An engineered N-acyltransferase-LOV2 domain fusion protein enables light-inducible allosteric control of enzymatic activity

Abstract

Transferases are ubiquitous across all known life. While much work has been done to understand and describe these essential enzymes, there have been minimal efforts to exert tight and reversible control over their activity for various biotechnological applications. Here, we apply a rational, computation-guided methodology to design and test a transferase-class enzyme allosterically regulated by light-oxygen-voltage 2 sensing domain. We utilize computational techniques to determine the intrinsic allosteric networks within N-acyltransferase (Orf11/∗Dbv8) and identify potential allosteric sites on the protein’s surface. We insert light-oxygen-voltage 2 sensing domain at the predicted allosteric site, exerting reversible control over enzymatic activity. We demonstrate blue-light regulation of N-acyltransferase (Orf11/∗Dbv8) function. Our study for the first time demonstrates optogenetic regulation of a transferase-class enzyme as a proof-of-concept for controllable transferase design. This successful design opens the door for many future applications in metabolic engineering and cellular programming.

Genetically encoded light-responsive proteins, broadly referred to as optogenetics, show great potential in the fields of metabolic engineering, cellular therapeutics, and cellular computing among many others (1, 2, 3). Expanding the available optogenetic tools allows scientists to push the development of innovative engineered systems and applications.

A strategy to design optogenetically controlled proteins without affecting other potential binders was outlined previously (4) and later expanded to nanocomputing agents, capable of performing cellular computation (5, 6, 7). Such strategy is based on allosteric control, whereby the positioning of the light-sensitive unit is fused distal to the functional site of a protein (8, 9, 10). By strategically leveraging light-sensitive proteins using this or similar strategies, researchers have achieved control over the activation and interaction of a constantly expanding suite of proteins including ion channels, kinases, phosphatases, GTPases, caspases, reductases, guanine nucleotide exchange factors, and transcription factors (3, 11, 12, 13, 14). However, there have been no studies to date demonstrating optogenetic regulation of a transferase-class enzyme.

Transferases are critical for proper cellular function: they intermediate the exchange of a functional group (R) between a donor (D) and a substrate (S). These enzymes regulate many essential cellular pathways ranging from protein biosynthesis to the quenching of reactive oxygen species (15). Transferases are also used in various industrial applications, including synthesis of oligosaccharides, nucleic acid modification, and other biochemical reactions (16, 17).

We have selected a unique transferase, N-acyltransferase (NAT) (Orf11/∗Dbv8), as the focus of this proof-of-concept study. NAT (Orf11/∗Dbv8), or NAT Orf11, is distinct from typical NAT enzymes due to its promiscuous affinity for large and bulky substrates (18). It has previously been shown that NAT Orf11 has the capacity to acylate teicoplanin, a last line of defense antibiotic for drug-resistant bacteria, including methicillin-resistant staphylococcus aureus and vancomycin-resistant enterococcus (18). Establishing control over the enzymatic activity of NAT Orf11 has numerous potential applications in metabolic engineering and the biosynthesis of potent acyl-teicoplanin compounds.

Here, we have engineered a novel NAT Orf11 that is optogenetically regulated by blue light, referred to as oNAT. NAT Orf11 has a binding pocket that is naturally flexible and displays nonspecific substrate interactions. In our design of an optically regulated NAT Orf11, we aimed to leverage the highly structured dark state of light-oxygen-voltage 2 (LOV2) sensing domain to lock the enzyme into a conformation that occludes the active site and prevents interaction between the enzyme and its substrate. We built this construct by utilizing the allosteric networks within NAT Orf11 to propagate the structural force of LOV2 to the binding pocket region. Upon blue-light irradiation, LOV2 undergoes an order-disorder transition due to unfurling of the Jα-helix (19), relaxing NAT Orf11 and allowing it to assume a more open structure. The cessation of LOV2-mediated inhibition increases the availability of the binding pocket and enhances enzymatic activity.

