Engineering microbes to have desired production phenotypes requires optimization of many physiological constraints (Pontrelli et al., 2018). These include the ability to express non-native enzymes, supply of required cofactors, precursor availability, oxygen tolerance, and accumulation of toxic intermediates or products. While rational design strategies prove useful in addressing many of these factors, there often remains a need for further optimization to improve production.

Metabolomics, a technology that involves the global quantitative assessment of metabolites in a biological system, can detect complex biological changes using statistical multivariate pattern recognition methods (chemometrics) (Putri et al., 2013). The system-wide analysis provided by metabolomics can be used to gain a deeper understanding of how various genetic modifications affect production phenotypes. Observations made with metabolomics have led to strain modifications that successfully increased biofuel production titers (Nitta et al., 2017; Ohtake et al., 2017; Fathima et al., 2018). For improvement of 1-butanol production in E. coli, a metabolomics-based approach was used to identify an imbalance of Coenzyme A caused by a Clostridial butanol production pathway. Further genetic modifications based on observations made using metabolomics allowed a production increase from 15 to 18.3 g/L of butanol (Ohtake et al., 2017). Similarly, an orthogonal partial least squares/projections to latent structures (PLS) model-based metabolomics approach was used to identify acetyl-CoA as the rate limiting step of 1-butanol production (Nitta et al., 2017).

In this work, we employed a metabolomics-based approach to identify targets for improving production of 1-propanol in E. coli. 1-Propanol is recognized as an important compound that is industrially valuable for both energy applications and as a C₃ chemical feedstock (Liao et al., 2016). Production of 1-propanol has been achieved via the keto acid pathway (Atsumi and Liao, 2008; Shen and Liao, 2013), which offers several advantages owing to its universality and compatibility with the host. 2-Keto acids are intermediates in amino acid biosynthesis pathways that can be found in almost all living organisms. The theoretical yield of 1-propanol was enhanced by combining the native threonine pathway with a heterologous citramalate pathway using glucose as a substrate (Shen and Liao, 2013). Different cofactor requirements are exhibited by each pathway individually. The threonine pathway requires 3 NAD(P)H and 2 ATP as inputs, while the citramalate pathway produces 1 NAD(P)H and 1 ATP as outputs (Figure 1). When the overall net reactions of each pathway are combined, the individual cofactor requirements can complement one another, resulting in a synergistic effect that improves the theoretical maximum yield. The efficacy of this synergistic approach has proven successful in improving the observed 1-propanol yield and productivity (Shen and Liao, 2013).

With this production strategy, each pathway converges into a common intermediate, 2-ketobutyrate (2KB), which is then converted into 1-propanol through reactions catalyzed by alpha-ketoisovalerate decarboxylase (KivD) and alcohol dehydrogenase (YqhD) (Figure 1). 2-Ketobutyrate is also a native precursor in the E. coli leucine biosynthesis pathway and subsequently converted into 2-ketovalerate. Because of the promiscuous nature of KivD and YqhD, 1-butanol is also formed as a side product. However, in this work, we focus solely on improving production of 1-propanol.

It has been observed that overexpression of the citramalate pathway in conjunction with the native threonine pathway exhibits an increased productivity (hereafter referred to as strain SYN3). Yet, there is no observed change in titer when the threonine pathway is further overexpressed over basal metabolic levels (hereafter called strain SYN12) (Shen and Liao, 2013). One recent study uses snapshot metabolomic profiling and dynamic metabolic turnover analysis to determine why the presence or overexpression of the citramalate or threonine pathways contributes to this change in productivity (Putri et al., 2018). The strains used in their study were SYN2 (citramalate pathway overexpressed and native threonine pathway deleted), SYN6 (threonine pathway overexpressed), and SYN12 (both pathways overexpressed). Notably, this study identified the importance of the PEP/pyruvate ratio, which affects glucose consumption, and how ammonia assimilation affects the rate of the threonine pathway. Extending these findings, we here build on this study by increasing 1-propanol titer in SYN12. First, derivatives of 1-propanol producing strains were cultivated under the optimal condition for production. Snapshot non-targeted metabolome analysis using gas chromatography/mass spectrometry (GC/MS) and widely targeted metabolome analysis using ion-pair LC/MS/MS were performed. The data sets obtained from GC/MS and ion-pair liquid chromatography (LC)/MS/MS (LC/MS/MS) were subjected to multivariate analyses to extract the most meaningful metabolites corresponding to the phenotype. Following the analysis of important metabolites, the corresponding metabolic pathways were examined, and gene targets were systematically selected to construct strains with improved production phenotypes.

The synergistic combination of both the citramalate pathway and threonine pathway to produce 1-propanol has been used previously to demonstrate improved theoretical yields and productivity. However, no change in titer has been observed with the overexpression of the threonine pathway over native levels. Here, we applied metabolomics-based strategies to understand the limitations of the synergistic pathway and to engineer SYN12 for increased 1-propanol titers.

