Computational design of catalytic dyads and oxyanion holes for ester hydrolysis

Abstract

Nucleophilic catalysis is a general strategy for accelerating ester and amide hydrolysis. In natural active sites, nucleophilic elements such as catalytic dyads and triads are usually paired with oxyanion holes for substrate activation, but it is difficult to parse out the independent contributions of these elements or to understand how they emerged in the course of evolution. Here we explore the minimal requirements for esterase activity by computationally designing artificial catalysts using catalytic dyads and oxyanion holes. We found much higher success rates using designed oxyanion holes formed by backbone NH groups rather than by side chains or bridging water molecules and obtained four active designs in different scaffolds by combining this motif with a Cys-His dyad. Following active site optimization, the most active of the variants exhibited a catalytic efficiency (k(cat)/K(M)) of 400 M(-1) s(-1) for the cleavage of a p-nitrophenyl ester. Kinetic experiments indicate that the active site cysteines are rapidly acylated as programmed by design, but the subsequent slow hydrolysis of the acyl-enzyme intermediate limits overall catalytic efficiency. Moreover, the Cys-His dyads are not properly formed in crystal structures of the designed enzymes. These results highlight the challenges that computational design must overcome to achieve high levels of activity.

Figure 1

Schematic representation of the reaction catalyzed by the de novo designed esterases and of the employed substrates. A) Representation of the programmed reaction scheme and B) representation of the substrates: tyrosyl ester 1 is the computationally designed substrate, whereas the fluorogenic coumarin ester 2 and the chromogenic p-nitrophenyl ester 3 are utilized to facilitate the activity screens.

Figure 2

Snapshots of the computation design process. A) Representation of the calculated theozyme of the ester substrate framed by the catalytic dyad (Cys-His) and the backbone NH – oxyanion contact. Note that in this case, the backbone NH contact is made by the cysteine itself. B) Image of the conformer ensemble of the tyrosyl ester as created by the software Omega (OpenEye).

Figure 3

Experimental characterization of the active designs and their respective knockout variants. The progress curves of the parental designs are depicted in blue, the traces of the single knockout variants (cysteine) are shown red, traces of the double knockout variants (cysteine and histidine) are illustrated in green, and the histidine knockout variants for ECH13 and ECH19 are shown in black: A) FR29; B) ECH13; C) ECH14; D) ECH19. The enzymes (5 µM) were tested with the coumarin ester 1 (FR29 design: 20 µM; ECH designs: 50 µM) and the reaction progress was monitored by measuring the appearance of the fluorescent coumarin product (excitation wavelength: 340 nm; emission wavelength: 452 nm). For each graph, the amount of substrate that was converted by the active designs after 70–100 sec was set to 100 % and then used to normalize the entire data set. The background was subtracted in all cases and the linear fits were extrapolated to zero substrate conversion.

Figure 4

Inhibition of a FR29 and a ECH19 variant by tyrosyl ester 1. The IC₅₀ of tyrosyl ester 1 was determined by incubating A) FR29 A44S/T112L/V151L ([E] = 2 µM) and B) ECH19 K354P/P364W ([E] = 2 µM) with increasing amounts of the inhibitor ([c] = 0.001 – 300 µM) before 50 µM of coumarin ester 2 was added and the release of coumarin was monitored. The IC₅₀ was determined by curve fitting and converted into the corresponding K_i using the Cheng-Prusoff equation.

Figure 5

Kinetic analysis of the designed esterases: A) Two-phase progress curves of selected de novo designed ester hydrolases. The conversion of 100 µM coumarin ester 1 by 10 µM of FR29 (red), ECH13 (black), ECH19 (green), FR29 T112L (purple) and FR29 A44S/V151L (orange) consists of a initial fast phase followed by a second, slow phase. B) Michaelis-Menten plots of the hydrolysis of coumarin ester 2 and by the in silico designed ester hydrolases and their best evolved variants (red: FR29; blue: ECH14; green: ECH19; black: ECH13; pink: FR29 A44S T112L V151L: purple: ECH19 K354P P364W). Only the slopes of the fast phases were considered for the determination of k₂ and K_M.

Figure 6

Crystal structures of the 4 active designs. In each case, the design model is shown in purple (ligand in cyan) and the crystal structure in green. The theozyme residues and the ligand are shown in stick representation, and selected other active site residues in line representation. A) ECH13: The Cα RMSD between design model and crystal structure is 0.97 Å over the 15 active site residues. The catalytic histidine, His100, is in a rotameric conformation different from the design model, and instead of pointing towards the ligand and Cys45, it makes a hydrogen bond with Asp10. This alternative conformation is facilitated by a small backbone shift between residues Pro99 and Phe103, and the observed close interaction between His100 and Asp10 would not be possible with the scaffold backbone conformation that served as the template for the design. B) ECH19 P364W: The design was based on the closed conformation of a periplasmic binding protein, but the apo protein crystallized in the open form, with an RMSD of 4.1 Å to the design model but an RMSD of only 1.6 Å to the open form of the scaffold protein (PDB 2uvg). The designed active site is mostly located in one of the scaffold’s two domains, close to the interdomain cleft. When superimposing design model and crystal structure based solely on the active-site containing domain, the resulting RMSD is 1.5 Å. However, the catalytic His226 does not interact with Cys161 as designed, but adopts a different rotameric conformation to interact with the sidechain-hydroxyl of Tyr250 and the backbone oxygen of Phe221. The stretch from Tyr218 to Lys230 that contains His226 has high relative B-factors, suggesting that it is fairly flexible. C) ECH14: the crystal structure has an RMSD of 1.4 Å to the design. The catalytic dyad is not formed, as the Cys132 containing loop-helix stretch between residues 127 and 140 moves upward away from the active site and His104 reorients around chi2. This unexpected movement may result from the W130S mutation, since W130 stacks against the PLP cofactor of the wild type scaffold and thus locks this backbone segment into the conformation used as the design template. D) FR29 A44S/T112L/V151L: The apo-structure of FR29 is more similar to the unliganded, more open conformation of the scaffold (PDB 1D2R, 0.86 Å RMSD) than to the ligand-bound structure (PDB 1mau, 2.7 Å RMSD) which was used as the template for the design. The catalytic dyad is not formed, since in the apo form of the scaffold the helix-turn-helix motif between residues 106 and 132 that contains the catalytic His125 moves outward relative to the catalytic Cys9, leading to shift in His125 Cα – Cys9 Cα distance from 10.9 Å to 12.4 Å. As the backbone of most of the designed active site residues shifts between the liganded and the apo structure, the active site is generally more open than in the design model.