Proof of Concept
Lux Brightness Enhance

Experiment Design: Transforming BL21(DE3) with two different plasmids: pET28a-luxCDABE , pET28a-luxCDABEGF. The two plasmids are differentiated by the existence of luxF and luxG. After that, two kinds of BL21(DE3) would be cultured for 16 hours under 23℃ after being induced by 0.2 mM IPTG. Then the data was accessed by photograph and luminous plate reader.

Results: In our design, the adding of luxF and luxG would increase the intensity of light to validate the feasibility of LAMPS (See project description)[1][2]. And our results showed the design worked really well! The light enhancement could be directly observed by naked eyes (Figure 1) and luminescence assay showed that the light could be boosted by an extraordinary magnitude of 58% (Figure 2).

Figure 1 Comparing luminescence of BL21(DE3) with pET28a-luxCDABE or pET28a-luxCDABEGF. Images were taken by an ordinary phone camera. The left three tubes were BL21(DE3) with pET28a-luxCDABE and the right three tubes were BL21(DE3) with pET28a-luxCDABEGF. It was clear that luminescence can be enhanced by additional luxF and luxG. All BL21(DE3) underwent a 16-hour culture under 23℃ after being induced by 0.2 mM IPTG.

Figure 2 The result of luminescence assay for BL21(DE3) with pET28a-luxCDABE and BL21(DE3) with pET28a-luxCDABEGF (n=6). The result showed that the addition of luxF and luxG increased the luminescence by 58%. All BL21(DE3) underwent a 16-hour culture under 23℃ after being induced by 0.2 mM IPTG. And the assay was conducted by a plate reader.


To make the bacterial LuxCDABE fluorescence system brighter and more useful for practical applications, we utilized biofluorescence resonance energy transfer (BRET) to increase its brightness and alter its color. [3]We created a new fusion fluorescent protein called LuxB:cp157Venus by attaching the yellow fluorescent protein cp157Venus to the C-terminal of LuxB using a linker (Glu-Leu). Since the emission spectrum of the LuxA-LuxB complex overlaps with that of the excitation spectrum of cp157Venus to a certain extent. When the two are close enough to each other (≤10 nm), the LuxA-LuxB complex in the excited state can undergo dipole-dipole resonance with cp157Venus, transferring its energy to the latter in a non-radiative manner, causing the latter to emit light with different frequencies and amplitudes. Since the efficiency of BRET is related to the sixth power of the distance between the two, only when the distance between the fluorescence donor and the fluorescence acceptor is appropriate, it is able to change the wavelength of the light while significantly increasing the brightness. We constructed two plasmids, pET-28a_lacO-LuxA-LuxB and pET-28a_lacO-LuxA-LuxB:cp157Venus, then transfected them into BL21(DE3) strain after plasmid amplification by DH5α strain, cultured at 37℃ and induced expression at 24℃. E. coli transfected with pET-28a_lacO-LuxA-LuxB successfully luminesced, but E. coli with the fusion protein LuxB:cp157Venus did not luminesce due to mutations in the target gene.


Although our first validation of BRET failed, as relevant literature has demonstrated its effect on enhancing fluorescent proteins, we directly turned our attention to our ultimate goal: to find brighter fluorescent proteins after BRET. Here is a picture extracted from relevant literature.

Original caption: The 2 ml cultures of JM109(DE3) expressing luxA + luxB, luxB:Venus + luxA and luxB:cp157Venus + luxA after the addition of decanal. The photograph was taken by SONY `\alpha`7s, ISO 5000, exposure time 2 s.

We decided to skip this step and turned to directly find or create brighter fluorescent proteins after BRET. We queried and screened the FPbase database for the two brightest proteins with fluorescence data and excitation light between 490-520 nm and named them A1 and A2. For the proteins with only amino acid sequences but no fluorescence data, we first trained the Long Short-Term Memory (LSTM) system with the existing proteins with complete information, constructed the mapping from amino acid sequences to fluorescence data, and then applied the system to predict the fluorescence intensity of the proteins with only amino acid sequences, and screened out the brightest two named B1, B2 from them.

