Chemoenzymatic Cascade Synthesis of Optically Pure Alkanoic Acids by Using Engineered Arylmalonate Decarboxylase Variants

Abstract Arylmalonate decarboxylase (AMDase) catalyzes the cofactor‐free asymmetric decarboxylation of prochiral arylmalonic acids and produces the corresponding monoacids with rigorous R selectivity. Alteration of catalytic cysteine residues and of the hydrophobic environment in the active site by protein engineering has previously resulted in the generation of variants with opposite enantioselectivity and improved catalytic performance. The substrate spectrum of AMDase allows it to catalyze the asymmetric decarboxylation of small methylvinylmalonic acid derivatives, implying the possibility to produce short‐chain 2‐methylalkanoic acids with high optical purity after reduction of the nonactivated C=C double bond. Use of diimide as the reductant proved to be a simple strategy to avoid racemization of the stereocenter during reduction. The developed chemoenzymatic sequential cascade with use of R‐ and S‐selective AMDase variants produced optically pure short‐chain 2‐methylalkanoic acids in moderate to full conversion and gave both enantiomers in excellent enantiopurity (up to 83 % isolated yield and 98 % ee).


Introduction
Enantiopure2 -methyl-substitutedc arboxylic acids are widely used as active pharmaceutical ingredients, building blocks, and fragrance and aroma compounds. [1,2] For instance, 2-methylbutanoic acid derivatives are present in aw ide range of fermented products such as bread, cheese, and severala lcoholic beverages, contributing to their complex flavor. [3] (S)-2-methylbutanoic acid is aprecursor for the synthesis of the cholesterol-lowering drug pravastatin. [4] (R)-2-Methylbutanoic acid is part of the sex pheromone of the invasive species Acutaspis albopicta. [5] (S)-2-Methylhexanoica cidi saconstituent of the cytotoxic marine natural compoundpalau'imide. [6] An efficient access to both pure enantiomerso fs hort-chain 2-methyl-substituted alkanoic acids would therefore pose as ignificant contribution to the development of numerous fine chemicals. Although aw ide range of biocatalytic methodsf or the synthesis of arylaliphatic 2-substituted carboxylic acids have been developed, including esterases, dehydrogenases, amidases, and decarboxylases, [1] the synthesis of smallc hiral aliphatic carboxylic acids is challenging becauseo ft he difficulty to discriminate between two structurally very similars ubstituents. Lipase-catalyzed kinetic resolution of 2-methylalkanoic acids achieved moderate optical purities,b ut is inherently limited to 50 %m aximum yield. [7] Asymmetric hydrogenation over ah eterogeneous Pd 0 catalyst in the presenceo fc inchonad erivatives as the chirall igands led to moderate enantioselectivities. [8] By using BINAPa nd Ru II , as eries of tiglic acid derivativesw as reduced with optical purities ranging from 79 to 97 % ee. [9] These recent achievements underline the difficulty to prepare these challenging compounds.
Several AMDasev ariants have been generated by protein engineering to alter the enzymatic performance such as the in-troduction of artificialr acemase activity, [18] complete inversion of enantioselectivity, [19] and activity improvemento fb oth Rand S-selective AMDasevariants. [20][21][22] Owing to these investigations, AMDased ecarboxylation is now applicable for the asymmetric synthesis of both enantiomerso f2 -substituted propionates with high opticalp urity.O krasa et al. reportedt hat wildtype AMDase catalyzes the asymmetric decarboxylation of methylvinylmalonic acids with outstanding R enantioselectivity. [15] This substrate specificity of AMDase suggestst he potential applicability of AMDasea nd its variants to produce both enantiomers of short-chain 2-methylalkanoic acids. Considering that AMDase does not accept substrates without ad elocalized p-electron system, [10,15] we envisioned ac ascade by combining the enzymatics ynthesis of intermediary 2-methylalk-3-enoic acids followed by ac hemicalC =Cd ouble bond reduction.
