Evolution of a highly active and enantiospecific metalloenzyme from short peptides

See allHide authors and affiliations

Science  14 Dec 2018:
Vol. 362, Issue 6420, pp. 1285-1288
DOI: 10.1126/science.aau3744

Evolution trains a from-scratch catalyst

Metal-bound peptides can catalyze simple reactions such as ester hydrolysis and may have been the starting point for the evolution of modern enzymes. Studer et al. selected progressively more-proficient variants of a small protein derived from a computationally designed zinc-binding peptide. The resulting enzyme could perform the trained reaction at rates typical for naturally evolved enzymes and serendipitously developed a strong preference for a single enantiomer of the substrate. A structure of the final catalyst highlights how small, progressive changes can remodel both catalytic residues and protein architecture in unpredictable ways.

Science, this issue p. 1285


Primordial sequence signatures in modern proteins imply ancestral origins tracing back to simple peptides. Although short peptides seldom adopt unique folds, metal ions might have templated their assembly into higher-order structures in early evolution and imparted useful chemical reactivity. Recapitulating such a biogenetic scenario, we have combined design and laboratory evolution to transform a zinc-binding peptide into a globular enzyme capable of accelerating ester cleavage with exacting enantiospecificity and high catalytic efficiency (kcat/KM ~ 106 M−1 s−1). The simultaneous optimization of structure and function in a naïve peptide scaffold not only illustrates a plausible enzyme evolutionary pathway from the distant past to the present but also proffers exciting future opportunities for enzyme design and engineering.

Metal ions are ubiquitous in nature, playing structural and/or catalytic roles in nearly half of all proteins. This dual functionality conceivably fostered the emergence of primordial metalloenzymes from simpler peptidic precursors by an evolutionary pathway involving metal-mediated assembly, followed by polypeptide fusion and diversification (Fig. 1A) (16). In mimicry of this process, protein designers have successfully used metal ions to template binding of weakly interacting peptides and generate supramolecular structures that display modest catalytic activities at their interfaces (714). Here, such complexes are shown to be excellent starting points for the design and evolution of highly active, globular metalloenzymes.

Fig. 1 Emulating metalloenzyme biogenesis from peptides.

(A) Zinc-mediated assembly of helix-turn-helix fragments, followed by fusion and asymmetric diversification, afforded MID1sc10, an efficient metalloesterase. (B) Simplified schematic showing the specific steps performed in the diversification process.

To explore metalloprotein biogenesis, we chose MID1, a homodimeric peptide containing two interfacial Zn(II)His3 sites that was computationally designed from a monomeric, 46–amino acid–long, helix-turn-helix fragment (11). The zinc ions originally served as prostheses for peptide assembly but also provided serendipitous activity for ester bond hydrolysis thanks to a small hydrophobic binding pocket adjacent to an open metal coordination site (12). Adopting nature’s fusion and diversification strategy, we connected adjacent N and C termini of the dimer subunits via a short Gly-Ser-Gly linker and removed the zinc site farthest from the linker by replacing metal-binding residues with noncoordinating amino acids suggested by computation. The resulting single-chain MID1 variant, MID1sc, binds a single Zn(II) ion and hydrolyzes p-nitrophenyl acetate at similar rates as MID1 (fig. S1).

For protein evolution, we developed a robust and sensitive screening assay based on the racemic fluorogenic ester 1 (Fig. 2A). MID1sc hydrolyzes ester (±)-1 with a turnover number (kcat) of 0.011 ± 0.001 s−1 (mean ± SD) and an apparent second-order rate constant (kcat/KM) of 18 ± 2 M−1 s−1. It also exhibits a twofold preference for cleavage of the (R)-configured substrate enantiomer (fig. S2 and table S1). We optimized this initial catalytic activity over nine rounds of laboratory evolution, exploiting both focused and random mutagenesis (Fig. 1B and fig. S3). Single residues close to the zinc center, lining the primitive binding pocket, and around the former zinc site were targeted by cassette mutagenesis, and the most promising mutations were shuffled. In addition, the full-length gene was randomized by error-prone polymerase chain reaction (PCR) to identify beneficial mutations distant from the active site. Over the course of evolution, self-acylating residues were replaced by arginine (Lys68 and Lys78) or targeted for randomization (Arg80) to prevent catalyst inactivation by covalent modification with the substrate (figs. S4 and S5). An average of one to two mutations were introduced per round of evolution to afford MID1sc9, which has a total of 20 substitutions distributed nearly equally over the N- and C-terminal helix-turn-helix fragments (Fig. 2B and fig. S6).

Fig. 2 Directed evolution of MID1sc.

