Biological approaches to address climate change and to sequester carbon dioxide (CO2) have focused on the development of microbial strains engineered to produce chemicals and fuels derived from renewable sources of sugar (1). Despite considerable success at engineering these strains at a small scale and the availability of systems biology based approaches for engineering strains, their success at the commercial scale has been hindered by financial limitations, particularly in the face of low oil prices and expensive feedstocks (1,2). In response, non-sugar feedstocks have emerged as alternatives. While CO2 has itself been considered as a feedstock, other alternative feedstocks typically consist of compounds derived from CO2 or from industrial waste materials (1,3). These are commonly one-carbon (C1) compounds such as methane (CH4), methanol (CH3OH), formaldehyde (C2HO), formate (HCOO−), carbon monoxide (CO), and syngas (predominantly CO and H2) (1,3–6). Notably, methane and syngas fermentations are currently under intense study and are the focus of commercial development (7–9).
A number of CO2 conversion technologies exist, including biological, chemocatalytic, thermochemical, photochemical, and electrochemical methods (4,10). From these, the electrochemical conversion of CO2 has been identified as promising, owing to several advantages (10). Such advantages include its ambient operating conditions, selectivity and scalability, as well as its potential to integrate with bioprocesses and its ability to couple with renewable energy sources (3,6,10). Furthermore, while other CO2 conversion technologies are limited in their product scope, electrochemical conversion can be used to produce a wide range of one-carbon (C1), two-carbon (C2) and three-carbon (C3) compounds. Namely, in addition to the C1 compounds previously listed, this includes C2 and C3 compounds such as ethylene (C2H4), ethanol (C2H5OH), acetate (CH2COO−), ethylene glycol (C2H6O2), and propanol (C3H7OH) (10–15). Not surprisingly, biological processes have been used to produce many of these same chemicals, which are conventionally produced by the petrochemical industry (16–18). Our work here is motivated by the observation that renewable chemicals, such as those derived by the electrochemical reduction of CO2, are feasible growth substrates for biological processes and merit their consideration as alternative feedstocks for bioprocesses (5).
In evaluating substrates as potential replacements for glucose, it is important to recognize that many cannot be naturally catabolized by traditional industrial workhorses. Hence, it is necessary to consider the substrate toxicity and biocompatibility, as well as the development of appropriate metabolic pathways for substrate utilization. Furthermore, while certain substrates may be biologically feasible, technical limitations in their own production may render them unusable downstream. In the case of electrochemically-derived products, this includes limitations in electrochemical performance, as measured by parameters such as the faradaic efficiency and selectivity (10,11,19–22). While production efficiency and bio-toxicity are more easily assessed, evaluating the feasibility of a new substrate for bio-based chemical production is complicated by how its utilization is linked to the highly interconnected metabolic network. Indeed, refactoring large metabolic pathways in heterologous hosts has proven challenging in the past (23). One method that may help to explain why a new substrate performs poorly examines the metabolic pathway that supports a substrate for chemical production in relation to the cell’s entire native metabolism (24).
In an earlier study (24), we characterized this relationship by calculating the interactions between two competing objectives of cellular systems; growth and chemical production. The theory laid out how the underlying network structure controls whether chemical production is independent of growth. That relationship was captured by a mathematical framework using elementary flux modes (EFMs) to measure the interconnectedness of the cell system and the desired objectives (25). Hence, we defined a metric to measure the orthogonality of the chemical production pathways with respect to the biomass production. We found that the organization of ideal metabolic structures, designed to minimize cell-wide interactions, had a characteristic branched topology. This type of orthogonal structure could be exploited for two-stage fermentation, as it lends itself to the design of metabolic valves for dynamic control (26,27). Dynamic control is a strategy employed to increase control over chemical production, often through the temporal segregation of bioproduction from cellular growth (26). Because of their characteristic branched topology, highly orthogonal pathways often have a key enzymatic step, or metabolic valve, which can be used to control the division of flux for cell growth and chemical production. Various strategies can be used to exert control over these metabolic valves, such as process conditions (pH, temperature, oxygen) or the chemical stimuli of genetic circuits (quorum sensing, inducers, internal metabolite concentration) (26–30). It seems natural then, that the design of orthogonal pathways, metabolic valves and dynamic control strategies would go hand-in-hand, particularly for the design of two-stage fermentations.
