GLUCOSE-XYLOSE GROWTH EXPERIMENTS

The experimental design

To investigate the effectiveness of our xylose-utilizing module, we constructed and transformed the plasmid containing the native XylR gene and that containing the mutated XylR (XylR*) into E. coli BL21*. Since XylR acts as a co-activator of the xylose operon, we hypothesized that overexpressing native XylR would also help to enhance xylose utilization. Also, the effect would be more pronounced when XylR* was expressed, as it has a significantly higher binding affinity. BL21*, BL21*-XylR, and BL21*-XylR* were grown in 0.2% glucose, 0.2% xylose, as well as a mixture of 0.1% glucose and 0.1% xylose. Growth, as an indicator of sugar substrate utilization, was measured via absorbance at 600 nm over 8-hours in a microplate reader.

XylR* was indeed overexpressed

To verify the expression of XylR and XylR*, SDS-PAGE was conducted for samples taken after induction. Thick bands at approximately 45 kDa were observed for induced cells, with no corresponding band for the WT and uninduced samples, showing that XylR and XylR* were overexpressed respectively (Figure 1). We moved on to examine whether overexpression of these proteins translated to faster growth (Figure 2).

BL21*-XylR* grew much faster!

Our growth experiments displayed rather interesting results. BL21*-XylR* had noticeably the highest growth in all three conditions, while BL21*-XylR displayed considerably little or even no growth (Figure 2a, b, and c). Most importantly, overexpressing XylR* seemed to increase xylose utilization, where the growth of Bl21*-XylR* was significantly higher than the wild-type in xylose (Figure 2e), as well as in a mixture of glucose and xylose (Figure 2f). Our xylose-utilization module presumably works!

Surprisingly, XylR* also appeared to enhance growth in glucose (Figure 2a). We expected it to have similar growth as the wild-type instead. Hence, it was difficult to ascertain whether the augmented growth in xylose is a result of improved utilization or other mechanisms, a possibility given XylR’s role as a metabolic regulatory factor.

Nonetheless, BL21*-XylR* exhibited comparable growth rates in glucose and the mixture of glucose and xylose, suggesting co-utilization of sugar substrates occured (Figure 2f). In contrast, this was not observed in the wild-type, where having a mixture of sugars resulted in lower growth rates as compared to glucose only (Figure 2d). This was significant as we are one step closer to our vision of utilizing lignocellulosic waste as feedstock, which requires simultaneous utilization of glucose and xylose.

..not so for BL21*-XylR

As for BL21*-XylR, there was little to no growth observed regardless of the sugar substrates (Figure 2e). Not only did XylR overexpression fail to improve xylose utilization as hypothesized, there appeared to be a growth inhibitory effect. It was possible that other metabolic processes were compromised.

Our xylose-utilizing module likely works, but what's next?

Our XylR*-overexpressing xylose utilizing module is likely to be functioning as intended, but further tests would be necessary to confirm it. To obtain a more definitive indication of glucose and xylose utilization, HPLC analysis could be carried out on the medium taken at certain time points. If xylose utilization was not the contributing factor to enhanced growth, further tests could be conducted to elucidate the mechanism.

On the other hand, if XylR* overexpression did increase xylose utilization as expected, there are plenty of possibilities to be explored! To further demonstrate the applicability of our xylose utilizing module, crude lignocellulosic waste extracts could be fed as well. We could co-transform our current module with our biosynthesis plasmids, demonstrating the production of naringenin or even luteolin from xylose, bringing us closer to our final system. By varying inducer concentrations, the optimal level of XylR* expression could be determined. As such, other strategies could be employed to attain the level of expression without using chemical inducers - one of the key objectives of Coup Dy'état. This includes constitutive or light-regulated expression or even metabolic engineering. More excitingly, a xylose-based nutrient-sensing module could conceivably be developed, allowing for dynamic regulation via light induction!

