Construction of producing strain

In order to enable Escherichia coli to better express purple bacillin synthesis genes[1], we performed codon optimization on these genes and synthesized purple bacillin synthesis genes suitable for expression in Escherichia coli. Using homologous recombination technology, we constructed plasmids containing five enzymes for the synthesis of purple bacillin.

Figure 1 Construction of two plasmids containing vioA, vioB, vioC, vioD, vioE. a.The plasmid containing vioA, vioB, vioD; b..The plasmid containint vioC, vioE

Two plasmids above were transferred into Escherichia coli K12 by heat shock conversion, and the results showed that the colonys turned out to be purple after 12 hours(Fig 2A). In order to understand the growth status of E. coli after expressing heterologous protein, we measured its growth curve. It was found that the high expression of these five proteins inhibited the growth of E. coli to some extent (Fig. 2B).

Figure 2 Phenotypic observation of Escherichia coli carrying five enzymes. a.The color of the strain with the five enzymes; b.The growthcurve of the strain with the five enzymes compared with the wild-type,

Semirational design of enzymes

Our team identified the conserved domain of the vioE gene family and we found at least four conserved domains on the vioE sequence derived from Purplebacterium(Fig. 3 a). The docking simulation between vioE and substrate molecules is also an important basis for us to determine the amino acid mutation site library. It can be seen from the docking details(Fig. 3 b) that ASP at positions 27 and 28 provides charge forces for substrate catalysis, and PRO at position 52 provides rigid forces for the stability of the substrate pocket. Based on the above considerations, we obtained a library of amino acid mutation sites with high transformation potential(Fig. 3 c)

Figure 3. Screening of amino acid mutation sites. a. Identification of the homologous domain of vioE from multiple species and construction of an evolutionary tree reveal the conservation of the modified site; b. Detailed diagram of the docking between vioE and the substrate, showing directly interacting amino acid residues and hydrogen bonds; c. Final determination Amino acid mutation site library.

Considering the catalytic mechanism of vioE[2], we selected two sites with high design potential from the point mutation site library: 151Asp and 187Phe. And saturated point mutation primers were designed based on the pJUMP29-1A plasmid (Fig 4 a): Mu-F: TCCGAGCACCCTGTATCTGNNKGCAGCGTCCGGCGAACC Mu-R: CGTTAACGNNKAGCGGCAAACACAGTCCGGGATCTCC

By ligating the product obtained by PCR with this pair of primers to the original plasmid, we successfully constructed a saturated mutation library at the 151Asp and 187Phe sites (Figure 4 a, b, c). We introduced the CE plasmid and ABD plasmid with the saturated mutation library into E. coli K12 (Figure 4 d). Starting from the growth of the first strain, we used time-lapse photography to continuously capture images for 7 hours, recording the previous The growth process of 42 single colonies.

Figure 4. Construction of protein mutant library. a. Plasmid map of the double-point saturation mutation library; b. Point mutation followed by amplification primers and sequences; c. Point mutation sequencing results, with red boxes marking the successfully mutated protein coding sites at position 151 and position 187; d. The plasmid of the mutant library was successfully transferred into E. coli and grown. Produce violacein.

Subsequently, we transferred it to a 1.5 ml EP tube containing 1 ml of liquid LB medium for preliminary incubation. After 5 hours of incubation, measure the OD600 of the bacterial solution. We transfer the incubated bacterial solution to a shaking tube containing 5 ml of LB medium. After 13 hours of fermentation, dissolve it with methanol and incubate it for 5 hours. We measure the purple color in the solution. Concentration of bacteriocin - OD578. Repeating this three times, we obtained the following results (Figure5 a).

Through heterogeneous ensemble learning, we constructed the enzyme activity from the enzyme molecular structure property data (including but not limited to molecular properties and energy landscape data) obtained from the two-site saturation mutation library (in this experiment, we used the crude protein of violacein. Yield represents the causal relationship diagram of enzyme activity) contribution. Then a baseline is constructed by rearranging the graph relationship, and the graph model constructed in the previous step is tested; the data is recombined based on the model that least violates the local markov condition in the second step, and multiple rounds of single Bayesian sampling was performed to generate the expectation-multifactor distribution map (Figure 5 b).

