Engineering

Biosensors and high-pass bacteria

Cycle 1 Ensure the feasibility of diffusion experiments

Design

In order to ensure the feasibility of our AHL diffusion experiments, we used the plasmids LasR-deGFP and LuxR-deGFP from 2024 WHU_China to construct high-pass engineered bacteria with fluorescence induced by 3-oxo-C12 HSL and 3-oxo-C6 HSL.

In this cycle, the expression of the downstream corresponding AHL receptor proteins (LuxR and LasR) are activated by the T7 promoter and lac operator. The lac operator ensures that LuxR and LasR are only expressed when needed, thereby reducing the metabolic burden on the engineered bacteria. The deGFP protein is induced by the promoters Plux and Plas. (Figure 1-1) When adding different concentrations of AHL, the bacteria will express different level of GFP.

Build

We transformed the plasmids LasR-deGFP and LuxR-deGFP from 2024 WHU_China into E. coli BL21 (DE3) strain. (Figure 1-2)

Test

To verify the response of E. coli to the two AHL molecules, 3-oxo-C12 HSL and 3-oxo-C6 HSL, we implemented liquid induction and solid diffusion experiments.

In liquid induction experiments, we mixed IPTG-induced bacterial cultures with serially diluted AHLs in 96-well plates. OD600 and fluorescence intensity were measured using a microplate reader once an hour (Figure 1-3).

Through liquid culture and fluorescence detection, we observed that in the first 2 hours, the biosensors could detect different concentrations of AHL. However, as time went on, the bacterial proliferation became excessive, and the detection effect declined (Figure 1-3).

In agar plate colony experiments, we dispensed the culture and different concentration 3-oxo-C12-HSL at the positions shown in Figure 1-4A. After 12h incubation, we detected the fluorescence of the colonies with inverted fluorescent microscope (Figure 1-4).

Learn

Biosensor of 3-oxo-C6 HSL does not show a good decline curve, which may be due to the high concentration of AHL in the short diffusion distance, which disturbs the metabolism of engineered bacteria or inhibits the growth of colonies, while the fluorescence intensity in the far distance may be caused by leakage expression.

Cycle 2 Replace T7 promoter with constitutive promoter J23106

Design

To eliminate the induction step required before using the biosensor and to avoid the excessive expression caused by the T7 strong promoter, we replaced the T7 promoter and lac operator with the constitutive promoter PJ23106. (Figure. 1-5)

Build

We performed PCR and Gibson Assembly to build plasmid PJ23106-LasR-deGFP. (Figure 1-6)

Test

Sequencing confirmed transformation success.

Referencing the work of Fedorec et al. (2024), we employed agar plate colony experiment to validate the “highpass”. We dispensed the culture and different concentration 3-oxo-C12-HSL at the positions shown in the map (Figure 1-7A) onto the surface of the agar plate, detecting the fluorescence with inverted fluorescence microscope.

Other concrete details can be found on the experiments page.

Learn

In the agar plate colony experiment, colonies closer to the central point, which had access to higher concentrations of AHL, did not express more GFP. Instead, colonies slightly farther from the centre point exhibited the highest fluorescence. (Figure. 6A) Observing the colony morphology revealed that the colonies located farther from the centre grew better. (Figure. 6B) We speculated that overmuch 3-oxo-C12 HSL may affect the colony growth and thereby influence the fluorescence intensity.

Cycle 3 Use medium-copy-number backbone

Design

We hypothesized that the primary reasons for poor bacterial growth may include two aspects: First, the constitutive expression of the receptor protein on the high-copy plasmid (ori is pColE1) significantly increases the metabolic burden on the engineering bacteria. Second, a high concentration of the AHL ligand strongly induces the promoter to drive excessive expression of GFP, further exacerbating metabolic stress. The combined effect of these two factors may lead to plasmid loss under sustained high metabolic load. To mitigate these issues, we propose the following strategy: replacing the ori with the medium-copy p15A origin and adopting the Rhl quorum sensing system that is orthogonal[3] to more systems than Las to construct a more stable biosensor bacteria. The following circuit is designed accordingly. (Figure.1-8)

Build

We obtained RhlR and PRhl fragments from Shi-shen Du's Lab. Using Gibson assembly, we inserted these fragments into the iGEM JUMP collection plasmid pJUMP26-1A (sfGFP) to construct the RhlR-sfGFP plasmid, which was then transformed into E. coli DH5α. (Figure 1-9, 1-10)

Test

We performed colony PCR and sequencing to verify successful plasmid construction. Sequencing confirmed transformation success. (Figure. 1-11)

We conducted the agar plate colony experiment on the engineered bacteria to determine whether they could emit green fluorescence, thereby achieving biosensor functionality, and plotted a series of dose response curves. (Figure. 1-12)

Learn

Through spiral dispensing in agar plate colony experiment, we confirmed that this design generated consistent biosensor response curves, demonstrating that the ability to detect C4-HSL concentrations in solid media. The obtained data display well-defined trends, making them suitable for subsequent fitting and computational mathematical modeling by the Model Group. Furthermore, Rhl biosensor bacteria carrying the RhlR-sfGFP plasmid exhibits a highpass filtering behavior, which supports their use as independent logic gate modules. The relevant verification can be viewed on the Results page.

Building upon this methodology, we subsequently succeeded in constructing the LasR-sfGFP plasmid and CinR-sfGFP plasmid, whose experimental result meet our expectation. However, we encountered difficulties during the construction of the TraR-sfGFP plasmid (corresponding to Cycle 4).

Build & test of LasR-sfGFP and CinR-sfGFP

Based on the successfully verified RhlR-sfGFP plasmid, we obtained the LasR and Plas fragments from the T7-LasR-deGFP plasmid using PCR. We then replaced the Rhl system in the RhlR-sfGFP plasmid and constructed the LasR-sfGFP plasmid using Gibson assembly. (Figure. 1-13A)

Concurrently, the CinR_AM and Pcin_AM sequences, derived from the iGEM Distribution Kit, were utilised to substitute the Rhl system in the RhlR-sfGFP plasmid. The CinR-sfGFP plasmid was then constructed through Gibson assembly. (Figure. 1-13B)

We conducted the agar plate colony experiment on the engineering bacteria to determine whether they could emit green fluorescence, thereby achieving biosensor functionality, and plotted a series of dose response curves. (Figure. 1-14A, Figure. 1-14B)

Cycle 4 Optimize Tra biosensor bacteria

Design

Based on the successfully validated RhlR-sfGFP plasmid construct, we utilized the RhlR-sfGFP plasmid as backbone and incorporated the synthesized sequences of TraR and Ptra from GenScript to construct the TraR-sfGFP plasmid, thereby establishing the biosensor bacteria for the Tra system. The following circuit was designed accordingly. (Figure. 1-15)

Build

To construct the TraR-sfGFP plasmid, we used Gibson assembly to replace the RhlR and Prhl parts in the RhlR-sfGFP plasmid backbone with the synthesised TraR and Ptra sequences. After that, we transformed it into E. coli DH5α. (Figure. 1-16, Figure. 1-17)

Test

We performed colony PCR and sequencing. The plasmid was successfully constructed. But in the subsequent agar plate colony experiment, we did not observe significant fluorescence.