Xylose would undergo a long metabolic pathway and eventually be converted to amino acids such as tyrosine, but the fun does not stop there.. (see: De novo Biosynthesis)

NARINGENIN BIOSYNTHESIS

Full de novo biosynthesis of compounds such as naringenin is far from easy. It requires the expression of many heterologous enzymes to convert it from tyrosine. Hence, we decided to use coumaric acid, the first intermediate of the naringenin biosynthetic pathway, as the starting point, and thus constructed a “de novo” plasmid harbouring the remaining three enzymes, 4CL, MCS, and OsPKS (Figure 3). By investigating the production of naringenin using the current construct, we would be able to gain insights which might prove useful for the eventual realization of complete de novo biosynthesis.

To test our construct, we supplemented the culture with coumaric acid as the starting compound. As such, we would expect that only samples supplemented with coumaric acid would produce naringenin. Additionally, we wanted to investigate if the addition of malonic acid, a precursor to one of the intermediates in the pathway, would further bolster naringenin production (Figure 3).

We produced naringenin in E. coli Acella..

We first tested our construct in E. coli Acella, which is a BL21-derivative strain and possesses complete deletions of endA and recA genes, thus reducing degradation and recombination of externally introduced DNA. Enzyme expression was determined using SDS-PAGE (Figure 4) whereas naringenin production was confirmed via HPLC (Figure 5).

SDS-PAGE confirmed the overexpression of MCS, as seen from the thick band corresponding to 55 kDa. However, there was no observed overexpression of both 4CL (59 kDa) and OsPKS (43 kDa) as evident in the lack of thick protein bands of their respective sizes (Figure 4). While this was expected of 4CL, since it was placed under a constitutive promoter, we did not expect to have no overexpression of OsPKS which was IPTG-inducible. Nevertheless, we proceeded to check for naringenin production (Figure 5).

As expected, coumaric acid was necessary for naringenin synthesis, achieving a production of 0.0397 mg/L. Interestingly, the addition of malonic acid boosted production tremendously, reaching a 10-fold increase in concentration to 0.330 mg/L (Figure 5)!

Despite not observing any overexpression of OsPKS and 4CL, it could be deduced that these enzymes were still expressed due to high levels of naringenin production. This also hinted at their superior enzymatic activities, as high amounts of expression did not appear to be necessary.

.. as well as BL21*

Encouraged by the results, we repeated the experiment in E. coli BL21*, our primary biosynthetic strain (Figure 6).

We observed a similar HPLC profile as that obtained using Acella. Naringenin production was the best when both coumaric and malonic acids were added, achieving a yield of 0.508 mg/L. It appeared that BL21* represents a strain improvement from Acella! Strangely, De Novo UI CA+MA (uninduced, coumaric and malonic acids added) also produced a substantial amount of naringenin at 0.282 mg/L even though it was not induced (Figure 6). This might be a result of leaky expressions of OsPKS and MCS, which were under the control of IPTG-inducible P_lac promoter.

Towards full de novo biosynthesis

Subsequent integration of PAL into the de novo plasmid would allow the cell to utilize tyrosine to produce naringenin without any addition of a starting substrate, thus achieving complete de novo biosynthesis. More ideally, we hope to replace the P_lac promoter with a constitutive or light-sensitive promoter of similar strength to completely remove any need for any chemical inducers. This would further strengthen our environmentally conscious design and reduce manufacturing costs at the same time.

Moreover, we have found that the addition of malonic acid increased the naringenin production to a large extent. Any future work should consider enhancing intracellular malonic acid concentrations via metabolic engineering. This could be achieved by channeling the metabolic flux to drive greater production of malonic acid via overexpressing native enzymes that are involved in malonic acid production, and/or engineering of these enzymes to improve their enzymatic activity.

With eventual full de novo biosynthesis of naringenin, we have the basis for production of any flavonoids, including our target compound, luteolin. But there's a catch..

