3. Results and discussion

A cascade reaction enables fluorogenic determination of ADH activity

As shown in Figure 1, the catalytic activity of ADHs in an oxidation reaction can be expressed as the rate of forming NAD(P)H. Although it is possible to quantitate NAD(P)H by directly measuring its light absorbance (UV 340 nm), this method suffers from low sensitivity and high background noise (R. Zhang et al., 2015). Alternatively, measuring the fluorescence of NAD(P)H Ex/Em=340/445 nm) can be used to quantify NAD(P)H and has shown a greater sensitivity compared to the absorbance-based approach (Held, 2007). However, our experiments showed that although the linear correlation between NADH concentration and its fluorescence can be obtained when NADH was serially diluted with deionized water, buffer or fresh media (Figure S1 A-D), such a correlation cannot be established when spent media was used as the diluent (Figure S1 E-F). These results suggest that metabolites released by growing cells can substantially interfere with the fluorescence detection of NADH, and therefore such fluorescence assay is not suitable for determining the NADH level in crude samples such as cell lysate or culture supernatant.
To tackle this issue, we employed an indicator enzymatic reaction in which NADH was used by diaphorase to reduce resazurin into resorufin, which is highly fluorescent (Figure 1). The main advantage of this system is that the long excitation/emission wavelengths of resorufin (Ex/Em=535/588 nm) can reduce the interferences from cellular metabolites, because most fluorescent molecules in cell lysates are excited in the UV range (Simeonov & Davis, 2004). As expected, our results showed that fluorescence of the formed resorufin responded linearly to NADH concentration across different medium backgrounds, including spent media (Figure S1 G-L). Meanwhile, this reaction system offers a much more extended linear detection range (0.1-700 μM) compared to NADH fluorescence-based approach (0.1-200 μM when the medium background is deionized water). Additionally, through transforming the unstable NADH into the more stable resorufin, this reaction system provides a more durable and accurate fluorescent signal, allowing a reliable and continuous fluorogenic determination of NADH concentration.
Given these advantages, we may couple an ADH-catalysed reaction with this indicator system to form a coupled enzyme assay for the ADH (Figure 1). To examine whether the ADH activity can be determined by this cascade reaction, we employed an enantioselective ADH, (R )-2-octanol dehydrogenase (PfODH) as the model enzyme. In the cascade reaction mixture, the purified enzyme PfODH was supplied with cofactor NAD+ and substrate to initiate target reactions. Concurrently, excessive amount of resazurin and diaphorase were provided in the mixture to stoichiometrically convert NADH into fluorescent resorufin. The enzymatic activity of ADH was thus proportional to the rate at the red fluorescence increased. As shown in Figure 2, the catalytic activity of PfODH toward various substrates, including its native substrate ((R )-2-octanol) and other non-native substrates (4-Fluoro-α-methylbenzyl alcohol, 1-(3-Methylphenyl)ethanol and 1-Phenylethanol), were determined by following the respective increases in fluorescence of resorufin over time using a plate reader. In contrast, the negative controls without adding PfODH all exhibited no fluorescence increase over time. These results showcased that the coupled enzyme assay can enable a sensitive and continuous, time-dependent monitoring of the ADH-catalysed reaction.

