Biocatalytic production of bicyclic β-lactams with three contiguous chiral centres using engineered crotonases

There is a need to develop asymmetric routes to functionalised β-lactams, which remain the most important group of antibacterials. Here we describe biocatalytic and protein engineering studies concerning carbapenem biosynthesis enzymes, aiming to enable stereoselective production of functionalised carbapenams with three contiguous chiral centres. Structurally-guided substitutions of wildtype carboxymethylproline synthases enable tuning of their C-N and C-C bond forming capacity to produce 5-carboxymethylproline derivatives substituted at C-4 and C-6, from amino acid aldehyde and malonyl-CoA derivatives. Use of tandem enzyme incubations comprising an engineered carboxymethylproline synthase and an alkylmalonyl-CoA forming enzyme (i.e. malonyl-CoA synthetase or crotonyl-CoA carboxylase reductase) can improve stereocontrol and expand the product range. Some of the prepared 4,6-disubstituted-5-carboxymethylproline derivatives are converted to bicyclic β-lactams by carbapenam synthetase catalysis. The results illustrate the utility of tandem enzyme systems involving engineered crotonases for asymmetric bicyclic β-lactam synthesis.

β -lactams are vital antibiotics and are finding new therapeutic applications [1][2][3][4] . Most bicyclic β-lactams (e.g. penicillins and cephalosporins) are produced by fermentation, or modification of fermentation-derived materials. Carbapenems, which are used for treatment of infections, including multidrugresistant bacteria 5 , are an exception. Carbapenems, which have at least three chiral centres, are produced by synthesis with consequent cost implications and limitations on derivatives that can be produced. The carbapenem substitution pattern affects their activities and pharmacokinetic profiles 6 . All clinically used carbapenems have the (6R)-hydroxyethyl sidechain (Fig. 1a) and most of them are C-1 substituted, in order to increase potency and avoid hydrolysis by dehydropeptidases 7,8 . There is a need to develop efficient asymmetric routes for antibiotic production, where cost of goods is important. With a view to enabling routes to functionalised bicycle β-lactams, in particular C-1/C-6-functionalised bicyclic β-lactams as in carbapenems, we are investigating engineering of carbapenem biosynthesis enzymes [9][10][11] .
The C-6 sidechain of natural C-1/C-6-functionalised carbapenems is likely introduced at a late stage during biosynthesis, making the engineered production of C-6 carbapenem analogues challenging 21,22 . Thus, there is interest in biocatalytic systems for stereocontrolled synthesis of carbapenem precursors functionalised at the C-1 and C-6-equivalent positions.
These results provide further insights into CMPS selectivity. Consideration of the non-observed potential of CMPS products (Fig. 5a) in the light of crystallographic analyses implies a role for steric clashes in determining product outcomes. Thus, a clash between the methyl group of (4R)-4-methyl-L-P5C and the methyl group of the (E)-enolate (or a precursor of) may be responsible for their apparent lack of reaction with the tested CMPSs (Fig. 5b). The stereoselectivity of CarB variants with a βbranched residue (Val, Ile) at the OAH-forming residue-108 for formation of products with either (6R)-or (4S)-stereochemistry (Fig. 4, green-shaded boxes) can be rationalised on steric grounds, i.e. a clash between the methyl group of the (E)-enolate and the methyl group of (4R)-4-methyl-L-P5C (Fig. 5b), or with the βmethyl of the 108-valine/isoleucine residue is disfavoured (Fig. 5c). Thus, CarB variants with a β-branched 108-residue favour formation of the (4S,6R)-stereochemistry; this stereoselectivity is improved by Q111 substitution with Asn or Ala (Table 1, entry 5, Fig. 4, green-shaded boxes), possibly due to enhanced productive binding of the (4S)-4-methyl-L-P5C stereoisomer. On the other hand, we propose ThnE variants without a β-branched residue (Met, Leu and Ala) at residue-153 to favour the formation of (4S,6S)-stereochemistry products (Table 1, entry 6, Fig. 4, grey-shaded boxes), because of a preference to productively bind (4S)-4-methyl-L-P5C 10 and hence form an (E)-enolate 25 .
In addition to mechanistic implications (Fig. 5), these results demonstrate the capacity of engineered CMPSs to catalyse formation of 4,6-alkyl-substituted t-CMP derivatives in high stereoselectivity. Although our 'isolated' yields are relatively low, given the micro-scale and non-optimised nature of the reactions, there is likely scope for improvement.
The hydrolytic stability of unsubstituted carbapenams/carbapenems is reportedly low, to the extent that their isolation in the free form (rather than as ester derivatives) has not been readily possible [35][36][37][38] . We found that 1,6-disubstituted carbapenams are hydrolysed more slowly than their unsubstituted or monosubstituted analogues 10 , which undergo hydrolysis during LC-MS -guided purification/lyophilisation as evidenced by NMR. By contrast, the t 1/2 of the (1S,3S,5S,6S)-1,6-dimethyl carbapenem c Methoxymalonyl-CoA formation as catalysed by MatB and its one-pot reaction with 4-methyl-L-GHP, as catalysed by a CMPS to give the three shown stereoisomers. We propose that the nascent methoxymalonyl-CoA product of MatB catalysis is either epimeric or has the (2R)-stereochemistry, analogous to other MatB reactions (as in c) 24,34 , but undergoes relatively rapid epimerisation, consistent with the observed t-CMP products was~42 days by NMR (4°C, sodium formate pH ∼7), revealing the stabilising effects of C-4/C-6 substitution.

