Ultrafast Dynamics and Catalytic Mechanism of Fatty Acid Photodecarboxylase

Abstract

Fatty acid photodecarboxylase is a newly discovered flavin photoenzyme that converts a carboxylic acid into a hydrocarbon and a carbon dioxide molecule through decarboxylation. The enzymatic reactions are poorly understood. In this study, we carefully characterized its dynamic evolution with femtosecond spectroscopy. We observed initial electron transfer from the substrate to the flavin cofactor in 347 ps with a stretched dynamic behavior and subsequently captured the critical carbonyloxy radical. The dominant process following this step was decarboxylation in 5.8 ns to form an alkyl radical and a carbon dioxide molecule. We further identified the absorption bands of two carbonyloxy and alkyl radical intermediates. The overall enzymatic quantum efficiency determined by our obtained timescales is 0.81, consistent with the steady‐state value. The results are essential to the elucidation of the enzyme mechanism and catalytic photocycle, providing a molecular basis for potential design of flavin‐based artificial photoenzymes.

Fatty acid photodecarboxylase converts a carboxylic acid into a hydrocarbon and a carbon dioxide molecule through decarboxylation. Two critical carbonyloxy and alkyl radical intermediates in the catalytic reactions were captured by deep UV detection, and their dynamics were determined with actual reaction timescales, revealing the molecular mechanism of the enzymatic reaction and mapping out the catalytic photocycle of this efficient photoenzyme.

Introduction

Fatty acid photodecarboxylase (FAP) is a light‐driven photoenzyme discovered recently in microalgae Chlorella variabilis.[ ¹ , ² ] Under blue light, FAP catalyzes the decarboxylation of fatty acids, producing CO₂ and alkane or alkene products, which can be utilized to efficiently convert cheap microbial biomass into affordable biofuel.[ ³ , ⁴ ] FAP has recently been biochemically engineered for enhanced catalytic efficiency, ^[5] broadened substrate selectivity,[ ⁶ , ⁷ ] and integration into catalytic cascades,[ ⁸ , ⁹ ] and its industrial application looks promising. Moreover, after DNA photolyase ^[10] and protochlorophyllide oxidoreductase, ^[11] FAP is the latest discovered enyzme of the only three natural photoenzymes. Unlike the other two, owing to the recent discovery of FAP, the light‐driven mechanism of this enzyme is poorly understood.

FAP belongs to the glucose‐methanol‐choline oxidoreductase (GMC) family.[ ¹² , ¹³ ] Contrary to many GMC enzymes, since FAP cleaves a carbon‐carbon bond, the electron flow in the reaction is cyclic and consequently unlike GMC no oxygen is required. Crystal structure (Figure 1)[ ¹ , ² ] reveals a hydrophobic tunnel connecting outside solvent and a relatively hydrophilic active site. The hydrocarbon tail of a stearic acid substrate binds in the tunnel, while the carboxylic group of the substrate dwells at the active site. Meanwhile, a flavin adenine dinucleotide (FAD) cofactor is also non‐covalently embedded at the active site, which serves as the light‐absorbing chromophore. The tricyclic ring of the cofactor faces directly toward the carboxylic group of the stearic acid substrate (shortest distance ≈3.7 Å), whereas the adenine side chain in a stretched configuration points in the opposite direction. The x‐ray structure clearly shows direct interactions between the cofactor and the substrate. In addition, the active site is surprisingly spacious. A few amino acid residues in the nearest vicinity between the substrate carboxylic group and FAD tricyclic ring include A171, R451 and N575, which are not conventionally considered for good electron or proton transfer. Therefore, the involvement of protein residues in the catalytic mechanism seems unlikely.

Figure 1.

Crystal structure of Chlorella variabilis fatty acid photodecarboxylase (pink), in complex with catalytic cofactor FAD (green) and substrate stearic acid (teal) (PDB ID: 6YRU ^[2] ). A close‐up look at the active site shows the carboxylic group of the substrate orients directly toward FAD. A water molecule connects a hydrogen bonding network involving FAD, R451 and the substrate. All numbers in the figure indicate distances measured in Å. Inset shows a kinetic scheme of proposed catalytic mechanism.