Incorporating the coupled enzyme assay with a protein secretion system

High-throughput screening used in enzyme evolution has relied heavily on the use of cell-based approach (Zeymer & Hilvert, 2018). However, there are two major defects with this method: (i) It requires cell lysis that could introduce interference to the screening assay; (ii) It relies on the false assumption that the amount of enzyme being assayed is uniform across all mutants, therefore resulting in an unfair comparison that consequently increases false positive or negative rates. To further develop our coupled enzyme assay into a better screening method, we set out to address these two limitations.
In particular, we proposed to combine the cascade reaction with a protein secretion system which is based on MsfGFP. The addition of such a secretion system can potentially offer two unique features to our assay. First, it has been reported that when recombinant proteins were fused with MsfGFP, it can lead to the auto-secretion of the fusion protein without disturbing their conformation and function (Z. Zhang et al., 2017). Such a non-specific auto-secretion system is highly desirable because cell lysis can therefore be exempted, and the secreted target enzyme can be assayed in the culture broth, which is a much simpler matrix than cell lysate. Second, MsfGFP can serve as a protein expression reporter for its passenger protein. It has been demonstrated that the fluorescence of sfGFP (non-mutated superfolder GFP) fusion protein is linearly proportional to the protein quantity (Pédelacq, Cabantous, Tran, Terwilliger, & Waldo, 2006). This feature can potentially be used to normalize the enzyme activity of each mutant during high throughput screening, since the fluorescence of MsfGFP fusion protein can be easily and efficiently measured using fluorescence plate reader.
To evaluate whether the MsfGFP-guided secretion system is generally applicable to ADHs, we fused the MsfGFP or eGFP (enhanced GFP, a control which would not lead to protein secretion) to N-terminus of five ADHs, including PfODH from Pichia finlandica, (S )-1,3-butanediol dehydrogenase (CpSADH) from Candida parapsilosis, glycerol dehydrogenase (GlyDH) from E. coli , Meso-2,3-butanediol dehydrogenase (BDH_KP) from Klebsiella pneumoniae and 2,3-butanediol dehydrogenase (BDH_BS) from Bacillus subtilis . The protein expression of the MsfGFP/eGFP-ADH fusion proteins was induced by IPTG. The grown cell cultures were then centrifuged to obtain the supernatants for examining whether the fusion proteins were secreted out of the cells. As shown in Figure 3, all the supernatants of the cells expressing the MsfGFP fusion proteins exhibited fluorescence to various degrees, whereas all the controls did not show any fluorescence. Meanwhile, the presence of secreted MsfGFP fusion proteins and the absence of eGFP fusion proteins in supernatant were further confirmed by SDS-PAGE analysis (Figure 3). Consistent with the fluorescence observation, the respective protein bands of MsfGFP fusion proteins showed varying degrees of staining, further indicating different levels of secretion for the MsfGFP fusion proteins. These results suggest that secretion of five ADHs can be realized by using this MsfGFP-guided auto-secretion system, although with various degree of secretion efficiency.
After we confirmed MsfGFP can enable the secretion of ADHs, we moved forward to examine whether the secreted ADHs’ activity can be determined by the coupled enzymatic assay. Assay reaction components were directly added to the culture supernatants of each fusion protein. As shown in Figure 4, the supernatant samples of most MsfGFP-guided ADHs (GlyDH, CpSADH, PfODH, and BDH_KP) generated various fluorescence increases over time through the assay reaction, whereas there was no fluorescence change for all the supernatant sample of eGFP fusion proteins. However, unlike other ADHs, the rate of red fluorescence increase was low for MsfGFP-BDH_BS, which may be mainly due to its low secretion level. Nonetheless, these results demonstrated that the MsfGFP-guided secretion system can be combined with the coupled enzyme assay to establish a secretion-based assay for determining activity of some ADHs.

Normalizing ADH activity by the fluorescence of MsfGFP

We next assessed the feasibility of using the fluorescence of MsfGFP to normalize the catalytic activity of ADHs among different mutants in high-throughput screening. To mimic the actual screening condition of the secreted ADHs, we purified the fusion protein MsfGFP-PfODH and serially diluted it into different concentrations with cell-free spent medium. The green fluorescence (Ex/Em=488/525 nm) of the diluted enzyme samples were first determined by fluorescence plate reader. The green fluorescence intensity exhibits a strong linear correlation to the concentration of MsfGFP-PfODH (Figure S2A). Subsequently, the catalytic activity of the MsfGFP-PfODH samples with various concentrations were determined by measuring their respective red fluorescence (Ex/Em=488/525 nm) after adding coupled enzyme assay reagents. When (R )-2-octanol was used as the substrate, the samples with higher concentrations of MsfGFP-PfODH exhibited higher red fluorescence intensities (Figure 5). Apparently, without considering the different amounts of enzyme in the assay, the specific activity of these enzyme samples, which was supposed to be the same, can be misrepresented based on the red fluorescence measurement. To address this limitation, we used the green fluorescence intensity, which reflected the total amount of the ADH in the assay, to normalize the red fluorescence intensity of each sample. The normalized signal of each sample was very close to each other (Figure 5), indicating similar specific activity. Moreover, Figure S3 showed that the same normalizing effect of MsfGFP’s fluorescence can also be applied when a different substrate ((R )-2-butanol) was used in the assay reaction. In addition, our further experiments also proved that this normalizing feature of MsfGFP also worked for another ADH (Figure S4). These results demonstrated that the apparent activity of ADHs can be normalized by the fluorescence of MsfGFP fusion proteins to enable a fairer comparison of enzyme mutants, thus reducing the false positive and false negative rates during screening.

