Exploring the substrate scope of ferulic acid decarboxylase (FDC1) from Saccharomyces cerevisiae

Ferulic acid decarboxylase from Saccharomyces cerevisiae (ScFDC1) was described to possess a novel, prenylated flavin mononucleotide cofactor (prFMN) providing the first enzymatic 1,3-dipolar cycloaddition mechanism. The high tolerance of the enzyme towards several non-natural substrates, combined with its high quality, atomic resolution structure nominates FDC1 an ideal candidate as flexible biocatalyst for decarboxylation reactions leading to synthetically valuable styrenes. Herein the substrate scope of ScFDC1 is explored on substituted cinnamic acids bearing different functional groups (–OCH3, –CF3 or –Br) at all positions of the phenyl ring (o−, m−, p−), as well as on several biaryl and heteroaryl cinnamic acid analogues or derivatives with extended alkyl chain. It was found that E. coli whole cells expressing recombinant ScFDC1 could transform a large variety of substrates with high conversion, including several bulky aryl and heteroaryl cinnamic acid analogues, that characterize ScFDC1 as versatile and highly efficient biocatalyst. Computational studies revealed energetically favoured inactive binding positions and limited active site accessibility for bulky and non-linear substrates, such as 2-phenylthiazol-4-yl-, phenothiazine-2-yl- and 5-(4-bromophenyl)furan-2-yl) acrylic acids. In accordance with the computational predictions, site-directed mutagenesis of residue I330 provided variants with catalytic activity towards phenothiazine-2-yl acrylic acid and provides a basis for altering the substrate specificity of ScFDC1 by structure based rational design.

Styrenes are valuable building blocks for the synthesis of fine chemicals, polymers and pharmaceutically active compounds. Accordingly biotechnologies for their synthesis continuously emerged, styrene production from glucose through engineered Escherichia coli cells, or Saccharomyces cerevisiae cells with improved phenotypes have been successfully developed 1,2 . These methodologies rely on the activity of ferulic acid decarboxylase (FDC1) on cinnamic acid, a metabolic intermediate of the shikimate pathway. However, the synthesis of differently functionalized styrenes through metabolic pathways still remains challenging, due to the limited substrate specificity of metabolic enzymes and the complexity of constructing such artificial metabolic pathways 1,2 . Through more convenient biocatalytic procedures, by the use of non-oxidative decarboxylases as whole cells, cell free extracts or isolated biocatalysts, bioproduction of styrene derivatives can also be approached [3][4][5][6] . The accessibility of commonly available cinnamic acid derivatives as starting materials, the mild and environmentally friendly reaction conditions render the decarboxylation approach appealing.
Regardless to the mechanism of action and the nature of the employed cofactor, currently, four distinct types of non-oxidative decarboxylases acting on aromatic acids have been described. Phenolic acid decarboxylases form Enterobacter sp., Bacillus pumilus and Lactobacillus sp. do not require cofactor and have strict substrate specificity to cinnamic acid derivatives possessing the 4-OH functional group [7][8][9] . Members of amidohydrolase superfamily (AHS), a diverse group of metallo-dependent enzymes, with broad range of catalytic diversity, hydrolyzing C-O, P-O, P-S, C-N, C-S, and C-Cl bonds [10][11][12][13][14] , were also shown to act on C-C bond cleavage of substituted benzoic acids. 5-Carboxyvanillate decarboxylase (LigW) was reported to catalyze the nonoxidative C−C bond cleavage of 5-carboxyvanillate (5-CV) 15 , while benzoic acid decarboxylases, showing high protein sequence homology with amidohydrolases 16 , decarboxylate diverse hydroxybenzoic acids 17  Phenylacrylic acid decarboxylases (PADs) are flavoproteins with a non-covalently bound flavin mononucleotide. Their most known representative is PAD1 from E. coli, also known as UbiX, which catalyse the decarboxylation of 3-octaprenyl-4-hydroxybenzoate in the ubiquinone biosynthesis 18 . In E. coli besides UbiX, another decarboxylase, UbiD, is also known to be involved in ubiquinone biosynthesis 18 . The homologues of UbiX and UbiD in Saccharomyces cerevisiae are PAD1 and FDC1, respectively, which were found to be employed in the decarboxylation of aromatic carboxylic acids, like ferulic acid, p-coumaric acid or cinnamic acid, both pad1 and fdc1 genes being required for the decarboxylation activity 19 .
