1 Introduction
The use of metabolic engineering to develop efficient microbial cell factories has been proved to be an attractive and powerful way to produce valuable chemicals and materials that are important for our society (Chae et al., 2017; Liu & Nielsen, 2019). However, the inherent complexity of cellular metabolism and the corresponding difficulties in balancing the trade-off between product formation and cell growth and viability have greatly slowed down the design-build-test (DBT) cycles (Nielsen & Keasling, 2016). This challenge motivates the need for new methods to accelerate the design and optimization of biosynthetic systems (Bowie et al., 2020; Bundy et al., 2018).
In recent years, cell-free systems have developed rapidly and gradually showed their strengths for speeding up DBT cycles (Dudley et al., 2015; Moore et al., 2018; Morgado et al., 2016; Sun et al., 2014). Cell-free systems are not constrained by the requirement of maintaining cellular viability and growth, thereby allowing the full allocation of carbon and energy resources to the product formation. Moreover, the openness of the cell-free systems allows direct access to the reaction conditions and cellular contents, providing great flexibility and freedom in the design and adjustment of biosynthetic reactions (Rasor et al., 2021; Vilkhovoy et al., 2020). Given the above superiorities, purified enzyme systems, which are the most common examples of cell-free biochemical synthesis, have been widely used to study enzymatic pathways and inform cellular expression (Bogorad et al., 2013; Dudley et al., 2015; Zhu et al., 2014). On the other hand, crude cell lysates have increasingly gained popularity for prototyping metabolism because they provide the endogenous metabolism for cofactor recycling and energy regeneration (Dudley et al., 2016, 2019; Jewett et al., 2008), which is limited in purified enzyme systems. Additionally, crude lysates also have the capability to build biosynthetic pathways by expressing functional catalytic enzymes directly in vitro by cell-free protein synthesis (CFPS) (Dudley et al., 2020; Grubbe et al., 2020; Karim et al., 2020; Rasor et al., 2022). The hybrid approach of CFPS driving metabolic engineering (CFPS-ME) has been successfully used to prototype the synthesis of polyhydroxyalkanoate (Kelwick et al., 2018), styrene (Grubbe et al., 2020), indole alkaloids (Khatri et al., 2020), valinomycin (Zhuang et al., 2020), and acetone (Rasor et al., 2022). In addition, the Jewett group developed an elegant approach termed in vitro Prototyping and Rapid Optimization of Biosynthetic Enzymes (iPROBE) in the context of CFPS-ME. In iPROBE, dozens of enzyme variants in hundreds of pathway combinations were rapidly tested to improve the productivity of butanol and limonene (Dudley et al., 2020; Karim et al., 2020). To build and assess different pathway combinations, the amounts of enzyme homologs that were produced by CFPS needed to be determined by the incorporation of14C-leucine in iPROBE. However, the procedure of radioactive incorporation is laborious, and radioactive 14C-leucine is unavailable for many laboratories. These constraints significantly limit the usefulness of iPROBE. Thus, there remains a great demand for an approach that enables testing and screening many enzyme homologs in a fast and technically simple manner.
β-Nicotinamide mononucleotide (NMN) is a key intermediate in nicotinamide adenine dinucleotide (NAD+) biosynthesis and exists in all living species. NMN has been demonstrated to have effective pharmacological activities in the treatment of various diseases, such as obesity, Alzheimer’s disease, and high fat diet-induced type 2 diabetes (Poddar et al., 2019; Yoshino et al., 2011). However, the current high price of NMN hampers the widespread use and practical implementation of this molecule. While there have been many efforts to improve the production of NMN by engineering biosynthetic pathways in vivo (Marinescu et al., 2018; Shoji et al., 2021) or in vitro (Qian et al., 2022; Zhou et al., 2022), these efforts have typically explored only a small set of enzyme homologs in their optimization strategies. Hence, productive enzyme homologs and combinations for efficient synthesis of NMN are still required.
The self-complementing split GFP, engineered from superfolder green fluorescent protein (sfGFP), was first developed by Waldo and his coworkers for protein tagging (Cabantous et al., 2005). In this system, sfGFP was asymmetrically split between β-strands 10 and 11 into a large (GFP1–10) and a small (GFP11) fragment. The two fragments were not individually fluorescent, but they could spontaneously interact with each other to form a functional GFP. By fusing GFP11 fragment on a target protein and detecting its association with GFP1–10 fragment, this system has been used in numerous biological studies including protein solubility assays (Cabantous & Waldo, 2006), screening of enzyme mutant libraries (Santos-Aberturas et al., 2015), and imaging protein localization in living cells (Kamiyama et al., 2016; R. Tamura et al., 2021). In addition, Karim and colleagues recently showed the possibility for quantification of protein produced in vitro by split-GFP (Karim & Jewett, 2018). However, no one to our knowledge has yet practically applied the split GFP system to prototype enzyme homologs.
In this work, a novel strategy, which combined CFPS with split GFP, was developed for prototyping enzyme homologs (Figure 1). The key idea was that the most productive enzyme homolog for each step in the candidate pathway was rapidly identified by using a normalized screening procedure. In this procedure, enzyme homologs were produced in parallel by CFPS in a few hours, and the expression level and activity of each homolog were determined simultaneously by using the split GFP assay. As a proof of concept, the capacity of this strategy was demonstrated by optimizing a three-step pathway for synthesizing NMN. By using this strategy, the time for testing 10 enzyme homologs of each catalytic step was reduced from a few weeks to 72 hours. Additionally, NMN biosynthesis was further optimized by improving physiochemical conditions, tuning enzyme ratios and cofactor concentrations, and decreasing the feedback inhibition to reach a 12-fold improvement over our initial setup. As a result, it was expected that this strategy would accelerate the timeline of DBT cycles and enhance efforts to optimize the production of desired products in cell-free systems.