Another important finding from our earlier study (24) was that glucose, while a common substrate for industrial fermentation, is not ideally suited for chemical production objectives due to the significant overlap between the pathways for biomass synthesis and chemical production. Instead, substrate selection should be based on the chemical targeted for production. Among the various substrates and products that we evaluated, we identified that ethylene glycol (EG) was a highly promising substrate for orthogonal production of a variety of chemicals because it minimized the interactions between biomass and chemical producing pathways. Today, EG is produced primarily by the petrochemical industry from ethylene, however, renewable alternatives are currently in the early stages of development (31,32). In particular, EG can be produced from the electrochemical conversion of CO2 (11,13), as well as from the chemocatalytic conversion of cellulosic materials and glycerol (a common waste in industrial biofuel and soap production) (31,32). Thus, though unconventional as a feedstock, EG could serve as a replacement for glucose in the modern bioprocess.
Though not commonly reported in metabolic engineering applications, there are two main types of naturally existing pathways that allow microorganisms to consume EG as a carbon source (33–36). The first pathway utilizes a diol-dehydratase resulting in the dehydration of EG to acetaldehyde. Acetaldehyde is then activated to acetyl-CoA by an acetaldehyde dehydrogenase enzyme, which provides the cell with the key pre-cursor metabolite to support growth via the tricarboxylic acid (TCA) cycle and gluconeogenic pathways. This pathway is most commonly found in some Clostridium species and a few other anaerobic organisms owing to the oxygen sensitivity of the diol-dehydratase (34,36). In the second pathway, EG is successively oxidized using nicotinamide cofactors and oxygen to produce glyoxylate (33,35). Glyoxylate, which is a gluconeogenic carbon substrate, can then be used as the growth metabolite as it enters lower glycolysis at the 2-phosphoglycerate node as well as the TCA cycle via the glyoxylate shunt. This oxidative pathway has been shown to exist in a variety of different bacteria (35).
Wildtype Escherichia coli (E. coli) MG1655 cannot naturally grow on or degrade EG. However, it is possible to select for a strain that does, and to our knowledge, only one study has ever reported EG utilization by E. coli. (37). That strain was selected from derivatives of propylene glycol utilizing mutants. Researchers identified increased activities of propanediol oxidoreductase, glycolaldehyde dehydrogenase and glycolate oxidase as the necessary components required for its assimilation. More generally, a survey of the literature shows that enzyme promiscuity is an essential element of the utilization of alcohols (38,39). In this specific case, enzymes regarded as being essential for propanediol or even glycerol utilization across many organisms have shown activity on EG and are regarded as the key methods for degradation, irrespective of the dehydratase route or the oxidative route via glyoxylate (33–35). Hence, in this study, EG assimilation was engineered in E. coli by overexpressing two genes: fucO (encoding propanediol oxidoreductase) and aldA (encoding glycolaldehyde dehydrogenase). This synthetic pathway is similar to the second natural EG utilization pathway previously introduced: EG is sequentially oxidized to glyoxylate thereby providing a gluconeogenic carbon substrate for growth. More specifically, the promiscuous activity of propanediol oxidoreductase converts EG to glycolaldehyde, which is subsequently converted to glycolate by glycolaldehyde dehydrogenase. The native glycolate oxidase then transforms glycolate to glyoxylate to support cell growth and maintenance.
Motivated by the prospect of utilizing EG as a renewable and alternative feedstock, we sought to compare EG with more conventional feedstocks for the production of select chemicals of industrial significance. In particular, formate, glucose, and xylose were selected as the comparative feedstocks, while succinate, ethanol, glycolate and 2,3-butanediol, were selected as the products of interest. Formate was selected as a feedstock as it is another product of electrochemical CO2 reduction (eCO2R) that has been well-studied, and has already been successfully employed within biological systems (3,6,40–42). Meanwhile, glucose and xylose were selected as typical renewable sugar feedstocks, considering that they comprise the largest fraction of sugars in lignocellulosic biomass (43). The four products of interest were selected as they are well-known bioproduction targets that have industrial significance (1,44–46).
