Engineering the protein dynamics of an ancestral luciferase

Abstract

Protein dynamics are often invoked in explanations of enzyme catalysis, but their design has proven elusive. Here we track the role of dynamics in evolution, starting from the evolvable and thermostable ancestral protein Anc^HLD-RLuc which catalyses both dehalogenase and luciferase reactions. Insertion-deletion (InDel) backbone mutagenesis of Anc^HLD-RLuc challenged the scaffold dynamics. Screening for both activities reveals InDel mutations localized in three distinct regions that lead to altered protein dynamics (based on crystallographic B-factors, hydrogen exchange, and molecular dynamics simulations). An anisotropic network model highlights the importance of the conformational flexibility of a loop-helix fragment of Renilla luciferases for ligand binding. Transplantation of this dynamic fragment leads to lower product inhibition and highly stable glow-type bioluminescence. The success of our approach suggests that a strategy comprising (i) constructing a stable and evolvable template, (ii) mapping functional regions by backbone mutagenesis, and (iii) transplantation of dynamic features, can lead to functionally innovative proteins.

Fig. 1. Illustration of the strategy for semi-rational engineering of protein dynamics.

Exploratory phase—1 A thermostable ancestral protein, Anc^HLD-RLuc, provided a robust and evolvable template that can withstand the destabilizing effects of protein backbone engineering. 2 Libraries of single triplet insertion and deletion variants were created following the TRIAD method based on the use of engineered transposons (TransIns and TransDel)^,. 3 Screening of the libraries led to identification of the improved insertion mutant AncINS. 4 Twenty-five mutants with significant changes in luciferase (LUC) and haloalkane dehalogenase (HLD) activities were characterized using bioinformatics, microscale techniques (nano differential scanning fluorimetry—nanoDSF), and microfluidics. 5 Structure-function relationships were described employing partial least squares (PLS) multivariate statistics. 6 Dynamic elements required for efficient catalysis were identified by structural, kinetic, biophysical and computational characterization (molecular dynamics (MD) simulations) of AncINS. Validation phase—7 Knowledge obtained during the exploratory phase was validated by transplanting a relevant dynamic fragment from the specialized descendant into the ancestor, yielding an enzyme, AncFT, with 7000-fold higher catalytic efficiency than Anc^HLD-RLuc and 100-fold longer glow-type bioluminescence than the flash-type Renilla luciferase RLuc8. Key mutants discussed in this study are highlighted in yellow.

Fig. 2. Quantitative structure-function relationships and analysis of the effects of template and mutation type on activity.

a Crystal structure of the thermostable Anc^HLD-RLuc (PDB ID 6G75) showing the positions of the L9 loop (light blue), the α4 helix (salmon) and the L14 loop (pale green), where InDels (spheres) resulted in an increase of luciferase (LUC) activity. Correlated motions of L3 (marine) and L18 (yellow) loops carrying two of the five catalytic amino acids (spheres) were identified as significant contributors to dehalogenase (HLD) activity by the partial least squares (PLS) regression analysis. b Weighted coefficients quantifying contributions of variables indicated in the PLS models to explain the variance in HLD (green) and LUC (blue) activity. Note that different directionality of these coefficients for all variables, except loop L9, suggests different mechanistic and structural requirements for the two enzymatic functions studied. PLS generated models that explained substantial amounts of the variation in both LUC activity (R² = 0.73, Q² = 0.67, n = 25 is the number of InDel variants) and HLD activity (R² = 0.63, Q² = 0.54, n = 25 is the number of InDel variants). To obtain Q² the model was recalculated 999 times with a randomly re-ordered dependent variables. c Comparison of frequencies of variants with their activities relative to the ancestral protein in the first-round insertion (R1I, blue) and deletion (R1D, red) libraries. The template for R1I and R1D was Anc^HLD-RLuc. The best insertion variant AncINS is indicated by an arrow. d Comparison of frequencies of variants with indicated activities relative to their respective templates observed in the first round of insertion library (R1I, blue) and the second round of insertion library (R2I, yellow). Note that the starting template Anc^HLD-RLuc for R1I and R1D is less active than the starting template AncINS for R2I. The template for R2I was AncINS. Source data is available as a Source data file for Fig. 2.

Fig. 3. Steady-state and transient kinetic analysis of CTZ conversion.

The kinetic models consist of an enzyme E, a substrate S, an enzyme-substrate complex in two conformations (E.S and E*.S), an enzyme-product complex E.P, and a product P. a Steady-state kinetic parameters (Michaelis constant K_m, turnover number k_cat, enzyme-product complex dissociation constant K_p) were determined with the substrate CTZ in 100 mM phosphate buffer at pH 7.5 and 37 °C by global analysis of triplicates of full progress curves recorded with at least five concentrations of CTZ. b Results of pre-steady-state kinetic analysis of the CTZ substrate binding in 100 mM phosphate buffer at pH 7.5 and a lower temperature (15 °C). This enabled identification of the induced fit substrate binding mechanism, involving initial collision of the enzyme and the substrate (described by forward rate constant k₊₁ and reverse rate constant k₋₁), followed by a conformational change of the enzyme induced by the bound substrate (described by forward rate constant k₊₂ and reverse rate constant k₋₂). A simple binding mechanism including only the first step was observed for Anc^HLD-RLuc. The kinetic parameters were determined by global fitting of tryptophan fluorescence traces obtained with at least 10 concentrations of CTZ and 10 concentrations of each tested enzyme, in each case with seven replicates (Supplementary Note 6). The data are presented as best fit values ± standard errors (S.E.) calculated from the covariance matrix during nonlinear regression. Source data is available as a Source data file for Fig. 3.