Computer model employed molecular dynamics simulations with Python and GROMACS to model and analyze the luminescence efficiency within experimental BRET systems. System preparation, molecular structure acquisition, and GROMACS formatting were followed by simulation setup, energy minimization, and equilibration. Subsequent production runs captured BRET system dynamics, while Python scripts processed trajectory data for key parameters, enabling the calculation of BRET efficiency. These simulations provided insights into donor-acceptor interactions, distances, and orientations, shedding light on the dynamic nature of BRET under various conditions. This integrated approach enhances our understanding of molecular interactions in biological systems and holds promise for diverse research applications.

To conduct molecular dynamics simulations, we need the corresponding pdb format file for each sequence under investigation. Therefore, we use AlphaFold 2 for protein structure prediction.

Here are the structures of four sequences that are tested in the lab.

The genes of A1, A2, B1 and B2 were synthesised and fused to the C-terminal of LuxB in LuxCDABEGF with the same linker (Glu-Leu) to form four different fusion fluorescent proteins. The constructed plasmids were transferred into BL21(DE3) and cultured at 37℃, and the expression was induced by adding IPTG at a final concentration of 0.2 mM at 24℃, and finally, the emission spectra were examined, and it was found that the four LuxCDABEGF fused with the new proteins, A1, A2, B1, and B2, were emitting different coloured light, which proved the feasibility of the BRET system. And from the analysis of spectral data, all four fusion fluorescent proteins showed a double peak trend or a flat head peak trend due to the close frequency of excitation and emission light, and the difference in brightness was not big, which proved the success of the prediction of the LSTM system. However, since the BRET efficiency is related to the sixth power of the distance, the brightness of the fusion proteins is weaker compared with the original LuxCDABEGF in the case of non-optimal distance (all four proteins used the same linker as before).

The spectrum of LuxCDABEGF fused with A1

The spectrum of LuxCDABEGF fused with A2

The spectrum of LuxCDABEGF fused with B1

The spectrum of LuxCDABEGF fused with B2

In order to obtain fluorescent proteins with brighter excitation light between 490-520nm, we decided to break the constraints of nature and create fluorescent proteins by computer technology. We used the Generative Adversarial Network to generate the amino acid sequences of the candidate fluorescent proteins, and then used the LSTM system to filter the generated sequences, and finally obtained the two brightest fluorescent proteins and named them as C1 and C2. Since the effectiveness of the LSTM system has been proved in the previous experiments, the fluorescent proteins of C1 and C2 have been created with a certain degree of credibility.


We started with the plasmid pUC57-NS3-2-lacUV-cscB-lacI-KanR-NS3-1[BBa_K4115045] left by the shanghaitech_China 2022 team as the basis and used Gibson Assembly to construct our transformation plasmid, as shown in the diagram.

original plasmid pUC57-NS3-2-lacUV-cscB-lacI-KanR-NS3-1

Our plasmid construction underwent two rounds of Gibson Assembly. We replaced CscB with sfGFP containing a degradation tag, and at the same time, we replaced the lacUV promoter with the final output promoter PKaiBC, which is an intrinsic rhythm promoter in cyanobacteria. Finally, we constructed the plasmid pUC57_NS3-2-PKaiBC-sfGFP-lacI-KanR-NS3-1[BBa_K4594019].

constructed plasmid pUC57_NS3-2-PKaiBC-sfGFP-lacI-KanR-NS3-1.

According to ODE modeling and related literature[4], PKaiBC can well output oscillating expression signals within cyanobacteria.In the first model, we used odes to simulate the concentration changes of the protein KaiABC in cyanobacteria and its circadian rhythm mechanism. It involves the phosphorylation of polyphosphate sites and the inactivation rules of interactions between the RapA promoter and KaiABC, and was simulated using the ODE45 algorithm in Matlab. After we used the logistic function to fit the ATP content and introduced the regulation of transcription and phosphorylation by ATP content, the four forms of KaiC protein appeared alternately, showing a stable oscillating change with a period of 24 hours.