(A) MID1sc was evolved for the hydrolysis of fluorogenic ester 1 to give 2-phenylpropionate 2 and coumarin 3. The * indicates the chiral center. (B) Crystal structure of MID1sc10, showing the zinc ion as an orange sphere and the coordinating histidines as green sticks. Linkage of two polypeptides via a Gly-Ser-Gly sequence (orange) and removal of a second zinc site present in the original MID1 design (yellow spheres) afforded MID1sc, which was subsequently optimized by mutagenesis and screening. The locations of beneficial mutations (magenta spheres) and residues replaced to prevent competitive zinc binding modes (cyan spheres) are highlighted. (C) Michaelis-Menten plots for MID1sc (yellow and inset) and MID1sc10 (green) show a 70,000-fold improvement in hydrolysis efficiency for (S)-configured 1 after optimization. v0/[E]0, initial rate divided by total enzyme concentration. (D) The evolved variant MID1sc10 is highly enantioselective as a consequence of a 2200-fold specificity switch from the modestly (R)-selective starting catalyst MID1sc. All error bars represent the SD of at least three independent measurements.

Because 21% of the protein was mutated by design and directed evolution, possible changes in Zn(II) coordination were probed by sequential replacement of each histidine in the original zinc binding site (His39, His61, and His65) by alanine (fig. S7 and table S2). Surprisingly, substitution of His39 had little effect on catalytic activity, whereas replacement of His61 and His65 led to a greater than fivefold decrease in turnover number. Based on the sequence of the evolved protein, we identified another histidine (His35) and two glutamates (Glu32 and Glu58) as possible alternative metal-binding residues in close proximity to the original zinc site. Whereas substitution of Glu32 and Glu58 with glutamine had little effect on catalytic efficiency, alanine substitution of His35 reduced activity >1000 fold, strongly suggesting that His35, together with His61 and His65, binds the catalytic zinc ion. This change in coordination sphere occurred midway along the evolutionary trajectory because His35 could still be replaced after the third round of mutagenesis without loss in esterase activity (table S1). With the goal of eliminating potentially competitive zinc binding modes, we incorporated the E32Q (Glu32→Gln), H39A (His39→Ala), and E58Q (Glu58→Gln) mutations to give the final optimized construct, MID1sc10, which had >10,000-fold higher activity than its MID1sc progenitor at subsaturating concentrations of racemic ester 1.

MID1sc10 is a highly active esterase. It preferentially catalyzes the hydrolysis of (S)-1 with a kcat of 1.64 ± 0.04 s−1 and a kcat/KM of 980,000 ± 110,000 M−1 s−1 (Fig. 2C and table S1). These steady-state parameters attest to notable catalytic proficiency [1/KTS = (kcat/KM)/kuncat = 9.3 × 1010 M−1, where KTS is the apparent transition-state binding affinity and kuncat is the rate constant for the uncatalyzed reaction (15)]. In this respect, MID1sc10 is similar to typical natural enzymes (16) and outperforms other artificial esterases, including catalytic antibodies (17), computationally designed enzymes (1821), and engineered zinc metalloproteins (9, 12, 13, 22), by two to five orders of magnitude (table S3). It is also superior to the natural zinc metalloenzyme human carbonic anhydrase (hCAII), which promiscuously hydrolyzes similarly activated p-nitrophenyl acetate with a kcat/KM of 2500 M−1 s−1 (23). Even for its natural activity, the mechanistically related hydration of carbon dioxide, hCAII, a nearly perfect zinc enzyme, has a catalytic proficiency that is 100-fold lower than that of MID1sc10 (15).

Given the importance of stereochemical control for industrial biocatalysis, the high enantiospecificity achieved by MID1sc10, manifest in a 990-fold kinetic preference for cleavage of the (S)-configured ester (Fig. 2D and table S1), is particularly notable. As the entire screen was performed with racemic substrate, this property was never subject to direct selection pressure. However, active site mutations introduced in the third round of evolution fostered the initial switch from the (R)-specific starting scaffold, and the new stereochemical preference was subsequently enhanced in step with specific activity (table S1).

Zinc is absolutely required for MID1sc10 catalysis; removal inactivates the enzyme. Nevertheless, it binds relatively weakly with an apparent dissociation constant (Kd) of 26 μM (fig. S8). Consistent with weak binding, zinc does not stabilize the evolved protein. Its thermal denaturation is unaffected by Zn(II) (Fig. 3A), whereas the metal ion increases MID1’s melting temperature by 24°C (11). When Zn(II) is added to apo-MID1sc10, signal broadening is observed in the 1H-15N-heteronuclear single-quantum coherence (HSQC) nuclear magnetic resonance (NMR) spectrum (Fig. 3B and fig. S9), suggesting the presence of several states that interconvert on an intermediate time scale. Together, these results indicate that design and evolution converted the zinc ion from an essential structural element into a dedicated catalytic cofactor.