Previous studies have proposed the photo‐Kolbe mechanism for FAP catalysis.[ ¹ , ² , ¹⁴ ] In this mechanism, upon blue‐light illumination, photoexcited cofactor (FAD*) abstracts one electron from the anionic palmitic acid RCOO⁻ (R denotes the hydrocarbon tail C₁₅H₃₁), producing anionic semiquinone FAD⋅⁻ and a carbonyloxy radical RCOO⋅ (k ₁ in Figure 1 inset). The latter proceeds to decarboxylate, yielding an alkyl radical R⋅ and a CO₂ molecule (k ₃). Subsequently, the alkyl radical becomes the alkane product RH upon electron return from FAD⋅⁻ and proton transfer from an unknown donor (k ₄). As oxidized FAD and semiquinone FAD⋅⁻ have distinctive absorption spectra, ^[15] the interconversion between the two flavin states can be detected to infer the ET reaction dynamics. The previous studies[ ¹ , ² , ¹⁴ ] mainly detected the initial excited‐state decay and CO₂ formation and have not been able to clearly resolve various intermediates and related elementary steps including the functional decarboxylation. Therefore, mechanistic details on an ultrafast timescale remain elusive. We propose a modified reaction scheme in Figure 1 (inset) based on the previous studies by adding a back electron‐transfer step (k ₂). Here, with femtosecond spectroscopy we resolve the molecular mechanism by capturing various intermediates and determine the catalytic photocycle.

Results and Discussion

The purified FAP sample has all FAD cofactors in the oxidized state. The absorption spectrum (Figure 2A) is significantly red shifted by about 25 nm compared to many other flavoproteins.[ ¹⁶ , ¹⁷ , ¹⁸ ] The red shifting is caused by the electrostatics of the binding pocket and/or the slightly bending structure of the cofactor. ^[2] Upon continuous UV or visible light illumination of the sample solution with adding reducing agent under anaerobic conditions, the oxidized state sample without substrate was completely converted to the anionic semiquinone FAD⋅⁻ state. The radical FAD⋅⁻ is very stable in the active site and remains for a long time without O₂ in the sample. The absorption is further red shifted till 600 nm with an unusual tail (Figure 2A). The conversion is reversible when the photo‐reduced sample is exposed to oxygen. Clearly, the reducing agent and O₂ can fluctuate into the tunnel to enable such conversion between FAD and FAD⋅⁻ upon blue‐light illumination. Such processes are not on ultrafast timescales. Since FAD⋅⁻ is usually prone to protonation forming FADH⋅,[ ¹⁶ , ¹⁹ , ²⁰ ] the fact that FAD⋅⁻ is stabilized in FAP suggests either lack of proton donors at the active site, even though the hydrophobic tunnel (35 Å long) and active site are full of water in the absence of substrate, or the site is very basic. Thus, the water molecules alone at the active site and the neighboring residues of N575 and R451 seem difficult to donate a proton to the FAD⋅⁻ radical.

Figure 2.

Characterization of steady‐state and excited‐state spectroscopic properties. A) Absorption spectra of oxidized FAD and anionic semiquinone FAD⋅⁻ in FAP. B) Steady‐state fluorescence quenching with increasing concentrations of substrate (numbers on the left). Inset shows bound fraction of protein converted from fluorescent intensity, which deviates from the trend (dashed line) as substrate gets saturated in solution. C) Time‐resolved fluorescence at 550 nm with (red line) and without substrate (green solid line). The former has a fast (blue dashed line) and a slow decay component (green dashed line), corresponding to 80 % FAP bound to substrate and 20 % unbound, respectively. D) Transient absorption of FAP probed at 900 nm with and without substrate. The bound ratio is lower in this experiment due to higher glycerol concentration in the buffer.