Recently, FDC1 from Aspergillus niger and S. cerevisiae was shown to possess a novel prenylated flavin mononucleotide cofactor (prFMN), while PAD1 was found to play role in the formation of the catalytically active, modified FMN-cofactor of FDC1 6 . The mechanism of the FDC1 catalysed decarboxylation was the first example for an enzymatic 1,3-dipolar cycloaddition 6,20 . Importantly, from biocatalytic point of view, several differently substituted cinnamic acid derivatives were shown to be good or moderate substrates of ScFDC1 21 . The presumably broad substrate tolerance nominates ScFDC1 as potential biocatalyst for decarboxylations leading to synthetically valuable styrenes. Moreover the availability of high quality, atomic resolution structures with the bound prFMN and various inhibitors 6 enables structure-based rational modification of ScFDC1. The tedious isolation process of the enzyme, requiring the bacterial co-expression with truncated ScPAD1 necessary for the production of the active prenylated FMN cofactor of FDC1, can be avoided by using as biocatalyst E. coli whole cells harbouring only the fdc1 gene, since UbiX of the host E. coli substitutes ScPAD1 in its role to provide the prFMN 22 .
Herein, in our aim to develop biocatalytic routes to various styrene derivatives we explored the substrate scope of ScFDC1, using whole-cells of E. coli harbouring the fdc1 gene of S. cerevisiae as biocatalyst. Since earlier studies focused mostly on cinnamic acid derivatives with functional groups at the 4-position of the phenyl group 21 , we investigated whether ScFDC1 accepts differently (o-,m-,p-) substituted phenyl-, bulky heteroaryl-or biaryl-analogues of cinnamic acid.

Results and Discussion
Generation of substrate library and decarboxylation activity assay. The tested substrate library was obtained through Knoevenagel-Doebner reaction (see further details in ESI) and included (i) cinnamic acid analogues with functional groups (-Br, -OCH 3 and -CF 3 ) in different positions (o−, m−, p−) of the aromatic ring (1a-j), or (ii) substrate analogues with extended alkenyl or alkyl chains (styrylacrylate 1k and 5-phenylpent-2-enoic acid 1l) as well as (iii) several biaryl and heteroaryl analogues of cinnamic acid (1m-x) (Fig. 1). While p-bromo, and m-, p-methoxy-cinnamic acids (1d,f,g) are known substrates of ScFDC1 21 , to our best knowledge no reports exist on the activity of ScFDC1 with other bulky biaryl-or heteroaryl analogues of cinnamic acids or with the extended alk(en)yl chain-containing cinnamic acid analogues (1a-c,e,h-x).
Whole cells of E. coli BL21 (DE3) pLysS expressing ScFDC1 from plasmid pTfdc1Sc 1 after proper induction were used as biocatalyst to perform the biotransformations of 1a-x to 2a-x (Fig. 1). The conversions in ScFDC1-catalysed biotransformations were determined by monitoring substrate depletion using reversed-phase HPLC, with o-anisol as internal standard calibrated to authentic cinnamic acid derivatives 1a-x (Figs S1-S24). Formation of the corresponding styrene derivatives was confirmed by GC-MS (Figs S25-S60).
