Site-selective enzymatic C‒H amidation for synthesis of diverse lactams

See allHide authors and affiliations

Science  10 May 2019:
Vol. 364, Issue 6440, pp. 575-578
DOI: 10.1126/science.aaw9068

This article has been retracted. Please see:

The right bond in the right place

Enzymes excel at specificity because of their constrained active sites. With appropriate evolutionary pressure, they can be made to differentiate between similar substrates or between positions on a single substrate. Cho et al. used directed evolution to generate cytochrome P450 variants that target different C–H bonds in substrates, forming lactam rings of varying size (see the Perspective by Hepworth and Flitsch). The enzyme directs amidation to the desired position and simultaneously prevents other side reactions.

Science, this issue p. 575; see also p. 529


A major challenge in carbon‒hydrogen (C‒H) bond functionalization is to have the catalyst control precisely where a reaction takes place. In this study, we report engineered cytochrome P450 enzymes that perform unprecedented enantioselective C‒H amidation reactions and control the site selectivity to divergently construct β-, γ-, and δ-lactams, completely overruling the inherent reactivities of the C‒H bonds. The enzymes, expressed in Escherichia coli cells, accomplish this abiological carbon‒nitrogen bond formation via reactive iron-bound carbonyl nitrenes generated from nature-inspired acyl-protected hydroxamate precursors. This transformation is exceptionally efficient (up to 1,020,000 total turnovers) and selective (up to 25:1 regioselectivity and 97%, please refer to compound 2v enantiomeric excess), and can be performed easily on preparative scale.

Functionalization of unactivated carbon‒hydrogen (C‒H) bonds represents a sought-after strategy to construct and diversify molecules, markedly reshaping the logic of synthetic chemistry (15). Because organic molecules often contain multiple, similar C–H bonds, a major challenge is to control the site at which a reaction takes place. Current approaches to targeting specific C‒H bonds rely mainly on substrate control, wherein the substrate guides the catalyst with directing groups (6, 7) or bears one C–H bond that is inherently more reactive (812). The former necessitates extra synthetic steps to manipulate directing groups, and both approaches limit the breadth of potential substrates. An appealing approach is to use a modular catalyst platform that could be tuned to deliver different site selectivities, thereby enabling regiodivergent C‒H functionalization of a broad range of substrates (1316).

Enzymes, the catalytic machinery of the biological world, exert exquisite control over selectivity in biochemical transformations (17). This is exemplified by cytochrome P450 enzymes, which can hydroxylate a specific C‒H bond within a complex molecule (18, 19). Furthermore, this site selectivity can be tuned by protein engineering techniques such as directed evolution (2022). In the past few years, our group and others expanded the boundaries of biocatalytic C‒H functionalization by repurposing native cytochrome P450s to perform carbene and nitrene transfer reactions (2325). These enzymes have not yet been engineered to control the site selectivity of the nonnatural reactions in a manner comparable to what has been achieved with P450-catalyzed hydroxylation.

C‒H insertion of sulfonyl nitrenes can be catalyzed by variants of cytochrome P450BM3 having the axial heme-ligating cysteine amino acid residue substituted with serine (so-called cytochrome P411s) (26, 27). As analogs of sulfonyl nitrenes, carbonyl nitrenes are synthetically more useful owing to their ease of manipulation and wide applicability of the resulting amide products. However, carbonyl nitrenes are difficult to use for C‒H functionalization due to their facile decomposition to isocyanates through a well-known Curtius-type rearrangement (28). Chang and co-workers reported that an iridium-based catalytic system could harness carbonyl nitrenes for intramolecular C‒H amidation; however, these catalysts were not enantioselective and formed only the thermodynamically favored products (29). We hypothesized that cytochrome P450s and other hemeproteins could be engineered to generate reactive carbonyl nitrenes, direct them to C–H amidation over competing decomposition pathways, and precisely control the enantioselectivity and site selectivity of the reaction.