Results and discussion

Engineering of the NAT/LOV2 fusion protein

To achieve light-inducible allosteric regulation of NAT Orf11/∗Dbv8In (PDB ID: 4MFJ), we introduced the LOV2 domain from Avena sativa (PDB ID: 2V0U) into NAT. We hypothesized that the 10 Å spacing between the N and C termini of LOV2 will result in minimal perturbation at the region of insertion and constrain NAT in an inactive conformation. Blue light exposure results in unfolding of the C-terminal Jα helix of LOV2, leading to a conformational shift of NAT that will activate the protein.

We first employed the Ohm online platform (9) (https://dokhlab.med.psu.edu/ohm/#/) to calculate the allosteric networks within NAT Orf11 (Fig. 1A). Ohm utilizes physics-based perturbation propagation algorithm to distinguish allosteric pathways and networks (9). In tandem with Ohm’s allosteric predictions, we analyzed the regions of NAT that would facilitate insertion of LOV2 while preserving enzymatic function. We limited our search for potential insertion sites to regions with high surface exposure and low sequence conservation. Wang et al. have also shown that a powerful metric for determining the essentiality of a residue for protein function is surveying the evolutionary landscape of similar proteins and quantifying the conservation, or maintained presence, of that residue across related proteins. In order to calculate surface exposure of each amino acid residue in NAT Orf11, we estimated the solvent accessible surface area (SASA) and identified surface exposed loop regions with SASA ≥40 Å (Fig. 1B). We determined a likely insertion site with high surface exposure and low conservation (Fig. 1B), located at residues K187-D188 in NAT Orf11. We defined this location as Loop1 (L1), indicated in red in Figure 1B. Ohm results also indicated allosteric coupling, indicating a high likelihood that introducing LOV2 at the L1 site will exert significant functional allosteric regulation of NAT Orf11 (Fig. 1C).

Figure 1.

Computational design and DMD simulation of N-acyltransferase/LOV2 fusion protein.A, ribbon diagram of wildtype N-acyltransferase orf11 protein, enzymatic binding pocket and active site (V197/S236/H196) indicated with a red arrow. B, amino acid conservation score (top), residue exposure (left), and contact map of N-acyltransferase Orf11. Conservation Score is from 1 (Highly Conserved) to 10 (Not Conserved). Selected insertion site (L1) is indicated in red, at a nonconserved and exposed loop. Bottom right, ribbon diagram, and exploded view of selected insertion loop (K187-D188). C, allosteric coupling of NAT residues, calculated by the Ohm online server (https://dokhlab.med.psu.edu/ohm/#/). Selected insertion site (L1) is indicated in red. D, visualization of the predicted allosteric pathway of the selected insertion site, K187-D188. E, final structure of LOV2-inserted N-acyltransferase fusion protein, oNAT. DMD, discrete molecular dynamic; LOV2, light-oxygen-voltage-sensing domain; NAT, N-acyltransferase.

Molecular dynamics simulations offer an efficient approach to estimate protein interactions and structural characteristics without requiring in vitro cloning and experimentation. We used discrete molecular dynamic (DMD) simulations to analyze conformations, specifically the availability of the active site of a NAT-LOV2 fusion protein containing a mutation in the LOV2 domain that locked the protein into a lit state (oNAT-L, LOV2: I510E/I539E) or a dark state (oNAT-D, LOV2: C450A), inserted at loop 1. We defined the external residues G155 and N429 as gatekeeper residues, with the distance between the two gatekeeper residues providing a metric for how much steric interference is being exerted by LOV2 in each conformation. Assessing the frequency of gatekeeper residue distance for each time step, we found that NAT-LOV2 had significantly more active site availability in the lit state than in the dark state (Fig. 2B) (p-value < 0.05). Wildtype NAT (NAT Orf11) has a resting state binding pocket availability of 6.4 Å. We observed that the lit state of oNAT was driven toward a more open conformation, with the active site readily available (11.7 Å), while in the dark state, the active site is occluded (5.7 Å) (Fig. 2C).

Figure 2.

