Exploring the Catalytic Promiscuity of Phenolic Acid Decarboxylases: Asymmetric, 1,6‐Conjugate Addition of Nucleophiles Across 4‐Hydroxystyrene

Abstract The catalytic promiscuity of a ferulic acid decarboxylase from Enterobacter sp. (FDC_Es) and phenolic acid decarboxylases (PADs) for the asymmetric conjugate addition of water across the C=C bond of hydroxystyrenes was extended to the N‐, C‐ and S‐nucleophiles methoxyamine, cyanide and propanethiol to furnish the corresponding addition products in up to 91% ee. The products obtained from the biotransformation employing the most suitable enzyme/nucleophile pairs were isolated and characterized after optimizing the reaction conditions. Finally, a mechanistic rationale supported by quantum mechanical calculations for the highly (S)‐selective addition of cyanide is proposed.


Introduction
Thea symmetric addition of water across C=Cb onds was denoted as a" dream reaction", because it allows one to convert ap rochiral alkene with 100%a tom efficiency into an on-racemic sec-alcohol. [1] However, industrial-scale production of simple bulk alcohols from olefins via hydration in the gas phase suffers from limited productivity [2] anda symmetric variants are only rarely found. [3] An attractive alternative to the use of chemo-catalytic protocols is the use of lyases (EC 4.2.X.X), which catalyze the addition of nucleophiles ontoe lectrophilic acceptor molecules.B esidesa minases [4] and carbolyases [5] formingC -N and C-C bonds,r espectively,h ydratasese ngagew ater as nucleophile and constitute al yase-subgroup (EC 4.2.1.X) of which ca. 170 were discovered up to now. [6] Hydratases catalyze the electrophilic addition of water onto isolated double( e.g.,s tearate and oleate hydratase) [7] and triple bonds (e.g.,a cetyleneh ydratase). [8] Alternatively,h ydration occurs via nucleophilic (conjugate) addition of water ontoe lectron-deficient a,b-unsaturated carbonyls ubstrates (e.g.,m aleate and aconitate hydratase [9] or Michael-type hydratase [10] ). Unfortunately,m anyo ft hese latter enzymes are encountered in primary metabolism and hence show an arrow substrate tolerance,w hich limits their applicability for synthesis.
Control experiments in the absence of biocatalyst or using E. coli expression host cells lacking the respectived ecarboxylase gene proved the requirement of FDC_Esf or product formationa nd excluded spontaneous backgroundreactions (except for 2q).
Among the first group of nucleophiles (Table 1, entries 1-5), methoxyamine (2a), cyanide (2b)a nd npropanethiol (2c)w ere readily accepted by FDC_Es (Table 1), while 2d and 2e were less promising due to low reproducibility( 2d)o rd ue to spontaneous background reactions (2q). Hence,t he former were chosen Scheme 2. Nucleophile screening with FDC_Es. for the enzyme screening(Scheme 3, for the screening with 2c see the Supporting Information,T able S4). All tested decarboxylases catalyzed the formation of nucleophile adducts 3a and 3b beside the minor hydration product 4 ( Table 2) with the single exception of PA D_Mc and 2a ( Table 2, entry 6). However, distinct variations in conversion and optical purity of the products( S)-3a and (S)-3b were noted.
Similarly,a ddition of cyanide (2b)p roceeded with moderate to good conversion, but stereoselectivities were generally much better with all enzymes ( Table 2, entries1-7 vs. entries8-14). FDC_Esa nd PA D_Ps performed best in terms of conversion ( Table 2, entries 8a nd 14) and the latter enzyme also exhibited superior stereoselectivity in the addition of cyanide (91% ee,T able 2, entry 14) thusp romoting it as ap romising candidate for further investigations. Again, with 2b hydration was only am inor side reaction (max. 8% with PA D_Mc).
Thea bsolutec onfiguration of products 3a-3c was determined by comparison with authentic reference material (3a), comparison of optical rotation values (3b)a nd CD spectroscopy (3c)a sd escribed in the Supporting Information (Table S5). Overall, as trong preference for the formation of the (S)-product is congruent for all enzymes.
Given the comparatively high sequence identity of 73% between PA D_Psa nd FDC_Es( other PADs show 48-52% identity,S upporting Information, Table S1)i ti sp lausible that these two enzymes also performsimilarly in the addition of nucleophiles.
In addition to wild-type enzymes,F DC_Esv ariants ( Figure 1) were tested with nucleophiles 2a and 2b (Table 3). Mutants designed to provide more space in the active site led to considerablyl ess conversion    .Active-site of FDC_Es( PDB-ID:4 UU3) [28] with p-vinylphenol (1)d ocked (green, docking performed with UCSF Chimera); [29] residues targeted by mutagenesis for improvement of catalysis are highlighted in orange.( Putative) catalytic key residues are markedwith an asterisk.
In order to evaluate the relevance of tyrosine residues 27 and 39, which are flanking the substrate,t he corresponding Phe-variantsw erep repared (Y27F, Y39F).W ith 2a,t he activity and selectivity were not affected( entries6 and 7) but with 3b conversion droppeds ignificantly going in hand with enhanceds electivities (entries14a nd 15).
Changing Glu72 (responsiblef or CO 2 -activation in the carboxylation reaction) to alanine or either of Ty r19 and Ty r21 (for deprotonation of the phenolic OH) [19b] to phenylalanine completely abolished the activity for nucleophile addition as wella sh ydration ( Table 3, entries 7-9 and1 6-18), which underlines their crucial role in catalysis.