The metabolomics strategies that were applied in this work contributed to a streamlined and logical series of strain improvement steps that ultimately allowed for increased production titers. Initial metabolomics analysis compared SYN3 and SYN12 to reveal an increase in concentrations of 2-aminobutyrate, norvaline, and 2-ketobutyrate in SYN12. SYN12 is additionally modified to overexpress the native threonine pathway. While no increase in titer was observed, the physiological consequences of the pathway overexpression became apparent in the accumulation of these upstream metabolites and byproducts.

We therefore sought to focus efforts on channeling the metabolic flux toward 1-propanol production. However, gene deletions of avtA, aimed at decreasing unwanted byproducts 2-aminobutyrate, had adverse effects on the production titers. Additional experiments aimed at increasing flux through glycolysis by overexpression of tpiA, and subsequently through the 1-propanol pathway, also resulted in a 2KB accumulation with a significant decrease in 1-propanol titers. 2KB is a short-chain alpha keto acid and a structural analog of pyruvate. As an indispensable metabolic intermediate in biological metabolism, 2-KB is found in the biosynthesis of L-isoleucine (Park et al., 2012; Yin et al., 2012), 2-aminobutyrate (Xu et al., 2019), 1-butanol (Atsumi and Liao, 2008), L-homoalanine (Zhang et al., 2010), and so on. However, because of its structural similarity with pyruvate, 2KB competitively inhibits enzymes used in pyruvate metabolism at high concentrations, which causes cellular toxicity (Zhang et al., 2016; Fang et al., 2021). As such, this suggests that enhancing the turnover rate of 2-KB to 1-propanol with KivD and YqhD may relieve toxicity caused by 2KB accumulation.

We then focused on optimizing flux of KivD and YqhD to relinquish this bottleneck. We thus constructed RBS libraries to vary the expression of each enzyme and isolated a variant. While SYN12-KivD34 increased KivD34 specific activity by 69%, it did not show increased titers (Supplementary Figures 3B,C). This might be due to the toxic intermediate propyl aldehyde, which can generate membrane lipid peroxidation (Pérez et al., 2008), thus suggesting that the turnover rate of propyl aldehyde to 1-propanol catalyzed by YqhD was not sufficient. SYN12-YqhD18, with a 33% increase in YqhD specific activity, exhibited a 38% increased titer and a 29% increased yield in 72 h compared to the base strain SYN12, which was the best 1-propanol producing strain constructed in Shen and Liao (2013). These results may imply that there was an insufficient turnover rate of toxic intermediates 2KB and propyl aldehyde into 1-propanol in SYN12. This highlights two potential future directions for further strain improvement. First, since propyl aldehyde is a toxic intermediate, it may be beneficial to simultaneously fine-tune expression of both of these pathway enzymes. However, generating RBS libraries to optimize both enzymes would result in a combinatorial library with a much greater number of variants, and thus, would require screening methods with larger throughput. Second, since turnover and toxicity of 2KB is improved with YqhD18, a future target for further engineering may be to increase flux through the pathway, possibly with overexpression of tpiA as demonstrated in SYN17. These future possibilities further highlight the value of metabolomics in the design-test-build cycle that is endemic to metabolic engineering and strain improvement.

Previous works have aimed at improving 1-propanol production in E. coli with various strategies. In one example, E. coli was modified to contain a keto-acid pathway to convert 2-ketobutyrate to propionyl-phosphate, which is then sequentially converted to propionate, propionyl-CoA, propionyl-aldehyde, and finally to 1-propanol (Choi et al., 2012). A separate study extended succinate dissimilation under anaerobic conditions (Srirangan et al., 2014). The titer reported here is much higher compared to other hosts such as in Corynebacterium glutamicum [12 mM (721 mg/L) from glucose] (Siebert and Wendisch, 2015), Propionibacterium freudenreichii (0.51 g/L from glycerol) (Ammar et al., 2013), and Saccharomyces cerevisiae (180 mg/L from glucose) (Nishimura et al., 2018). The efficient genetic tractability of E. coli allows for more complex genetic modifications than other species, and therefore E. coli has potential to be further engineered for higher 1-propanol production.

The successful strain improvement in this work demonstrates the application of metabolomics-guided metabolic engineering strategies. More importantly, the data generated from metabolomics highlights metabolic perturbations that occur from pathway bottlenecks that may be relevant to the improvement of other production pathways.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.

KN, NK, and WL: metabolite measurement and data curation and analysis. TO: data curation and interpretation, and writing the manuscript. SPu: metabolite measurement, data analysis and interpretation, supervision, and writing the manuscript. CS: strain construction, data analysis and interpretation, and revising manuscript. SPo: strain construction, data analysis and interpretation, and writing the manuscript. EF and JL: supervision, project management, and resources. All authors have read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.