Learn

Consult the extant literature on the subject found that, wild type Ptra promoter is nonfunctional in E. coli. To solve the problem of the Tra system's poor functionality, researchers optimized and modified the Ptra promoter and TraR protein sequences.[1] (Figure. 1-18)

Therefore, we referenced the literature and changed the promoter to Ptra* and performed point mutation to the TraR.

Re-design & Build

We synthesized Ptra* using DNA oligo annealing and constructed Ptra*-TraR-sfGFP plasmid based on TraR-sfGFP plasmid (Figure 1-19A). Simultaneously, we substituted the amino acid at position 192 of TraR with Trp using site-directed mutagenesis , constructing Ptra*-TraR(W)-sfGFP plasmid (Figure 1-19B).

Test

We performed colony PCR and sequencing. The plasmids were successfully constructed. However, in the ensuing agar plate colony experiments, no significant fluorescent signal was detected, and increasing the concentration of 3-oxo-C8-HSL did not yield any noticeable improvement.

Learn

Upon consulting the literature, we discovered that the overall fluorescence response of Tra system was lower than that of the Las one. [1] (Figure 1-20) Therefore, we hypothesize that enhancing GFP expression can be achieved by increasing the plasmid copy number or replacing with a stronger constitutive promoter.

Redesign & build

On one hand, we replaced PJ23106 in the Ptra*-TraR-sfGFP plasmid with PJ23119, a stronger constitutive promoter, to construct J23119-Ptra*-TraR-sfGFP plasmid. (Figure 1-21A) On the other hand, we replaced the medium-copy p15A origin with the high-copy ColE1 origin to construct Ptra*-TraR-ColE1-sfGFP plasmid. (Figure 1-21B)

Test

We performed colony PCR and sequencing. The plasmids were successfully constructed.

Subsequently, we conducted the agar plate colony experiment, and found that Tra biosensor bacteria with Ptra*-TraR-ColE1-sfGFP plasmid could successfully respond to 3-oxo-C₈-HSL, and emit green fluorescence. Construction of the Tra biosensor bacteria was successful. (Figure 1-22)

Learn

Following increasing the plasmid copy number, green fluorescence was observed from the Tra biosensor bacteria, indicating that the functionality of both TraR and the Ptra* part was successfully confirmed.

Band-pass engineered bacteria

Cycle 1 Conduct a proof-of-concept using the IPTG system

Design

Similar to the “high-pass” circuit, we refered to the work of Fedorec et al. (2024) [2] to design a “band-pass” genetic circuit. The character of a band-pass filter is that the downstream circuit cannot be activated at too high and too low input signals, only the appropriate concentration of signal molecules can activate downstream gene expression. Bacteria with a band-pass circuit can be used as an XOR gate in our project, and can be combined with the “high-pass” bacteria to form a half-adder or full-adder under the condition of three inputs.

Refer to the work of Fedorec et al. (2024) [2] , we designed a band-pass genetic circuit induced by IPTG and the final effect is presented by the expression of sfGFP. The expression of sfGFP is promoted by T7 RNA polymerase(T7 RNAP) and inhibited by PhlF at the same time, only when the expression of both is at an appropriate level can sfGFP be expressed. This requires the input of IPTG to be at an appropriate intermediate concentration. When the input IPTG concentration is too low, sfGFP cannot be effectively expressed due to the lack of T7 RNAP ; and when the concentration of IPTG is too high, the expression of GFP will be inhibited by PhlF.

In our final design, sfGFP will be replaced by enzymes that can produce other inducible molecules like N-acyl homoserine lactones (AHLs).

We have designed the following circuit (Figure 2-1) to achieve the effect in the literature. We constructed two plasmids and transformed them into BL21(DE3) to have all the components in our bacteria. To regulate the relative expression levels of different components, we used vectors with different copy numbers.

Build

Based on the vector pJUMP26-1A(sfGFP) and pJUMP27-1A(sfGFP), we conducted PCR to linearize vectors and inserted promoter sequences through overlapping ends. Then we conduct Gibson Assembly to construct the two plasmids. (Figure 2-2) We conduct co-transformation on BL21(DE3), which already has T7 RNAP gene initiated by lac UV5 promoter on its genome.

Test

To test the IPTG band-pass circuit, we used two different patterns to inoculate colonies and added a series of concentrations of IPTG solution as input.

We expected colonies closest to and farthest from the input point both display low fluorescence expression, and other colonies display high fluorescence expression. We used two different patterns to inoculate colonies and give a series of concentrations of IPTG as input (Figure 2-3). The spiral pattern is used to verify the character of band-pass bacteria, and the logic gate pattern is used to find an appropriate combination of input distance and concentration to make the colony work as an XOR gate.

The result was very consistent with expectation: The colonies closest and farther away from the input point hardly emit fluorescence, and the colonies in the middle distance emit obvious fluorescence (Figure 2-4).

Further, we found some groups already possessing the properties of XOR logic gates(fig.27).

Learn

The successful validation of the IPTG band-pass circuit demonstrates that our approach to building a band-pass circuit is feasible. We will imitate the construction method to build a quorum sensing band-pass circuit.

Cycle2: Las band-pass circuit

Design

We first tried to construct the Las band-pass circuit firstly (Figure 2-6). The operating principle of Las circuit is similar to that in IPTG circuit in Cycle 1: when the 3-oxo-C12 HSL concentration is too low, GFP cannot be effectively expressed due to the lack of T7 RNAP; when the concentration of 3-oxo-C12 HSL is too high, the expression of GFP will be inhibited by PhlF. Only when the expression of both is at an appropriate level can GFP be expressed.