Using the secretion-based dual fluorescent assay in protein directed evolution

Considering the aforementioned advantages, we combined the resorufin-based cascade reaction system and the MsfGFP-guided secretion system to establish the secretion-based dual fluorescent assay (SDFA) for directed evolution of ADHs. On the one hand, the cascade reaction system offers a sensitive and continuous determination of ADHs activity through a red fluorescence measurement. On the other hand, MsfGFP-guided secretion system provides an easy and efficient way to translocate the target ADHs to a simpler reaction matrix and to normalize the enzyme’s activity through a green fluorescence measurement.
In the workflow of SDFA (Figure 6), the mutated genes of an ADH were first fused to the C-terminus of msfgfp to create a plasmid library. The host cells (E. coli ) were then transformed with the constructed plasmids and were spread on selection agar plates. Fully grown mutant colonies were inoculated in liquid culture medium and induced for enzyme expression. Subsequently, the cell cultures were centrifuged down to obtain the supernatants that contained the secreted enzymes for further enzyme screening. The green fluorescence of the supernatants from the enzyme mutants was first determined by a plate reader. Afterwards, assay reagents were added to the supernatant samples to monitor the red fluorescence. Finally, the catalytic activity of each mutant was normalized (red/green fluorescence) to identify the true positive mutant hits.
As a case study for demonstrating the applicability of SDFA in high-throughput screening, we employed PfODH as the model enzyme and sought to engineer its stereoselectivity. In oil industry, alcohol oxidation is one of the most extensively studied reactions as it plays a key role in converting petroleum-based feedstock into essential building blocks for a wide variety of chemicals (Parmeggiani, Matassini, & Cardona, 2017). PfODH belongs to the short chain dehydrogenase/reductase (SDR) superfamily. It has a very broad substrate specificity and can catalyse the oxidation of alcohols to the corresponding aldehydes or ketones, which are the ubiquitous chemical precursors in pharmaceutical and fine chemical industries (H. Yamamoto & Kudoh, 2013). However, PfODH has a substrate preference towardR -enantiomer of alcohols, thus resulting in incomplete conversion of racemic alcohol, i.e. S -enantiomer of alcohols cannot be fully utilized (H. Yamamoto & Kudoh, 2013). From an atom economy point of view, enzymes that can oxidize both enantiomers of alcohol are more desirable for this application.
Therefore, we set out to engineer the stereoselectivity of PfODH through screening enzyme variants with enhanced activity towardS -enantiomer of alcohol. To achieve this goal, we constructed a combinatorial library, PfODH-G99VWA-H150WTA-Y193RBT (containing 60 genotypes), on critical sites G99, H150, and Y193 (Figure 7), which are involved in the formation of the active cavity and extend their side chains into the cavity of PfODH. These corresponding sites were suggested as the switch to tune the stereoselectivity of the enzymes in the superfamily of SDRs (Qin et al., 2018). Variation of steric hindrance and polarity on sites of G99, H150, and Y193 might influence the stereo-preference of PfODH (Qin et al., 2018). Therefore, the mentioned library was constructed as the demonstration to test the applicability of SDFA as a high throughput screening method in directed evolution of ADHs.
To ensure full coverage of the design space for our constructed library, 307 clones were picked for enzyme expression. These 307 mutants were assayed by SDFA in 96 well plate format (Figure S5). Through screening, top 10 variants that exhibited the highest catalytic activity against (S )-2-octanol were selected for sequencing to identify their genome type. Sequence analysis revealed that 8 out of the 10 selected mutants were identified as PfODHG99L/A155L/Y193A, whereas the remaining two mutants are PfODHG99Q/A155L/Y193A and PfODHG99V/A155L/Y193A. Given the fact that PfODHG99L/A155L/Y193A was dominant in the top 10 positive hits and also showed higher catalytic activity than the other two variants, PfODHG99V/A155L/Y193A was identified as the best mutant from this library screening. To cross-validate the screening result of SDFA, PfODHG99L/A155L/Y193A was expressed in shaking flask cultures and purified for further determining its kinetic parameter. Our results showed that PfODHG99L/A155L/Y193A had a 197-fold higher kcat/km value toward (S )-2-octanol in comparison with wild-type PfODH. These results demonstrated the capacity of SDFA to identify improved enzyme variants from a library of mutants, illustrating the usefulness of this assay in high-throughput screening of ADH mutants.