NARINGENIN GROWTH EXPERIMENTS

Before conducting biosynthesis using naringenin as the substrate, we needed to ensure that it does not exert a growth-inhibitory effect on our bacteria. Naringenin has been shown to compromise growth and proliferation of cyanobacteria and the amoeba Dictyostelium discoideum ^{[1] [2]}, but we were unsure of its effects on E. coli. Hence, we investigated the effects of varying concentrations of naringenin on growth of BL21*. Concentrations of 0.1, 0.2, and 0.4 mM were used, and growth was tracked over a period of 10 hours by measuring OD₆₀₀ using a microplate reader (Figure 7).

How did naringenin affect BL21* growth?

Our results showed that addition of naringenin caused a slight reduction in cell growth compared to the control, which was expected. Nonetheless, proliferation was sustained over a period of 10 hours, suggesting that naringenin did not cause significant toxicity to E. coli BL21*. We thus concluded that the working concentration of 0.2 mM was permissible for luteolin bioproduction. Following this, we were interested to determine the optimal naringenin concentration and OD₆₀₀ for introduction of naringenin, hence we also made use of the results to build a model for optimal bioproduction (see: Modelling).

Naringenin production: check. But greater challenges lies ahead.. (see: Luteolin)

LUTEOLIN BIOSYNTHESIS

While producing naringenin, we also investigated its conversion to luteolin, our target compound, from naringenin. This requires the expression of two enzymes F3’H and FNS. To achieve this, we constructed various luteolin-producing plasmids where the enzymes were placed under different light-sensitive or chemical-inducible promoters (details in Figure 8).

Adapting biosynthesis protocols from literature, cell cultures were induced by being kept in the dark and by adding the relevant chemical inducers, arabinose and ATc. 0.2mM naringenin was supplied after induction. We also examined the expression of our constructs via RT-qPCR (Figure 9), and verified luteolin production using HPLC (Figure 10).

F3'H and FNS genes were expressed

F3’H and FNS genes were shown to be expressed. F3’H, being under the control of the P_BLrep promoter which essentially functions constitutively in the dark, was expressed as expected. FNS, on the other hand, was under the ATc-inducible promoter P_Tet and arabinose-inducible promoter P_BAD respectively. Characterizations of P_BLrep-F3'H and P_BAD-FNS have also been carried out. The expression of FNS without the presence of inducers was likely due to leakiness. While ATc induction resulted in a significant increase in expression levels, the effect of arabinose induction was less pronounced (Figure 9).

And luteolin was produced!

As we harvested the samples, we could not contain our excitement when we observed that the medium had turned noticeably yellow! The colour became even more distinct after centrifuging. After carrying out our safety protocol (see: Safety), we analyzed our samples via HPLC. Much to our delight, small amounts of luteolin were detected! We have successfully produced 0.155 to 0.201 mg/L of luteolin (Figure 10b). This was, in our opinion, a more than respectable yield, surpassing that of 0.09 mg/L previously reported, albeit in Streptomyces albus ^[3]. Another group managed to attain 4 mg/L in E. coli BL21*, but significant metabolic engineering was involved ^[4].

Out of the different combinations of constructs used, Brep-F3'H+Brep-FNS had the best performance (Figure 10b). This suggested that the strength of the P_BLrep may rival chemically-inducible promoters, thus it has the potential to break the stranglehold of chemical inducers in biomanufacturing!

We were also interested to find out how much naringenin was left at the end of biosynthesis. Indeed, a higher luteolin production corresponded to lower residual naringenin in our samples (Figure 10b and c). Not all naringenin was converted to luteolin. Some might have reacted to form apigenin, which had yet to be converted to luteolin, in the final samples. Naringenin could have also been degraded or converted to other flavonoid compounds. Subsequent HPLC analysis could encompass more compounds downstream of naringenin in the biosynthetic pathway.

