plays such a significant role in the catalytic mechanism, the elucidation of its photochemical properties and dynamics benefits our understanding of FAP function, as well as flavoenzyme in general. So far, the anionic semiquinone FAD^•− in FAP has only been transiently formed and never been a subject to any investigation.

Among five flavin states under physiological conditions, the anionic radical FAD^•− is usually unstable and prone to protonation to FADH^• and further reduction to FADH⁻.^12–14 However, the glucose-methanal-choline oxidoreductase (GMC) superfamily is a rare exception. In many GMC enzymes, semi-reduced FAD^•− state can be acquired upon photoreduction and stabilized at the active site.^15–18 GMC enzymes catalyze the oxidation of a wide scope of substrates: a hydroxyl group is converted to corresponding aldehyde and in some cases subsequent carboxylic acid, at the sacrificial cost of O₂.¹⁹ The reaction involves directional transfer of two electrons and two protons from the substrate to FAD in oxidized state, which is hereby fully reduced to FADH₂, then from FADH₂ to O₂. Such functionality is independent of light.

Despite vast differences in function and mechanism, sequence and structural alignments determine that FAP belongs to the GMC superfamily.^1,20 Crystal structures of Chlorella variabilis FAP with stearic acid substrate and Aspergillus niger glucose oxidase (GOX) without substrate are aligned for comparison (Figure 1A).^10,21 Both proteins fold into an N-terminal FAD-binding domain and a C-terminal substrate-binding domain, which is characteristic in GMC superfamily. As GOX oxidizes glucose, a highly hydrophilic small molecule, the active site is fully solvated. In the absence of ligand, water molecules form a H-bonding network connecting to outside solvent. On the other hand, because fatty acids have lengthy hydrocarbon tails, FAP uniquely possesses several extra hydrophobic helices that bind the fatty tails, resulting in a deeper and presumably “dryer or structured water chain” binding pocket. When substrate binds, the hydrophobic tunnel is water-free. At the active site of GOX, a water molecule is trapped between H516, H559 and the N5-atom of FAD (Figure 1B). The histidine duo is highly conserved in GMC superfamily, vital for enzyme stability and functionality.^{19, 21–23} The positive charge on H559 has also been proposed to stabilize the negative charge on FAD^•− upon its formation.^24–26 At the active site of substrate-bound FAP, the carboxylic group of stearic acid orients towards the FAD cofactor (Figure 1B). Notably, the conserved histidine duo is absent in FAP. Instead, the substrate is flanked by R451, another positively charged residue. Previous studies conclude that both this positive charge and its position are critical to FAP activity.¹⁰ Similar to GOX, a water molecule simultaneously H-bonds with R451, substrate carboxylic group and N5-atom of FAD. Despite theoretical computations performed,¹⁰ the role of this water molecule is not fully understood. In summary, FAP generally resembles GOX in terms of overall topology and active-site structure, yet differences in detail are remarkable. Unfortunately, no structure of FAP without substrate or FAP in FAD^•− state has been published so far. It is still unknown how protein side chains and active-site solvent would react to the formation of the stable FAD^•− in FAP.

Figure 1.

(A) Crystal structures of Chlorella variabilis fatty acid photodecarboxylase (FAP, green; PDB code: 6YRU) with stearic acid substrate, overlayed with Aspergillus niger glucose oxidase (GOX, gray; PDB code: 1CF3) without substrate. Both proteins fold into an N-terminal domain binding the FAD cofactor (yellow) and a C-terminal domain binding the substrate (cyan). Compared to GOX, FAP uniquely possesses extra hydrophobic helices that bind the hydrocarbon tail of the substrate; these helices are highlighted. (B) Close-up look at the active site. Solvent molecules of FAP (red) and GOX (pink) within 4 Å of the substrate are displayed as spheres. Distances from a trapped water to corresponding counterparts in both proteins are measured in Å.