In this study, we began by comparatively evaluating EG as a feedstock by measuring the orthogonality of each substrate-product combination, using select bioconversion pathways. Consistent with our previous evaluation (Pandit, 2017), EG demonstrated the greatest orthogonality score for all four products considered. For the products investigated, it was determined that the EG-glycolate combination scored the highest based on this metric. Thus, as a case study we engineered and characterized E. coli as a biocatalyst capable of growth and glycolate production, using EG. This case study attempts to validate our orthogonal approach for chemical production, relating the network topology and two-stage fermentation.
Glycolate is an alpha-hydroxy acid used in the synthesis of a variety of different plastics and polymers, cosmetics and industrial detergents (44,45,47,48). Conventional approaches to produce glycolic acid in E. coli have focused on using glucose and/or xylose as a substrate, and have implemented genetic strategies that couple production to growth (43–45,49). Theoretical yields have been dependent on both the substrate selected as well as the biosynthetic pathway used for production. Examples of glycolate production from glucose in literature have primarily been demonstrated by the activation of the glyoxylate shunt (44,45), while glycolate production using xylose has been demonstrated by the use of a synthetic pathway for xylose assimilation in E. coli (43,49). In E. coli the highest of the identified reports achieves titers of 56.44 g/L and a yield of 0.52 g/g, using glucose as the feed material (50). More recently, Hua and colleagues reported achieving a glycolic acid titer of 110.5 g/L and a yield of 94.4% in Gluconobacter oxydans from EG, using an integrated production, separation and purification technology (48). To our knowledge, only five studies have examined EG conversion to glycolic acid as a biotransformation, none of which were in E. coli (48,51–54).
In this work, we used a combination of computational and experimental investigations to thoroughly characterize the metabolism and growth physiology of E. coli growing on EG (Fig. 1). First, we used orthogonality to demonstrate EG’s potential as a substrate, and selected the EG to glycolate production pathway as a case study. Next, we characterized the glycolate production system using flux balance analysis (FBA), and showed that the selected pathway supported cell growth through shake flask experiments. Subsequently, two sets of fed-batch growth experiments were performed to optimize growth and production, as well as the use of oxygen as a metabolic valve. Findings from the first growth experiment were combined with a computational study on the effect of oxygen to design strategies for the second set of growth experiments and thus improve pathway performance. Overall, we find that EG has the potential to replace glucose in industrial bioprocesses, particularly in applications where CO2 streams and renewable electricity are available. Further, we demonstrate that computational tools can successfully inform the design and optimization of production systems.
Figure 1. Summary of investigations performed in this study. (A) The orthogonality (OS) for various substrate-product pairs were evaluated. The most orthogonal pair (ethylene glycol (EG) to glycolate), was selected to demonstrate the use of orthogonality as a metric to establish successful production systems. (B) A stoichiometric metabolic model was built to characterize the production system and its metabolic behaviour using flux balance analysis (FBA) and flux variability analysis (FVA). Both aerobic and oxygen-limiting conditions were used to investigate the effect of oxygen level on growth and production. (C) Two strains employing the selected pathway, each with different enzyme mutants, were then tested in shake flasks to confirm that the pathway could indeed sustain cell growth and glycolate production. (D) The best performing strain was then tested in bioreactors, using a two-stage growth/production system. As per the identified metabolic valve (glycolate oxidase) and FBA results, a decrease in oxygen level was predicted to switch the system from growth to production. Thus, two secondary air flow rates were tested to evaluate the effect of oxygen level on the production stage. (E) Using data collected from the shake flask and bioreactor experiments, metabolic modeling was used to further characterize the production system and its response to oxygen level. (F) Finally, insights gained from modeling and earlier experiments were used to inform the design and testing of two strategies to improve glycolate production in the bioreactors. Colours indicate the type of analysis performed: blue for computational and green for experimental.