According to the previous study,[ ¹ , ² ] the quenching of the excited FAD* emission in the presence of substrate was attributed to the electron transfer from the substrate to excited FAD*. Figure 2B shows our titration curves of different substrate concentrations with 20 μM of FAP to observe the changes of fluorescence intensity. Using these data, we derived the dissociation constant (K _d) of the binding complex to be about 15 μM (Figure 2B inset). We have examined various buffer conditions to attempt to increase the substrate solubility such as by changing pH value, mixing different organic solvents, varying salt concentrations and so on. Under the best current conditions, we can achieve a maximum of binding complex percentage to be about 80 %. Although the binding affinity of palmitic acid to FAP is high, low substrate solubility prevents excessive substrate from mixing with FAP, but the binding complex concentration can be maintained constant throughout the entire experiments.

Figure 2C shows the excited FAD* dynamics in the absence and presence of substrate upon blue‐light illumination by detecting the fluorescence emission at 550 nm with temporal resolution of 30 ps. The FAP samples contain about 80 % binding complex and 20 % unbound protein. Without the substrate, the dynamics shows a lifetime decay in 4.7 ns (k ₀ in inset of Figure 1). With excessive substrate, the transient after subtracting the unbound protein contribution becomes much faster and the dynamics exhibits a single stretched exponential decay, ${\mathrm{e}}^{{-\left(\frac{t}{\tau }\right)}^{\beta }}$ , with a time constant of t=285±12 ps and a stretching parameter β=0.8 (Figure S1). Clearly, the dynamics becomes more than ten times faster than the unbound one, indicating that the quenching of the fluorescence emission results from the electron transfer (ET) between the substrate anionic palmitic acid RCOO⁻ and excited FAD* to form RCOO⋅ and FAD⋅⁻ as reported before with the dynamics being fitted as a single exponential decay in 300 ps.[ ¹ , ² ] Here, the stretched ET behavior shows that the ET reaction couples with the environment relaxations, that is, solvation, as observed in many protein ET systems.[ ²¹ , ²² , ²³ ] The protein active‐site solvation was observed in a few to hundreds of picoseconds in flavoproteins such as in flavodoxin ^[24] and photolyase. ^[25] The active‐site solvation of this protein FAP will be reported elsewhere. Thus, the ET in 285 ps here is strongly coupled with the active‐site fluctuations. Our 10‐ns MD simulations show a mobile active site, and several water molecules fluctuate between the substrate and FAD, probably due to the anionic substrate and the positive R451. We found that one water molecule stays in the cavity enclosed by the anionic substrate, R451 and FAD (Figure 1) for a significant long time, like a stable structural water. Such an immobile water in the active site was also reported recently. ^[2] Thus, ET between the substrate and FAD is probably mediated by the water molecule,[ ²⁶ , ²⁷ ] a bridge to enhance the electron tunneling between the donor and acceptor with the electron density more located at the N5 position of FAD⋅⁻. ^[28]

We also used the transient‐absorption method to detect intermediates to reveal the catalytic photocycle. Figure 2D shows an absorption transient detected at 900 nm and exhibits the same dynamics of the excited FAD* as the fluorescence detection above, again 4.7 ns for the unbound protein and 285 ps for the binding complex ET with a β=0.8. Such near‐infrared detection at 900 nm shows that except the excited state no intermediates have absorption around this wavelength. Finally, the average decay time can be calculated as $⟨\tau ⟩=\frac{t}{\beta }\Gamma \left(\frac{1}{\beta }\right)=$ 323±13 ps (k ₁ +k ₀ in Figure 1), resulting in the average ET dynamics in 347±17 ps (k ₁) with 93 % quantum efficiency for this first‐step ET reaction.

