Not content with simple fermentation to produce our target compounds, we sought to address the gaps in current biomanufacturing and incorporated various features into Coup Dy'état. Our system was designed with three main principles in mind: commercial viability, eco-friendliness, and emerging technology adoption (Figure 1).

In our project, we set out to demonstrate the possibility of converting waste to dyes in E. coli via dynamic and stress regulation, achieved using computer-assisted optogenetic control. The illustration below encapsulates how all the features are integrated together to complete the bioproduction process.

The various parts of our system are designed to be modular; they can be incorporated into another system and be expected to work. With the advent of more biomanufacturing technologies, additional features can be assimilated seamlessly. Coup Dy'état is thus a technological platform where microbial production of any chemicals can be optimized.

With Coup Dy'état, we hope to set the standard for biomanufacturing. Click on each segment of the illustration to find out more about the various features of our system!

We are certain that our approach is much less harmful to the environment than the production of synthetic dyes. But what if we compared our project to traditional methods of obtaining natural dyes? While synthetic biology opens up new vistas for the bioproduction of heterologous compounds, we can only consider ourselves a superior alternative to plant cultivation and extraction if the process of acquiring the sugar feedstock for fermentation consumes significantly less land, water, and other resources ^[1].

But that’s not all. From sucrose derived from sugarcane in Brazil to glucose extracted from corn in the USA, the demand for sugar sources in biomanufacturing competes with demands for animal feeds and human consumption ^[2]. Additionally, sugarcane farming takes a substantial toll on the environment, fuelling deforestation, soil degradation, water pollution, and biodiversity loss ^[3].

Our team realized that we needed to ensure that our bioproduction method is itself eco-friendly and sustainable. We thus turned to lignocellulosic waste. It is the underutilized, non-edible portion of plants found in empty fruit bunches, the major biomass byproduct of the palm oil industry ^[4]. This is incinerated in abundant quantities in plantations all over Indonesia, the largest palm oil producer in the world, contributing to air pollution and haze ^[4] which plague her neighbours, including Singapore.

All these burning issues can be alleviated if we can feed lignocellulosic waste to our bacteria. However, there is a major hurdle. Xylose and glucose are the major constituents of lignocellulose, and economically feasible bioconversion requires one to use both sugars at the same time. Unfortunately, this is hindered by carbon catabolite repression (CCR), a regulatory mechanism where expression of genes necessary for the utilization of secondary carbon sources is suppressed by the presence of a preferred substrate, which is often glucose (Figure 2a and b) ^[5].

Recently, CCR has been shown to be released by performing two substitution mutations in the xylose operon regulatory protein (XylR) gene in E. coli, which creates a strain that is able to co-utilize xylose and glucose efficiently. This mutant, XylR*, has a higher affinity to its DNA binding site, and thus activates its promoter independently of the co-activator, the cAMP receptor protein (CRP) (Figure 2c). As a result, expression of xylose metabolic and transport genes occurs even in the presence of glucose ^[5]!

This had huge implications for our project as converting waste into dyes is now viable. We thus constructed a plasmid which allows overexpression of XylR* upon induction by IPTG (Figure 3), and demonstrated its effectiveness in enhancing xylose utilization in E. coli BL21 Star (DE3) by conducting growth experiments.

We deviated from our initial plan of engineering a xylose-utilizing strain like what was accomplished in the study, as we found it easier to transform this construct into our dye-producing bacteria and allow xylose to serve as the feedstock. This also enables potential optimization of the expression levels of XylR* and balancing of metabolic flux, which may vary depending on the biosynthetic pathway or fermentation conditions.

Along the targeted pathway, we took special interest in a key intermediate compound, naringenin. Naringenin, a flavanone, is extremely valuable due to its antioxidant, anti-inflammatory, antiviral, and anti-tumour properties ^[6]. More intriguingly, naringenin is also the intermediate platform compound for the biosynthesis of other flavonoids, many of which are potential candidates for textile dyes (Figure 4).