Discrete molecular dynamics simulations of oNAT.A, discrete molecular dynamic simulations tracking the availability of the active site of NAT over time. The distance between surface residues G155 and N429 are used as a metric for active site availability. Gray represents variance in the repeated simulations. B, histogram plotting the frequency of active site availability throughout the duration of the simulation. A Gaussian curve was fit to demonstrate the shift in active site availability. C, representative images displaying the distance between gatekeeping residues of the binding pocket of wildtype N-acyltransferase (6.4 Å), dark-state oNAT (5.7 Å) and lit-state oNAT (11.7 Å). NAT, N-acyltransferase.

Discrete molecular dynamics simulations of oNAT quantify conformation shift

Performing molecular dynamics simulations is an efficient way to estimate protein interactions and structural characteristics without requiring in vitro cloning and experimentation. We utilized DMD simulations (20, 21, 22) to understand the effect the lit and dark conformations have on the NAT-LOV2 fusion protein, specifically the availability of the active site.

Following the selection of the LOV2 insertion site, we fused the corrected NAT Orf11 structure with either the lit- or dark-state of LOV2 at the L1 (K187/D188) insertion site. We curated NAT-LOV2 structural models as described in methods, proceeding with the corrected structure to conduct simulations. All reported simulation data were conducted using the optimized NAT-LOV2 structure. We conducted constant-temperature, all-atom DMD simulations of NAT-LOV2 in both the lit and the dark conformation (Videos S1 and S2). We then calculated the availability of the binding site throughout the duration of the DMD simulations (Fig. 2A). Assessing the frequency of gatekeeper residue distance for each time step, we found that NAT-LOV2 had significantly more active site availability in the lit state than in the dark state (Fig. 2B). Wildtype NAT (NAT Orf11) has a resting state binding pocket availability of 6.4 Å. We observed that the lit state of oNAT was driven toward a more open conformation, with the active site readily available (11.7 Å), while in the dark state, the active site was occluded (5.7 Å) (Fig. 2C).

Structural characterization of oNAT

After validating our design of the NAT-LOV2 fusion protein, we generated the engineered protein through standard molecular cloning techniques (Experimental procedures). Our optogenetic NAT-LOV2 fusion protein was produced in Escherichia coli and purified on a His-trap column before conducting subsequent testing. For validation purposes, we also made versions of the NAT-LOV2 fusion protein with a point mutation that locks it into the dark conformation (oNAT-D) and a mutant that locks it into the lit confirmation (oNAT-L).

We performed size-exclusion chromatography (SEC) to determine synthesized protein purity and size. SEC indicated that protein size was approximately correct. The fraction (mL) at which the nonaggregated protein of interest eluted was used to estimate protein size by referencing known standards separated by the SEC column. Using this method, we estimated that NAT Orf11 eluted at 35 kDa, oNAT at 68 kDa, oNAT-L at 71 kDa, and oNAT-D at 70 kDa (Fig. 3D). While there was no observable degradation or aggregation of the wildtype or engineered protein, both oNAT-L and oNAT-D demonstrated substantial aggregation after His-Trap purification. Fractions containing only the nonaggregated protein were collected and used for subsequent protein characterization. We next performed a melting curve (T_M) analysis of all generated proteins. Melting curve data suggested that oNAT, oNAT-L, and oNAT-D had increased stability over NAT Orf11 (Fig. 3C). oNAT is approximately 50% larger than NAT Orf11 (36.7 kD versus 53.3 kD calculated size) due to the addition of the LOV2 domain. oNAT’s larger size may have a stabilizing effect, as it has been indicated that larger proteins are more stable due to the higher number of intermolecular bonds (23). Lastly, we conducted circular dichroism (CD) measurements of NAT Orf11 and oNAT and lit/dark mutants (Fig. 3A). Following spectral deconvolution, we found that NAT Orf11 and oNAT had similar structural compositions (Fig. 3B). Based on these structural characteristics of our engineered oNAT protein, we concluded that we had a correctly folded and functional protein that we could take forward to in vitro experimentation.