Initial screening for the decarboxylation of the substrate library. Initial screening of the substrate panel for decarboxylation activity of ScFDC1 was performed with whole cell suspensions reaching optical density (OD 600 ) of 1 at 30 °C, pH 7.0 and 1 mM substrate concentration. ScFDC1 could decarboxylate a broad range of variously substituted cinnamic acid analogues (1a-j) as biaryl (naphthyl-or biphenyl-) 1m,s,t or heteroaryl (quinolinyl, benzofuranyl and 5-phenylthiophenyl) 1n-r,u acrylates (Table S1). Styrylacrylate 1k, with extended conjugation and chain length was also transformed with high conversion by ScFDC1, while in case of its non-conjugated analogue 1l was inert. However, ScFDC1 could not decarboxylate bulky phenothiazine-2-ylacrylate 1x, 3-(5-(4-bromophenyl)furan-2-yl)acrylic acid 1v, or (2-phenylthiazol-4-yl) Optimization of the whole-cell biotransformations. Next, the reaction conditions of whole-cell biotransformations were optimized focusing on the effect of pH, temperature and biocatalysts/substrate ratio upon conversion, using (3-(3-(trifluoromethyl)phenyl)acrylic acid (1i) as model substrate. The study for biotransformations in buffers of various pH values ranging from 6.0-8.0 revealed the highest degree of conversion at pH values of 6.5 and 7.0 ( Fig. S70), in accordance with the reported pH optimum for the purified FDC1 enzyme 21 .
The optimal temperature was found to be 35 °C, at lower temperatures conversion values significantly decreased, while at 45 °C no product formation was observed (Fig. 2). The effect of biocatalyst/substrate ratio on conversion values was tested by using different amounts of whole-cell biocatalysts (OD 600 of 1,2 or 3) at a fixed, 2 mM substrate concentration of 1i. Expectedly, shorter reaction times were achieved with increasing amount of whole-cell ScFDC1 biocatalyst (the reaction after 4 h at OD 600 = 1, 2 and 3 resulted in conversion of 42%, 76% and ~100%, respectively). In further experiments, comparison of the ScFDC1 activity with different substrates was performed at moderate cell densities (OD 600 ≤ 2) to provide sufficiently long reaction time for precise monitoring of the time-course profile.
Finally, the FDC1-catalyzed reactions of the entire substrate panel (2a-x) were performed under the optimal reaction conditions (100 mM sodium phosphate buffer pH 7.0, cells OD 600 of 1, 35 °C), monitoring the conversions over longer time period (Table 1, Fig. 3a-d).
While substrates 1a,c,d,f,g,k,m,p,r were fully converted within short (<24 h, Fig. 3a, Table 1) or moderate (<72 h, Fig. 3b, Table 1) reaction time, in case of substrates 1b,e,i,j,q,t,u the reactions ceased at high, but not full conversions after 72 h (Fig. 3c, Table 1). Moderate or low conversion values were obtained with substrates 1h,n,o,s, (Fig. 3d, Table 1) while using substrates 1l,v,w,x neither substrate depletion or product formation could be detected (Table 1).
To study the effect of cell viability loss over long reaction times, to the reactions not reaching full conversion within 72 h (Fig. 3c,d, Table 1) were added additional batch of fresh cells (OD 600 = 1) when the reaction ceased. In this way the reactions, initially providing moderate to good conversions (substrates 1b,e,h,i,j,q,s,t,u), proceeded with complete transformation within additional 24 h after supplementing the reaction with fresh cells. This behaviour indicated cell viability issues over prolonged reaction times. However, in case of substrates 1n,1o supplementation with further cells led to only moderate increase of conversion indicating product inhibition or product toxicity upon cells.
FDC1-catalyzed decarboxylation follows a so far unprecendented enzymatic 1,3-dipolar cycloaddition mechanism involving the formation of a covalent substrate-prFMN adduct. While the inductive effect of substituents of the dipolarophile substrate are known to increase the 1,3-cycloaddition reaction rate of nitrogen ylide dipoles, such as prFMN, in case of FDC1 catalyzed reaction it was shown that the presence of an extended π-system associated with the aromatic ring of the substrate also has an important role in stabilization of the transition state 21 , supporting the assumption that π-stacking interaction exists between the planar cofactor and the aromatic moiety of substrate. Therefore, proper binding interaction implies the location of the α,β-double bond of the substrate in proximity to the C1′ and C4a atoms of the cofactor (Fig. 4a). The substrate orientation is further facilitated by R175, interacting with the carboxyl group of the substrate, while the other key residues are E285, acting as acid-base in the reaction mechanism, and E280, which presumably tunes the pKa of R175 and in turn E285 ( Fig. 4a) 25 . Accordingly, the reaction rates are influenced by multiple substrate-related factors, such as inductive effects of substituents, presence of extended conjugation, substrate orientation related to the prFMN and within the catalytic site, influenced by both size and planarity of substrate. Therefore, to rationalize the different rates of conversion for substrates 1a-x, the possibility of planar arrangement of their molecular skeleton, their LUMO energies and orientation in the active site of FDC1 were computationally explored.