Discussion
The stereocontrolled synthesis of heterocycles, such as bicyclic βlactams, with contiguous stereocentres is a challenge in development of natural products/natural product like drugs. Our results highlight the utility of engineered crotonases, and more generally enzyme-catalysed reactions proceeding via enolate intermediates, including when coupled with malonyl-CoA-forming enzymes, in addressing aspects of this challenge. We have described reactions with engineered CMPS enzymes with L-P5C giving CMP products substituted at C-6 [23][24][25] . Introducing an epimeric methyl substituent at C-4 of L-P5C 10 , with a view to selectively preparing (4,6)-disubstituted-t-CMP derivatives with the (4S)-stereochemistry, which are potential precursors of 1β-methyl-carbapenams, increases the number of potential products to four stereoisomers (assuming conservation of (5S)-stereochemistry) 10,11,13,25 . The results (Fig. 3, Table 1) reveal the potential of engineered CMPS catalysis for stereocontrolled production of (4,6)-disubstituted-t-CMP derivatives, not only with the desired (4S,6R)-stereochemistry, as in most clinically used carbapenems, but for C-4/C-6-trisubstituted products (i.e. mono-alkylated at one of C-4 or C-6 and dialkylated at one of C-4 or C-6).
In the case of CMPS-catalysed reaction of C-2 epimeric alkylmalonyl-CoA with C-4 epimeric 4-methyl-L-P5C (Fig. 2c), of the four possible stereomeric products, one was not observed   Fig. 8 Conversion of 4,6-substituted-t-CMP derivatives into carbapenams. See Table 2 for substrates, products and conversions under standard conditions, i.e. the (4R,6S)-product. We propose that this is due to a steric clash involving the (E)-trisubstituted enolate and the methyl group of (4R)-methyl-L-P5C (Fig. 5b). This proposal implies scope for further engineering or expanding the scope of CMPS catalysis. Interestingly, substituting one of the oxyanion hole-forming residues (108 CarB /153 ThnE ) has a major impact on C4/C6 stereocontrol; variants with a β-branched residue at this position favour formation of (4S,6R)-products, while ThnE variants lacking a β-branched residue favour formation of (4S,6S)-products (Fig. 5).
The results also reveal the capacity of the tandem MatB/CMPS system to enhance stereoselective formation of certain (4S,6S)disubstituted-t-CMP derivatives, and to expand the range of accepted substrates. Thus, the stereoselectivity of CMPS-catalysed process can be enhanced by coupling an appropriately engineered CMPS with a malonyl CoA synthetase starting from a P5C derivative and an achiral C2-alkylated malonic acid derivative. Except for the case of 2-methoxymalonic acid, coupling MatB catalysis to that of engineered CMPSs enabled stereoselective formation of (4S,6S)-disubstituted-t-CMP derivatives, in some cases with high stereocontrol at C-4 and C-6. Similarly, coupling Ccr to engineered CMPSs enabled stereoselective formation of (4S,6R)-disubstituted-t-CMP derivatives, again with high stereocontrol at C-6 and > 75% stereocontrol at C-4. The range of substrates transformed by the MatB/CMPS pairs, including some with a heteroatom at C-6 is substantial. Some of these were converted by CarA into bicyclic β-lactams demonstrating the viability of the MatB-CMPS-CarA process for production of 1βmethyl-substituted carbapenams. Notably some of these products manifested improved hydrolytic stability compared with the unsubstituted 1β-carbapenams [35][36][37][38] . Thus, although challenges remain in developing the methods described here for the large-scale preparation of useful carbapenems, the results clearly demonstrate that engineering of biosynthesis enzymes has potential for the stereocontrolled production of functionalised bicyclic β-lactam derivatives.

Methods
Preparation of enzymes and variants reported. For details, see Supplementary Methods. All proteins were prepared and purified to > 95% by SDS-PAGE analysis. Mutagenesis of the plasmid-bearing carB or thnE genes was performed according to the QuikChange Site-Directed Mutagenesis Protocol (Stratagene). Supplementary Table 1 gives the oligonucleotide primers used for carB double-variants preparation. For a full list of the variants prepared and tested, see Supplementary Structural assignment of reported catalytic products. A combination of (high)resolution MS and 2D-NMR analysis was employed, as fully detailed within the Supplementary Methods. Stereochemistries were assigned through combined analysis of 3 J HH coupling constants and 2D NOESY, assuming that the (S)-stereochemistry at C-2 is maintained during the acid-mediated deprotection of amino acid semialdehydes and product formation, as has been already confirmed 17 (see Supplementary Figs. 2-38).
Quantification of yields and diastereomeric ratio of the products of CMPS and CarA catalysis. Yields of different products of CMPS and CarA catalysis were calculated using a combination of LC-MS and 1 H NMR spectroscopy, as detailed within the text (Fig. 3, Table 1 and as previously reported 10,11,25 ).

Data availability
Data are available from the corresponding author on reasonable request.