Based on this relationship, we were able to screen out the sample group with the greatest increase in enzyme activity from a larger virtual mutant library. In the predicted region (pocket size 600ű100Å or 1200±100Å, channel length 6±2Å), it is most likely to obtain mutants with higher enzyme activity (Figure 5 b), so we conducted further experimental verification and repeated it twice The measurement results are shown in Figure 5 c. The correctness of the software was confirmed by the experimental results. We obtained mutants that increased the enzyme activity by up to 32% in the predicted region (pocket size 600±100Å or 1200±100Å, channel length 6±2Å).

Figure 5. vioE semi-rational design. a. Double-point mutation library enzyme activity data; b. Bayesian sampling process of the enzyme activity prediction model; c. Larger virtual Samples screened from the mutant library were re-verified for yield.

We performed MD simulations[3] on Glu-His mutants to elucidate the molecular mechanism of increased enzyme activity. Compared with WT-ligand, the mutant significantly enlarged the substrate binding pocket (Fig.6 a,b). The larger substrate-binding pocket reduces the steric hindrance for ligand binding to the enzyme, thus facilitating the catalytic reaction. The Rg value of the mutant complex was significantly greater than the Rg value of the wild-type (WT) protein, indicating that after the substrate was docked with the mutant protein, the protein chain expanded and the enzyme activity increased relative to the WT protein (Fig.6c, d).

Calculation of the number of hydrogen bonds showed that the mutant complex formed more hydrogen bonds than the WT protein, thereby increasing the rate of the catalytic reaction (Figure 6e, f). Furthermore, we noticed an altered charge distribution in the mutant substrate binding pocket (Fig.6g, h). This results in a reduction in the number of positive charges in the active center and a shift in charge effect to neutrality. Improve the enzyme activity of vioE by eliminating the influence of electrostatic interactions.

Figure 6. Structural basis for the altered activity of the mutant. a, b. Analysis of Substrate Pocket Volume Differences; c、d. Calculating the number of hydrogen bonds; e, f. Rg calculation; g、h. Substrate pocket surface charge changes

RNA skeleton assists phase separation to improve yeild

We plan to co-localize vioABCDE with the help of specific interaction between RNA hairpin boxB and λ-peptide. To achieve this effect, we have to add λ-peptide onto each of the five enzymes, which raises the risk that the activity of enzymes may be influenced.

Firstly, we predict the structure of vioABCDE with or without λ-peptide at N-terminal with the help of Alphafold2 as shown in figure 1. Then we overlap corresponding proteins respectively to see if the structure of the enzyme itself is damaged with λ-peptide.

From the overlap results, we find out that though there’re still some changes in the structure of vioB and vioC, most domains don’t seem to shift much.

Figure 7. structure overlap of vioABCDE with and without λ-peptide

The sequence of RNA scaffold from the paper “Spatial engineering of E.coli with addressable phase-separated RNAs”[4]. To test if different number of boxB will work in different efficiency, we only retain two boxB with BcuI restriction site in between for further tests. By the way, we visualize the interaction of RNA scaffold and λ-peptide-sfGFP with Alphafold3. Figure 8 shows the interaction prediction of λ-peptide-sfGFP with a 74-nt RNA sequence with one boxB, indicating they’re jointed together by λ-peptide and a RNA hairpin. Figure 9 shows the interaction prediction of λ-peptide-sfGFP with the whole 47CAG-2boxB RNA scaffold. However, the predicted structure of whole RNA scaffold is low in credibility.

Figure 8. interaction prediction of λ-peptide-sfGFP with a one-boxB RNA sequence

Figure 9. interaction prediction of λ-peptide-sfGFP with the whole RNA scaffold

We finally added λ peptide and GS linker to the front of every enzyme, and we proved that the added λ peptide and GS linker won’t infulent their function and the strain with these enzymes that has been modified could turn to purple as expected(Fig 10).

Figure 10. Strain with modified vioA/B/C/D/E

Unfortunately, although we initially obtained a long fragment of multi-copy boxb through RCA reaction and overlap PCR(Fig.11), due to time reasons, we failed to finally construct an rna skeleton expression plasmid with multi-copy boxb so we couldn’t test the function of RNA skeleton.

Figure 11. gel image of multiple-boxB fragment yet to insert built with different methods

Metabolic modification of E. coli

Thanks to the help of modeling, we constructed a whole-genome model[4] of E. coli that introduced the heterologous synthesis pathway of violacein, and screened out the appropriate modified gene through the metabolic flux model - the gene tnaA encoding tryptophanase (responsible for l -tryptophan degradation) and sdaA involved in serine degradation (See modeling page).