Rather than using the carbonyl analogs of the sulfonyl azides to generate enzyme-bound nitrenes, we sought a more easily accessible and biocompatible nitrene precursor. Tsutsumi and co-workers identified a naturally occurring nitrene-transfer reaction catalyzed by a cytochrome P450 enzyme in which the nitrene is generated from an acyl-protected hydroxylamine (30). We reasoned that acyl-protected hydroxamates 1, which are prepared from carboxylic acids in one step (Fig. 1A), could act as nitrene sources in enzyme-catalyzed reactions (3133).

Fig. 1 Hemeprotein-catalyzed intramolecular C‒H amidation.

(A) Design of acyl-protected hydroxamate 1 as nitrene precursor, inspired by naturally occurring nitrene precursors in the biosynthesis of benzastatins (30) (Ac, acetyl; Piv, pivaloyl). (B) Reaction scheme and proposed catalytic cycle of intramolecular C‒H amidation catalyzed by the P411 variant. (C) Crystal structure of a variant closely related to P411 E10 (Protein Data Bank ID: 5UCW) (27), with mutated residues marked in blue. (D) Directed evolution of E10FA to LSsp3 for the synthesis of β-lactam 2a and further improvement of TTN by process optimization. Experiments were typically performed at analytical scale using suspensions of E. coli cells expressing E10FA variants [optical density at 600 nm (OD600) = 2.5] in KPi (0.1 M, pH 8.0) buffer, 50 mM substrate 1a, 5 volume % organic cosolvents, and 400 μl of reaction volume at room temperature under anaerobic conditions for 36 hours. See supplementary materials for further details. TTN is the total desired product as quantified by comparison to standards by gas chromatography or HPLC, divided by total cytochrome P411. Single-letter amino acid abbreviations (here and in Fig. 3): A, Ala; E, Glu; F, Phe; G, Gly; L, Leu; M, Met; N, Asn; P, Pro; R, Arg; T, Thr; V, Val; Y, Tyr.

Motivated by the extensive pharmaceutical applications of β-lactams (34), we challenged hemeproteins to construct these strained, four-membered rings through intramolecular C‒H amidation (3537). We chose as the model substrate hydroxamate 1a with two sets of reactive C‒H bonds, potentially affording β-lactam 2a through C(sp3)–H amidation or δ-lactam 3a through C(sp2)‒H amidation (Fig. 1B). We first tested whether free heme could catalyze C‒H amidation in biocompatible conditions and observed alkyl amine 4 as the sole product, presumably from Curtius-type rearrangement of the carbonyl nitrene followed by hydrolysis (table S1). We then tested a panel of hemeprotein variants, including P450s, P411s, cytochromes c, and globins, under anaerobic conditions. Most evaluated hemeproteins, including the P411 variants previously engineered for C‒H insertion of sulfonyl nitrenes (26, 27), only generated amide 5, presumably through reduction of the carbonyl nitrene intermediate. A few hemoproteins of various lineages, however, produced small amounts of the lactam products (tables S1 and S2). The P411 variant E10-V78F S438A (E10FA), originally engineered to catalyze carbene transfer to alkynes (38), was able to transform substrate 1a to the desired β-lactam (2a), giving only 1.5% yield but 90% enantiomeric excess (ee). The δ-lactam 3a was also detected, with 1.0% yield. In stark contrast to the reaction catalyzed by free heme, amine 4 was not observed in the reaction using this hemeprotein, supporting our hypothesis that an enzyme could divert highly unstable carbonyl nitrenes to the desired C–H amidation and bypass the Curtius-type rearrangement.