Crystal structures of FDC1 (PDB code: 4ZA7, 4ZA8) disclose that the ring system of prFMN adopts a planar conformation to facilitate substrate binding 6 . Consequently, the planarity of the substrates was explored computationally using symmetry/geometry constraints (Table 1). Imaginary frequencies confirmed that planar conformations of structures 1b,h,l,n,o,s,t,u,x correspond to transition states, while their lowest energy conformations deviate from planarity (e.g. at the carboxylic group in case of ortho-substituted ligands 1b,h or between the two aromatic rings in case of biphenyl acrylic acids 1s,t). These results were in agreement with the experiments indicating that maximal conversion could not be achieved with non-planar substrates and in some cases even no conversions could be observed (Table 1). On the other hand, substrates 1a,c-g,i,j,k,m,p,q,r with planar ground state conformation showed high decarboxylation rates, suggesting a good correlation between substrate planarity and reaction rates ( Table 1).
The LUMO energies of the dipolarophile substrates provided further insights into the FDC1 reaction. In case of substrates bearing similar heteroaryl moieties, such as benzofuranyl acrylic acids 1p,q,r, quinolin derivatives 1n,o or biphenyacrylates 1s,t, the conversion increased with decreased LUMO energies (Table 1). In case of mono-substituted cinnamic acid analogues 1a-j no significant correlation between LUMO energy levels and conversion rates could be observed, which correlated with the reported negative slope of the Hammet plot obtained for the FDC1-catalyzed reactions of para-substituted cinnamic acids 21 , supporting the significant influence of electron delocalization (besides the inductive effects) in transition state stabilization.
Docking results (Table 1) revealed the best binding affinities for substrates 1a-j containing one aromatic ring. Importantly, in case of bulky substrates inactive binding poses were also identified. In case of bulky ligands   Fig. S72). In case of compounds 1v,w higher energy active site binding poses were also obtained within the 14 Å grid box (Table 1), with distant arrangement of the catalytically important E285, R175 residues and the prFMN cofactor from the substrate's carboxyl group and the acrylic double bond, correspondingly (Fig. 4b). While in case of 5-(4-bromophenyl)furan-2-yl)acrylic acid 1v presumably the length of the substrate exceeds the limits of the catalytic site for active binding position, in case of (2-phenylthiazol-4-yl) acrylic acid 1w positioning with proper orientation of the carboxyl group towards residue E285 and R175 was also obtained, however the arrangement of the α,β-double bond with respect to the C1′ and C4a atoms of the prFMN cofactor was unfavourable (Fig. S73). In case of 1l, the lack of the extended π-system and the unfavourable arrangement of the acrylic double bond due to the non-planar molecular skeleton contributed to the negligible reactivity (Fig. S73).