We chose to use λRed recombination technology to complete the target gene replacement of chassis microorganisms (Fig. 12 a). For this purpose, we selected 500 bp upstream and downstream sequences of the target genes (tnaA and sdaA) as the homology arms of the knockout fragment, and selected FRT sequences at both ends. The KanR and CAT genes were used as resistance selection markers, and overlap PCR was used to connect the upstream and downstream homology arms and the selection genes to obtain a complete gene knockout fragment (Fig. 12 b). Then, electroporation was used to transform the gene knockout fragment into arabinose-induced E. coli K12-PKD46 electroporation competent cells. After incubation and growth, two single-gene knockout strains were finally obtained. We verified the success of the knockout by designing specific primers in the internal and external homology arm segments of the target gene (Fig. 12 c).

Figure 12. Verification of metabolic engineering and yield improvement of E. coli K12. a. Indicates the relative position of the target gene and the gene knocked out by homologous recombination; b. Indicates the PCR verification of the gene knockout E. coli colony

In order to verify the impact of this metabolic modification on the biological performance of the host bacteria, we measured the growth curve of the gene knockout strain and compared it with the wild type. We discovered that the maximum growth amount and growth rate of the modified strain decreased. After that, we verified the ability of the metabolically engineered strain to produce violacein. Finally, although the mutant's growth was somewhat affected, it can be seen that the violacein production of the △tnaA strain is significantly increased by 48% compared with the wild type, and the violacein production of the △sdaA strain is significantly increased by 23%.

Figure 13. Growth curve of metabolically engineered strain measured for yield verification. a. Growth curves of two gene knockout types; b. Determination of violacein production of two gene knockout types.

Communication module

After reading a literature[4], we found that the production of final color can be changed by modify the expression of vioC and vioE. While vioA, vioD, vioB and vioE are expressing normaly without vioC, the produced pigment will be green one(oxychromoviridans). Nevertheless, if the five genes are expressing togather, the pigment will be the purple one(violacein). Therefore, we hope to use two orthogonal quorum sensing systems to regulate the expression of vioC and vioE respectively to get different colors.

characterization of receptors sensitivity

We first constructed the promoters corresponding to RhlR[5] and nahR[6] in front of sfGFP, which are induced by C4-HSL and salicylic acid respectively. Meanwhile, we constructed a plasmid containing corresponding receptors.

Figure 14. Plasmid containing RhlR and nahR

In order to verify the induction ability of salicylic acid and C4-HSL on the promoters we used, and to avoid the situation that excessive concentration of salicylic acid would lead to poor bacterial growth in subsequent experiments, we used the fluorescence microplate to detect the fluorescence intensity under different signaling molecule concentrations. The results showed that the fluorescence intensity increased gradually with the salicylic acid concentration increasing, and then decreased because the salicylic acid concentration is too high, which inhibits the growth of bacteria. And the higher the concentration of C4-HSL, the higher the fluorescence intensity. And we found that salicylic acid promoter has certain leakage expression. (fig 15)

figure 15. RhlR and nahR sensitivity detection. a.nahR sensitivity detection. ; b.RhlR sensitivity detection

Construction and verification of producer

Since the AHL synthetase may produce different AHL molecules in different species of bacteria, we chose the C4-HSL synthetase(pagI) of Pantobacter, which is closely related to Escherichia coli. We also found out pchBA can control sal synthesis in E.coli. By genes synthesis and homologous recombination, We constructed plasmids respectively contain AHL synthetase and salicylic acid synthetase gene.

Figure 16. Two plasmids containing PagI and pchBA respectively.

The control strains containing pagI gene were fermented and extracted to obtain methanol solution containing AHL molecules, and the AHL molecular characteristics were determined by mass spectrometry. As can be seen in the figure below, heterogenous expression of pagI in Escherichia coli can produce a common characteristic peak of AHL molecules (m/z=102), and this AHL molecule peaks at 4.3min to 4.6min. As shown in Figure B, further mass spectrometry showed that the mass charge ratio of the AHL molecule in cationic mass spectrometry was 172. Therefore, it can be shown that this gene can heterogeneously produce AHL molecules in Escherichia coli, and the AHL molecules are C4-HSL and it is exactly C4-HSL molecules.(Fig. 17)

Figure 17. Mass spactrum of AHL produced by PagI. a. Graph extracted out from m/z=102.055. b. Graph showing the different m/z peaks that appear at 4.5 minutes, the pink lines represent the map of strains with pagI, and the blue lines represent the map of wild-type E. coli

Functional verification of signal transduction

In order to further verify that the signaling molecules can be accepted by the corresponding receptor strains after diffusion outside the cell, and that the two quorum sensing systems do not interfere with each other, we separately point different receptor strains under different control strains, and the results show that our two quorum sensing systems have strong orthogonality. The signaling molecules produced by the control strain can spread outside the cell and be sensed by the corresponding receptor strain, but due to the limited concentration of the generated signaling molecules, the receptor strain that is far away from the control strain will have difficulty in accepting the signaling molecules(Fig.18a,b). And we further verify that these two quorum sensing systems do not interact with each other(Fig. 18c,d).