While commercial naringenin can be added as the substrate during the biomanufacturing process, we find it desirable to achieve de novo biosynthesis given its strategic location in the biosynthetic pathway, which provides us with the basis for the synthesis of many other high-value flavonoids. Besides reducing costs, this would eliminate the need to introduce expensive chemical precursors ^[7], which is consistent with our team’s vision of involving less chemicals in the biomanufacturing process.

Serendipitously, there was a recent breakthrough involving the production of naringenin from xylose via a synergistic co-culture system comprising E. coli and S. cerevisiae ^[6]. Subsequent literature review also revealed the advantages of co-culture biosynthesis, with each independent strain devoted to the functional expression of different stages of the biosynthetic pathway, allowing for reduced metabolic burden, flexibility of employing suitable hosts, and adjustable biosynthetic strength ^[8]. Inspired by these revelations, we decided to adopt a co-culture strategy as well, especially considering the length of the biosynthetic pathway and how we intended to use xylose in our feedstock.

Therefore, we sought to produce naringenin from tyrosine in a single E. coli strain, which could then be co-cultured with cells responsible for the synthesis of our final compound, thus completing the pathway. This process involves catalysis by four enzymes - phenylalanine ammonia lyase (PAL), 4-coumarate-CoA ligase (4CL), Oryza sativa polyketide synthase (OsPKS), and malonyl-CoA synthetase (MCS) ^[9] put together in a single plasmid (Figure 5a).

Currently, our de novo construct carries three of the enzymes necessary for naringenin production. While PAL is placed under a constitutive promoter, OsPKS and MCS are strategically put under the control of a P_lac promoter instead (Figure 5b). This enables overexpression of the two more crucial enzymes involved in the latter stages of the pathway, thus helping to drive the reaction forward.

By supplementing the medium with coumaric and malonic acid as substrates, we demonstrated the production of significant amounts of naringenin in E. coli BL21 Star (DE3) cells carrying the de novo construct. The production of naringenin was verified using High Performance Liquid Chromatography (HPLC), while the expression of the enzymes was confirmed via SDS-PAGE.

With the eventual complete construct, de novo biosynthesis of naringenin can be achieved with just the introduction of this plasmid. This design complements the use of xylose as feedstock to achieve a much more environmentally-friendly bioproduction of naringenin, and ultimately, dyes.

The choice of target compound was given careful consideration as its production is the raison d’être of Coup Dy’état, and the design of all the features in our system would revolve around it. We explored the production of various flavonoids, all derived from the intermediate platform compound, naringenin. Chrysanthemin (pink), callistephin (orange), and theaflavin (reddish-brown) were our original candidates as many enzymes are shared across their biosynthetic pathways. A range of colours could then theoretically be produced from a mixture of these compounds, whose relative concentrations could be varied by controlling the expression of key enzymes (Figure 6). However, the promiscuity of the enzymes and inherent leakiness of the pathway necessitate extremely tight control of gene expression which was technically challenging to accomplish.

The major turning point occurred when we interviewed designers and synthetic dye companies and learned that such colours were not commercially viable. Primary colours are preferred, as they could be mixed to produce the desired hue. Following this, we opted to focus our efforts on the synthesis of luteolin, a yellow flavone located downstream of naringenin too (Figure 6).

Notwithstanding the warmth and optimism that it evokes, the colour yellow has a murky history. The vivid Indian Yellow dye was derived from the urine of cows which had been fed only mango leaves and were thus kept in a perpetual state of near-starvation, yellow to be dubbed the “cruellest” colour ^[10]. Presently, the synthetic yellow compound PCB-11, which possesses a plethora of toxic effects, is found ubiquitously in textiles and other consumer products ^[11]. Luteolin, which is perhaps the oldest European dye, enjoys a much better reputation. It is deemed to possess the best light fastness amongst natural yellow dyes, and is still used to dye silk, wool, cotton, and leather ^[12]. A bright future for yellow seemed to be in the cards with luteolin.