Preparative-Scale Biotransformation
In order to fully characterize products 3a and 3b and to evaluate the applicability of this biotransformation for the preparative scale,r eactions were performed with 50 to 100 mg substrate after optimization of the reactionc onditions.P romising initial results ( Table 1) and conversions of up to 73% in an enzyme screening (Supporting Information, Table S4) encouraged us to include also propanethiol (2c)i nthe up-scales.
Given the heterogeneity of pK a values of the nucleophilesa nd theirp H-dependent reactivity,adetailed pH-study was performed and the maximum of substrate-andn ucleophile-loading was evaluated. Dimethoxyethane (DME) was identifieda sasuitable cosolvent for water-insoluble nucleophile 2c (Supporting Information, Section 5). Optimalr esults are listed in Table 4. After isolation andpurification of the nucleo-phile adducts( 3a-c) ( Table 4), the absolute configuration of all products was determined to be (S)( Supporting Information, Table S5) and hence nicely matchedt he stereoselectivity of the biocatalytic hydration. [17a] QuantumMechanical Mechanistic Investigations To shed more light on the mechanism and the origin of stereoselectivity in the asymmetric nucleophile addition,D FT calculations were undertaken with cyanide as representative nucleophile using the activesite modelb ased on PAD_Bs (Figure 2a, amino acid numbers were adjusted by + + 8t of it the sequence of FDC_Es).
Thet otal size of the model comprised3 09 atoms and the overallc harge was À1. In analogyt othe previous studies, [19] the hydroxy group of p-vinylphenol was assumed to be deprotonated uponb inding to the tyrosine residues Ty r19 andT yr21, while Glu72 was modeled in its protonated state.S ince the substrates can bind to the active site in many different ways, al argen umber of structures of the enzyme-substrate complex( > 40) have been optimized. Thes tructure with the lowest energy (called React)i ss hown in Figure 2b. [a] Conditions: see Table 2; n.d. = not determined due to low conversion.  Interestingly,i nt he geometry optimization of the enzyme-substratec omplex, cyanide was found to spontaneously abstractaproton from Glu72 to form HCN (Figure 2b). Thec alculations suggest that cyanide addition involves aq uinone-methide intermediate in analogy to that proposedf or hydration. [19a] The reactions tarts with ap roton transfer from hydrogen cyanide to the b-carbon of the substrate,f orming the quinone-methide.T he second step is an ucleophilic attack of the resulting cyanide at the a-carbon to generate the product (Scheme 4).
Theo ptimized geometries of twot ransitions tates are given in Figure 3( geometries of other stationary points are given in the Supporting Information, Figure S46) and the calculated energy profile is showni n Figure 4. In order to ensure that the lowest energy barrier is obtained, we followed the reactionp aths starting from the six lowest-energy Michaelis complexes.
Proton transfer onto Cb was calculated to be the rate-limiting step with ab arrier of + + 17.2 kcal mol À1 , and the resulting quinone methide (Int)i s4 .1 kcal mol À1 higher than React.T he subsequent nucleophilic attack has ar ather low barrier of + + 7.9 kcalmol À1 relative to Int,a nd the entire reactioni sc alculated to be exothermic by 17.5 kcal mol À1 .Inthe optimized transition state structure for proton transfer (TS1), the bond distanceso ft he breaking and the forming H-C bonds are 1.35 and1 .43 ,r espectively.C yanide was located in the vicinity of Ty r27 in TS1.Ahydrogen bond is formed between cyanide and Ty r27 in Int, and this interaction is maintained during the nucleophilic attack. Interestingly,w eo btained also optimized structures of stationary points in which the hydroxy group of Ty r27 points away from cyanide (see the Supporting Information, Figure S47 for geometries and relative energies), however, the barriers for the twos teps are higher by 2-4 kcalmol À1 ,s howing that Ty r27 is important for the reaction( cf.  Scheme4.Proposed catalytic mechanism for the stereoselective addition of cyanide (2b)a cross 4-hydroxystyrene(1). entry 14). Thei mportance of Ty r27 in the (de)carboxylation has been addressedinp reviousstudies. [19] Experimentally,t he ee of (S)-3b is 64-91%, which corresponds to an energy difference of 1-2 kcalmol À1 between the barriers leading to the enantiomeric products. Ther eactionp athway with the lowest energy barriers discussed abovef avors the generation of the (S)-product, which nicely corroboratest he experimental data in Table2andT able 3.
To investigate the origin of the observeds electivity, we have also optimized the geometries of the transition states for the pathway leading to the (R)-enantiomer (termed TS1' ' and TS2' ',F igure 4). Indeed, both transitions tates were found to have higher energies comparedt ot hose of the (S)-pathway.T he barrier for proton transfer in the lowest-energy pathway leading to the (R)-enantiomer is 18.7 kcal mol À1 ,w hich is 1.5 kcalmol À1 higher than that for the (S)-enantiomer, while the barrier for the C-C bondformation is calculated to be 13.1 kcal mol À1 ,c ompared to 12.0 kcal mol À1 for the (S)-enantiomer.T he calculations confirm the experimental stereoselectivities very well.
Analyzing the geometries of the transitions tates leading to (S)-and (R)-enantiomers (Figure 3), we note that the phenoxide group of the substrate is anchoredb yT yr19 and Ty r21. On the other hand, the cyanide nucleophile is positioned by Wat1, which in turn formsh ydrogenb onds with Thr76a nd Thr106. Them ain difference between the transition states in  two pathways is which face of the substrate is exposed to the cyanide nucleophile,t hat is,w here the methylene group of the substrate is pointing. In TS1 and TS2 [favoring (S)-3b], the methylene group points toward the side-chains of Va l78a nd Ile93, while in TS1' ' and TS2' ' [favoring (R)-3b]t he methylene points toward am ore crowded area where the Ty r39 side chain is located. Similar interactions were concluded to be responsiblefor the stereoinduction in the hydration reaction. [19a]
Quantum mechanical calculations revealed details on the mechanisma nd identify steric interactions responsible for the stereochemical outcome.