Build

To build the full circuit, we construct two intermediate plasmids first. Based on the Las half adder 1 in high-pass circuit, iGEM distribution kit and plasmids from 2024 WHU-China, we conducted PCR to linearize vectors and get fragments, then we conduct Gibson Assembly to construct the two intermediate plasmids. After that, we conduct the whole Las band-pass plasmid also with PCR and Gibson Assembly (Figure 2-7).

Test

We successfully obtain the "band-pass" bacteria. Similar to "high-pass" engineered bacteria, we first performed simplified Agar plate colony experiments. (Figure 2-8) Unfortunately, in the Agar plate colony experiments, our "band-pass" bacteria did not show any fluorescence signal.

Learn

We encountered many difficulties during the construction process of Las band-pass circuit, especially the issue of fragment and plasmid loss. We considered the plasmid may be too large and there may be too many duplicated fragments. Due to the success of cycle1, next step was to try the construction of IPTG band-pass circuit to build Las band-pass circuit.

Cycle 3: Las band-pass circuit improvement

Design

Due to the difficulty of the construction of Las system, we decided to reduce the engineering difficulty by constructing a circuit without T7 RNAP. The new circuit uses Plas to initiate expression of sfGFP directly (Figure 2-9).

Build

To build the full circuit, we conducted overlap PCR to get fragments, then we conduct restriction cloning to construct the whole plasmid. (Figure 2-10).

Test

We successfully obtain the "band-pass" bacteria. Similar to IPTG band-pass band-pass circuit, we used two different patterns to inoculate colonies and give a series of concentrations of IPTG and 3-oxo-C12 HSL at the same time in different proportions as input(Figure 2-3). Unfortunately it still did not show any fluorescence signal.

Learn

The reason for the lack of fluorescence signal might be excessive expression of phlF. T7 RNAP is a strong factor in initiating downstream gene expression and it is necessary in our circuit. So next we planned to add T7 RNAP back into our band-pass circuit.

Cycle 4 Las band-pass circuit improvement based on IPTG band-pass circuit

Design

Based on the experience we have gained from the previous two cycles, we decided to to gradually replace each component with which in Las band-pass circuit based on the IPTG band-pass circuit in cycle 1. We replaced lacI with lasR and lac UV5 promoter with Plas.

Theoretically, when both 3-oxo-C12 HSL and IPTG are used for induction, the engineered bacteria will exhibit a band-pass effect.

Build

Based on the plasmid IPTG-band-pass-p15A and IPTG-band-pass-pSC101 in cycle 2-1, we conducted Las dual plasmid system. Through high fidelity PCR and Gibson Assembly, we get fragment needed and engineered plasmid IPTG-band-pass-p15A to plasmid Cm-p15A-LasR-PhlF. The final dual plasmid system is shown below (Figure 2-11).

Test

Similar to IPTG band-pass band-pass circuit, we used two different patterns to inoculate colonies and give a series of concentrations of IPTG and 3-oxo-C12 HSL the same time in different proportions as input(Figure 2-3). The result is very consistent with expectation: The colonies closest and farther away from the input point hardly emit fluorescence, and the colonies in the middle distance emit obvious fluorescence(Figure 2-12).

However, due to the limited number of experiment groups, we did not obtain particularly ideal results in logic gate test(Figure 2-13).

Learn

The most difficult point is how to adjust the ratio of IPTG and 3-oxo-C12 HSL to achieve band-pass effect. We originally planned to fully replace the system, but due to time constraints we did not implement it.

Although we did not obtain obvious XOR logic gate due to time limitations, the band-pass effect showed by Las-IPTG double-plasmid bacteria has already indicated that it is likely to be an appropriate concentration ratio of 3-oxo-C12 HSL and IPTG to make it a XOR gate.

Cycle 5 Rhl band-pass circuit

As our ultimate goal is to use AHL as the "wire", we have further designed Rhl band-pass circuit based on Rhl quorum sensing system.

Design

The circuit of Rhl band-pass circuit is very similar to which in IPTG band-pass circuit. We use RhlR as a receptor to receive C4-HSL signals and activate T7 RNAP expression, which can further promote the expression of sfGFP; and we use phlF as a repressor to inhibit the expression of sfGFP. Maintaining an appropriate ratio between phlF and T7 RNAP can create a band-pass effect(Figure 2-14).

Build

Based on the Rhl high-pass plasmid and iGEM distribution kit, we conducted PCR to linearize vectors and get fragments, then we conduct Gibson Assembly to construct Rhl band-pass plasmid (Figure 2-15).

Test

Similar to IPTG band-pass circuit, we used two different patterns to inoculate colonies and added a series of concentrations of C4-HSL as input.

We expect that colonies near the input point show lower fluorescence expression, however the results did not fully meet our expectation.

Learn

The implementation of XOR gate effect strongly depends on the mutual coordination of PhlF, T7 RNAP and their relative components. A slight imbalance in the effects of the two elements can lead to disappointing results. Additionally, we found that the fluorescence intensity of colonies in all experimental groups was relatively low, we think it is an issue with the C4-HSL response threshold. In the next cycle, we will make improvements to address these two issues.

Cycle 6 Rhl band-pass circuit improvement based on IPTG band-pass circuit

Design

For the two issues in the previous cycle, we have come up with a solution for each. For the weak overall fluorescence brightness, we found a Rhl promoter(Part:BBa C0071 - parts.igem.org) from Team Tokyo Tech 2016 whose promoter activity was stronger than that of wild type (previously used)(Figure 2-16).

At the same time, another group of us tried a different way to improve the Rhl band-pass circuit. We transformed two plasmids into BL21(DE3) strain just like what we did in IPTG band-pass circuit. Actually, we replaced one of the plasmids in IPTG band-pass circuit to gradually replace IPTG with C4-HSL (the left one in Figure 2-18).

Build

Based on the plasmid Rhl-band-pass in the previous cycle, we conduct PCR and changed the promoter sequences through overlapping ends. Both Prhl were replaced by Prhl(NM)(Figure 2-17).

For the dual plasmid: based on the vector pJUMP26-1A(sfGFP), we conducted PCR to linearize vectors and inserted sequences of Rhl band-pass circuit through overlapping ends. Then we conduct Gibson Assembly to construct the plasmid. Using this plasmid and plasmid IPTG-band-pass-pSC101 in cycle 2-1, we conduct co-transformation on BL21(DE3), which already has T7 RNAP gene initiated by lac UV5 promoter on its genome. The final dual plasmid system is shown below (Figure 2-18).

