Terminal Alkenes from Acrylic Acid Derivatives via Non‐Oxidative Enzymatic Decarboxylation by Ferulic Acid Decarboxylases

Abstract Fungal ferulic acid decarboxylases (FDCs) belong to the UbiD‐family of enzymes and catalyse the reversible (de)carboxylation of cinnamic acid derivatives through the use of a prenylated flavin cofactor. The latter is synthesised by the flavin prenyltransferase UbiX. Herein, we demonstrate the applicability of FDC/UbiX expressing cells for both isolated enzyme and whole‐cell biocatalysis. FDCs exhibit high activity with total turnover numbers (TTN) of up to 55000 and turnover frequency (TOF) of up to 370 min−1. Co‐solvent compatibility studies revealed FDC's tolerance to some organic solvents up 20 % v/v. Using the in‐vitro (de)carboxylase activity of holo‐FDC as well as whole‐cell biocatalysts, we performed a substrate profiling study of three FDCs, providing insights into structural determinants of activity. FDCs display broad substrate tolerance towards a wide range of acrylic acid derivatives bearing (hetero)cyclic or olefinic substituents at C3 affording conversions of up to >99 %. The synthetic utility of FDCs was demonstrated by a preparative‐scale decarboxylation.


Introduction
The production of organic building blocks from renewable carbon sources is a current trend in synthetic organic chemistry. [1][2][3][4] The major primary intermediates of traditional industrial-scale synthesis are light alkenes such as ethylene, propylene and butadiene which are produced from crude oil via steam-cracking, which has been described as the single most energy-demanding process in the petrochemical industry. [5,6] In view of the fact that biocatalytic transformations are operational under mild and environmentally-friendly conditions and proceed with high chemo-, regio-and stereoselectivity, [7] there is an increasing interest in expanding the scope and efficiency of enzymatic reactions. [8][9][10][11][12][13] Biological routes towards alkenes are rare and have been investigated only recently. [14][15][16][17][18][19][20] For instance, oxidative decarboxylation of (saturated) fatty acids by the P450 mono-oxygenase OleT [21][22][23] and the non-heme oxygenase UndA [24] yields terminal alkenes on a small scale. [25] In order to avoid the requirement for sophisticated and sensitive electron-transfer proteins, redox-neutral decarboxylation of p-hydroxycinnamic acids ('phenolic acids') derived from the breakdown of lignin catalysed by phenolic acid decarboxylases was investigated. [7] The latter enzymes act via simple acid-base catalysis, [26] which requires the presence of a phenolic 'activating' group in the substrate, which severely limits their applicability. Furthermore, the electron-rich p-hydroxystyrenes thus obtained are not very stable and are prone to (spontaneous) oxidation and polymerisation.

Optimisation of Biotransformation Conditions
In order to assess the biocatalytic potential of FDCs, three previously described representatives [28] from Aspergillus niger (AnFDC), Saccharomyces cerevisae (ScFDC) and Candida dubliniensis (CdFDC) were each co-expressed with the native E. coli UbiX in E. coli to produce the holo-enzymes AnFDC UbiX , ScFDC UbiX and CdFDC UbiX . In this system, the FDCs were fused with a polyhistidine tag, whereas UbiX was co-expressed untagged to enable in vivo production of prFMN, allowing for the purification of the prFMN-bound FDC to homogeneity by Ni affinity chromatography.
Using purified AnFDC UbiX as the catalyst, biotransformation conditions were optimised for the decarboxylation of 20 mM 1 a as a model reaction. The enzyme displayed a broad pH window (pH 6.0-9.0) with highest conversions of > 99 % achieved at pH 7.5 (phosphate buffer) and pH 8.0 (Tris-HCl buffer) (Supporting information Section S1.1). AnFDC UbiX showed high activity between 20 and 45 8C with highest rates obtained at 37-42 8C, however protein precipitation was observed upon incubation at ! 37 8C for 1 h. Hence, subsequent reactions were performed at 30 8C. Under the optimised conditions, biotransformations were performed with i) freshly purified enzyme preparations (snap-frozen or lyophilised), ii) E. coli whole cells containing AnFDC either as fresh resting whole cells or in lyophilised form, and iii) using fresh cell-free extract (snap-frozen or lyophilised). In all cases, conversions of > 80 % were achieved highlighting the suitability of FDCs in isolated form or as whole cell biocatalyst. Similarly, lyophilised wholecell ScFDC showed a broad temperature optimum between 30 8C and 45 8C, with a sharp drop beyond this value, while the pH-profile peaked at 6.0 (Supporting information Section S1.3).
Monitoring ScFDC-catalysed decarboxylation of (aromatic) ferulic and (non-aromatic) sorbic acid over time revealed a typical hyperbolic decline of the substrate concentration, wherẽ 90 % conversion was reached within~8 h, and the reaction was complete after~16 h (Supporting Information, Figure S5). Control reactions featuring all reaction conditions but containing E. coli whole cells harbouring an empty pET vector revealed no conversion of 1 a.