The development of chemoenzymatic cascade reactions, that is, one-pot consecutive multiple reactions combining enzymaticr eactions with chemical catalysis,h as recently received increasing attention. [23][24][25] This reaction concept allows to combine the strengths of biological and chemical catalysts in a one-pot reaction system,s ave downstream work-up processes, and addresses challenges such as the instability of reactioni ntermediates. Especially,v ariousc atalytic reactions with use of transition-metal catalysts have been successfully combined with enzymatic reactions under aqueousorn onaqueous conditions. [26][27][28][29] The chemocatalytic reactions incorporated in such chemoenzymatic cascade systemss of ar,s uch as cross coupling [26] and metathesis, [29] greatlye xpand the catalytic scope of biocatalysis. Indeed, the hydrogenation of nonactivated C=C double bondsi se xtremely difficult for biocatalysis:F or example, ene-reductases catalyze an asymmetric hydrogen transfer towardsactivated2 ,3-unsaturated compounds, but are inactive towardsn onactivated alkenes. [30] The gut bacterium Lactobacillus plantarum possesses ap olyunsaturated fatty acid metabolism pathway,i nw hich as ingle nonactivated double bond of linoleic acid is reduced by using four different enzymesi ns ix reactions teps. [31] On the contrary,t he hydrogenation of non-polarizedC =Cd oubleb onds is frequently performed with use of transition-metal catalysts.
Herein, we report ac hemoenzymatic cascade reaction to produce optically pure short-chain 2-methylalkanoic acids by combining the enzymatic asymmetric decarboxylation of methylvinylmalonic acid derivatives 1 and the chemical reduction of the nonactivated C=Cd ouble bond (Scheme 1).

Results and Discussion
For the construction of this cascade system, several facts needed to be considered. First, although the activity of AMDasew ild-type (AMD-WT) had already been reported in the synthesis of (R)-2,t he catalytic performance of other AMDase variantsr emained to be confirmed. Especially the mutations in the S-selective AMDase variants with as hifted catalytic cysteine residue (G74C/C188G) were expected to significantly affect substrate recognition in the active site. Second, the chemicalr eduction should not show any side reactivity.I ti s known that transition-metal catalysts can cause double bond migrationt of orm the a,b-unsaturated compound, [32] which may lead to ap artial racemization of the optically pure intermediates 2.T hird, the reactionc onditions should be compatible with both reactions teps. Therefore, ap otential inhibition of the chemical reduction had to be carefullye valuatedi nt his designedc ascade.
To evaluate the effect of mutations within the hydrophobic pocket (V43I/A125P/V156L/M159L) and/or the catalytic cysteine residue (G74C/C188G), the specific activities of four AMDase variants,t wo being R-selective (AMD-WT and V43I/A125P/ V156L/M159L, AMD-IPLL [22] )a nd two being S-selective (G74C/ M159L/C188G, AMD-CLG, [20] and V43I/G74C/A125P/V156L/ M159L/C188G,A MD-CLGIPL [21] ), towardst he substrates 1 were determined (Figure2 and Ta ble S1). AMD-CLGIPL (with mutations on the hydrophobic pocket and the catalytic cysteine residue) showed a2 5-to 65-fold activity decrease compared to AMD-IPLL (with mutationso nt he hydrophobic pocket). This fact suggestst hat the enantioselectivity-invertings hift of the cysteineresidue in the active site critically disrupts the catalytic performance of AMDaset owards the alkenyl substrates 1.N evertheless, both R-a nd S-selective variants with mutationsi n the hydrophobic pocket, AMD-IPLL and AMD-CLGIPL, showed 3.5-10, respectively 1.7-3.2 times higher activity than their corresponding prior generationsA MD-WTand AMD-CLG. The hydrophobic pocket causes ground-state decarboxylation of the pro-R carboxyl residue of the substrates 1 (Figure 1). This implies that astructuraloptimization of the hydrophobic environment, leading to enhanced conversion rates for arylmalonic acids, [20][21][22] also positively affectss ubstrate decarboxylation. It should be noted that the activity of AMD-IPLL and AMD-CLGIPL towards at ypicala rylmalonic acid derivative (up to 209 and 55 Umg À1 ,r espectively) [22] is 19-170 and 190-2200 times higher, respectively,t han that towards the here-discussed alkenyl substrates 1.T hism ay in part reflectt he lower capacity of the single C=Cd ouble bond to stabilize the enolate intermediate during the transition statec ompared to that of an aromatic system. Next, the enantioselectivity of the improved variants AMD-IPLL and AMD-CLGIPL wase valuated. Enzymatic reactions with use of crude cell extract including either of the AMDase variants showed full conversion of 1 at ac oncentration of 10 mm and generatedt he monoacids 2 with high optical purities with one exception (Table 1): S-selective AMD-CLGIPL produced (S)-2a with only 66 % ee (S). The same value was obtained from experimentsw ith purified enzyme, ruling out any racemizing side reactions from the cellular debris. In control reactions, neither significant spontaneousd ecarboxylation in reactions without enzyme nor significant racemization of product 2 during prolonged incubations in the presence of crude cell extract were observed.
Whilst (R)-2a is produced with high opticalp urity by both, the R-selective AMD-WT [15] and AMD-IPLL( entry 1i nT able 1), the stereo-determining double mutation G74C/C188G in variant AMD-CLGIPL can alter the enzyme-ligand binding. This may hypothetically result in ar educed discriminationb etween the methyla nd vinyl substituents of the substrates within the mutated actives ite, eventually leading to am arked decrease of enantioselectivity for the smalls ubstrate 1a.O ther possible explanations for this unexpected result may be rearrangement processes of the ligand in the transition state or reprotonation by an alternate, unwanted, proton donor such as water.
Havingc onfirmed the possibility to produce enantiomerically enriched 2,w et hen investigated the subsequentr eduction. The nonstereoselectiver eduction of nonpolarized C=C double bondsi st ypicallyaccomplished with heterogenous catalysts such as Pd 0 ,R h 0 ,R aney nickel, or PtO 2 under aH 2 atmosphere. [33] For the reduction of optically pure 2,t he tendency of transition metals to promote isomerization reactions [32] poses a risk. After completed ecarboxylation of 1d (10 mm)c atalyzed by crude cell extract containingA MDase IPLL, addition of Pd/C and H 2 resulted in complete reduction of the intermediary (R)-2d within 1.5 h. The resulting (R)-3d could be isolated in 83 % yield, showing that the combination of the reducing catalyst with crude cell extracts is straightforward. As expected, however,t he Pd 0 catalysti nduced partial isomerization of the stereocenter: While (R)-2d was formed as pure enantiomer (> 99 % ee), the final product (R)-3d had as ignificantly reduced optical purity (81 % ee).
As an alternative C=Cd ouble bond reduction approach, we investigated the in situ generation of diimide. Diimide is ah ydrogen donor that can selectivelyr educe nonpolarized unsaturated bondsb yaconcerted hydrogen transfer without isomerization. [34] Diimide itselfi su nstable and has to be generated in situ from its precursor hydrazine by using oxidation catalysts. Several catalysts have been reported fort he in situ generation of diimide, [35][36][37][38] and those reactions were performed in organic solventi nm ost cases. To circumvent extra work-up processes, aw ater-soluble Cu II catalyst, [39] CuCl 2 ,w as selected for hydrazine oxidation. The reduction activity wasc onfirmed by using am odel substrate, trans-3-hexenoic acid 4,u nder aqueous conditions ( Table 2, entry 4).