Test

Similar to IPTG band-pass circuit, we used two different patterns to inoculate colonies and added a series of concentrations of IPTG and C4-HSL the same time in different proportions as input(Figure 2-3).

The result is very consistent with expectation: The colonies closest and farther from the input point hardly emit fluorescence, and the colonies in the middle distance emit obvious fluorescence (Figure 2-19).

However, due to the limited number of experiment groups, we did not obtain particularly ideal results in logic gate test (Figure 2-20).

Learn

The most difficult point is how to adjust the ratio of IPTG and C4-HSL to achieve band-pass effect. We originally planned to fully replace the system, but due to time constraints we did not implement it.

Although we did not obtain obvious XOR logic gate activated by AHL due to time limitations, the band-pass effect showed by Rhl-IPTG double-plasmid bacteria has already indicated that it is likely to be an appropriate concentration ratio of C4-HSL and IPTG to make it a XOR gate.

Connector bacteria

Cycle 1 Use the same AHL for transmission (amplifier)

Design

As the signal of one bit may spread the previous digit to cause crosstalk, we use amplifying bacteria to pull two digit apart from each other. To construct an amplifying bacteria of Las signal, we designed the LasR-LasI plasmid using the T7-LasR-deGFP plasmid as the vector. The LasR gene is initiated by the T7 promoter and regulated by the lac operon, thus requiring IPTG to induce expression. The LasI gene is regulated by the las promoter. When the externally imported 3-oxo-C12AHL molecules bind to the LasR protein, they can activate the las promoter, express the LasI gene, and export the Las AHL molecules, thereby amplifying the Las signal, lengthening the distance a signal can transmit.(Figure 3-1)

Build

We obtained the LasR-vector and LasI fragments from LasR-deGFP and LuxR-LasI (from 2024 WHU-China) by PCR respectively, and then constructed the LasR-LasI plasmid through Gibson assembly.(Figure 3-2)

Test

Plasmids were transformed into E. coli BL21(DE3). Sequencing verified transformation success.(Figure 3-3)

We dispense samples as shown in Figure 3-4. to verify the function of amplifying bacteria. Biosensors can detect the Las AHL signal output from the amplifying bacteria through green fluorescence. The position of eight biosensors are arranged carefully to avoid having the front block the back. Figure 3-5 shows the result.

Learn

We found that amplifying bacteria can achieve the amplification effect of receiving input 3-oxo-C12 HSL signals and output 3-oxo-C12 HSL signals. However, in the absence of 3-oxo-C12 HSL input, downstream biosensor could still detect 3-oxo-C12 HSL signal and produce green fluorescence due to amplification by the amplifying bacteria.

These abnormal phenomena indicate that the amplifying bacteria have certain leakage expression problems, that is, they generate 3-oxo-C12 HSL output signals when there is no external 3-oxo-C12 HSL input.We speculate that positive feedback effect of amplifying bacteria and leaky expression of the T7 promoter result in the expression of the LasI gene in the absence of 3-oxo-C12 HSL. On the other hand,the positive feedback between lasR and lasI may also play a role. To further prevent the leakage, we change amplifier to connector bacteria in later experiment, which can receive one AHL and turn this signal to another.

Cycle 2 Use different AHL for transmission to avoid excessive positive feedback (connector)

Design

In serial computation, the carry signal generated from a lower-bit operation must propagate to the next higher-bit position to perform multi-bit computations. We designed the connector bacteria to transfer the carry output (Ci).

The carry output (Ci) phenotype in half-adder and full-adder bacteria can be enable by the highpass genetic circuits, which form the basis of our connector bacteria. Within the same connector bacterium, the two QS systems employed are mutually orthogonal[3] to prevent crosstalk between the upstream and downstream colonies. (Figure 3-6)

Build

In this build, we employed mutually orthogonal QS system, Rhl and Las systems, and each built upon the plasmid of its own respective biosensor bacteria.

For instance, Gibson assembly was employed to replace the sfGFP gene in the LasR-sfGFP plasmid from the Las biosensor bacteria with the RhlI sequence (Figure 3-7A), which has the capacity to produce C4-HSL. This allowed us to create the LasR-RhlI connector bacteria (Figure 3-7B), which transmit signals to downstream colonies. Similarly, we constructed the RhlR-LasI connector bacteria (Figure 3-6C) using the RhlR-sfGFP plasmid as the backbone.

Test

We performed colony PCR and sequencing. The plasmids were successfully constructed.

Subsequently, we conducted a agar plate colony experiment and found that the AHL ligand concentration expressed by connector bacteria followed the diffusion curve, but there was certain leaky expressions.(Figure 3-8)

Learn

The AHL ligands produced by the connector bacteria were detected by the biosensor bacteria, which emitted fluorescence in accordance with a diffusion curve. The functional validation of the connector bacteria was successful in the test.

Cycle 3 Build connectors for different AHL systems

Design

Similarly, we utilize three pairwise orthogonal QS systems[3]—Rhl, Tra, and Cin system—to construct a greater variety of connector bacteria which is capable of transmitting QS signals from different molecules. (Figure 3-9)

Build

We utilized Gibson Assembly to replace the sfGFP gene in the RhlR-sfGFP plasmid from the Rhl biosensor bacteria with TraI and CinI sequences, thereby constructing RhlR-TraI and RhlR-CinI connector bacteria. (Figure 3-10)

Test

We performed colony PCR and sequencing. The plasmids were successfully constructed.

However, in the ensuing agar plate colony experiments, no significant fluorescent signal emitted by downstream biosensor bacteria was detected.

Learn

The functional validation of the two constructed connectors bacteria was unsuccessful. This may be because the inducer proteins express an insufficient concentration of AHL ligands, or because the function of the heterologous proteins is impaired in the E. coli chassis. Further investigation is required.

Light-inducible AHL degradation enzyme

Cycle1 Verify the functions of AiiA and explore the experimental conditions

Design

In this cycle, we aim to test the function of AiiA to degrade acyl-homoserine lactone (AHL) molecules and to identify an appropriate method to quantify the degradation efficiency of AiiA. From a theoretical standpoint, measuring the enzymatic activity of purified AiiA would be an excellent approach; however, because screening split sites is required for the light-induced engineering of AiiA, protein purification at this stage would impose a substantial workload. Therefore, we plan to first employ a simpler assay to assess AiiA degradation efficiency, and proceed to purification thereafter. After drawing on the work of Koch et al.(2014)[4] we will express AiiA in E. coli under IPTG induction and incubate the cells in medium containing AHLs to degrade the AHLs. After a period of incubation and degradation, the residual AHL concentration in the culture supernatant will be detected using the biosensor bacteria. Detailed experimental procedures are provided in the Measurement section.