Substrate Tolerance of FDCs
To highlight the synthetic utility of FDCs, the substrate scope of AnFDC, ScFDC and CdFDC was investigated. An array of 60 different a,b-unsaturated carboxylic acids were tested in the decarboxylation direction encompassing substituted cinnamic acids and heterocyclic analogs thereof, as well as non-aromatic acrylic acid derivatives and a,b-acetylenic substrates (Scheme 2 & Figure 1). Initially, isolated enzymes were used for the substrate profiling study (Table 1). In addition, ScFDC was also applied as lyophilised whole cell preparation (overexpressed in E. coli) to evaluate its applicability on preparative-scale for potential industrial use. Overall, a broad set of substrates covering different structural motifs and electronical properties were employed (Table 1).
First, a range of cinnamic acid derivatives with various substituents at the p-position of the aromatic moiety (1 a-11 a) were examined. Substrates bearing weakly electron-withdrawing groups such as p-halogens (2 a-4 a) and weakly e À -donating groups such as p-methyl (5 a) were well tolerated by the enzymes affording > 84 % conversion (Table 1, entries 2-5). Strong e À -donating groups such as p-NH 2 6 a, p-OH 7 a and p-OMe 8 a were perfectly accepted by whole cells (c = 86-99 %, entries 6-8) while a drop in conversion was observed with purified enzymes as catalyst (c = 61-80 %). A strong e À -withdrawing p-NO 2 group (10 a) led to diminished conversions (c = 18-50 %, entry 10) using purified enzymes. Steric restriction seems to appear with a larger p-Ph group (9 a) which was only reasonably accepted by FDC from A. niger (c = 40 %, entry 9). Complete loss of activity was observed with an even larger substituent (p-OPh, 48, Figure 1). Substrate 11 a which carries two carboxyl groups was regioselectively decarboxylated yielding 4-vinyl benzoic acid (11 b) as sole product, albeit in low conversions of up 8 % (entry 11). Remarkably enough, in contrast to phenolic acid decarboxylases (PADs), the confining requirement for an activating p-hydroxy group proved to be dispensable which is in line with the proposed 1,3-dipolar cycloaddition mechanism of FDCs.  The influence of the substitution pattern at the aromatic ring on the enzyme's performance has been further evaluated applying mono-(o-or m-, for p-see above), di-, tri-and pentafunctionalised cinnamic acid derivatives. A NO 2 -substituent in m-position was similarly tolerated as the p-analogue (10 a versus 15 a, entries 10, 15) whereas a strong e À -donating group (such as OH) in m-position led to reduced reaction rates compared to the p-pendant (7 a versus 12 a, entries 7, 12). Disubstitution in p-and m-position was well accepted (p-OH and m-OMe, ferulic acid, 17 a, c up to > 99 %; p-and m-OMe, 19 a, c > 99 %, entries 17,19) as long as the m-substituent was not too e À -pushing (p-and m-OH, caffeic acid, 18 a) which led to a significant drop in conversion (c = 33 %, entry 18) correlating with the results from above. The p-naphthyl derivatives (30 a and 31 a) which formally correspond to a p-/m-di-substitution with weak e À -donating groups were excellent substrates, which were quantitatively decarboxylated (c > 99 %, entries 30, 31). The size as well as the electronic nature of the o-substituents seem to play a crucial role which were well tolerated as long as they were small (F, 20 a, c= 82 %, entry 20; F, 25 a, c > 81 %, entry 25; Me, 16 a, c > 99 %, entry 16). Sterically more demanding methoxy-(21 a, c= 36 %; 22 a, c= 31 %, entries 21, 22) and nitro-groups (14 a, c up to 10 %, entry 14) were less favoured which also applies to polar (strong e À -donating) o-substituents such as OH (13 a, c up to 8 %, entry 13) and led to a complete loss of activity in case of two polar (o-and p-OH) groups (49, Figure 1). Tri-substituted compounds with functional groups significantly larger than a F-atom were poor substrates (sinapic acid, 23 a, c= 3-15 %, entry 23; 24 a, c= 5-8 %, entry 24).
The substrate profiling was further extended to a,bunsaturated carboxylic acids containing O-, S-and N-heteroaromatic systems at C3. The enzymes were excellent catalysts for the decarboxylation of 2-furyl-(26 a) and 2-thienyl acrylic acid (27 a) furnishing the corresponding vinyl products in up to > 99 % conversion. AnFDC UbiX was also capable of decarboxylating the imidazole-derivative 28 a albeit with very low rate (c = 5 %, entry 28), which is presumably caused by the high degree of protonation (~90 %/100 %) at pH 6.0/7.5 creating a positive charge. The bicyclic indole-derivative (29 a) was reasonably well accepted (c up to 42 %, entry 29).
The results from Scheme 2, Table 1 and Figure 1 reveal a clear substrate structure-activity pattern of the FDCs enzymes: i) Minimal substrate requirements consist of an acrylic acid moiety with an extended p-system in the b-position, which is fulfilled by an aromatic system or a (minimal) second conjugated C=C bond.
ii) Compounds lacking an a,b-C=C bond, which is an essential requirement to undergo 1,3-dipolar cycloaddition with the prFMN cofactor, are unreactive, as well as acetylenic analogs.
iii) The (E) or (Z) configuration of the reactive C=C bond seems to be critical. iv) Sterically demanding groups impede reaction rates. v) Strongly electron-donating groups impede reaction rates.