To extend the computational investigations of substrates showing no conversion, the ground state of the substrate-cofactor covalent intermediate resulting after the 1,3-cycloaddition of various substrates to prFMN was computed in gas phase. The obtained gas phase geometries were overlaid on the prFMN crystal structure and clashes between the substrate and active site residues were evaluated. Comparison and agreement of this calculation method for trans-cinnamic acid 1a to the reported gas-phase 6 and QM/MM 23 results validated this mode of modelling. The result in case of 1v adduct indicated, that although the ground state conformation of the intermediate is favourable and is aligned along the linear catalytic site, the short distance between the bromine and enzyme backbone prevents the substrate from fitting into the active site (Fig. S74). Fitting the covalent intermediate forming from 1w and prFMN into the enzyme revealed severe steric clashes between the substrate sulphur atom and residue I330 (Fig. S75). Accordingly, it seems that I330 and Q192 narrows the active site in proximity to the substituent placed on the phenyl group of substrate, forming a gate, which can be passed by bulky, but only linearly oriented substrates, such as 3-([1,1′-biphenyl]-4-yl)acrylic acid 1s and 3-(4′-fluoro-[1,1′-biphenyl]-4-yl) acrylic acid 1t (Fig. 4c). The non-linear ligands such as (2-phenylthiazol-4-yl)acrylic acid 1w show steric clash either with the gate-forming residues (Fig. S75) or, in case of 3-(5-(4-bromophenyl)furan-2-yl)acrylic acid 1v, with a backbone carbonyl (Fig. S74). Interestingly, in case of the less bulky bicyclic aryl substrates 1m-r, the gate-forming residues (I330 and Q192) do not hinder the active orientation of the ligand (Fig. S76).
To validate the interactions observed in modelling, residues Q192 and I330 were replaced by smaller residues using site-directed mutagenesis. Since Q192 is also involved in cofactor binding through hydrogen bonding with the ribitol tail of prFMN (Fig. S77), mutation Q192N was envisaged. While catalytic activity of the Q192N mutant towards 1a was maintained (92% conversion after 24 h reaction time), this mutant remained inactive with substrates 1v,w ( Table 2). On the other hand, the I330V and I330A mutants displayed activity towards 1w, resulting in moderate conversions (5 and 15% with mutants I330V and I330A, respectively) after 24 h reaction time ( Table 2). Besides the improved binding energy of 1w to the active site of FDC1 I330A in comparison with the one obtained within the active site of wt-enzyme (E b (F133A FDC1) = -3.5 kcal/mol and E b (wt-FDC1) = -2.8 kcal/mol), the arrangement of the acrylic double bond compared to prFMN also changed to the favourable position (Fig. 4d). Similarly to the wt-FDC1, none of the mutants exhibited activity towards 3-(5-(4-bromophenyl)furan-2-yl)acrylic acid 1v and 3-(10-methyl-10H-phenothiazin-2-yl)acrylic acid 1x (Table 2), providing further support for docking results indicating that bulky 1v,x exceed the volume of the active site (Figs S72,S74).  The results demonstrate that ScFDC1 possess broad substrate tolerance. Substrate planarity is beneficial for the decarboxylation reaction and is imposed by the flat active site, but also by the formation of the 1,3-cycloaddition adduct with the prFMN cofactor. The revealed limits of substrate tolerance of FDC1 are dictated by the length and flatness of the active site, but also by the passage formed by residues Q192 and I330, which narrows the active site and provides access to its full length only for bulky substrates with planar and linear structure.
Instrumentation and analytical methods. 1 H-and 13 C-NMR spectra were obtained using Bruker (Billerica, MA, USA) Avance spectrometers operating at 400 MHz and 101 MHz/600 MHz and 151 MHz, respectively. All spectra were recorded at 25 °C in MeOD-d 4 , CDCl 3 or DMSO-d 6 . 1 H and 13 C NMR spectra were referenced internally to the solvent signal.
The production of styrenes was confirmed through gas chromatography-mass spectrometry (GC-MS) analysis. The samples were prepared by extracting the biotransformations (see section 3.4.) with n-hexane or tert-butyl methyl ether and drying the extract over anhydrous sodium sulphate. The GC-MS analyses were performed using a Shimadzu QP 2010 PLUS Mass Spectrometer coupled with Gas Chromatograph (Shimadzu). The mass spectrometry was performed in the electron impact mode (MS/EI) at 70 eV. Peak identification was carried out by analogy of mass spectra with those of the mass library (NIST 2.0 and Wiley). The system was configured as detailed in Table S2.
The HPLC determination of conversion was performed on Agilent 1200 and/or 1260 series high performance liquid chromatography (HPLC) using Gemini NX-C18 150 × 4.5 mm or Zorbax SB-C8 50 × 2.1 mm columns, flow rate: 1 mL/min. The quantification of the conversion values was based on the determination of the consumption of acrylic acid substrates 1a-x using anisole as internal standard. The details of the HPLC methods used to determine the conversions are described in Table S3.