Since many GMC enzymes can produce and stabilize FAD^•− state, we attempted previous photoreduction conditions on FAP and fortunately succeeded in converting the cofactor from the yellow FAD state to the pink FAD^•− state.¹⁵ The photoreduction is performed anaerobically without substrate. Continuous overnight blue light illumination fails to further convert FAD^•− to any other state, showing that FAD^•− is stabilized at the active site, presumably by the positively charged R451. When exposed to oxygen, FAD^•− is gradually re-oxidized to the yellow state, which shows nonpolar O₂ can slowly diffuse through the hydrophobic tunnel and arrive at the cofactor. We hereby show the FAP absorption spectra in both FAD and FAD^•− states (Figure 2A). Compared to GOX, the oxidized-state spectrum is red shifted by 30 nm, extending to 550 nm, whereas the anionic semiquinone-state spectrum is red shifted by 50 nm, extending to 600 nm. Such a red shift is attributed to different local electrostatic environments.²⁷ In addition, the FAD^•− spectrum has a shoulder at 550 nm; this feature is not seen in many flavoproteins but commonly observed in GMC family.^26,28 The excited oxidized state FAD* has a lifetime of 4.7 nanoseconds (ns) as usual with a strong fluorescence emission (Fig. S1), indicating no aromatic residues near FAD as confirmed by the structure, otherwise we would observe fluorescence quenching by ultrafast electron-transfer reactions.^13,14,29 For excited FAD^•−*, the emission is much weaker than that of oxidized FAD* (Figure 2B), indicating that FAD^•− has a short lifetime, i.e., FAD^•−* is mostly deactivated nonradiatively. Surprisingly, the emission peaks are red shifted with longer excitation wavelengths and the action spectrum of FAD^•− also deviates from the absorption spectrum, implying complication in the deactivation processes of FAD^•−*, i.e., the higher excited states are not internally back to the first excited state and thus probably deactivate via conical intersection(s) (CI) nonradiatively.³⁰

Figure 2.

(A) Steady-state absorption spectra of FAP in both oxidized FAD state (black) and anionic semiquinone FAD^•− state (red). (B) Steady-state fluorescent emission spectra of FAP in the FAD^•− state. Insets are absorption spectra overlayed with the fluorescence peak intensities at given excitation wavelengths (action spectrum).

To identify the cause of excited-state nonradiative decay, we first performed femtosecond-resolved fluorescence up-conversion experiments (Figure 3). We first excited the sample at 400 nm, a higher electronic state (doublet D₆).^27,31 The fluorescence transients gated at several wavelengths (460–680 nm) of the emission spectrum show multiphase exponential decay (Table S1). Clearly, the transients slow down from 460 nm to 540 nm but get faster from 540 nm to 680 nm, a typical solvation process with faster decays at the blue and red sides of the emission; the red-side decay implies the spectral shrinking during solvation. These processes mix with fast deactivation. They are globally deconvoluted with the dynamics in ~1.4 ps, 54–86 ps and a few ns. The nanosecond component is further characterized with time-correlated single photon counting and we observed the longtime decay in 1.2 ns and 4.7 ns (Figure S2A and Table S2). Considering oxidized state FAD* has a lifetime of 4.7 ns and the fitted amplitudes peak at 550 nm like the oxidized one (Figure S2B), we assign the 4.7-ns phase to residual oxidized-state molecules in the sample (a few percent at most), and the other components (54–86 ps and 1.2 ns) unambiguously to FAD^•− (Figure S1C). In parallel, we also performed fluorescence up-conversion at 550-nm excitation of the electronic D₃ state^27,31 where oxidized FAD does not absorb. We observed the dynamics of fluorescence emission (600–660 nm) dominantly in ~1.6 ps with a minor component in 72–96 ps (Table S3). The long component of 1.2 ns is absent. Clearly, the excitation-dependent emission spectra and the ultrafast multiphasic FAD^•−* dynamics suggest the presence of conical intersections (CIs) as observed in many large organic molecules.^{12,27,31–33} The multiphasic feature of FAD^•−* dynamics indicates diverse pathways on the potential energy surfaces (PES) to reach the CIs.

Figure 3.