Preparation of Biocatalysts:Cloning and Heterologous Expression
Theg enes encodingf or the respective PA Ds and FDC were transformed in E. coli BL21(DE3)a nd heterologously expressedasd escribed previously. [17a] Site-Directed Mutagenesis Site-directed mutagenesis was carried out with the Quik-Change PCR mutagenesis kit from Stratagene using the respective primers equences listed in the Supporting Information, Table S3.

Nucleophile Screening
Lyophilized E. coli whole cells containing heterologously expressed FDC_Es( 20 mg) were rehydratedf or 30 min at 700 rpm shaking in KP i buffer (887 mL; 100 mM, pH 7.6) in 2.0 mL reaction vials.Astock solution containingt he respective nucleophile in either buffero r1 ,2-dimethoxyethane (DME),d epending on the compoundss olubility (100 mL, 1M), was added to the cell suspension followed by short mixinga nd addition of the substrate 4-vinylphenol 1 (13.4 mLo fa8.6% w/w solution in propylene glycol). The mixture was incubated for 24 ha t3 08 8Ca nd 700 rpm in an Eppendorf Thermoshaker. Then the mixture was split into two equal aliquots µ 500 mL. One aliquot was extracted with ethyl acetate (2 500 mL) and after drying over MgSO 4 the organic phase was subjected to GC-MSa nalysis.T he other half was diluted with acetonitrile containinga nisolea si nternal standard (10 mM) for quantification of substrate 1 and hydrate 4 using HPLC. Putative adducts were identified by means of MS fragmentation patterns.

QM Calculation Details
All the calculationsw ere performed using the B3LYP density functional method, [30] as implemented in the Gaussian 09 program. [31] Geometries were optimized with the 6-31G(d,p) basis set and more accurate energies were obtained by single-point calculationso nt he optimized structures with the larger basis set 6-311 + + G(2d,2p). Single-point solvation energies with SMDm ethod [32] were calculateda tt he same level as the geometry optimization using e = 4. Frequency calculations were performed the same levela st he geometry optimization to obtain zero-point energies (ZPE). Dispersion corrections were added using the DFT-D3(BJ) method. [33]