Synthesis of (E)-arylacrylic acids. The synthesis of acrylic derivatives 1a-j,l-z was performed by using the corresponding aldehydes 3a-j,l-z as starting material with the Knoevenagel-Doebner reaction (Fig. S69). Synthesis of styrylacrylate 1k was performed as earlier reported 29 .
Synthesis of compounds 1g,m. The corresponding aldehyde 3g,m (3 mmol) and malonic acid (6 mmol, 2 equiv.) were dissolved in pyridine (20 mL) and the mixture was heated under reflux for 6 h, and further stirred at 80 °C overnight. After cooling, the reaction mixture was concentrated to dryness and the remained solid was washed with 5% aqueous HCl solution (20 mL), diethyl-ether (20 mL) and finally dried to afford pure acrylic acids 1g,m (yields: 82% for 1g and 78% for 1m).
The synthesized known compounds 1a-x were characterized by 1 H and 13 C NMR measurements (for NMR data see ESI, Chapter 3), while further HPLC measurements (see ESI, Chapter 2) confirmed their high purity.
Biotransformation with whole-cell FDC1. General procedure. Seed cultures of E. coli BL21(DE3) harbouring the corresponding recombinant plasmids, were prepared in 100 mL LB broth and grown overnight. Shake flasks (2 L) containing 500 mL of LB were inoculated with 5 mL of seed culture. Cultures were grown at 37 °C until an OD 600 of ~0.6 was reached, at which point the cultures were induced by IPTG addition at a final concentration of 0.2 mM. Cultures were incubated for an additional 4.5 h (resulting in an OD 600 of ~3) before cells were collected and centrifuged at 6000 rpm for 15 min. The pellet was washed with 100 mM sodium phosphate buffer, pH 7.0, followed by resuspension to a final OD 600 of ~1, aliquoting, centrifugation and storage at -20 °C. Expression of ScFDC1 protein was confirmed by the SDS-PAGE analysis (see Fig. S78).
ScFDC1 activity assays were performed in 1.5 mL glass vials sealed with PTFE septum, with a reaction volume of 1 mL using buffers with different pH (see section "The effect of pH on biotransformation"). Stock solutions of substrates were prepared in DMSO. Assay contained substrates at various final concentrations between 0.5-2 mM. The reactions were incubated at different temperatures (see section "The effect of temperature on biotransformation"). The samples were taken after 24 h.
Samples from biotransformations were diluted with acetonitrile and analysed by HPLC and GC-MS as described in ESI, Chapter 2.
The effect of temperature on biotransformations. The influence of temperature (30,35,40, 45 °C) on the ScFDC1-catalyzed decarboxylation reaction was tested with 3-(3-(trifluoromethyl)phenyl)acrylic acid (1i). The enzymatic reactions were performed in 1.5 mL glass vials sealed with PTFE septum, with a reaction volume of 1 mL using 100 mM sodium phosphate buffer at pH 7.0., 2 mM substrate concentration and ScFDC1-containing whole cells with OD 600 of ~1. The reactions were incubated at different temperatures (30, 35, 40, 45 °C). The samples were taken after 15 h, and analysed by HPLC to determine the conversions.
Effect of cells quantity on biotransformation. The screening contained 3-(3-(trifluoromethyl)phenyl) acrylic acid (1i, 2 mM) and ScFDC1-containing whole cells whole cells with OD 600 of 1, 2 and 3. The enzymatic reactions were performed in 1.5 mL glass vials sealed with PTFE septum, with a reaction volume of 1 mL using 100 mM sodium phosphate buffer at pH 7.0. The reactions were incubated at 35 °C. The samples were taken after 15 h and analysed by HPLC to determine the conversions.