The RCOO⋅ radical produced through photodissociation of a neutral molecule and its subsequent decarboxylation have been studied in solution before.[ ²⁹ , ³⁰ , ³¹ , ³² ] One previous study showed a visible absorption band of an aromatic carbonyloxy radical in acetonitrile with a peak around 700 nm. ^[32] Thus, we first scanned our probe wavelengths from 500 to 900 nm to search the absorption of RCOO⋅ in FAP after the forward ET reaction. Surprisingly, we found a long component of an intermediate absorption in the region between 700 and 900 nm, besides the known excited‐state FAD* and triplet ³FAD dynamics. With a series of visible probe wavelengths (Figure 3), we determined the relative absorption band of the intermediate shown in Figure 3A together with a reported aromatic carbonyloxy radical absorption. ^[32] Clearly, this intermediate is RCOO⋅ because no other intermediate species have absorption in this region; the absorption of FAD⋅⁻ is located below 600 nm (Figure 2A) and the FADH⋅ absorption, if formed, is below 700 nm. ^[15] In the FAP active site, the RCOO⋅ absorption shifts to the red side with a peak at 800 nm, caused by the local electrostatics, as we also observed for FAD and FAD⋅⁻ absorption. In Figure 3B, we show the transient at 800 nm with deconvolution of the FAD* decay and RCOO⋅ formation and decay. The rise formation time of RCOO⋅ is the same of the decay dynamics of FAD* above, indicating the production of RCOO⋅ radical resulting from the forward ET reaction. The RCOO⋅ decay dynamics was best fit to occur in 5±0.5 ns. As shown in the scheme of Figure1 inset, the RCOO⋅ bifurcates and the decay rate is the sum of two steps (k₂+k₃ ): one is the back electron transfer (k₂ ) of a futile cycle to the original state, resulting in a loss in catalytic efficiency; the other channel is the functional decarboxylation (k₃ ) to form R⋅ and CO₂. From 700 to 600 nm, the signals are only from FAD* (or stimulated emission of FAD*); for example, the transient probed at 600 nm is a mirror inversion of the 900‐nm transient. From 600 to 400 nm, we found that the signals are all from FAD* and FAD⋅⁻. In Figure 3C, the transient probed at 550 nm can be deconvoluted into FAD* decay and FAD⋅⁻ formation and decay. For the latter, the formation time is the same as the dynamics of FAD* and the decay also bifurcates into two parts; one goes to the back electron transfer (k ₂) to the original state and the other part finally proceeds to the electron return (k ₄) to the R⋅ radical.

Figure 3.

Femtosecond‐resolved transient‐absorption dynamics of 473 nm excited FAP with substrate in the visible and near‐infrared regimes. Unbound protein and triplet state formation signals have been subtracted for clarity. A) RCOO⋅ (R=C₁₅H₃₁) absorbance constructed from fitting parameters (dashed line), in comparison with the reported spectrum of an aromatic carbonyloxy radical (solid line) derived by dissociation of bis(2,6‐dichlorobenzoyl) peroxide. ^[32] B) 800 nm probed transient decomposed into excited state FAD* (blue) and RCOO⋅ rise‐and‐decay (red) contributions. C) 550 nm probed transient decomposed into FAD* (blue) and FAD⋅⁻ (orange) components. The decay of FAD⋅⁻ reveals the back‐electron‐transfer pathway.

It has been reported that the electron return from FAD⋅⁻ to R⋅ occurs in more than a hundred nanoseconds (k ₄).[ ¹ , ² , ¹⁴ ] Thus, the decay of FAD⋅⁻ observed within 3 ns window should come from the portion of the back electron transfer from FAD⋅⁻ to RCOO⋅ and the other portion with a longtime (>100 ns) decay of the electron return. The FAD⋅⁻ transient in Figure 3C can be fit by two components (detailed in the Supporting Information and Figure S2): one decay is in 5 ns (k ₂ +k ₃), reflecting the decay rate of FAD⋅⁻ for the back electron transfer with a branching fraction k ₂/(k ₂ +k ₃) and the other part mainly decays in more than 100 ns, representing the decay dynamics of FAD⋅⁻ for the electron return with a branching fraction k ₃/(k ₂ +k ₃), also the fraction of functional decarboxylation (Figure 1 inset). Thus, knowing the amplitude ratio of two components (k ₂/k ₃) by fitting the FAD⋅⁻ transient and the total decay rate (k ₂ +k ₃) of (5 ns)⁻¹, we can derive the back electron transfer of k ₂=(36±4 ns)⁻¹ and the functional decarboxylation of k ₃=(5.8±0.6 ns)⁻¹, leading to 86 % efficiency of functional decarboxylation caused by the RCOO⋅ bifurcation. With the determined forward ET efficiency of 94 %, the final total catalytic quantum efficiency is 80 %. Remarkably, this observed catalytic efficiency (0.8), obtained from the measured dynamic timescales, is in good agreement with the steady‐state measured overall quantum yield reported previously, ^[1] further confirming our accurate determination of all reaction times for elementary steps and indicating all subsequent steps after decarboxylation proceeding without any efficiency loss.