Typical femtosecond-resolved fluorescence transients of FAP in excited FAD^•− state at excitation at 400 nm (A) and 550 nm (B), respectively. Inset shows the transients at early delay times.

To implement our understanding of FAD^•−* decay dynamics, we performed a series of ultrafast transient-absorption experiments to probe various intermediates. At 800 nm, where none of the flavin ground states absorbs, the transient at 400-nm excitation exhibits a triple-exponential decay in 2.6 ps, 37 ps and 3.2 ns (Figure 4A), consistent to the pattern observed in the fluorescence decay. Thus, the signal probed at 800 nm is mainly attributed to FAD^•−* (Figure S2D). The slightly different time constants may result from different accessible regions of PES for the two methods. Similarly, at 550-nm excitation, the transient only showed two decay components of 3.7 ps and 37 ps, whereas the nanosecond component is absent, also agreeing with the fluorescence detection. These results indicate that a population with a nanosecond lifetime can only be produced when FAD^•−* is excited to a higher D₆, not D₃, electronic state. This fractional population, trapped in a valley on the PES, cannot get access to a CI, thus must deactivate in nanoseconds with fluorescence emission.

Figure 4.

Excitation-dependent femtosecond-resolved absorption transients of FAP in FAD^•− state, excited and probed at respective labelled wavelengths (A-C). Panels on the right show the absorption transients at short delay times.

We then tuned to shorter wavelengths and observed the similar pattern: multiphasic dynamics from a few picoseconds to several nanoseconds at 400-nm D₆ excitation and 2-phasic behaviors from a few to tens of picoseconds for 500/550-nm D₃ excitation (Tables S4–S6, Figure S3). Specifically, for the probing wavelength at 550 nm, the absorption transient at 400-nm excitation is positive, showing an initial faster decay in 1.3 ps, then a rise in 36 ps and finally decays in 310 ps and 3.1 ns, indicating an intermediate formation. Similarly, at 500-nm excitation we observed a similar initial decay in 1.7 ps and then a rise in 35 ps, but the total signal is negative due to bleaching of the ground state (Figure 4B). As we further decreased probing wavelength to 450 nm, we observed a rise component in 1.1–2.1 ps and then a decay in 36–38 ps (Figure 4C). Interestingly, even at 400-nm excitation, we did not observe any long components at 450–480 nm. On the other side, we probed the dynamics from 520 to 660 nm at 400-nm excitation (Figure 5A) and mostly observed long components from hundreds of picoseconds to a few nanoseconds (Table S4). These behaviors reflect the excited state evolving along different regions of PES, probed by various probing wavelengths, resulting in the observation of various combinations of rise and decay from a few picoseconds to tens of picoseconds for 500/550-nm D₃ excitation or even extending to a few nanoseconds for 400-nm D₆ excitation, displaying the complexity of energy landscapes. Finally, we also probed the dynamics in the UV range, especially in the vicinity of 351 nm at 550-nm excitation (Tables S6 and S7). According to our previous studies,¹³ the contributions of the excited-state positive signal and the ground-state negative bleaching are cancelled out around 350 nm and the resulting signal is dominantly from the intermediates. Figure 5B showed the changes from the dominant negative signal to the positive one (Table S7), very sensitive to the probe wavelengths from 351.4 nm to 350 nm. The dynamics initially decay very fast in 1.1–1.3 ps, subsequently rise in 4.9–8.4 ps and then decay again in ~30 ps, further showing the complexity of excited-state deactivation.

Figure 5.

Absorption transients of FAP in FAD^•− state, (A) excited at 400 nm and probed at 520–570 nm and (B) excited at 550 nm, probed around 351 nm, where dramatic changes in signal trends are observed.