Structural and Mechanistic Aspects
Azomethine ylides have been characterised as dipoles with pronounced nucleophilic character. [46] Due to their inherent reactivity, they are usually prepared in situ, for example by ringopening of aziridines. [47,48] Initial cycloadduct formation in the reaction mechanism of FDC is expected to proceed through interaction between the HOMO of prFMN and the substrate's LUMO. [49] Thus, potential substrates must show a somewhat ambiguous character: the a,b-unsaturated carboxylic acid molecule must be electrophilic enough to allow cycloadduct formation with the nucleophilic cofactor in the first place. However, after decarboxylation, the cycloadduct should dissociate easily into the olefinic decarboxylation product and cofactor, allowing a new catalytic cycle to initiate. This suggests that decarboxylation itself (the loss of one EWG as CO 2 ) is the crucial step that raises electron density in the substrate-cofactor adduct, promoting it to undergo cyclo-elimination. Strongly electron-deficient dipolarophiles are potent mechanistic inhibitors of FDC enzymes, which has been demonstrated experimentally. [35] Additionally, the enzyme only accepted substrates with an extended p-system conjugated to the acrylic acid moiety. This preference ensures diffuse electron density in both cofactor and substrate, which allows enhanced matching orbital energy levels according to HSAB and FMO principles. [50][51][52][53] These considerations are in excellent agreement with the observed substrate preference of FDC enzymes. An analysis of the AnFDC active site architecture provides a rationale for FDC tolerance to cinnamic acid residues bearing small substituents (Figure 2a, R 1 = F/Me) at the a-carbon to the carboxylate ( Figure 2). The orientation of the substrate in the active site positions R 2 and R 3 substituents at a water filled cavity (Figure 2a), indicating that large groups can be accommodated at the m-and p-positions of the aromatic ring. In contrast, the AnFdc1 structure highlights potential steric constraint with large R 1 substituents and o-substitutions of the aromatic ring (R 4 ). These predictions are in excellent agreement with biotransformation data presented in Table 1.
In order to prove the applicability of this method on preparative scale, the decarboxylation of ferulic acid (17 a) was performed. The substrate load was increased from 10 to 16.8 mM in 20 mL reaction volume. HPLC-analysis revealed incomplete conversion of the starting material (48 %). The product was isolated by extraction of the aqueous phase with EtOAc and was purified by flash chromatography yielding 19 mg (38 % yield) of 17 b. Product identity and purity were confirmed by NMR spectroscopy (see Supporting Information).