We further tuned probe wavelengths into the UV region below 400 nm till 240 nm. Several transients are shown in Figure 4. Clearly, we need to fit FAD* and all intermediates of FAD⋅⁻, RCOO⋅ and R⋅. The product CO₂ has no electronic absorption at longer than 210 nm and thus these transients do not contain any CO₂ product, ^[33] which can be detected by the time‐resolved IR spectroscopy. ^[34] Using the kinetic analyses (Supporting Information), we found another UV absorption band of RCOO⋅ peaking at 350 nm with more absorption extending to deep UV. The radical of R⋅ starts its absorption below 300 nm with a shoulder at 275 nm with large absorption also toward deep UV. In our current setup, we can detect the signal till 240 nm. The clear long rising signals in the UV transients must be from the formation of the R⋅ radical. For example, at 280 nm the transient shows all intermediates of FAD⋅⁻, RCOO⋅ and R⋅ (Figure 4B). The formation time of the R⋅ radical is 5 ns as the total decay dynamics of RCOO⋅ while its decay dynamics by the electron return is in more than a hundred nanoseconds beyond our time window. Especially, at 240 nm the long rise of the positive signal obviously represents the R⋅ formation (Figure 4C). The two absorption bands of RCOO⋅ in both the visible (Figure 3A) and UV regions (Figure 4A) agree with those reported in literatures, ^[32] but are red shifted in the active site of protein FAP. The absorption of a primary alkyl radical in gas phase was reported before ^[35] and our obtained absorption band of the long‐chain alkyl R⋅ radical is also red shifted again (Figure 4A).

Figure 4.

Transient‐absorption dynamics of FAP with substrate in the ultraviolet regime. Excited‐state stimulated emission and ground‐state FAD bleaching are indistinguishable, and thus built into FAD* and FAD⋅⁻ dynamics. Unbound protein and triplet‐state formation signals have been subtracted for clarity. A) RCOO⋅ (red dashed line) and R⋅ (green dashed line) absorption spectra constructed from fitting parameters, in comparison with reported carbonyloxy and alkyl radical spectra.[ ³² , ³⁷ ] B), C) Decomposed transients probed at 280 and 240 nm. All transients from IR to UV are fitted with a consistent kinetic model.

One mutant done by others ^[14] was C432S, 5.5 Å away from the R⋅ radical carbon, and inactivated the enzyme activity, suggesting C432 to be the proton donor. However, our experiment of such the same mutation showed no fluorescence quenching at all, indicating no forward electron transfer between the substrate and FAD*, probably due to the active‐site structural change induced by the mutation of C432S (Figure S3). ^[2] Another experiment proposed the final proton source from buffer solution ^[14] but even in absence of the substrate, the proton seems difficult to fluctuate into the active site through a long hydrophobic tunnel of 35 Å, as we also did not observe the FADH formation under the reduced stable FAD⋅⁻. It is not convinced that the R⁻ is stable and exits out of the long tunnel to capture a proton in buffer solution. In the very recent report, ^[2] the multiple methods were used to reveal the FAP catalytic mechanism. In the paper, they reported many detailed observations and proposed a sophisticated mechanism involving two water molecules with a series of proton‐transfer steps, including a cyclic proton transfer with R451 assisted by Q486. After decarboxylation, 75 % of the product CO₂ reacts with one trapped water molecule to form bicarbonate (HCO₃ ⁻), finally leaving one H⁺ and the R⋅ to form RH with electron return. For the other channel (25 %), the product CO₂ directly exists out of the active site in less than 100 ns and R451 donates one H⁺ to R⋅ to form RH with electron return, but it is unknown how the deprotonated neutral R451 returns to the original state. The paper also showed the product CO₂ formation through decarboxylation by transient infrared spectra in about 270 ps. The dynamics is very similar to the forward ET in 300 ps. In addition, the published IR data did not show enough sampling from 450 ps to 15 ns. ^[2] If the decarboxylation occurs ultrafast in much less than 300 ps, the observed 270 ps could represent the forward ET process. However, with our detection sensitivity, any kinetics of decarboxylation slower than 40 ps and faster than 270 ps, that is, an inverted kinetics due to the 300‐ps forward ET, should be detected. In our work reported here, we detected various intermediates of FAD⋅⁻, RCOO⋅ and R⋅; the decay dynamics of FAD⋅⁻ and RCOO⋅ clearly show their timescales of 5 ns, not 270 ps at all. With the new model (Figure 1), we also determined the reaction branching for decarboxylation and its dynamics in 5.8 ns, and the final enzyme efficiency of 0.80 from the measured reaction timescales, the same as the steady‐state value, further confirms our obtained dynamic timescales.