Figure 3.

Spectroscopic characterization of oNAT.A, circular dichroism measurement of wildtype NAT (black), oNAT (blue), oNAT-L (green), and oNAT-D (gray). B, secondary structure composition of wildtype NAT (black), oNAT (blue), calculated wildtype NAT + LOV2 (white), oNAT-L (green), and oNAT-D (gray). C, Tm measurement of wildtype NAT (black), oNAT (blue), oNAT-L (green), and oNAT-D (gray). D, size-exclusion chromatography of wildtype NAT (black), oNAT (blue), oNAT-L (green), and oNAT-D (gray). All measurements were collected in an optically isolated environment. CD, circular dichroism; NAT, N-acyltransferase.

Colorimetric assay confirms blue light–regulated enzymatic activity of oNAT

To quantify the enzymatic activity of NAT Orf11 and oNAT, we developed a chemical assay that detects reaction product formation with a visible-light color change. This in-house colorimetric assay was rigorously tested to ensure that enzymatic reaction products were in the linear range of detection with minimal background signal (Fig. S1).

Following validation of the colorimetric assay, we quantified the enzymatic activity of oNAT-D and oNAT-L (Fig. S2). After differences in NAT-LOV2 dark and lit mutant enzymatic activity were consistently observed, we proceeded to quantify oNAT activity with and without blue-light stimulation. We conducted enzymatic reactions on a custom-built 96-well plate illuminator for 30 min under pulsed blue light stimulation (lit) or unlit (dark). NAT Orf11 had significantly greater reactivity with 2-aminophenol (30.3 μM) than 4-chloroaniline (10.8 μM) (Fig. 4B). For all tested substrates, oNAT illuminated by blue light exhibited significantly more enzymatic activity than unilluminated oNAT (Fig. 4C). Statistical calculations were performed in R and compared using a Student’s t test, alpha value set to 0.05. Total replicates of each condition were ≥12.

Figure 4.

Colorimetric assay of N-acyltransferase and oNAT enzymatic activity.A, graphical depiction of the transferase reaction catalyzed by N-acyltransferase. This reaction, which we computationally predicted to occur preferentially when exposed to 420 nm light, results in the generation of free coenzyme A. Free coenzyme A reacts with DTNB to form 2-nitro-S-mercaptobenzoic acid, a compound with a dark yellow color. B, measured enzymatic activity of wildtype N-acyltransferase with 2-aminophenol and 4-chloroaniline. C, measured enzymatic activity of engineered oNAT with 2-aminophenol or 4-chloroaniline exposed to either 420 nm blue light (blue) or no light (gray). ∗ indicates statistical significance, p-value <0.05. Box plots depict the median and interquartile range. Shaded areas are the kernel probability density of all data points. Tails are trimmed to the range of the data.

Here, we have successfully demonstrated a computationally designed achieved significant light-regulated allosteric control over NAT (Orf11/∗Dbv8). Additionally, oNAT retains broad specificity for both a readily acylated substrate (2-aminophenol) as well as a poorly converted substrate (4-chloroaniline). This finding indicates that inserting LOV2 at our predicted site (L1) does not overstabilize NAT to a degree that diminishes substrate specificity, while maintaining the structural integrity of the protein such that it retains functional enzymatic activity.

In future studies we plan to improve the dynamic range of opto-proteins and apply them to specific biological systems. Potential applications of oNAT include metabolic engineering of desirable metabolites, regulated acylation of bioactive compounds for localized drug activity. Another potential avenue for investigation is the regulation of oNAT in live cells to modulate metabolic activities. While our oNAT system has demonstrated robust regulation of transferase activity, it is at a reduced level when compared to wildtype enzyme. There are few potential techniques we can apply to improve the allosteric control of our engineered system. One method, developed by McCormick et al. (24), involves strategically mutating specific residues in the protein of interest to enhance or reduce its strength the allosteric network. This work demonstrated a four-fold enhancement of their enzyme’s dynamic range in the lit versus dark states. We could dramatically improve the dynamic range of oNAT by applying this technique to our system. One potential application of this work is in the generation of acyl-teicoplanin. Teicoplanin is a last line of defense antibiotic for drug-resistant bacteria, including methicillin-resistant Staphylococcus aureus and vancomycin-resistant enterococcus (18). Efficient synthesis of acyl-teicoplanin, either in a commercial setting or in an in vitro application, could be a path forward to combatting drug-resistant bacteria.