Build

To evaluate whether AiiA is prone to generating inclusion bodies, we constructed two plasmids, pGEX-AiiA and pGEX-GST-AiiA, which enable IPTG-inducible expression of AiiA and the GST–AiiA fusion protein, respectively.(Figure 4-1)

We amplified the DNA fragments used for plasmid construction by PCR, with the AiiA sequence synthesized by GenScript. The fragments were assembled with the linearized vector via Gibson Assembly to generate the final plasmids. The assembled plasmids were first transformed into E.coli DH5α strains; single colonies were picked and validated by colony PCR and Sanger sequencing. (Figure 4-2) After verification, plasmids were isolated and transformed into E. coli BL21(DE3) strains.

Finally, BL21 strains containing pGEX-AiiA or pGEX-GST-AiiA plasmids were induced with IPTG, and Coomassie Brilliant Blue staining was performed to confirm proper protein expression.(Figure 4-3)

Test

The experiment for measuring degradation efficiency can be simplified to induction–degradation–detection. Induction was carried out at 25°C with 0.5 mM IPTG for approximately 12 h. Cells were then incubated with AHL in a shaker at 37°C for about 2 h, after which the cells were removed by centrifugation. Fresh medium and the biosensor bacteria were added, and the mixture was incubated at 37°C in a plate reader with OD and fluorescence recorded every 4 min. Detailed experimental procedures and data processing are provided in the Measurement section.

Figure 4-4 displays the most representative result. We also evaluated degradation efficiency at terminal AHL concentrations of 0.5 mM and 0.05 mM, which yielded results consistent with those observed at 1 mM. The outcomes for the GST–AiiA–expressing strain were likewise consistent with those for the AiiA-expressing strain.

Learn

The results indicate that, even without IPTG induction, leaky expression of AiiA is sufficient to completely degrade C4 in the medium within 2 h. Under expression conditions of 25°C with 0.5 mM IPTG, enzymatically active AiiA was produced; consequently, the GST tag will not be used in subsequent experiments to avoid inclusion body formation. AiiA exhibits strong degradation efficiency, and our assay was progressively refined during this process. As a next step, we plan to proceed directly with the light-induced engineering of AiiA.

Cycle 2 AiiA computational split sites selection and rational design

Design

To gain in-depth structural insights into the AiiA protein and identify suitable split sites, we employed molecular dynamics simulations and molecular docking to facilitate a more rational selection of split loci. Given that our AiiA sequence differs slightly from the AiiA structures available in the PDB, we plan to use AlphaFold3 to obtain the structural model of our AiiA protein.[5] Subsequently, molecular dynamics simulations and associated analytical methods will be applied to characterize the flexible regions, active sites, and principal components of AiiA, using the CHARMM36 force field. The SPELL algorithm will also be utilized to determine split sites based on split energy and solvent accessible surface area information.[6]

After selecting the split sites, we constructed fusion proteins of split AiiA with VVD[VVD is a light-oxygen-voltage (LOV) sensory domain that dimerizes in response to blue light, thereby facilitating the self-assembly of split proteins. Please refer to the introduction in Cycle3 for detailed information]. The structural models of these fusion proteins were also generated using AlphaFold3, and molecular dynamics simulations were performed to assess the stability of the fusion constructs. We then employed protein–protein docking methods, including AlphaFold3, Z-Dock, and Rosetta, to predict the conformation of the VVD dimer. Using the dimeric protein structure, molecular docking experiments were conducted with the AHL molecule, followed by further molecular dynamics simulations to evaluate the enzymatic activity upon dimerization.

Through the above computational approaches, we aim to assist in the screening of optimal split sites in AiiA, while also refining the in silico methodology based on wet-lab experimental results, thereby contributing to the improvement of algorithms for identifying protein split sites.

Build

We began by determining the three-dimensional (3D) structure of the AiiA protein using AlphaFold3.(Figure 4-5)

The necessary scripts for molecular dynamics (MD) simulation and analysis were developed. MD simulations were subsequently conducted employing the GROMACS for CPU-2021 package.

echo -e "1\n1" |gmx_mpi pdb2gmx -f A1.pdb -o A1.gro
gmx_mpi editconf -f A1.gro -o A1-box.gro -c -d 1.2 -bt dodecahedron
gmx_mpi solvate -cp A1-box.gro -cs spc216.gro -o A1-sol.gro -p topol.top
gmx_mpi grompp -f ions.mdp -c A1-sol.gro -p topol.top -o ions.tpr
echo "13" |gmx_mpi genion -s ions.tpr -o A1-sol-ions.gro -p topol.top -pname NA -nname CL -neutral

gmx_mpi grompp -f em.mdp -c A1-sol-ions.gro -p topol.top -o em.tpr
gmx_mpi mdrun -v -deffnm em
echo "10 0" |gmx_mpi energy -f em.edr -o potential.xvg

gmx_mpi grompp -f nvt.mdp -c em.gro -r em.gro -p topol.top -o nvt.tpr
gmx_mpi mdrun -v -deffnm nvt
echo "16 0" |gmx_mpi energy -f nvt.edr -o temperature.xvg

gmx_mpi grompp -f npt.mdp -c nvt.gro -r nvt.gro -t nvt.cpt -p topol.top -o npt.tpr
gmx_mpi mdrun -v -deffnm npt
echo "18 0" |gmx_mpi energy -f npt.edr -o pressure.xvg
echo "24 0" |gmx_mpi energy -f npt.edr -o density.xvg

gmx_mpi grompp -f md.mdp -c npt.gro -t npt.cpt -p topol.top -o md_0_1.tpr
gmx_mpi mdrun -v -deffnm md_0_1