Conclusion

Based on these findings and previously reported results,[ ¹ , ² , ¹⁴ ] we propose a catalytic photocycle of photoenzyme fatty acid photodecarboxylation (Figure 5). The key findings in this study are the successful detection of various intermediates of FAD⋅⁻, RCOO⋅ and R⋅ and determination of their reaction timescales, thus revealing the involved elementary steps and elucidating the catalytic mechanism and photocycle. The initial injection of an electron to the cofactor is 347±17 ps, and the subsequent decarboxylation time is about 5.8±0.6 ns. The slow back electron transfer in 36±4 ns ensures the dominant pathway of decarboxylation. The final quantum yield determined from these dynamic timescales, consistent with the direct steady‐state measurement, further confirmed the enzymatic reactions and the related dynamics, that is, the catalytic mechanism and photocycle. One unresolved question is where the proton comes from in the final formation of RH, that is, the reaction of R⋅+e⁻+H⁺ to form RH, although more efforts have been made to search for the proton source.[ ¹ , ² , ¹⁴ ]

Figure 5.

Catalytic photocycle of FAP. An ultrafast forward electron transfer coupled with local environmental relaxation occurs upon FAD photoexcitation, forming an RCOO⋅ radical. The RCOO⋅ dynamics are bifurcated into a relatively slow back electron transfer resulting in a futile cycle, and a dominant decarboxylation pathway forming an R⋅ radical and a CO₂ by‐product. Experimental evidence for both proposed radicals has been captured in this study. Electron return and proton transfer to the R⋅ radical to form RH occur on timescales beyond our observation window.

The intricate mechanism involving major bicarbonate formation deserves more consideration. ^[2] All experiments were done in vitro and the proposed mechanism may not be the same as that in the living cell. As substrate acidity heavily depends on aggregation states and environment polarity, literatures have reported fatty acid pK _a values ranging from 4.5 to 10.1.[ ³⁶ , ³⁷ , ³⁸ ] In vivo, especially in a hydrophobic environment near membrane, the palmitic substrate could have a large pK _a value, and thus it may not deprotonate to form anionic RCOO⁻, that is, the substrate binds in the form of a carboxylic acid (RCOOH) at the active site. This alternative possibility is very appealing; with RCOOH/RCOO⋅≈2.24 V vs NHE (normal hydrogen electrode), ^[39] the free energy for this ET and PT is ≈0 eV, also a feasible reaction. The resulting RCOOH⁺ by ET can quickly deprotonate to form a carbonyloxy radical RCOO⋅ and the proton can interact with the trapped water molecules, such as a Zundel structure (H₂O)₂⋅H⁺; this proton later recombines with the returned electron and R⋅ to form final hydrocarbon RH. In this way, the mechanism in vivo does not need R451 to donate a proton, does not consume one trapped water molecule and does not form bicarbonate (HCO₃ ⁻). Such a new in vivo mechanism is much simpler and clear, both the electron and the proton from the substrate of carboxylic acid RCOOH.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

R. Wu, X. Li, L. Wang, D. Zhong, Angew. Chem. Int. Ed. 2022, 61, e202209180; Angew. Chem. 2022, 134, e202209180.

Data Availability Statement

The data that support all findings of this work are available in the main paper and the Supporting Information of this article.