We chose E10FA as the parent for directed evolution of lactam synthases (LS). To improve catalytic activity as well as site- and enantioselectivity of C‒H amidation for the synthesis of β-lactam 2a, we targeted amino acid residues located near the heme iron in a closely related crystal structure (P411 E10) for sequential rounds of site-saturation mutagenesis and screening (Fig. 1C). In each round, the libraries of enzymes were expressed and screened as whole-cell catalysts in 96-well plates for increased product formation and/or enantioselectivity. Two rounds of site-saturation mutagenesis and screening introduced the mutations F78L and A264G, leading to a 28-fold improvement in total turnover number (TTN) to 52,000 for β-lactam 2a while almost eliminating formation of δ-lactam 3a (Fig. 1D). Double site-saturation mutagenesis and screening subsequently added mutations T327R and V328M, which further increased TTN to 71,000 and ee to 96%. Another round of double site saturation led to the discovery of the final variant, LSsp3, containing mutations L263N and A268F, which affords the desired β-lactam 2a with 223,000 TTN (96% yield) and 96% ee. Adding both the E. coli expressing LSsp3 and precursor 1a portion-wise to a reaction mixture boosted the TTN for β-lactam 2a to 1,020,000 (see supplementary materials for details).

With variant LSsp3 in hand, we evaluated a collection of hydroxamates for enantioselective β-lactam synthesis (Fig. 2A). Various aryl substituents were accepted to furnish β-lactams 2b2g with excellent TTN (up to 180,000) and ee (up to 94%); only trace amounts of δ-lactams through C(sp2)‒H amidation were detected. Introducing two methyl groups adjacent to the carbonyl group did not hamper the reaction, leading to the formation of product 2h. Olefin-containing substrates could also undergo enantioselective C‒H amidation (2i, 2j) with no competing aziridination observed, highlighting the enzyme’s complete chemoselectivity. An α-alkynyl C‒H bond was also readily amidated (2k). Substrates with heteroatom substituents, such as amide and silyl groups, reacted well to give the desired products 2l and 2m.

Fig. 2 Scope of β-lactam products.

(A) Synthesis of β-lactams from corresponding hydroxamate substrates using LSsp3. Experiments were typically performed at analytical scale using LSsp3-expressing E. coli cells resuspended to OD600 = 2.5 or 3.0 in KPi (0.1 M, pH 8.0) buffer, 25 mM or 50 mM substrate 1, 5 volume % organic cosolvents, and 400 μL of reaction volume at room temperature under anaerobic conditions for 36 hours. (B) Gram-scale synthesis of β-lactam 2s using cell lysate of E. coli expressing LSsp3. See supplementary materials (table S5) for details and more scale-up reactions. In all cases, rr > 20:1.

The reactive C–H bonds in all the preceding substrates were weakened by adjacent substituents. This is not required, however, as evidenced by the ester-substituted β-lactam 2n, for which the turnover was not diminished. Substrates with secondary and tertiary aliphatic C‒H bonds were also accepted, leading to products 2o2p. The site selectivity of the enzyme was highlighted by the formation of product 2q from a substrate with multiple similar aliphatic C‒H bonds. Even aliphatic C‒H bonds adjacent to a trifluoromethyl group were amenable to C‒H amidation, affording product 2r. The transformation could be applied to preparative-scale synthesis without sacrificing yield or enantioselectivity (Fig. 2B). Isolation of most products was straightforward: product 2s, for example, was simply filtered from the aqueous reaction mixture, resulting in 86% yield [1.62 g, 92% high-performance liquid chromatography (HPLC) purity] and 92% ee.

Having achieved β-lactam synthesis by intramolecular C(sp3)‒H amidation, we next investigated whether other types of C‒H bonds could be amidated selectively. The parent enzyme E10FA showed trace reactivity toward C(sp2)‒H bonds; only two rounds of site-saturation mutagenesis and screening generated variant LSsp2, with mutations F78A and T327P, that exclusively functionalized C(sp2)−H bonds (Fig. 3A). Instead of forming the corresponding β-lactams (2a, 2s2u), LSsp2 diverted the substrates to the corresponding δ-lactams (3a, 3s3u) with comparable TTNs (Fig. 3B).