# Analysis Scripts
echo -e "1 \n 0" | gmx_mpi trjconv -s md_0_1.tpr -f md_0_1.xtc -o md_0_1_noPBC.xtc -center -pbc mol -ur compact

echo -e "4 \n 4" | gmx_mpi rms -s md_0_1.tpr -f md_0_1_noPBC.xtc -o rmsd.xvg -tu ns
echo -e "4 \n 4" | gmx_mpi rms -s em.tpr -f md_0_1_noPBC.xtc -o rmsd_xtal.xvg -tu ns

echo "1" | gmx_mpi gyrate -s md_0_1.tpr -f md_0_1_noPBC.xtc -o gyrate.xvg

echo -e "4 \n 4" | gmx_mpi trjconv -f md_0_1_noPBC.xtc -s md_0_1.tpr -o avg.pdb -dump 0
echo -e "4 \n 4" | gmx_mpi rmsf -f md_0_1_noPBC.xtc -s avg.pdb -o rmsf.xvg -res

echo -e "1 \n 1" | gmx_mpi sasa -f md_0_1_noPBC.xtc -s md_0_1.tpr -o sasa.xvg -tu ns

echo -e "1 \n 1" | gmx_mpi hbond -f md_0_1_noPBC.xtc -s md_0_1.tpr -num hbnum.xvg

echo "1" | gmx_mpi do_dssp -f md_0_1_noPBC.xtc -s md_0_1.tpr -o ss.xpm -sc ss_comp.xvg -tu ns
gmx_mpi xpm2ps -f ss.xpm -o ss.eps

echo -e "4 \n 4" | gmx_mpi covar -f md_0_1_noPBC.xtc -s md_0_1.tpr -o eigenvalues.xvg -v eigenvectors.trr -xpma covara.xpm

echo -e "4 \n 4" | gmx_mpi anaeig -f md_0_1_noPBC.xtc -s md_0_1.tpr -v eigenvectors.trr -proj proj.xvg -first 1 -last 2

echo -e "4 \n 4" | gmx_mpi mdmat -f md_0_1_noPBC.xtc -s md_0_1.tpr -mean dm.xpm -frames frames.xpm -no -n
gmx_mpi xpm2ps -f dm.xpm -o dccm.eps

Based on the subsequent use of AlphaFold3 for the folding and docking of the fusion protein, the folding results were satisfactory. However, the docking results provided by AlphaFold3 exhibited low confidence scores and lacked consistency across multiple runs. Consequently, we employed an alternative strategy utilizing the Z-Dock rigid-body docking method followed by Rosetta for flexible refinement to predict the dimeric conformation of the fusion protein. Unfortunately, the final model showed a significant discrepancy compared to the VVD dimer conformation reported in the literature. We conducted some preliminary analysis and discussion within the Learn part. Ultimately, however, as these experiments did not yield conclusive or meaningful results, this section was omitted from the final manuscript for brevity.

Test

We performed 60 ns molecular dynamics simulations using GROMACS 2021.1. The system was constructed based on the AiiA structure provided by AlphaFold 3, employing the CHARMM36 force field and the TIP3P water model. The protein was placed in a dodecahedral water box with a 1.2 nm clearance from the protein surface. The system underwent energy minimization, followed by 100 ps each of NVT and NPT equilibration. Production simulations were conducted under the NPT ensemble, with temperature maintained at 300 K (V-rescale) and pressure at 1 bar (Parrinello–Rahman). A 2 fs integration time step was used, along with the LINCS constraint algorithm. Trajectory analyses included RMSD, RMSF, radius of gyration, hydrogen bonding, and principal component analysis to evaluate the structural stability and dynamic properties of the protein. The PDB file of AiiA used in our study was submitted to the SPELL web server (https://dokhlab.med.psu.edu/spell/login.php) to obtain the corresponding split protein output. Additional relevant parameters are provided in the Supplementary Materials.

The selection of split sites was primarily based on RMSF profiles and SPELL results(Figure 4-6, Figure 4-7), the selection was also partially informed by the results of Principal Component Analysis (PCA). (see Supplementary Materials) and visual inspection of the molecular dynamics trajectories. Some trade-offs were made during the selection process. For instance, while residue 154 exhibits relatively low flexibility, it was still chosen as a split site due to its favorable Split Energy and solvent accessible area (SAA) characteristics, and was subsequently validated in experimental assays. These wet-lab experiments also provide insights into the relationship between local flexibility and splitting efficiency. Based on the initial round of experimental results, our strategy will be adaptively adjusted to explore alternative split sites. Following the finalization of split sites, the structures of the resulting split fusion proteins were predicted using AlphaFold 3 and subjected to molecular dynamics simulations.

A detailed analysis of the simulations revealed several insightful findings. For instance, the choice of split site appeared to have a minimal impact on the Solvent Accessible Surface (SAS) of the VVD-cAiiA fusion protein but a more pronounced effect on the SAS of the nAiiA-VVD fusion, particularly around residues 154-181. Furthermore, splitting at 154-181 significantly altered the RMSD fluctuations of nAiiA-VVD, leading to a more compact and stable protein conformation. Consequently, the nAiiA-VVD construct split at site 181-182 exhibited very low RMSD fluctuations during the simulation, indicating high structural stability. Additionally, the RMSD fluctuations in VVD-cAiiA were primarily traced to the region encompassing residues 181-208aa.(Figure 4-8)

Integrating these observations, we hypothesize that the region 154-208 constitutes a crucial hydrophobic core in AiiA (consistent with the SAS profile from SPELL analysis), which interacts with other hydrophobic patches. From this perspective, residue 208 represents a more favorable split site.

Learn

The molecular dynamics simulations provided valuable insights that aided in our screening of potential split sites. The wet-lab experiments further demonstrated that the self-assembly activities of the proteins split at the three selected sites were nearly identical, supporting the reliability of the SPELL predictions. However, the VVD domain did not exhibit its expected function in the wet-lab experiments, which prevented us from leveraging these experimental results to refine our computational methodology.

For protein-protein and molecular docking studies, the dimerization of VVD is known to require a ligand and involves light-induced conformational changes, which falls beyond the scope of our current technical capabilities. Conversely, docking simulations conducted solely with AiiA would provide insights predominantly into the self-assembly of the split fragments, rather than informing on the functional interaction within the full fusion protein system.

Cycle 3 AiiA light-inducible split and functional validation

Design

Our strategy for constructing light-induced AiiA is to split AiiA and fuse the fragments to photosensory proteins; upon light induction, the photosensors dimerize and thereby promote reconstitution of the split fragments into full-length, functional AiiA. In the absence of light, the photosensors gradually dissociate, driving disassembly of the reconstituted AiiA and consequent loss of degradative activity.