Fig. 3 Biophysical characterization and crystal structure of MID1sc10.

(A) The thermal stability of MID1sc10 is similar in the presence (green) and absence (black) of zinc. [Θ], mean residue ellipticity. (B) Overlay of the 1H-15N-HSQC spectra of MID1sc10 in the presence (green) and absence (black) of zinc. For the full spectrum, see fig. S9. δ, chemical shift; ppm, parts per million. (C) Structural alignment of MID1sc10 (green) and MID1 (gray), illustrating the >30° tighter crossover angle. (D) The observed structural changes transformed the shallow binding site of MID1 (gray) into a deep, hydrophobic pocket in MID1sc10 (green). (E) Cut-away view of the active site, showing the snug fit of phosphonate 4 in the binding pocket. The zinc ion is shown as an orange sphere and the ligand is shown in space-filling representation (carbon, yellow; oxygen, red; phosphorus, black; sulfur, orange). (F) View of the MID1sc10 active site with phosphonate 4 (yellow) coordinating to the Zn(II)His3 complex (orange sphere and green sticks). Arg68 and Gln58 form mechanistically relevant hydrogen bonds to phosphonate 4 and the backside nitrogen of His61, respectively.

To elucidate the origins of these effects, we cocrystallized MID1sc10 with racemic phosphonate 4, a structural mimic of the esterolytic transition state that competitively inhibits the enzyme with an inhibition constant (Ki) of 1.1 ± 0.1 μM (fig. S10) and increases the enzyme’s affinity for Zn(II) more than 100-fold (fig. S8). The crystal structure, solved at 1.34-Å resolution (Figs. 2B and 3, C to F, and table S4), confirmed that the protein adopts a helical bundle fold, albeit with substantial structural changes compared with MID1 (11). In addition to the altered Zn(II) coordination sphere identified by mutagenesis, the crossover angle of the two helix-turn-helix fragments decreased to 47°, which is >30° tighter than in MID1 (Fig. 3C, fig. S11, and table S5) but still considerably larger compared with canonical four-helix bundles (typically 20°) (24). This dramatic conformational change was brought about by extensive remodeling of the protein interior to accommodate the large ester substrate. Five out of 13 residues lining the substrate binding pocket were mutated [M38W (Met38→Trp), K68R (Lys68→Arg), Q80S (Gln80→Ser), L83T (Leu83→Thr), and H84L (His84→Leu)], substantially deepening and reconfiguring the active site for shape-complementary transition-state recognition (Fig. 3, D and E).

Another early mutation, Q36P (Gln36→Pro), introduced a kink in the second helix, helping to form a tighter binding pocket for the substrate and facilitating replacement of His39 by His35 as a Zn(II) ligand (fig. S12 and movie S1). The resulting metal environment (Fig. 3F and fig. S7) resembles the zinc site in carbonic anhydrase (25). Introduction of a second-shell hydrogen-bonding interaction between Gln58 and the backside nitrogen of His61 is intriguing in this context, because natural zinc enzymes utilize similar interactions to tune metal ion reactivity (25, 26). Like carbonic anhydrase, MID1sc10 presumably exploits the Lewis acidity of Zn(II) to acidify a coordinated water molecule and generate a high local concentration of hydroxide for substrate cleavage. Fitting the pH-rate data for ester hydrolysis afforded a kinetic pKa of ~8 (fig. S13), which is higher than the value of 6.8 determined for ionization of zinc-bound water in carbonic anhydrase (25) but falls in the range of pKa’s observed for other peptides and model complexes (26), including MID1 (12).

Consistent with MID1sc10’s high enantiospecificity, only the (S)-enantiomer of phosphonate 4 is bound in the crystallized complex (fig. S14). The inhibitor adopts an extended conformation with the 2-phenylpropionyl group sitting snugly at the bottom of the hydrophobic pocket and the charged leaving group near the entrance of the active site (Fig. 3E). This orientation allows the phosphonate to coordinate Zn(II) via one of its oxygen atoms, as expected for a mechanism involving nucleophilic attack of a zinc-bound hydroxide on the ester substrate (Fig. 3F). The other phosphonyl oxygen forms a bidentate hydrogen bond with the side chain of Arg68, a residue introduced during evolution. Similar interactions have been observed in zinc enzymes like carboxypeptidase A (27) and contribute to electrostatic stabilization of the anionic transition state. In MID1sc10, the guanidinium group of Arg68 additionally makes productive cation-π interactions with the coumarin, which may assist departure of the leaving group.