Figure 18. Function and independence verification of two quorum sensing system

Controlled pigment synthesis

In order to control the formation of different color pigments, we modified the vioC-vioE plasmid, changing the promoter before vioC to P_RhlR and the promoter before vioE to P_nahR, and inserted a strong terminator between the expression modules of vioC and vioE to avoid the interaction between them.

Figure 19. Plasmid of vioC and vioE whose promoter have been changed to P_RhlR and P_nahR respectively

To verify the regulation of the produced pigment color by the presence of salicylic acid and C4-HSL, we transferred the plasmids with vioA, vioD, vioB, double receptor plasmid and the above plasmids into E. coli K12. They were treated with salicylic acid induction and co-induction of salicylic acid and C4-AHL by adding corresponding signal producing strains, and were fermented for 24 hours. The results showed that the strains can be induced by the control strains and can produce different color according to the the kind of signal sended by controller strains(Fig.20).

Figure 20. Bacterial fluid color induced by different control strains

Firstly, we constructed the split-cre-loxP system plasmid, this part contains pJUMP42-lac promoter-creC-promoter-loxP-TT-loxP-GFP (figure21.a)and pet15b-lacUV5-creN (figure21.b), For better comparison, we also constructed plasmids containing complete cre, pet15b-lacUV5-cre(figure18.c) and corresponding pJUMP42-promoter-loxP-TT-loxP-GFP(figure21.d).[7]

Figure 21.the plasmid design of memeory system

In our design, after co-transiting Escherichia coli K12, iptg was used to induce the expression of mitosome, and we can see it in Coomassie bright blue stain. After protein successfully induced, through colony PCR, we successfully observed that the combination of creN and creC showed activity, resulting in the same strip cutting as the positive control group, that is, the complete cre enzyme expression group, and a relatively complete editing level after a lone time induced culture, while the control group showed that no single split cre element could do this. However, we observed the weak cleavage activity in the presence of creC alone in the cutting activity of single creC components was observed in the experiment of culture for two days after one induction (figure22.b), but this phenomenon was not observed in subsequent repetitions, which may be a phenomenon of activity leakage with a low probability

Figure 22.Verification of split-cre-loxP gene editing activity only when two components are present at the same time

Further, considering the weak cleavage activity in the presence of creC alone, we tried to explore the influence of different cleavage sites on the activity after cleavage enzyme polymerization, aiming to develop a split-cre-loxP system with lower background leakage level and higher cleavage activity.We selected a total of 3 candidates through comprehensive literature reports [1][2][3]and software algorithm prediction[4], and observed that the division at 43AA had the best division efficiency and activity

Figure 23.Verification of gene editing efficiency at different division sites a. Simple information and screening basis of candidate fission sites b. Coomassie bright blue staining successfully induced the expression of various components of mitase *:creC **:creN ***:cre ****:GFP c. d. Two efficiency tests at three split sites in different repeat experiments e. cleavage efficiency test histogram

Next, we characterized split-cre-loxP. By changing the influence of different induction time and lighting time, we obtained the experimental results as shown in the figure below, which further proved its feasibility, indicating that 2-4 hours of light time may be the best time for light induction, while too long time of light induction may lead to the stagnation of cutting activity, it may be because of the inhibition of protein synthesis, the degradation of fission enzymes, and the failure of IPTG..