Fig. 4. Comparison of structural changes in the apo forms of the enzymes.

Anc^HLD-RLuc (white), AncINS (chain A light blue, chain B slate), AncFT (salmon), RLuc8 (chain A pale cyan, chain B light teal). a Top view of the active site. The putative catalytic pentad and α5 helix occupy the same position in all proteins, but the α4 helix adopts different conformations. b Crystal structure of the whole protein, with the main access tunnel (identified using Caver 3.02) shown as spheres. The outward/inward movement of the α4 helix causes opening/closing of the tunnel. c Front view of the main tunnel. The black arrow indicates the visible constriction due to closure of the α4 helix. d The size of the active site cavity is changed by repositioning of amino acids in the α4 helix. e The size of the main tunnel mouth is changed by movements of the α4 helix and L9 loop.

Fig. 5. Structural characterization of coelenteramide binding to the catalytically defective RLuc8-W121F/E144Q mutant.

a Cartoon representation of the overall structure of the RLuc8-W121F/E144Q mutant with coelenteramide (CEI) in its active site. CEI is the product of the LUC reaction and shown as cyan space-filling spheres. Residues of the conserved catalytic pentad are shown as purple spheres; the central eight-stranded β-sheet is coloured yellow; the α4 helix and L9 loop (L9-α4 element) are coloured violet, and the L14 loop is coloured orange. b Cutaway surface representation of the enzyme active site cavity with the bound CEI (shown as cyan sticks). The colouring is the same as in panel (a). Water molecules are shown as red spheres. c Close-up view of structural superposition of RLuc8-W121F/E144Q (green), RLuc8 (PDB ID 2PSF A; cyan) and RLuc8 (PDB ID 2PSF B; teal). CEI is shown as cyan space-filling spheres. Note the conformational sampling of the L9-α4 fragment. The bottom 4-hydroxyphenyl group connected to the CEI acetamide moiety is deeply buried in the active site cleft, where it is anchored in the slot tunnel through multiple hydrophobic (P224, I223 and I266) and aromatic π-stacking (W156) interactions. The 4-hydroxyphenyl group interacts with the indole NH group of W156 through a water-mediated hydrogen bond bridge. The acetamide moiety of CEI is positioned close to the conserved catalytic centre. In chain B, the top 4-hydroxyphenyl group linked with the CEI pyrazine ring interacts with a side chain of K189 through a water-mediated hydrogen bond bridge and forms a hydrogen bond with the carboxylate group of D162 from a symmetry-related enzyme molecule.

Fig. 6. Engineering conformational dynamics by backbone modifications.

a Conformational dynamics in the cap domains of crystal structures: AncINS (PDB ID 6S6E) has two monomers in the asymmetric unit differing in conformation of the α4 helix carrying the insertion. Chains A (light blue) and B (slate) are in the open and closed conformations, respectively. RLuc8 (PDB ID 2PSF) has analogous conformations, with an open chain A (pale cyan) and a closed chain B (light teal). AncFT (PDB ID 6S97) has one monomer in the asymmetric unit, in which helix α4 is more open than in Anc^HLD-RLuc, AncINS, and even RLuc8 chain B. b Conformational dynamics in the cap domains observed during molecular dynamics simulations. B-factors of backbone atoms standardized across all protein variants, ranging from −2 (blue) to 2 (red). B-factor values standardized for each protein are indicated by the thickness of the lines representing the protein backbone. Values were averaged per secondary structure element. c Dynamics and hydration of the cap domain based on HDX-MS assessments of peptide deuteration after 60 s. Anc^HLD-RLuc is most heavily deuterated in the L9 loop. The deuteration pattern of AncFT is most similar to that of RLuc8. d Main access tunnel geometries of the proteins: representative snapshots of the open (lighter shades, upper row) and closed (darker shades, lower row) conformations. Average bottleneck radii and standard deviations from n = 1000 snapshots from MD simulations were calculated by Caver 3.02. The grey area corresponds to closed conformational states with tunnel radii below 1.4 Å (the radius of a water molecule). Source data is available as a Source data file for Fig. 6b and c.

Fig. 7. Comparison of bioluminescence of enzymes purified from bacterial cultures and in lysates from mammalian cells.

a Full decay kinetics of the conversion of 2.2 μM CTZ by 50 nM of RLuc8, AncINS and AncFT purified from bacterial cultures. The data are presented as relative values to initial luminescence; Anc^HLD-RLuc is not plotted due to low activity leading to large signal scattering. Solid lines represent the best fit to the experimental data. Experiments were repeated independently three times with consistent results. b Bioluminescence signal steadiness in relative values in lysates from mammalian cells expressing AncFT and RLuc8. Experiments were repeated independently three times with consistent results. Activity was measured using the commercial Renilla luciferase assay kit (Promega) with 20 μL of cell lysates and 100 μL of assay buffer. Luminescence signal in lysates from mammalian cells expressing Anc^HLD-RLuc and AncINS was not detectable under tested conditions. Source data is available as a Source data file for Fig. 7.