Once we confirmed that we would be producing luteolin, we expressed the two enzymes required to produce it from naringenin in E. coli: flavone synthase (FNS) from Petroselinum crispum and flavonoid 3’-hydroxylase (F3’H) from Arabidopsis thaliana (Figure 6), as heterologous expression of these enzymes, albeit in Streptomyces albus, was successful in producing small amounts of luteolin ^[13]. E. coli BL21 Star (DE3) was selected to be our chassis as it is an RNase knockout strain, which enhances protein expression and dye production. It was also shown to be a suitable host for flavonoid synthesis ^[15]. The expression of the enzymes FNS and F3’H are optogenetically controlled (see: Blue Light Repressible System).

Through multiple rounds of biosynthesis, we were able to successfully demonstrate the production of luteolin via HPLC. The levels of mRNA and protein expression were also investigated using qPCR and SDS-PAGE respectively.

Where induction of gene expression is necessary, chemical inducers are staples owing to their capability to trigger high expression levels ^[16]. However, the drawbacks are numerous: potential cellular toxicity, non-reversible induction, delayed diffusion, and high costs ^{[17] [18]}. All these problems are exacerbated in large-scale biomanufacturing (Figure 7).

Enter optogenetics. The allure of being able to manipulate cellular behaviours using light is causing optogenetics to make waves in the field of synthetic biology. Light-based induction offers many advantages: non-toxicity, reversibility, precise spatiotemporal control, tunability, and availability ^{[19] [20]} (Figure 7). Hence, optogenetics offers a compelling alternative to traditional chemical induction, particularly in biomanufacturing, a view that has been corroborated by several experts in the field.

We adopted the EL222 blue light system, thus renouncing the use of chemical inducers. Originating from the marine bacterium Erythrobacter litoralis HTCC2594, EL222 is a photosensitive DNA binding protein, with a N-terminal light-oxygen-voltage (LOV) domain and a C-terminal helix-turn-helix (HTH) DNA binding domain. Irradiation by blue light of wavelength 450nm exposes the hitherto sequestered HTH, facilitating dimerization and subsequent DNA binding ^[21].

In Coup Dy'état, the repressible system was employed to facilitate dynamic gene regulation (see: Cell-Machine Interface). With energy conservation in mind, blue light is shone during the shorter growth phase, and switched off during the lengthier production phase to derepress the expression of enzymes. This is achieved by placing the DNA binding site of EL222 between the -35 and -10 hexamers of the consensus promoter in E. coli, creating the blue light repressible promoter P_BLrep ^[21]. As a result, EL222 acts as a repressor, blocking the binding of RNA polymerase and repress gene expression in the presence of blue light. In the dark, RNA polymerase can now bind, and gene expression takes place (Figure 8).

To elucidate the strength, sensitivity, and leakiness of P_BLrep, and its inducible counterpart, P_BLind, we performed many rounds of characterization experiments. This was essential in providing us with insights for its subsequent integration into bioproduction to drive target gene expression. The characterization involves the use of mRFP as a reporter placed under the control of the promoters. However, the output was not representative of the ON/OFF status of the promoter as a result of the accumulation of mRFP, which has a long half-life. To resolve this, we introduced degradation tags of varying strengths so that the mRFP produced would be targeted for quicker proteasomal degradation (Figure 9). This improves on our previous model for promoter characterization.

Through our characterization experiments, we also found P_BLrep to be better suited for our system due to its relatively lower leakiness, further justifying our choice of a repressible system. Hence, we replaced the chemically-inducible promoters, which were originally used to control the expression of our biosynthetic genes F3’H and FNS, with P_BLrep (Figure 10). To verify that our blue light repressible system works as intended, we demonstrated that cells that were exposed to blue light throughout the course of biosynthesis had a significantly lower yield.

Our project thus built upon the landmark study on optogenetically-regulated bioproduction in yeast ^[22] and demonstrated that the use of this technology with bacteria shows promise. We envision this to be the future of biomanufacturing.