We successfully produced our target compound, but there was more to be done

Based on our biosynthesis results, our enzymes and constructs were functional, and we have obtained preliminary indication of the potential of P_BLrep. As with all bioproduction, further enzyme and metabolic engineering could be performed to enhance the production of luteolin. What was more interesting to us, however, was how to best achieve tight regulation of protein expression.. (see: Blue Light Repressible System))

CHARACTERIZATION OF P_BLrep

The EL222 blue light repressible system was our choice of control over protein expression. However, we found that previous work done on the characterization of the promoter, P_BLrep, was insufficient. Hence, extensive efforts were put in to improve its characterization and to better understand its properties. Please refer to the Parts Improvement page for our detailed findings.

BLUE LIGHT REPRESSIBLE CONTROL OF LUTEOLIN PRODUCTION

Having thoroughly characterized the promoter, we integrated it into luteolin biosynthesis. We wanted to demonstrate that light can indeed be used to regulate enzyme expression in biomanufacturing, which would represent a significant milestone in Coup Dy'état. To do this, we had a light ON set-up where cells were subjected to blue light repression throughout the course of biosynthesis, and a light OFF set-up where cells were kept in the dark during the induction and production stages, allowing expression of enzymes. Both set-ups involved co-culturing of cells carrying the Brep-F3'H plasmid and those with the Brep-FNS plasmid. If our blue light repressible system did function as expected, we should observe higher luteolin production, as quantified using HPLC, for light OFF cells (Figure 11).

The control works!

The results turned up even better than what we dared hope! Light OFF (induced) cells had a production of 0.352 mg/L of luteolin, whereas light ON (repressed) cells showed no production at all (Figure 11b). This indicated that our blue light repressible system is functional.

What further intrigued us was the yield obtained, which was higher than our previous biosynthesis (see: Luteolin). This hinted at the superior productivity of co-culture biosynthesis.

Why was this so significant?

This results demonstrated the possibility of inducing protein expression without the use of chemical inducers. With this accomplishment, the possibilities are endless! Optogenetic control enables us to perform regulation of protein expression that is easily tunable and reversible, amongst other advantages which chemical inducers cannot offer (see: Blue Light Repressible System, Design). Through the use of P_BLrep, dynamic regulation can be achieved by irradiating cells with blue light and repressing enzyme expression during the growth phase, followed by the production stage where enzymes and compounds are synthesized when cells are kept in the dark. Moreover, other forms of regulation where induction needs to be reversed are also permissible. Stress regulation is one such example..

CHARACTERIZATION OF BURDEN-RESPONSIVE PROMOTER P_htpG1

To regulate stress via an external control, i.e. light, a reporter is necessary. To demonstrate that our stress reporter is sensitive to externally introduced plasmids which produce recombinant proteins, we built and tested various constructs transformed into E. coli as illustrated in (Figure 12). Cells were grown and samples were extracted at certain time points to measure fluorescence (GFP/RFP) and OD₆₀₀ in a microplate reader.

First, we set out to characterize the P_htpG1 promoter which was chosen to be the stress sensor. Using E. coli DH5α as the host, we introduced our stress reporter module along with a GFP-expressing construct (Figure 12b). The level of RFP expression was compared to cells transformed with the stress reporter module only, which served as a control (Figure 12a). Since the RFP levels for cells containing only the module should only correspond to the basal stress level and that induced by the module itself, additional levels of RFP detected would be a result of GFP expression. For this experiment, we included two replicates picked from different colonies (GFP+RFP A and GFP+RFP B) as our experimental strains ((Figure 13)).

There was an overall trend of increasing RFP/OD levels over time (Figure 13a), which was indicative of increased cell stress over time. By comparing RFP/OD of the control and experimental strains at the 24 h time point (Figure 13b), we demonstrated that GFP production in cells caused an approximately 0.5 fold increase in RFP/OD levels, suggesting an equivalent increase in cellular stress.

To further confirm that GFP production contributed to celluar stress and thus elevated RFP levels, we compared the production of both fluorescent proteins (Figure 13b and c). Indeed, GFP+RFP A, which had a higher GFP productivity, exhibited higher levels of stress as reported by its higher RFP expression.