Combining all experimental analyses, we proposed the following model for FAD^•−* deactivation in FAP (Figure 6). Upon 500/550-nm excitation, the ground-state (D₀) FAD^•− radicals are pumped to the doublet excited state (D₃),^27,31 and then bifurcate into two pathways that relax via distinct routes on the D₃ potential energy surface, until they finally reach a CI with D_n (n=2–0) in 1–2 ps and 50–95 ps, as shown in the fluorescence results. The former can be a direct entry into the CI and the latter needs to overcome a barrier to reach the CI. For simplicity, in Figure 6, we assume the CI of D₃ directly interacting with D₀. Upon 400-nm excitation, the molecules are excited to the higher excited state (D₆)^27,31 which quickly drops back to D_n (n=5–0) via various CIs. Similarly, we assume a CI with D₃ for simplicity. However, this reaction bifurcates again, with a small fraction reaching a potential valley previously inaccessible with direct excitation to D_n (n=5–0). This trapped fraction decays to the ground state in nanoseconds with noticeable emission. Such the complexity of the PES results in a wide range of observation of deactivation dynamics probed at various wavelengths by the transient-absorption detection.

Figure 6.

Schematic potential energy surface of FAP in FAD^•− state. Decay pathways of 400 nm-excited and 550-nm excited radicals are marked in arrows. When directly excited to D₃, the wave packet takes multiple routes to the conical intersections (CIs) of D_n (n=2–0), resulting in multiple ps decays finally to D₀. However, when directly excited to D₆, a fraction of the wave packet reaches a previously inaccessible potential well, represented by a unique ns decay component, and the majority passes through various CIs of D_n (n=5–0) and finally back to the ground state.

In the proposed catalytic mechanism of FAP, it takes 100 ns for intermediate FAD^•− to return an electron to R^•.^1,9,10 Our study provides a preliminary answer to what would happen if FAD^•− gets excited by another photon during such a slow process. Due to the various conical intersections of D₃ with the lower levels, the majority of excited FAD^•−* at D₃ excitation deactivates in less than 100 ps, which reduces the likelihood of undesired radical reactions. Still, at D₆ excitation the biological relevance of the minor nanosecond population is unclear, but it is still much shorter than 100 ns in catalysis.^1,10 Thus, the radical FAD^•− is overall “safer” within 100 ns, even with light excitation during the catalytic reaction. Nonetheless, C. variabilis lives 40 meters deep in coastal ocean water, where UV light does not penetrate,³⁴ and excitation of FAD^•− to D₆ or higher electronic states is unlikely in nature. In turn, the ultrafast deactivation of excited FAD^•− is necessity through evolution to ensure the long waiting time of FAD^•− in FAP during the catalysis to finally donate one electron back to the ground state.

During our preparation of this manuscript, we came across a recently published study on the FAD^•− state in several GMC enzymes, including GOX.²⁶ The recent work proposed that in about 100 fs, FAD^•−* ejects an electron to a positively charged nearby H559 upon photoexcitation with the FAD^•− itself temporarily oxidized to FAD. The electron returns in 20 ps to finish up a futile photocycle. The model is supported by absence of FAD^•−* fluorescent emission and resemblance of 20-ps decay associate spectrum (DAS) to the difference absorption spectrum between FAD^•− and FAD, as well as other experimental and computational supplements. Although GOX and FAP both belong to the GMC superfamily, the ultrafast charge-separation model clearly does not apply to FAP, as we observed the multiphasic fluorescence decay and transient-absorption dynamics above 600 nm, where neither FAD nor FAD^•− absorbs. Thus, FAD^•−* is evidently not quenched within 100 fs in FAP. Also, our experimental results do not support occurrence of electron transfer. Although their DFT calculations supported HisH₂⁺ and ArgH⁺ having a relatively high and close electron affinity,^26,35 we examined the structures (Figure 1) and the FAD^•− can also be at a distance of 5.8 Å with HisH₂⁺ in GOX and 5.6 Å with R451 in FAP, also both separated by a structured water molecule. It can be the structured water molecule to screen the positively charged R451 as well as the negative FAD^•− and to block an electron transfer in FAP, if any. Thus, the FAP is unique in the GMC superfamily. The absence of an electron transfer from FAD^•− to the positive R451 in FAP or the presence of the electron transfer to H559 in GOX deserves more investigation.