The sequential chemoenzymatic cascade reaction was then performed by combining AMDased ecarboxylation and C=C double bond reductiont hrough the in situ generated diimide by using ac opper salt. GC-FID analysiso ft he final product 3 presented moderate to full conversion over two steps and, more importantly,aconserved high enantiomeric excess (Table 3). This result implies that the reduction activity of diimide was not influenced by contaminants from E. coli cell free extract and that racemizing or inhibiting side reactions could  be circumvented with this approach. Interestingly,t he lowest reduction rate was observed for the productiono f( R)-and (S)-3c (Table 3, entries 5a nd 6). Diimide selectively reducest erminal olefins and is less reactive towards internal and especially multisubstituted double bonds. [38] With use of 1a on am illigram-scale (110 mg, 0.76 mmol), the reaction proceededt o complete conversion over both steps and allowed the isolation of (R)-3a with excellent optical purity (98 % ee)a nd 83 %y ield (64 mg, 0.63 mmol).

Conclusion
The feasibility to combine enzymatic decarboxylation and chemicalr eduction was demonstrated in this study.T he here developed chemoenzymatic reaction cascadec an be operated in sequential mannert og ive access to optically pure shortchain 2-methylalkanoic acids 3.
Our work presents ap otential synthetic applicability of AMDaseb iocatalysis in combination with metallo-and organochemicalr eactions in the same reactions ystem, buta lso points out difficulties with side reactivitiesa nd selectivity issues in the combinationo ft he different catalysts. The hydrogenationo fn onactivated C=Cd ouble bonds by using ah eterogeneous Pd 0 catalyst initially resulted in significantly decreased optical purity of the final products 3,w hich is most probablyd ue to partial isomerization. An alternative reduction approachb ya pplying in situ generated diimide as the reductant produced the final products 3 in moderate to full conversion and with conserved high enantiomeric excess. While partial racemization could be overcome by the choice of an appropriate reductiona pproach, the reduced selectivity of S-se-lectiveA MDasev ariants towards the shortest alkenyl substrate 1a may be addressed in future work.

Determination of enzymatic activity
Purified AMDase wild-type and its variants were used for specific activity measurement towards alkenyl substrates 1.T he enzymatic  Analytical scale chemoenzymaticone-pot two-step combination of enzymatic decarboxylation and C=Cr eduction The first-step decarboxylation was performed on 0.5 mL scale with AMDase variants IPLL and CLGIPL. The reaction conditions were: substrate 1 (10 mm),T ris-HCl (50 mm, pH 8.0 at 30 8C), 10 % v/v (IPLL) or 50 % v/v (CLGIPL) of the crude extract of AMDase variants (ca. 75 mg wetted cell pellet mL À1 )a t3 08C. Substrate conversion was followed by HPLC analysis as described above. After full conversion of the substrates 1,t he second-step C=Cr eduction was performed by addition of hydrazine monohydrate (20 equiv) and CuCl 2 (0.01 equiv). Preparative chemoenzymaticone-pot two-step reaction to produce3a 2-Methyl-2-alkenylmalonate 1a (110 mg, 0.76 mmol) was dissolved in Tris-HCl buffer (40 mL, 50 mm, pH 8.0 at 30 8C) containing 10 %v/ vo ft he crude extract of an AMDase variant IPLL (ca. 75 mg wetted cell pellet mL À1 ). The biocatalysis was performed at 30 8Ca ss tated above for 4h and full conversion of substrate 1a was confirmed by TLC control. Next, hydrazine monohydrate (20 equiv) and CuCl 2 (0.01 equiv) were added to the reaction solution and the mixture stirred at 30 8Cu nder aerobic conditions. After the addition of 2 m HCl, the carboxylates were extracted with diethyl ether.T he combined organic phases were dried with anhydrous MgSO 4 and the solvent was evaporated. The final product 3a was isolated by flash column chromatography (diethyl ether/n-pentane = 1:1). Isolated yield:83% [64 mg, 0.63 mmol, 98 % ee (R)].