1. Introduction

Many drugs and value-added chemicals need to meet stringent enantiopurity standards in pharmaceutical and fine chemical industries. During the last decade, there has been a growing interest in using enzymes for producing chiral chemicals. This is mainly driven by the inherent advantages of biocatalysis over chemical synthesis, such as high chemo-, regio- and enantio-selectivities, mild reaction conditions and using less toxic reagents or solvents (Nealon, Musa, Patel, & Phillips, 2015; Zheng et al., 2017).
In particular, alcohol dehydrogenases (E.C. 1.1.1.x; x=1 or 2, ADHs) are among the most investigated redox enzymes. ADHs are a group of NAD(P)(H)-dependent enzymes that catalyze the interconversion between alcohols and ketones or aldehydes. Because of their strict stereoselectivity, ADHs have been employed in the production of various chiral products with high optical purity (Chen, Liu, Lin, & Zheng, 2016; Kara et al., 2014; Savile, Gruber, Mundorff, Huisman, & Collier, 2014). However, the performance of wild-type ADHs is rarely adequate for industrial processes. In general, tailoring enzyme properties, such as catalytic activity, stereoselectivity, substrate specificity, thermostability and tolerance of organic solvent is often required to realize the practical application of ADHs (R. Zhang, Xu, & Xiao, 2015).
Directed evolution is one of the most effective methods to engineer ADHs. An ADH may be randomly mutated by using error-prone PCR to identify beneficial mutations. Putative critical residues of an ADH may also be identified based on docking the substrate/product into experimental-determined or predicted tertiary structure of the ADH. Beneficial mutations at these putative residues would be characterized through mutagenesis. Often, multiple beneficial mutations need to be combined and the aforementioned process may be iterated.
In directed evolution, screening the mutants is usually the rate-limiting unit operation, because the rate of hitting a beneficial mutation is low. The assay commonly used to quantify activity of ADH is to monitor the production or depletion of NAD(P)H based on absorbance at 340 nm as these enzymes couple their enzymatic reactions with the nicotinamide cofactors (Forrest & Gonzalez, 2000). Usually cells expressing an ADH was lysed prior to the assay. Unfortunately, this approach is not suitable for high-throughput screening, because of the strong absorption of UV light by cell lysates (Mayer & Arnold, 2002).
Alternatively, colorimetric assays that combine the redox reaction with a dye-forming reaction have been developed to reduce the noise, because the absorption of visible light by cell lysates was weaker than that of UV light. One of the most widely adopted colorimetric approaches is the nitroblue tetrazolium/phenazine methosulfate (NBT/PMS) assay (Fibla & Gonzàlez-Duarte, 1993). In the presence of the transferring agent PMS, NBT can be reduced by the NAD(P)H regenerated by an ADH-catalyzed reaction to produce an insoluble blue-purple formazan dye, whose absorbance can be measured at 580 nm. This colorimetric assay has been used in high-throughput screening to identify novel or improved ADHs, such as 6-phosphogluconate dehydrogenase (Huang, Chen, Zhong, Kim, & Zhang, 2016), glycerol dehydrogenase (H. Zhang, Lountos, Ching, & Jiang, 2010), halohydrin dehydrogenase (Fox et al., 2007), and phenylacetaldehyde reductase (Makino, Dairi, & Itoh, 2007). However, this colorimetric assay still needs improvement due to its background noise and limited detection range. Moreover, this assay also fails to enable a continuous measurement of ADHs activity as the final colour product is insoluble, resulting in precipitation that interferes with the absorbance quantification.
In addition, the aforementioned assays generally suffer from two common issues when they are used to screen cells that express enzyme. First, in order to release target enzyme from host cells for determining catalytic activity, cell lysis is often required by these methods. However, once the cells are lysed, together with the enzyme of interest, all cytosolic components are simultaneously released, resulting in a complex matrix that could reduce the sensitivity and accuracy of the assay implemented (Contreras-Llano & Tan, 2018; Mayer & Arnold, 2002). For example, cytosolic components, such as nicotinamide cofactor, autofluorescent or chromogenic metabolites may miscount for the catalytic activity of the target enzyme. Non-specific interactions with endogenous enzymes (e.g. endogenous ADHs) could also interfere with the target enzyme reaction if they share similar substrate specificity.
Furthermore, when these assays were applied in high-throughput format, such as microtiter plate screening, the catalytic activity was usually evaluated without normalizing by the amount of enzyme interrogated. Because, it is time-consuming and tedious to quantify the enzyme amount for each variant in a large-scale screening effort. Consequently, this inevitably leads to an unfair comparison among the mutants. The quantity of the target enzyme in each mutant sample could vary due to fluctuating protein expression level and nonhomogeneous cell growth rate. Overall, this may not only produce a considerable number of false positives, but also miss out some of the true positives. It will be desirable to have a simple and easy way to quantify the enzyme amount for each mutant so that their catalytic activity can be normalized for a fairer screening.
To tackle these limitations in this study, we developed a secretion-based dual fluorescence assay (SDFA) for high-throughput screening of mutated ADHs. This assay is based on a cascade reaction leading to the formation of a red fluorescent molecule. It also uses a protein secretion system translocating the target enzyme to a simpler reaction matrix than cell lysate.
As shown in Figure 1, in the presence of an enzyme (diaphorase), resazurin reacts with the NAD(P)H generated from the ADHs-catalysed reaction, leading to the generation of the highly fluorescent molecule—resorufin—whose emission signal can be easily detected at 588 nm when excited at 535 nm. Since the emission falls within the range of red light, in which fewer Escherichiacoli endogenous molecules emit fluorescence, there is less interference from the reaction mixture and spent culture media.
Meanwhile, we combined the cascade reaction with a protein secretion system which was enabled by a mutated superfolder green fluorescent protein (MsfGFP) (Z. Zhang et al., 2017). By fusing ADHs to MsfGFP, the fusion proteins can be efficiently secreted out of the host cells for enzymatic activity quantification. Therefore, cell lysis was not required by this assay. This led to a much simpler reaction matrix that was free of complex cytosolic components. As a result, the interference caused by intracellular metabolites and endogenous enzymes was eliminated. Also, this secretion system simplified the screening procedure and avoided the potential inconsistency of lysis efficiency.
Additionally, another merit of this assay is that the green fluorescence of the MsfGFP fusions was linearly proportional to its protein amount. This provides a simple and efficient way to quantitate the amount of target enzyme in the reaction matrix. Therefore, normalization of the catalytic activity of different mutants can be easily achieved, further improving the accuracy of the assay.
In this study, we have collected experimental data to support the claimed advantages of SDFA, and successfully demonstrated the usefulness of SDFA through engineering of an ADH, which led to substantial improvement of its catalytic efficiency.