Although the evolved catalyst shows good activity with p-nitrophenol and coumarin esters of 2-phenylpropionate, catalytic efficiency drops substantially for esters of simple aliphatic acids (table S6). For example, the kcat/KM for p-nitrophenyl acetate is similar to that observed for the starting catalyst. As for natural hydrolases, shape-complementary binding interactions between the enzyme and portions of the substrate distant from the scissile bond contribute substantially to catalytic efficiency (28), presumably by properly positioning the ester for effective reaction.

The extraordinary activities, efficiency, and specificities of modern-day metalloenzymes are the products of eons of evolution. The bottom-up construction of a zinc-dependent esterase by end-to-end doubling of the MID1 peptide and subsequent directed evolution shows that the putative historical roads taken by these natural catalysts are also fruitful avenues for producing new enzymes. The de novo generation of a highly active metalloesterase in this way compares favorably with computational enzyme design, which uses sophisticated software algorithms to equip the binding pockets of natural protein scaffolds with the catalytic functionality needed to accelerate a chosen target reaction and is one of the most promising approaches to tailored catalysts to emerge in the past few years (29, 30). Although computationally designed enzymes have been evolved to high activities for several reactions (3134), creation of efficient catalysts for the hydrolysis of esters like 1 has proved challenging (1821). Instead of a metal ion cofactor, computational designs have relied on a single nucleophile (18, 20) or embedded catalytic dyads (19) and triads (21) to cleave the substrate via a transient acyl-enzyme intermediate. However, even after laboratory evolution, the apparent second-order rate constants for protein acylation (k2/KS) have never exceeded 2000 M−1 s−1 (table S3), and slow deacylation limits overall turnover (kcat/KM < ~100 M−1 s−1).

The comparative ease of evolving a 10,000-fold more efficient zinc-dependent esterase is thus striking and speaks to the efficacy of metal ion catalysis. Even though no reaction-relevant chemical information was provided by design, the optimized MID1sc10 active site recapitulates the natural mechanisms of native zinc enzymes, suggesting that the intrinsic chemical potential of such systems is readily realizable once the metal ion is installed in an appropriate binding pocket. The flexibility of the helical bundle fold may have been advantageous in this respect, expediting the evolutionary search for a chemically and sterically complementary binding pocket that could effectively align substrate and metal-ion-bound water and lower the transition state barrier for ester hydrolysis.

MID1sc10 embodies the structural and functional properties that metals likely imparted to proteins long ago. Promiscuous binding of different substrate molecules and metal ions by primordial scaffolds would have been a potentially rich source of novel activities. Looking forward, our simple metalloprotein may similarly constitute an excellent system for exploring divergent evolution and functional diversification. By elucidating how sophisticated enzymatic functions emerge from naïve peptide scaffolds, such experiments have the potential to inform ongoing efforts to create new metal-dependent protein catalysts for chemical transformations unknown in nature (8, 3538).

Supplementary Materials

Materials and Methods

Figs. S1 to S15

Tables S1 to S7

References (3962)

Movie S1

References and Notes

Acknowledgments: The authors thank C. Stutz-Ducommun and B. Blattmann from the Protein Crystallization Core Facility at the University of Zürich, R. Arnold from the NMR Service of the Laboratory of Organic Chemistry at ETH Zürich, the Mass Spectrometry Service team of the Laboratory of Organic Chemistry at ETH Zürich, the Protein Analysis Group at the Functional Genomics Center of the University of Zürich, and the staff at the Swiss Light Source (Paul Scherrer Institute) for technical support. The authors also thank Y. Azuma, M.-O. Ebert, T. Edwardson, A. Vandemeulebroucke, and D. L. Niquille for helpful discussions. Funding: The work was supported by the Lithuanian-Swiss Research and Development cooperation program (CH-3-SMM-01/03), the Swiss National Science Foundation (31003A-156276), the NIH (R01GM073960 and T32GM008570), and a postdoctoral fellowship from the ETH (to D.A.H.). Author contributions: S.S., B.K., and D.H. designed the research; B.S.D. and S.L.G carried out the design simulations; S.S. performed enzyme evolution, characterization, and crystallization; D.A.H. and Z.L.P. performed chemical synthesis; A.D. performed acylation experiments and enzyme characterization; P.R.E.M. solved the crystal structure; and S.S., D.A.H., and D.H. wrote the manuscript. Competing interests: The authors declare no competing financial interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the main text or the supplementary materials. Plasmids encoding the enzymes reported in this study are available for research purposes from D.H. under a material transfer agreement with the ETH Zürich. Coordinates and structure factors have been deposited in the Protein Data Bank with the accession code 5OD1.
View Abstract

Stay Connected to Science

Navigate This Article