We used natural transformation, mixing the cyanobacteria with the plasmid, followed by incubation in the dark, and plate screening to obtain transformants.

To verify that we have inserted the target gene into the cyanobacteria's genome, we picked colonies and used PCR primers to verify the successful transformation. First, we conducted PCR using primers on both ends of NSIII. However, we only observed numerous non-specific bands and blank NSIII bands with lengths similar to the wild-type control group.

blank Neutral site III is about 1200bp

Considering that Synechococcus is often polyploid, and there may be untransformed cells, we attempted PCR using exogenous sequences, specifically the upstream and downstream primers for PkaiBC-sfGFP. Positive bands were also observed.

Almost every sample has a bright target band

Furthermore, we validated the successful integration of luxCDABEFGF gene cluster using the same method, as shown in the figure.

We examined the cyanobacteria under a fluorescence microscope. The cyanobacteria transformed into PKaiBC-sfGFP showed obvious green fluorescence. We used untransformed PCC7942 bacterial culture as a control group and found that the control group had basically no fluorescence.

Transformed PCC7942

wildtype control

The fluorescence photos further proved the successful transformation and also showed that the foreign protein we introduced could be expressed normally in cyanobacteria.

Flux balance analysis (FBA) is a computational method commonly used in systems biology to study biological networks, especially metabolic networks. The core idea is to predict the growth rate of cells or the rate of other biological processes by balancing input and output metabolite fluxes under given environmental conditions.

The basic principle is to exploit the specific rate or "flow" that exists for each metabolic reaction in an organism. Under steady-state conditions, we consider the production rate and consumption rate of each metabolite to be balanced, which means that in the metabolic network of an organism, the total input flow of each node is equal to the total output flow.

In addition to equilibrium constraints, there are other constraints, including but not limited to upper and lower limits on enzyme reaction rates, nutrient uptake rates, and artificially prescribed simulation conditions.

FBA often uses linear programming to find a flow distribution that satisfies all equilibrium constraints to maximize or minimize some objective function, such as maximizing biomass yield. In this model, we refer to the experimental method of Triana J. to maximize the flow rate of the biomass reaction.

It can be seen that if GlgC is silenced, the glycogen production rate, ATP production rate, and cell growth rate all decrease significantly during the day. According to the saving of materials, the sucrose yield is significantly improved, and our design can basically be realized.

This indicates that the target fragment has been successfully integrated into the genome of the cyanobacteria. However, the colonies picked were not monoclonal, and we believe this may be related to the insufficient selectivity of the solid agar medium. However, we didn't have any time to purify a monoclone and conduct characterization experiments.


[1] Brodl, Eveline et al. “The impact of LuxF on light intensity in bacterial bioluminescence.” Journal of photochemistry and photobiology. B, Biology vol. 207 (2020): 111881. doi:10.1016/j.jphotobiol.2020.111881

[2] Nijvipakul, Sarayut et al. “LuxG is a functioning flavin reductase for bacterial luminescence.” Journal of bacteriology vol. 190,5 (2008): 1531-8. doi:10.1128/JB.01660-07

[3]Kaku T, Sugiura K, Entani T, Osabe K, Nagai T. Enhanced brightness of bacterial luciferase by bioluminescence resonance energy transfer. Sci Rep. 2021;11(1):14994. Published 2021 Jul 22. doi:10.1038/s41598-021-94551-4

[4]Cohen SE, Erb ML, Selimkhanov J, et al. Dynamic localization of the cyanobacterial circadian clock proteins. Curr Biol. 2014;24(16):1836-1844. doi:10.1016/j.cub.2014.07.036

Back to Top