Time conversion profile using the optimized procedure. The ScFDC1-catalyzed reactions were performed with the entire substrate panel (1a-x) under the optimal reaction conditions. The enzymatic reactions were performed in 1.5 mL glass vials sealed with PTFE septum, with a reaction volume of 1 mL using 100 mM sodium phosphate buffer at pH 7.0, 2 mM substrate concentration and ScFDC1-containing whole cells with OD 600 of ~1. The reactions were incubated at 35 °C. Samples for the time conversion profile were taken after 1, 2, 3, 4, 8, 24, 30, 48, 72 hours. In case of the reactions of substrates 1b,e,h,i,j,n,o,q,s,t,u after reaching the stationary phase of conversions (Fig. 3c,d, Table 1) the reactions were supplemented with additional batch of fresh cells (OD 600 = 1) and were monitored by HPLC for an additional 24 h.
Site-directed mutagenesis. The site-directed mutagenesis was performed following the protocol described by Naismith and Liu 30 . The PCR reaction contained 2-4 ng of template DNA (recombinant plasmid pCDFDu-et_ScFDC1), 1 μM solution of primer pair (Table S4,  Computational studies. Ground state geometries of substrates 1a-x were obtained at the DFT level of theory, employing the B3LYP functional and the 6-311++G(d,p) basis set. Harmonic vibrational frequencies, obtained at the same level of theory, confirmed that the stationary points are true local minima. All DFT calculations were performed using the Gaussian 09 package 31 .
Molecular docking was performed using the structure of ligand-bound FDC1 from Aspergillus niger (PDB code: 4ZA7) 6 , based on the high structural similarity of the active residues of ScFDC1 and AnFDC1 (Fig. S71). The presence of prFMN cofactor and α-methyl cinnamate in AnFDC1 supports an active conformation, compared to the impaired ScFDC1 structures (i) without cofactor 32 (PDB code: 4S13), (ii) with the catalytically essential glutamate E285 in inactive conformation 6 (PDB code: 4ZAC) or (iii) of mutant E285D (PDB code: 6EVF) with 37 fold decreased k cat value 25 , and (iv) with mutation R175A (PDB code: 6EVE), altering hydrogen bonding network implied in substrate fixation and hence inactivating the enzyme 32 . The search space was defined as a cubic box centered at the binding site, with an edge length of 20 Å. This grid box also incorporates a large surface pocket, where the best binding modes of large substrates were observed. Therefore, smaller grid boxes ranging from 10 to 16 Å were also used in docking studies to restrict the search space to the catalytic site and find favourable, although energetically higher, binding poses. Autodock Vina 33 version 1.1.2 was used to perform rigid receptor docking. The center of the grid box was obtained from the central atom of the co-crystalized α-Me-cinnamic acid within the structure of AnFDC1 (PDB code: 4ZA7) 6 .
Cofactor-substrate covalent intermediates geometries were evaluated for substrates 1a,v,w using the B3LYP/6-31 G(d,p), including the D3 Grimme's dispersion correction with Becke-Johnson damping 34 to account for stacking interactions. In case of the covalent adduct between 1a and prFMN, the geometry was refined with the B3LYP/6-311++G(d,p) and the obtained result was in excellent agreement with the reported data 6 .

Conclusions
Our study exploring the substrate scope of ScFDC1 using different cinnamic acid analogues revealed that the large cavity of the enzyme active site accepts, besides the -OCH 3 , -CF 3 or -Br-substituted cinnamic acids, several bulky biaryl or heteroaryl substrates, as well as styrylacrylate. Computational studies indicated that substrate planarity is beneficial for the decarboxylation reaction and is determined by the narrow active site as well as by the formation of the 1,3-cycloaddition adduct with the prFMN cofactor. It was shown that substrate preference of ScFDC1 was further determined by a channel bottlenecked by gatekeeper residues residues Q192 and I330 and also by the limited volume of the substrate-binding pocket, restricting the access of bulky, non-linear substrates.
The results demonstrate that whole-cells of E. coli harbouring the fdc1 gene are efficient catalysts for the production of a wide variety of styrene derivatives, furthermore display the substrate profile of FDC1 and provide perspectives for the rational design driven expansion of its substrate tolerance.