Figure 24.Blue light induction duration experiment

On the basic split-cre-loxP, we try to extend the orthogonality to include the orthogonality of loxP sites in order to realize the regulation of orthogonal multiplex signals

Figure 25.simply test of orthogonal loxp site

in the read in part, we select a set of toehold switches characterized by iGEM to verify its specific ability to read data. At first, we tried to apply RNA trigger externally to observe the color change of single colonies, but the expensive price of RNA synthesis and high usage prevented us from making many attempts and did not get good results. We turned to the construction of the expression of trigger RNA in bacteria and ultimately simply obtained the activation of green fluorescence, and plan to design more complete phenotypic experiments in the future

Figure 26.simply test of toehold switch .The left to right co-rotation combinations are toe1，toe1+trigger3，toe1+trigger1，toe3，toe3+trigger1， toe3+trigger3 (BBa_K2621011,BBa_K2621011+BBa_K2621016,BBa_K2621011+BBa_K2621014,BBa_K2621013,BBa_K2621013+BBa_K2621014,BBa_K2621013+BBa_K2621016)

Acquire wild type and mutated HYER

Firstly, we acquire sequence of original HYER-1 from the paper which identified it. Based on its structure analysis, we focus on 3 sites: catalytic triad, J2/3 and 2nt bulge. After literature work, we decide to make 3 mutations: G567A[10], G338A[11], 584_585del[12].

To visualize structure of mutated HYER more clearly in case it loses basic conformation, we simulate their folding with DNAMAN in comparison with original HYER(Fig.27), and assume that they fold roughly in the same way.

Figure 27. predicted structure of HYER and mutated HYER, from DNAMAN

We use point mutation kit to build mutated HYER. Since the recognition site of HYER widely exists in vectors, we don’t put a promoter upstream of HYER in the beginning, and finish point mutation at this stage. Later, the promoter is added. We choose lacUV5, for its transcription initiation site is relatively clear(6-8 bp downstream from the Pribnow box).

Figure 28. lacUV5 upstream of HYER, recognition site for transcription initiation is shown in green

To make HYER’s recognition more specific, we add 14-nt recognition site referring to the article identifying it. Probably because of its catalytic activity, we only get two mutated HYER in this way(plasmids shown below).

Figure 29. the general view of lacUV5-J2/3-RS and lacUV5-2nt-RS

Construct sfGFP readout plasmid for function characterization

To test if mutated HYER works as we want, we construct readout plasmids containing one loxp site with TRS/RS on both sides, simulating controlled sequence with recognizable sites. The bacteria containing readout plasmids appear green without potential blocking effects of HYER.

For tests of blocking function, TRS/RS sequence are added to the sequence between promoter and RBS as shown in figure 30.

 Figure 30. readout plasmid(left) and the recognized blocking part(right)

We’ve planned to characterize the blocking activity by quantifying the intensity of sfGFP. After co-transformation with mutated HYER with/without RS, we find that colonies from both control group and co-transformation group appear white instead of green, and none of the several modifications in readout plasmid work it out. This may be due to the addition of recognition sites too close to RBS, resulting in the formation of toe like structures that inhibit the translation of GFP.

Confirm transcription level with qRT-PCR

We co-transform readout plasmid and each of the two plasmids with constitutive expression of mutated HYER(J2/3 mutation and 2nt bulge mutation). After ~24h of culturing on agar plate(Amp+ Spec+ for co-transformation, Spec+ for readout only), they’re kept in -70℃ fridge temporarily and refreshed when we’re ready to test the expression of sfGFP.

For qRT-PCR, we prepare 4 groups: control group with readout plasmid only(negative control), readout plasmid with RS-inserted J2/3 mutated HYER, readout plasmid with RS-inserted 2nt bulge mutated HYER, and readout plasmid with 2nt bulge mutated HYER(conditional negative control).

For HYER expression detection, it’s clear that this RNA is expressed in all three co-transformation groups, as shown in figure 31.

Figure. 31 qRT-PCR results of HYER expression

Since there’re some accidents in readout plasmid building, and we’re out of time to use the right plasmid for formal repetition experiments to check the transcription level of sfGFP, we currently lack the data of sfGFP mRNA expression.

Construction of plasmid vectors

The plasmid we chose was pJUMP29, all of our vectors were accomplished using Gibson Assembly.

Figure32 .The plasmid for biosafety

Illumination experiments

We chose KillerRed that had been used in the iGEM Parts ( BBa_K4628047) to commit suicide. To test whether the killerre system is still applicable in E. coli, we did a simple and interesting light-induced experiment.

Under the same bacterial concentration, we performed white light irradiation on four groups with the same bacterial patterns for different durations. It can be seen that there is a clear trend of bacterial clearance after 3 hours of irradiation.

Figure.33  scavenging effect of KillerRed. From left to right are the dark control group, the 1-hour light exposure group, the 2-hour light exposure group, and the 3-hour light exposure group.