First, for the choice of the light-controlled module, we selected VVD. Among LOV-domain proteins, VVD has the smallest molecular mass and is sensitive to blue light, undergoing rapid dimerization upon blue-light illumination. In addition, numerous VVD variants have been developed that enhance its oligomerization propensity in both intracellular and extracellular contexts.

Second, in plasmid design, drawing on the work of Sheets et al. (2020) [7] we fused VVD to the N- and C-terminus of the split AiiA fragments, respectively, with flexible linkers between domains. For convenience, the two proteins are expressed from a single promoter (Figure 4-9).

Finally, for split-site selection, our basic principles were: (1) the site should reside in a relatively flexible region so as not to disrupt the native secondary structure; (2) the split fragments should not expose hydrophobic surfaces; and (3) each fragment should preserve its structural integrity. Moreover, to achieve light-induced activity, the intrinsic self-complementation of the split fragments must be tuned to an intermediate level—namely, they should exhibit minimal spontaneous reassembly into the full-length protein without VVD, yet efficient reconstitution under VVD dimerization. To select the split sites and tune reconstitution propensity, we primarily referred to the SPELL algorithm described by Dagliyan et al.(2018) [6] and employed molecular dynamics simulations and molecular docking to evaluate local flexibility, hydrophobic/hydrophilic regions, and predicted degradative activity after reconstitution as reference metrics for candidate sites.

Build

In this section, guided by the SPELL predictions, we constructed plasmids encoding split AiiA variants with split sites at amino acid positions 154, 181, and 208.(Figure 4-10)

Similar to Cycle 1, the fragments used for plasmid construction were obtained by PCR, with the AiiA and VVD coding sequences synthesized by GenScript. The fragments were assembled with the linearized vector using Gibson Assembly to generate the final plasmids. The assembled plasmids were first transformed into Escherichia coli DH5α; single colonies were picked and verified by colony PCR and Sanger sequencing. (Figure 4-11) After confirmation, plasmids were extracted and the corresponding plasmids were transformed into E. coli BL21(DE3).

Finally, we also performed Coomassie Brilliant Blue staining to confirm that the proteins were expressed correctly.(Figure 4-12)

Test

During this cycle, we first optimized our experimental methodology and the biosensor bacteria (details can be found in the Measurement section). In the new round of experiments, the relative fluorescence unit (RFU) of the background group remained stable, unlike the decline observed over time in Cycle 1.

Under the new experimental conditions, we also used BL21 strains containing the pGEX-nmCherry-VVD-VVD-cmCherry plasmid for degradation incubation as a control. (Initially, we used BL21 without any plasmid as the control. However, we constructed the aforementioned engineered strain in Cycle 4 and considered it a more rigorous control. Therefore, we included this additional experiment during Cycle 4. The plasmid map for this engineered strain is provided in the Cycle 4 section.)

Although the response curve of the biosensor bacteria exhibited some changes—such as an extended overall response duration and the response to the AHL started earlier—the final expression intensity remained unchanged. Furthermore, across multiple experiments, the response curve of the "None" degradation group demonstrated strong consistency, indicating that the optimized experimental method offers enhanced reproducibility and yields reliable results.(Figure 4-13)

We plan to conduct degradation experiments at both 37°C and 25°C. The higher temperature of 37°C promotes AiiA enzyme activity, while 25°C is more conducive to VVD polymerization and the assembly of the split AiiA fragments. Due to the lack of specialized optogenetic equipment, we cannot precisely control the intensity of blue light illumination. For the light-exposed group, samples were incubated in a shaker equipped with blue light sources. The dark control group was incubated in a separate shaker, with EP tubes wrapped in aluminum foil to block ambient light.

Results for other cleavage sites were similar to those for site 208 and are not shown here. As can be observed, detectable activity was present at both 37°C and 25°C, which we attribute to the self-assembly of the split protein fragments. However, blue light illumination did not enhance the degradation efficiency. Furthermore, BL21 strains containing the pGEX-AiiA plasmid showed effective degradation under identical light conditions, confirming that blue light exposure does not impair AiiA activity. Experimental results for other cleavage sites consistently supported these findings.

The three cleavage sites tested showed similar degradation efficiency. We hypothesize this result stems from their comparable splitting energies (as referenced in the SPELL), leading to similar self-assembly activities and consequently similar degradation efficiencies.

Learn

Unfortunately, despite multiple repetitions of the blue light-induced degradation experiments under various blue light illumination conditions, the results consistently indicate that our design was unsuccessful. Based on the collective experimental evidence, we conclude that the failure to achieve light control is primarily due to either the inability of VVD to form dimers or the lack of a promoting effect of VVD dimerization on the reassembly of the split protein fragments.

In this cycle, our approach was somewhat rushed. For the next steps, we plan to first utilize the split mCherry to validate VVD function and identify suitable blue light illumination conditions.

Cycle 4 Light-induced Engineering and Functional Characterization of the VVD Domain

Design

Given the complexity and non-intuitive nature of the VVD-AiiA experimental procedures in Cycle 3, which hindered efficient troubleshooting, we designed Cycle 4 to independently validate the light-induced function of VVD.The overall design is similar to that of Cycle 3: we split mCherry into N-terminal (n mCherry) and C-terminal (c mCherry) fragments, with VVD domain attached to the internal end of each fragment , whereby blue light irradiation induces VVD dimerization, facilitating the reassembly of the two mCherry fragments into a functional protein that emits red fluorescence.

With the exception of the mCherry component, all other elements in the design remain consistent with those in Cycle 3. The split sites for mCherry were selected primarily based on the design described in Fan et al. (2008)[8]. Furthermore, we extended this design by constructing additional plasmids to investigate whether the failure in Cycle 3 could be attributed to elements such as the linker or RBS. These include variants with altered linkers, replacement of the RBS with a linker[9], and constructs where both VVD domains are fused to the N-terminus of split mCherry.

Build

For this cycle, we constructed the following four plasmids:

Details of the linkers used are provided below:

linker1: -GSGSGSGSGS- linker2: -GSAGSAAGSGEF- linker3: -GSGSGSGSGSGSGSGSGSGS- linker4: -GGGGSGGGGS

Following a strategy similar to that in Cycle 1, the DNA fragments required for plasmid construction were obtained by PCR. The mCherry gene fragment was sourced from the 2024 iGEM contribution kit, while the VVD gene fragment was synthesized by GenScript. The resulting fragments were assembled with a linearized vector using Gibson Assembly to generate the final plasmid constructs.