Stress, and the cell’s response to it, is a universal phenomenon. In large industrial bioreactors, stress is exacerbated owing to the large variation in the local environment as compared to small bench-top fermenters or shake-flask cultures. This is the result of a longer global mixing time, causing significant heterogeneity in the medium. Besides hydrodynamic shear stress, cells are also subjected to stress induced by fluctuations in the level of oxygen, pH, substrate, and other parameters ^[23]. As Dr Nic Lindley said, “take a sample point in a bioreactor and you get an image, take a look of the whole bioreactor and you get a film.”

All these compound the difficulty in translating the biological potential of the cell as demonstrated in the laboratory into actual production. One type of stress, however, applies to bench-top experiments in the laboratory and large-scale processes in the industry, and across the bioproduction of any compounds. This opens up the possibility of investigating and regulating it in bioproduction.

This particular stress is that induced by the expression of recombinant proteins. The depletion of finite cellular resources during the expression of synthetic constructs constitutes an unwanted burden, hampering the growth and expected performance of engineered cells in an unpredictable manner ^[24]. Stress regulation was accomplished by decreasing the expression of a synthetic construct in response to increasing burden. Regulated cells outperformed controls in protein yield ^[25], a discovery with significant implications for the future of biomanufacturing.

Therefore, we strove to improve our yield by incorporating a stress reporter module into our system, which studies the level of stress induced by our constructs and facilitates stress regulation. The module consists of the mRFP gene placed under the control of a burden-responsive promoter P_htpG1 (Figure 11a). This particular promoter showed high promise as a biosensor for burden, as it displayed the best ON/OFF characteristics compared to other promoters that were similarly upregulated upon the induction of synthetic constructs ^[25].

Using this part, we were able to quantify the level of stress induced by GFP production by measuring the fluorescence level, thus demonstrating proof of concept (Figure 11b). Our constructs were subsequently tested and shown to impose a substantial metabolic burden (Figure 11c). Characterization of the P_htpG1 promoter will hopefully pave the way for stress regulation and the ensuing optimization of the bioproduction process.

Biomanufacturing involves the introduction of synthetic constructs often with the aim of maximizing flux towards heterologous pathways, placing tremendous burden on cells, attenuating their growth and curtailing their production (see: Stress Reporter). Cells’ natural propensity to grow and multiply is thus directly in conflict with synthetic biologists’ desire to maximize their production.

However, prioritization of both growth and production could conceivably be accomplished using dynamic control. This separates fermentation into two phases: a growth phase where cells are cultivated to high densities, followed by a production phase where expression of non-native pathways takes place ^[26]. Dynamic control can be achieved through these means - chemical induction, autoregulation, and computer-assisted feedback control. As compared to the other two, computer-assisted feedback control is tunable, versatile, modular - it is easier to incorporate additional modules e.g. nutrient- and substrate-sensing modules ^[27], can perform more complex logic, and places less metabolic burden on the cell. These advantages that come with giving control over to the user will only become more significant as our system scales up.

Computer-assisted feedback control enjoys a particularly excellent synergy with optogenetics, as light sources are easily programmed to turn on/off. However, using light to control transcription remains a challenge in itself. Even state-of-the art methods are unable to simultaneously achieve precision and repeatability, nor can they withstand changing operating culture conditions ^[28]. Furthermore, based on literature review, custom optogenetic tools have only been designed for well plates, which are small-volume, high-throughput systems. These tools are inadequate for our project as our objective is to scale up to a large-volume, low-throughput bioreactor.

Our designs’ first focus was then to create an in silico automated feedback control system, a theoretically effective and feasible strategy for overcoming current challenges in light-controlled gene expression ^[29] as compared to typical open-loop feedback control systems for gene expression ^[28]. Our second focus was to accommodate larger-volume standard lab equipment. We thus produced solutions based on these design considerations.

We designed two complementary hardware systems. The first fulfils users’ needs to characterize their optogenetic circuit in scaled-up systems. The second enables optogenetic biomanufacturing. See our Hardware page for our design iterations, how we developed our final designs, and how we validated these designs through experiments.