These showed that our stress reporter module works as expected! It is not only able to report cell stress but is also sensitive and responsive to the presence of externally introduced constructs. With that, we moved on to bigger things: testing our module with the bigger constructs used in our biosynthesis.

DEMONSTRATION OF STRESS INDUCED BY BIOSYNTHETIC CONSTRUCTS

Higher recombinant protein expression leads to more stress

To justify the need for our stress reporter module, we performed the same tests on our biosynthetic constructs. As the level of RFP expression corresponds to the level of stress, we hypothesized that larger constructs which express more proteins would induce a heavier burden on the cells, resulting in higher RFP levels. To examine this, we used our de novo plasmid, a significantly larger construct expressing three larger enzymes (Figure 12c). Again, two replicates picked from different colonies (de novo A and de novo B) were the experimental strains (Figure 14).

The results were as expected! Cells harbouring the de novo constructs exhibited higher RFP levels, indicating that the stress promoter, P_htpG1 was activated to a greater extent (Figure 3a and b). Hence, we were able to tentatively conclude that the level of stress, manifested in RFP expression, is correlated to higher recombinant protein expression. When asking cells to do our bidding, the stress that they experience is indeed something not to be taken lightly!

Our stress reporter module is robust

Thus far, the investigation of our stress reporter module has been carried out in DH5α. Hence, to demonstrate the robustness of the module, we tested it in a different genetic background, i.e. BL21*, which was also our choice strain for biosynthesis. For this experiment, we included more biosynthetic constructs: P_Brep-FNS and P_BAD-FNS (Figure 12d and e) along with the previously-tested de novo plasmid. The results were as follows (Figure 14).

Our stress reporter module appeared to work in BL21* as well! The strain possessing the larger de novo plasmid had a noticeably higher level of stress and RFP expression compared to the control strain, similar to what was observed previously in DH5α (Figure 15b and Figure 13b). As for our other biosynthetic plasmids, P_Brep-FNS and P_BAD-FNS, the RFP levels induced were relatively lower and closer to that induced by the GFP-expressing construct. This did not come as a surprise, as only a single enzyme, FNS, was produced.

To further investigate our stress reporter's versatility, we subjected it to a different temperature (25°C) on top of the usual temperature (37°C) used in all our previous experiments involving the reporter (Figure 16).

Similar trends were observed despite varying the temperature. The cells grown in 25°C, however, had lower RFP expression across all the samples (Figure 16). This could be attributed to lower rate of protein synthesis, as evident in the lowered GFP production. Another possible reason is P_htpG1's involvement in the heat-shock response, hence it is likely to be induced by higher temperatures. Nonetheless, our results demonstrated that the promoter retains its sensitivity within this temperature range.

Stress regulation: the way forward

We have investigated the many properties of our stress reporter module, helmed by the burden-responsive promoter P_htpG1. We have shown that it was able to reflect cell stress, and its level of activation corresponded to level of recombinant protein expression. We have also demonstrated that it was robust in different temperatures and genetic backgrounds. It has certainly withstood all our tests, without so much a hint of pressure indeed!

Subsequently, we aimed to further improve on our characterizations. For example, varying levels of GFP could be expressed either using constitutive promoters and RBS of different strengths, or varying levels of induction using inducibe promoters. This would allow us to draw a better correlation of the output signal, i.e. RFP expression, with the level of recombinant protein expression.

But what's the purpose of doing all these? By studying the level of stress induced by foreign constructs, we could conceivably achieve stress regulation, which was shown to improve protein yield. Incorporating stress regulation into biomanufacturing could thus enhance the production of the target compound. This is our end goal, towards which the characterization of the burden-responsive promoter P_htpG1 was the rudimentary, but crucial, step.