The assembled plasmids were first transformed into E. coli DH5α cells. Single colonies were selected and validated by colony PCR and DNA sequencing. After confirmation of the correct sequence, the plasmids were extracted and subsequently transformed into E. coli BL21(DE3) cells for downstream applications.

Test

The functional validation of split mCherry was carried out in three main steps: protein expression, blue light induction, and fluorescence measurement. Experiments were conducted in both liquid and solid media.

In liquid culture, protein expression was induced overnight at 25°C, followed by illumination with a blue LED array. Fluorescence intensity was measured using a microplate reader before and after light exposure.

On solid media, a method similar to diffusion assay was employed: IPTG was spotted at the center of the plate, and bacterial suspension was applied in a spiral gradient around it. After about 12 hours of incubation, the plates were exposed to blue LED light, and fluorescence was visualized before and after illumination using an inverted fluorescence microscope. All procedures were performed under light-shielding conditions, with blue light intensity gradually increased to identify optimal activation conditions.

However, under neither condition—liquid or solid media—did split mCherry produce detectable fluorescence before or after blue light illumination.

Learn

A series of Split mCherry assays revealed that our experimental design and conditions failed to effectively activate the VVD domain. This issue, which could stem from several factors, is currently under further investigation. Previous results have demonstrated that AiiA possesses intrinsic self-assembly activity. We believe that achieving light-induced degradation of AHL by AiiA is feasible, provided that the light-induced dimerization function of VVD is properly restored. However, due to time constraints, this particular challenge remains unresolved.

Light-inducible protein degradation

Cycle 1 Explore degradation conditions and effects of LOVdeg

Design

To test the blue-light-inducible protein degradation effect of LOVdeg [6], we fused LOVdeg tag to mCherry and induced the fused protein expression with IPTG.(Figure 5-1)

Because of the blue light induction, to avoid the quenching effect of blue light, we use the red fluorescent protein mCherry instead of GFP. Throughout this part, we performed experiments using low copy number plasmids pSC101 that result in moderate levels of a given protein of interest, for the reason that degradation as the sole mode of gene expression control may be limiting when proteins are at very high levels.

Build

We utilized the pJUMP27-1A (sfGFP) vector from the iGEM distribution kit to insert the LacI fragment and mCherry-LOVdeg expression circuit into it via PCR and Gibson assembly, constructing the pSC101-LacI-mCherry-LOVdeg plasmid. Concurrently, we constructed the control pSC101-LacI-mCherry plasmid. After construction, we transformed these plasmids into E. coli. (Figure 5-2, 5-3)

← Swipe left/right to view full content →

Test

We performed colony PCR and sequencing to verify successful plasmid construction. Sequencing confirmed transformation success.

With these bacteria and 1M IPTG induction, we conducted the agar plate colony experiment to determine whether they could emit red fluorescence, and to validate the LOVdeg tag-mediated degradation of the fluorescent protein.

At first, the plates in both the light and dark groups were cultured covered with aluminum foil to block light. At 14 hours, the light group was exposed to blue light for 3 hours, while the dark group continued light-blocked incubation. Imaging was performed at 18 hours to generate fluorescence-input distance curves. (Figure 5-4)

Learn

We observed a significant brightness difference between the LOVdeg-dark and LOVdeg-light group. However, no such difference was detected between the control-light and LOVdeg-light groups. One possibility is that the blue-light illumination time was insufficient. Alternatively, blue light might have arrested bacterial growth [2], so that 3 h of irradiation was not enough to let the fluorescence of the control and LOVdeg light groups diverge.

Therefore, we plan to extend the lighting time in order to further observe the degradation effect.

Re-test

Based on the previous cycle, we find an appropriate input distance and subsequently conducted agar plate colony experiments with blue light induction for 10 hours.(Figure 5-5)

Learn

It was successfully verified that LOVdeg tag has a degradation effect under the 10h induction of blue light.

We plan to replace the plasmid promoter with the QS system promoter to verify whether it can accept AHL ligands and emit fluorescence, and whether the fluorescent protein can be degraded upon exposure to blue light.

1. Scott SR, Hasty J. Quorum Sensing Communication Modules for Microbial Consortia. ACS Synth Biol. 2016 Sep 16;5(9):969-77. doi: 10.1021/acssynbio.5b00286. Epub 2016 May 19.
2. Fedorec AJH, Treloar NJ, Wen KY, Dekker L, Ong QH, Jurkeviciute G, Lyu E, Rutter JW, Zhang KJY, Rosa L, Zaikin A, Barnes CP. Emergent digital bio-computation through spatial diffusion and engineered bacteria. Nat Commun. 2024 Jun 8;15(1):4896. doi: 10.1038/s41467-024-49264-3.
3. Jonkergouw C, Savola P, Osmekhina E, van Strien J, Batys P, Linder MB. Exploration of Chemical Diversity in Intercellular Quorum Sensing Signalling Systems in Prokaryotes. Angew Chem Int Ed Engl. 2024 Jan 8;63(2):e202314469. doi: 10.1002/anie.202314469. Epub 2023 Nov 10.
4. Abramson, J., Adler, J., Dunger, J. et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630, 493–500 (2024). https://doi.org/10.1038/s41586-024-07487-w
5. Dagliyan, Onur, Andrey Krokhotin, Irem Ozkan-Dagliyan, Alexander Deiters, Channing J. Der, Klaus M. Hahn, and Nikolay V. Dokholyan. "Computational design of chemogenetic and optogenetic split proteins." Nature Communications 9, no. 1 (2018): 4042.
6. Nathan Tague, Cristian Coriano-Ortiz, Michael B Sheets, Mary J Dunlop (2024) Light-inducible protein degradation in E. coli with the LOVdeg tag eLife 12:RP87303. https://doi.org/10.7554/eLife.87303.3
7. Baloban, M., Manoilov, K.Y., Karasev, M.M. et al. Photoswitchable intein for light control of covalent protein binding and cleavage. Nat Commun 16, 8263 (2025). https://doi.org/10.1038/s41467-025-63595-9
8. Fan, J.Y., Cui, Z.Q., Wei, H.P., Zhang, Z.P., Zhou, Y.F., Wang, Y.P., Zhang, X.E. Split mCherry as a new red bimolecular fluorescence complementation system for visualizing protein-protein interactions in living cells. Biochem Biophys Res Commun 367(1), 47-53 (2008). https://doi.org/10.1016/j.bbrc.2007.12.101