(a) Production and Preparation of Biocatalysts
Cloning, expression and purification of AnFDC UbiX , ScFDC UbiX and CdFDC UbiX were performed as previously described. [26,30] The purified enzymes were either snap-frozen or stored at À80 8C until when needed or lyophilised and stored at À20 8C. For the preparation of the whole cell biocatalysts, cultivation was performed in 500 mL LB broth medium with kanamycin (30 mg mL À1 ) and ampicillin (50 mg mL À1 ). Cultures were initially incubated at 37 8C with shaking at 200 rpm. At an optical density (OD 600 ) between 0.6 and 0.8, isopropyl b-D-1-thiogalactopyranoside (IPTG) was added to a final concentration of 0.3 mM to induce protein expression and MnCl 2 to the final concentration of 1 mM was added. Incubation was continued at 20 8C and 250 rpm for 18 h. Cells were then harvested by centrifugation and suspended in sodium phosphate buffer (100 mM, pH 7.5). The harvested cells were used as fresh resting cells or lyophilised preparation.

(b) General Procedure for Isolated Enzyme Decarboxylation
For FDC UbiX -catalysed decarboxylation reaction using purified enzyme preparation, a 500 mL reaction mixture contained carboxylic acid substrate (5 mM), 2-10 % (v/v) DMSO, purified FDC UbiX (0.2 mg mL À1 ) in sodium phosphate buffer (100 mM, pH 7.5). Reaction mixtures in 2 mL tightly-closed glass vials were incubated at 30 8C with 180 rpm shaking for 18 h, after which the enzyme was inactivated by the addition of an equal volume of MeCN and vigorously mixed. The reaction mixtures were centrifuged (4 8C, 2,831 rcf, 5 min); the clear supernatant was filtered and analysed by reverse phase HPLC. Where analysis of biotransformation was performed on the GC-MS, an equal volume of EtOAc (containing a known concentration of an internal standard where necessary) was added to biotransformation mixture, vigorously mixed, centrifuged and the organic layer was extracted twice. The aqueous layer was then acidified to a pH of~2 and further extracted with EtOAc with centrifugation (4 8C, 2,831 rcf, 5 min) to improve the separation of phases. The organic layers were combined and dried over anhydrous MgSO 4 and samples were analysed by GC-MS.

Co-solvent Studies
Stock solutions of 36 a (200 mM) were prepared in MeCN, acetone, 1,4-dioxane, MeOH, EtOH, i-PrOH, t-BuOH, DME, DMF, DMSO, THF, DCM, chloroform and EtOAc. Lyophilised cells were rehydrated in 800, 900 or 950 mL phosphate buffer (100 mM, pH 6.0). 50 mL of the corresponding stock solution was added to the mixture and pure co-solvent was added to achieve a reaction volume of 1 mL, followed by incubation. For water-miscible co-solvents, sample workup and analysis was performed as described above. For immiscible solvents, partial evaporation of the organic layer was observed and therefore, only the aqueous phases were analysed using HPLC.
Substrates standards and product markers, and the resulting biotransformation products were analysed by reverse phase chiral HPLC using isocratic methods with different solvent ratios of MeCN and H 2 O, with 0.1 % TFA as additive. The flow rate was maintained at 1 mL min À1 and elutes were detected by the UV detector at a wavelength of 245 nm (except for pyrrole which was monitored at 210 nm). To account for the variation in UV response between the starting material and the product, relative response factors were experimentally determined. Correction factors were calculated from the ratio of the slopes of standard curves plotted for varying concentrations of both the acid and the corresponding alkene at a UV detection wavelength of 245 nm.

Associated Content
Data on pH study, co-solvent compatibility study, analytical protocols including HPLC, GC-MS analyses and representative traces of biotransformation products are available in the Supporting Information.