This proof-of-concept work demonstrates that current methodology and design for optogenetic regulation of protein activity can be applied to transferases, opening the door for future applications.

Experimental procedures

Computational design

Potential allosteric sites of NAT Orf11/∗Dbv8 were determined using previously established methodology (4). Briefly, we downloaded the crystal structure of acyltransferase (PDB: 4MFJ) (https://doi.org/10.1021/ja504125v) from protein databank and relocated all missing atoms and residues using Modeller-9v14 (25). We considered the X-structure of LOV2 domain (PDB ID: 2V0U) (26) to model lit- (Jα helix in unfolded conformation) and dark- (Jα helix in folded conformation) LOV2 states. To resemble light-induced unfolding of Jα helix of LOV2, we employed repulsive potentials and performed short DMD simulations on LOV2 (27). To locate suitable position in NAT for LOV2 insertion, we estimated the SASA and identified surface exposed loop regions with SASA ≥40 Å. Concurrently, we calculated the conservation scores of each residue using the ConSurf online server (https://consurf.tau.ac.il/overview.php). After determining likely insertion sites, we cut the peptide bond between the K187 and D188 residues of NAT and inserted dark- and lit-LOV2 states (27). We performed molecular modeling of NAT-LOV2 complexes using Modeller-9v14 (25). During the model generation, we imposed harmonic restraints on NAT to avoid unnecessary structural changes while generating NAT-LOV2 complex conformations. We analyzed 20 models generated for NAT-LOV2 complex in both dark and lit states and selected the best structural conformation with minimal clash score and least modeler objective function. We improved the quality of NAT-LOV2 complex structural data using Chiron (28) and Gaia (29) tools. We further optimized NAT-LOV2 structural models through several high-temperature DMD simulations (0.7 kcal/mol·k_B, where k_B is the Boltzmann constant). We estimated allosteric communications between insertion site and target site using Ohm server and assured that both sites are coupled through strong allosteric interactions (9). Subsequently, we employed constant-temperature, all-atom DMD simulations at 0.5 kcal/mol·k_B on NAT-LOV2 complexes for 5 × 10⁶ DMD time steps. We monitored the convergence of energy distributions for the protein during the simulations to evaluate its equilibration. To acquire statistically significant simulation data, we repeated each DMD simulation four times. All dynamic, structural, and energy parameters were extracted from the last 4 × 10⁶ DMD time steps of trajectories. We used PyMOL (https://pymol.org/2/) [Schrödinger, L.L.C. The PyMOL Molecular Graphics System, Version∼1.8 (Schrödinger, Inc, 2015)] and Visual Molecular Dynamics (30) to generate structural figures and movies.

Plasmid assembly and cloning

We cloned the NAT orf11 constructs following standard procedures. We first amplified the genes of interest from NAT (Orf11/∗Dbv8) and LOV2 vectors. NAT (Orf11/∗Dbv8) was synthesized by Genscript (Genscript Biotech Corp), LOV2 vector was purchased from Addgene (Addgene plasmid #87356). Next, we inserted the LOV2 sequence into the computationally determined loop with flanking glycine linkers using a modified megaprimer method, previously described (4). Briefly, NAT Orf11 vector and LOV2 insert fragments were generated with PCR amplification, adding homologous segments to the 5′ and 3′ ends of each fragment. The purified PCR products were assembled in subsequent ‘megaprimer’ PCR reaction, inserting LOV2 into NAT Orf11 between the K187 and D188 codons (in-frame). We treated the assembled PCR product with KLD enzyme mixture (NEB) for 5 min at room temperature, then transformed the treated construct into chemically competent DH5α E. coli strain (Zymo Research Corporation). Sanger sequencing was performed on single-colony clones to confirm correct insertion of LOV2 into NAT Orf11. oNAT-L (LOV2: I510E/I539E) and oNAT-D (LOV2: C450A) mutants were generated with the Q5 site-directed mutagenesis kit (NEB) following the manufacturer’s protocol. All primers and sequences can be found in the Supplemental Material (Tables S1 and S2).