Our stress promoter could work in tandem with the blue light repressible system to regulate enzyme expression based on level of stress detected. The cell, however, could not achieve this on its own. Some external help is needed.. (see: Cell-Machine Interface)

To support our biomanufacturing platform, we prototyped a small-scale bioreactor with a 2-in-1 OD and fluorescence sensor, a peristaltic pump and a blue LED system. We investigated and verified each component’s functionalities to calibrate them for use in our system. Our findings are presented below.

FlUORESCENCE SENSING: Characterized, Functional

The graph showed a strong linear correlation between the light sensed by the TSL235R and RFP levels as measured by the NanoDrop. By using this calibration curve, an equivalent RFP value could be calculated from an output frequency and vice versa. It is then a simple matter to set a threshold RFP value indicating maximum allowable metabolic burden, convert this to a sensor frequency, and program a microcontroller to turn on blue light sources to repress production once that RFP level has been reached. One may also turn off blue light sources to allow the cells to begin production again when RFP levels are acceptable.

OD SENSING - Characterized, Functional, Room For Improvement

LED brightness is a function in the Arduino library for the LEDs. It may take on a value between 1 and 255, with 255 corresponding to the highest possible brightness for the LED.

The graphs for LED brightnesses between A and B displayed behaviour similar to that of Figure 18, up to an approximate OD value of 1.57. We can conclude that for high OD, there is again a strong linear correlation between the light sensed by the TSL235R and OD levels as measured by the NanoDrop. For OD values higher than 1.57, higher LED brightnesses caused a greater change in sensor output frequency per unit change in OD. In other words, the gradient of the graph between OD 3.26 and OD 1.57 increased with increasing LED brightness.

However, sensor output frequencies plateaued after OD = 1.57, indicating that the sensor is not sensitive enough to detect changes in light intensity when OD is sufficiently low. This is possibly because the LED is so bright at that point that the maximum light intensity the TSL235R can detect was reached.

Isolating the graphs of lowest LED brightnesses on Figure 2b, all graphs had a positive gradient and the R2 values were close to 1, and this behaviour remained consistent throughout the whole range of ODs. At these brightnesses, no plateaus for sensor outputs were observed. However, as the gradients of these graphs were relatively small and the TSL235R outputs frequency values as whole numbers only, a slight fluctuation in sensor readings may cause the user to calculate an equivalent OD that does not reflect the “true” OD. From these 5 lowest-brightness graphs, the sensor will most likely perform best and cause the least false conversions if we pick the brightness level which corresponds to the graph with the highest gradient. From Figure 2b, this is the graph at which LED brightness = 1.

For now, as a temporary measure, our recommendation is to write a code with an additional command such that the LED brightness will decrease once the readings begin to plateau, to always maintain the best possible sensor performance. In this case, the LED brightness should decrease to 1 at OD = 1.57.

For future work, a more suitable sensor could be selected and troubleshooting to extend the effective range of the sensor can continue.

PUMPING - Characterized, Functional

From the graph, it is clear that there was a strong linear correlation between mass and time. This proved the functionality of our pump as it showed that the pump was able to displace fluid at a steady rate. The mass flow rate in g/s is the gradient of the best-fit line.

After obtaining graphs at other RPMs, we plotted the mass flow rate as a function of RPM and were able to characterize our pump.

The graph showed a linear correlation between the mass flow rate and revolutions per minute. This can be used as a calibration curve. The microcontroller can be programmed such that the user inputs her desired mass flow rate, and using the equation of the best fit line obtained from such a calibration, the RPM of the stepper motor can be set.

For future work, more measurements can be carried out to improve the characterization of the pump.

RUNNING SMALL BIOREACTOR

With all these individual components ready, we assembled them into a small bioreactor system to perform cell-machine interface and automation. First, RFP cell with low OD (0.01) is transfered into the bioreactor. The bioreactor system is placed in the incubator at RPM 225 and 37 degree.

Blue light is turned on to repress the expression of RFP. After a few hours, our OD sensor detects that OD increases from 0.01 to 0.6. Blue light is turned off. We see a steady increase in RFP reading in our sensor over the next couple of hours.

CLOSE