BL21 is commonly used for protein expression and characterization. Typically, when we engineer an unfamiliar protein, the initial procedure is conducted in BL21, followed by transferring the results to the final microbe we intend to use, such as EcN in our design. However, for the following reasons, we decided to engineer within EcN directly:
- Limited time: BL21 and EcN have the Csg gene in their genomes. Knocking out both genes would consume too much time for us.
- Human Practices from Neel Joshi: At the onset of our project, we contacted Professor Joshi via email, seeking advice on engineering CsgA. It was suggested that, as a protein, CsgA is difficult to purify and characterize. We should focus more on the design rather than the protein itself.
- The Design of our project: Our project aims to display tetrazine on the surface of EcN. Therefore, characterizing CsgA on EcN is more straightforward and relevant to our goals.
We used the way of CRISPR-Cas9 to delete the gene. Firstly, we transfer plasmid pHCY-25A, and then newly constructed plasmid pgRNA-Csg-1 (Figure 2), which we inserted the selected gRNA (Figure 1) and N20 sequence.(Reference: https://chopchop.cbu.uib.no/#).
By inducting and controlling of culturing temperature, we successfully acquired EcNΔCsg with no plasmid inside. (Figure 3) The results were verified by colony PCR and growth on multiple plates with different antibiotics.
Curli fibers serve as an excellent bacterial surface display platform. Praveschotinunt, Pichet, et al.1 inserted p-azido-l-phenylalanine (pAzF) into CsgA, leveraging a copper-free click reaction between the azide group and dibenzocyclooctyl (DBCO) to enable the tracking of engineered bacteria in vivo. However, such click reaction isn't strictly bioorthogonal. DBCO would react with the sulfhydryl group, which means off-target.
Compared to the former reaction, the click reaction between tetrazine and trans-cyclooctenes (TCO) displays a faster reaction rate and has less problem in off-target. Group Mehl developed several unnatural amino acids with tetrazine, including Tet v1.0 to 4.0. Among the four, Tet v2.02 was the most used and developed amino acid, so we chose it as our unnatural amino acid. We synthesized Tet v2.0 through a 2-step route (Figure 5) and verified through mass spectrometry (Figure 6).
We tested the expression of protein with Tet v2.0 insertion with GFP-151TAG (which is, changing site 151 of GFP into TAG so that we can insert unnatural amino acids in it). We transferred pUltra-Ambrx (Figure 7), which has an RS-tRNA pair to insert Tet v2.0, and a plasmid carrying GFP151TAG to BL21, and purified the protein. The tetrazine would quench the fluorescence of GFP, thus making it yellow rather than green (Figure 8). This could verify that our system of Tet v2.0 could insert the amino acid into a protein.
1. Pre-selection of the Insertion Site
One of the most important issues of our design is choosing the insertion site of Tet v2.0. Better insertion efficiency and displaying of tetrazine could improve the effectiveness of our design.
According to previous work, S89TAG (which is, changing Serine at site 89 into amber codon where unnatural amino acid inserts) has been reported as one possible option. However, according to our investigation, S89TAG is at the end of the second β sheet, which may affect the structure of CsgA, and it's not the perfect place to display tetrazine for reaction.
The insertion site selection may largely contribute to the effectiveness of our design. The wrong site may lead to misfolding of protein, and reduced reaction rate. On such issues, we communicate with both our PI and Dry lab, hoping to get a better site. Here's some of the advice that we got:
Dry lab: they simulate the insertion by displaying Tet v2.0 at different sites of the protein. As a result, F97TAG is recommended to be a better option, since firstly, it's on the loop of CsgA; secondly, the structure of phenylalanine is similar to Tet v2.0, so Tet v2.0 wouldn't affect the protein structure; Thirdly, F97TAG performs better in reaction rate simulation.
For more information, please refer to dry lab.As the result, we chose S89TAG and F97TAG as our initial selection of Tet v2.0 insertion.
2. Generation of Curli Fiber
The first question we meet is,will the insertion of Tet v2.0 affect the expression rate of CsgA, therefore affecting the formation of Curli Fiber. We investigate such a problem by using Congo Red. It's what people usually use to quantify the formation of the curli fiber. By adding Congo Red after induction and expression of EcN, we found that both S89TAG and F97TAG can express the fiber with high efficiency, even higher than WT. (Figure 9)
The problem of the experiment is that the CR binding of CsgAWT didn't display a positive control, after several rounds of experiment. Referring to previous works, they didn't either have a significant positive control result.
Another strange thing that bothers us is that, without adding Tet v2.0, after induction, the bacteria can still exhibit a high efficiency in curli fiber formation. It was later verified by scanning electron microscope (Figure 10), which we did the same treatment to the bacteria, with or without adding Tet v2.0 into the expressing system.
After discussion, we thought it might be because of the expression of truncated protein. Although the expression stops at the amber codon, there are still two β sheets left, which might form the structure of amyloid protein, and generate curli fiber. We inquired our PI about our guesses, and it was recommended that the insertion site be placed at the beginning of the protein. Under such conditions, the protein stops its initial expression and won't form curli fiber. So, we add Y48TAG and Y50TAG into our Congo Red assay (Figure 11).
The change of insertion site doesn't significantly change the result. This proves that our previous hypothesis might be wrong, and the formation of curli Fiber without Tet v2.0 might be due to the insertion of other natural amino acids like Tyrosine or Phenylalanine. This is because the RS we used won't just insert Tet v2.0, it may also have a false positive which inserts amino acids with similar structures. However, we stopped our investigation here and paid more attention to other aspects of our design. We assumed that CR assay was a reference rather than a concrete characterization of curli generation. In the aspect of curli fiber generation, S89TAG and F97TAG perform equally.
