Multi-level rebalancing of the naringenin pathway using riboswitch-guided high-throughput screening

Highlights

• Multi-level rebalancing at both transcription and translation stages by riboswitch.

• A threefold improvement of naringenin titer over the transcription-optimized strain.

• Multi-level optimization strategy uncovered previously inaccessible solution space.

• The highest naringenin concentration, 260.2 mg/L, for glycerol-based production.

• This strategy can be broadly applicable for maximizing metabolite production.

Abstract

Recombinant microbes have emerged as promising alternatives to natural sources of naringenin—a key molecular scaffold for flavonoids. In recombinant strains, expression levels of the pathway genes should be optimized at both transcription and the translation stages to precisely allocate cellular resources and maximize metabolite production. However, the optimization of the expression levels of naringenin generally relies on evaluating a small number of variants from libraries constructed by varying transcription efficiency only. In this study, we introduce a systematic strategy for the multi-level optimization of biosynthetic pathways. We constructed a multi-level combinatorial library covering both transcription and translation stages using synthetic T7 promoter variants and computationally designed 5′-untranslated regions. Furthermore, we identified improved strains through high-throughput screening based on a synthetic naringenin riboswitch. The most-optimized strain obtained using this approach exhibited a 3-fold increase in naringenin production, compared with the parental strain in which only the transcription efficiency was modulated. Furthermore, in a fed-batch bioreactor, the optimized strain produced 260.2 mg/L naringenin, which is the highest concentration reported to date using glycerol and p-coumaric acid as substrates. Collectively, this work provides an efficient strategy for the expression optimization of the biosynthetic pathways.

Introduction

Naringenin is a key molecular scaffold in the production of flavonoids (Fowler and Koffas, 2009; Wang et al., 2019). Flavonoids, including naringenin, are high value-added substances with diverse pharmacological properties for the treatment of obesity, diabetes, viral infection, hypertension, and cancer (Alam et al., 2014, Nahmias et al., 2008; Forkmann and Martens, 2001; Hollman and Katan, 1998). Extraction from natural sources yields very small amounts of flavonoids and requires a large number of solvents. Therefore, recombinant microbes have emerged as robust and promising factories for the economical production of flavonoids (Chemler and Koffas, 2008) owing to their scalability and the availability of renewable resources.

A number of efforts have been made to construct recombinant strains by expressing heterologous biosynthetic enzymes using Escherichia coli or Saccharomyces cerevisiae with metabolic engineering strategies (Table 1). For naringenin-producing recombinant strains, three heterologous enzymes have been typically introduced: p-coumaric acid-CoA ligase (4CL); chalcone synthase (CHS); chalcone isomerase (CHI) (Fig. 1). A simple overexpression of these enzymes, however, often lead to low productivity (Wang et al., 2019; Jones et al., 2016, Marschall et al., 2016, Zhou et al., 2019), reinforcing the need of expression optimization.

An efficient experimental approach for pathway optimization consists of the construction of a combinatorial expression library and the adaptation of a high-throughput screening method. Ideally, the library should cover a large solution space and the screening should be able to identify as many variants in the library as possible (i.e. high-throughput). Although gene expression control can be conducted at various levels by the modulation of gene copy numbers, transcription efficiency, and translation efficiency, the construction of the combinatorial library has focused on controlling gene expression at a single expression stage, i.e., transcription or translation (Table 1). This offers limited coverage of the enormous possible solution space due to the limitation in the screening method to evaluate each cell based on its naringenin-producing capability.

The screening of naringenin-producing microbial strains relies on the direct measurement of the metabolite concentration (Jones et al., 2015, 2016), on a colorimetric assay based on the color of a pathway intermediate (Alberstein et al., 2011), or on additional chemical reactions (Gao et al., 2020, Zhou et al., 2019), all of which are low-throughput methods. Recently, a synthetic naringenin riboswitch was developed as a high-throughput screening device in individual cells (Jang et al., 2017, Xiu et al., 2017). With this biosensor, naringenin production can be quantitatively analyzed as a fluorescence signal. Accordingly, it can be readily coupled with various high-throughput methods, enabling efficient screening of large expression libraries. Notably, riboswitch-based biosensors do not require regulatory protein expression, unlike transcription factor-based biosensors (Wang et al., 2019). Moreover, the cis-acting mechanism of the riboswitch should minimize the off-target effects often associated with trans-acting transcription factor-based biosensors (Yang et al., 2013).

In this study, we introduced a systematic strategy for the multi-level expression optimization of biosynthetic pathways based on synthetic biology tools. This approach involved the generation of a wide and balanced expression library both at the transcriptional and translational levels and the subsequent high-throughput screening using the naringenin riboswitch (Fig. 2). We used the final strain from a previous single-level transcription optimization (Jones et al., 2016) with the increased intracellular malonyl-CoA availability (Xu et al., 2011) as a parental strain to evaluate the performance of our strategy. In the study, the parental strain showed the highest production of naringenin with glycerol as the substrate with high economic advantages (da Silva et al., 2009; Martínez-Gómez et al., 2012). Previous studies (Table 1) mainly focused on transcriptional optimization (Gao et al., 2020; Wang et al., 2019, Xu et al., 2012, Zhou et al., 2019), and additional chemical reactions were required for screening or naringenin production (Gao et al., 2020; Palmer et al., 2020). In addition, some studies for naringenin production were not suitable for economic bioprocess due to the use of cerulenin (Dunstan et al., 2020; Leonard et al., 2008; Santos et al., 2011). In fact, the high cost of cerulenin is a critical constraint to establish economical production processes (Supplementary Note 1). In this context, several studies attempted to establish glycerol-based naringenin production process without the use of expensive compounds (Xu et al., 2012; Jones et al., 2016; Yang et al., 2015), but further strain improvement is still needed to achieve economic feasibility. Collectively, the parental strain we used in this study enabled us to establish a naringenin production process using glycerol as an economical substrate without the addition of cerulenin, an expensive fatty acid inhibitor for increasing malonyl-CoA availability. Furthermore, the multi-level gene expression optimization was attempted for the first time in this study to improve naringenin production.

Collectively, we used T7 promoter mutants and in silico-designed 5′-UTR variants simultaneously to generate an expression library for pathway genes. Next, we transformed the naringenin riboswitch into the library and performed a fluorescence-based screening to identify improved mutants. Furthermore, the potential of the isolated mutants for industrial production of naringenin was confirmed by fed-batch fermentation. Finally, through analysis of protein expression of the screened mutants, this study demonstrated that optimized expression cannot be easily achieved with a single rational approach, thus emphasizing the importance of multi-level combinatorial strategy.