Fig. 3 Engineering lactam synthases for regiodivergent intramolecular C‒H amidation.

(A) Evolutionary trajectory of variants LSsp2, LSsp3, LSβ, and LSγ. (B) Selectivity and scope of LSsp2- and LSsp3-catalyzed intramolecular C‒H amidation. (C) Selectivity and scope of LSβ and LSγ. (D) Regiodivergent amidation of aliphatic, homobenzylic, and benzylic C(sp3)‒H bonds catalyzed by LSβ, LSγ, and LSsp3. Regioisomeric ratio (rr) indicates the mole ratio of major product to combined minor regioisomers. See supplementary materials (tables S6 to S8) for details.

To further demonstrate the tunability of enzyme-catalyzed C‒H amidation, we synthesized precursor 1v with two sets of C(sp3)‒H bonds, potentially affording β-lactam 2v or γ-lactam 6v (Fig. 3C). Variant LSsp3 gave β-lactam 2v as the major product, together with 6v in a 2.2:1 ratio. Sequential site-saturation mutagenesis and screening delivered variant LSβ bearing mutations R327P, P329L, and A330V, which produces β-lactam 2v with a regioisomeric ratio (rr) up to 25:1 without diminishing TTN and ee. As shown in the synthesis of β-lactams 2w2y, this variant enforces high selectivity for the β-lactam products despite the presence of weaker benzylic C‒H bonds and the option to form less-strained γ-lactams. Meanwhile, variant LSγ was also revealed, which has mutations P329V A330M and can selectively synthesize γ-lactam 6v (Fig. 3C). The site selectivity of LSγ was largely insensitive to aromatic substitution, and γ-lactams 6w6y were afforded with up to 25:1 rr, 209,000 TTN, and 97% ee (Fig. 3C).

Finally, we challenged the enzymes with nitrene precursor 1z, which has three sets of reactive C(sp3)−H bonds. Variants LSβ, LSγ, and LSsp3 selectively formed β-lactam 2z, γ-lactam 6z, and δ-lactam 3z, respectively. The site selectivity shown in this example further showcases the powerful regiocontrol of the enzymes that is tunable by directed evolution (Fig. 3D).

We have engineered a “lactam synthase” that can be tuned by directed evolution to convert individual substrates into different lactams through a catalyst-controlled C‒H amidation process. Reactivity trends due to bond strength, inductive effects, steric accessibility, or ring strain could be completely overturned in this catalyst-controlled process. With these results, we propose that genetically tunable enzymatic catalysis may provide a general strategy to address the challenge of site selectivity in C‒H functionalization.

Supplementary Materials

Materials and Methods

Tables S1 to S8

NMR Spectra

References (3958)

References and Notes

Acknowledgments: We thank K. Chen, S. Brinkmann-Chen, D. C. Miller, and D. K. Romney for comments on the manuscript; X. Huang for helpful discussions; K. Chen for the parent variant E10FA; and the Caltech Center for Catalysis and Chemical Synthesis and the Caltech Mass Spectrometry Laboratory for analytical support. Funding: This work was supported by NSF Division of Molecular and Cellular Biosciences grant MCB-1513007; the Joseph J. Jacobs Institute for Molecular Engineering for Medicine (I.C.); and the Deutsche Forschungsgemeinschaft (JI 289/1-1 to Z.-J.J.). Author contributions: I.C. and Z.-J.J. designed the research and performed the experiments with guidance from F.H.A. I.C., Z.-J.J., and F.H.A. wrote the manuscript. Competing interests: I.C. and Z.-J.J. are inventors on patent application CIT-8187-P submitted by the California Institute of Technology that covers the intramolecular and intermolecular enzymatic C–H amidation. Data and materials availability: All data are available in the main text or the supplementary materials. Plasmids encoding the enzymes reported in this study are available for research purposes from F.H.A. under a material transfer agreement with the California Institute of Technology.

Stay Connected to Science

Navigate This Article