Protein expression

E. coli BL21 (DE3) pLyseE strains were then transformed with the plasmid containing the protein of interest and cultured to an absorbance of 0.6 to 0.8 at 600 nm in LB broth. We induced gene expression with the addition of 0.5 mM isopropyl-β-d-thiogalactopyransoide to the bacterial culture. We incubated isopropyl-β-d-thiogalactopyransoide–induced cultures for 18 h at 18 °C. We then resuspended cell pellets in lysis buffer (20 mM Na2HPO4, 40 mM imidazol, 500 mM NaCl, pH 7.4) with protease inhibitors (1 mM phenylmethylsulfonyl fluoride and 1 μM pepstatin A) then lysed by sonication. We separated the supernatant containing protein components from precipitate by centrifugation at 15,000 rpm for 30 min at 4 °C.

Protein purification

All proteins were purified as previously described (31). We ran protein lysates through a His-trap column and eluted with 500 mM imidazole using an AKTA Pure FPLC machine. SEC was performed to confirm that minimal protein aggregation or degradation was occurring. We then concentrated the purified proteins with centrifugal filtration prior to storage at 4 °C.

Size-exclusion chromatography

We loaded the purified proteins (1 mg/ml protein in 20 mM Na2HPO4, 100 mM NaCl, pH 7.4) onto a Superdex 200 column (GE HealthCare) at a flow rate of 0.5 ml/min using an AKTA Pure FPLC machine. We monitored UV absorption at 280 nm. We ran standards (Gel filtration standard, Bio-Rad Laboratories) on the cleaned column prior to running the samples. The standards included five molecular weight markers: vitamin B12 (1.35 kDa), myoglobin (17 kDa), ovalbumin (44 kDa), γ-globulin (158 kDa), and thyroglobulin (670 kDa). We calculated the apparent molecular weight of designed proteins by plotting log of molecular weight versus elution volume for all the proteins.

Structural characterization

We measured the melting temperature (Tm) and CD of each construct using a Jasco J-1500 Spectrometer at the Penn State X-Ray Crystallography core facility. CD measurements were deconvoluted using Jasco CD software.

Enzymatic assay

We determined the acylation of our target compound with an in-house colorimetric assay, adapted from Tsirka et al. 2018 (32). We performed reactions in black clear-bottom 96-well plates [Greiner CellStar, Cat No 655090 (Greiner Bio-One)]. Reactions contained 1 μg of enzyme, 0.5 mM substrate [either 2-aminophenol (Sigma Aldrich; Part No. A71301) or 4-chloroaniline (Sigma Aldrich)] and 0.4 mM acyl-CoA donor (Avantilipids; Part No. 870710) diluted in 20 mM Tris-HCl. Final reaction volumes were 100 μl. Reactions were either stimulated by pulsed blue-light or kept in the dark for 30 min. Stimulation was administered with a custom-built illuminator (33). All reactions were performed at room temperature. Reactions were terminated with the addition of 25 μl of 5 mM DTNB (Ellman’s reagent) and 8 M urea diluted in 20 mM Tris-HCl. 405 nm wavelength absorbance was measured on an EnVision plate reader (PerkinElmer). A 0 to 200 μM CoA standard curve was generated. Reaction absorbance was related to free CoA to quantify enzymatic activity.

Data availability

All datasets generated and used in this study that are not present in the manuscript or supplemental information file are available from the corresponding author upon request.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare no conflict of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Joseph Jez