3. Display of Tetrazine
The next question we want to find out is the display efficiency of tetrazine on the surface of CsgA. We verify such display efficiency through several aspects:
3.1. Protein Level
To verify the insertion of Tetv2.0 into CsgA at the protein level, we attempted to purify CsgAS89TetV2.0-His and CsgAF97Tetv2.0-His with previously reported methods. Indeed, we successfully purified the Tetv2.0-incorporated CsgA variants, which we applied TCO-Cy5 for labeling, observing that only the Tetv2.0-incorporated CsgA proteins exhibited a strong fluorescent signal (Figure 12). Furthermore, CsgAS89TetV2.0-His and CsgAF97TetV2.0-His were validated by protein molecular weight mass spectrometry (Figure 13). Collectively, these results confirm the successful expression and purification of Tetv2.0-incorporated CsgA proteins.
3.2. Bacteria Level
On the Bacteria Level, we verify the displaying of tetrazine through three aspects: TCO-Cy5 assay, Flow Cytometry, and Confocal Microscopy Through Focusing.
For the TCO-Cy5 assay, we wanted to preliminarily find out the capacity of the overall tetrazine display. We labeled the bacteria with TCO-Cy5 for 5 hours after overnight induction and then washed them with PBS through centrifuging. S89TAG demonstrates a higher display of tetrazine on its surface after OD600 normalization. (Figure 14) Precisely, after comparison with the fluorescence generated by Cy5 itself, for OD600=1 of the bacteria, S89TAG displayed about 0.13μM Cy5 in average, while F97TAG displayed about 0.11μM Cy5 in average. It could also be directly observed through the color of centrifuged labeled bacteria. (Figure 15) However, this is the general effect of protein expression and the insertion and display of Tet v2.0. We couldn't identify which was the main factor, therefore we continued our experiment on single cell dimension.
The next experiment we conducted was Flow Cytometry so that we could compare S89TAG and F97TAG in single-cell dimensions. We first compared the percentage of TCO-Cy5 labeled cells. (Figure 16) S89TAG and F97TAG displayed a similar ratio of about 80%, much higher than that of CsgAWT which is about 10%.
However, it's the overall labeled ratio of bacteria, ignoring the fact that some bacteria are dead. Thus, we continued the experiment, identifying live/dead through DMAO/PI staining. In brief, DMAO could pass all cell membranes, while PI could only pass damaged cell membranes from dead bacteria, so live cells were only labeled by DMAO. Among all, about 36% of bacteria are alive. After analyzing 10,000 living cells, about 65% of F97TAG were detected to be labeled by TCO-Cy5, and was slightly better than the 63% of S89TAG (Figure 18), indicating that tetrazine on F97TAG may have a better displaying.
For confocal experiment, we also used TCO-Cy5, DMAO and PI to stain our engineered bacteria. We successfully obtained the fluorescent co-localization signal of DMAO and TCO-Cy5 in CsgAS89TAG and CsgAF97TAG, while there was no Cy5 signal in CsgAWT either under the same expression condition or without induction(Figure 18), which confirmed that Tetv2.0 was site-specifically inserted into CsgA and well displayed on bacterial surface. Moreover, from our fluorescence co-localization results, CsgAF97TAG performed better than CsgAS89TAG, no matter considering corresponding overall fluorescence intensity or the co-localization performance.
4. Influence of Tet v2.0 to Bacterial Growth
Safety is one of the most important issues when associated with humans. It's especially important in bacterial therapy, since the proliferation of bacteria in vivo may affect human health.
In our project, a total of 3 things may have safety issue: EcN, Tet v2.0, and prodrug (TCO-DOX). We initially thought it wasn't a big deal until experts and other iGEMers from Human Practices continued to express their worries about safety. For example, Professor Zhuang Liu expressed his worries about the toxicity of Tet v2.0, both to bacteria and cells. Thus, we decided to pay a bit of attention to safety issues, even though we still focus more on the effectiveness of our design.
For more information, please refer to Human Practice for the conversation with Zhuang Liu, and what we design to guarantee safety.As experiments progressed, we found that the growth of S89TAG and F97TAG was much worse than CsgAWT, though the expression of CsgA was all under the control of the Arabinose promotor. Three reasons might explain such phenomenon: the toxicity of Tet v2.0; the expression pressure of protein; and the toxicity of truncated CsgA expression.
To find out the main cause, we allowed bacteria to grow under various conditions at 37°C within a microplate reader. (Figure 18) The result is that Tet v2.0 wouldn't affect the growth; after induction, S89TAG and F97TAG both with or without Tet v2.0 grew in a much worse condition, while CsgAWT didn't have such a phenomenon. It might be concluded that the truncated CsgA expression might largely contribute to the reduced growth of EcN. Such a result may partially explain the safety issue of Tet v2.0, and it also gave us a hint about the purification of CsgA.
5. Final Selection of Insertion Site
According to previous experiments (and the next experiment of characterization of De-cage), we had three conclusions between the two sites:
- Both S89TAG and F97TAG wouldn't affect the generation of curli fiber.
- S89TAG expressed more CsgA with Tet v2.0 inserted.
- F97 displayed tetrazine better than S89TAG
Because of the need of capability of reaction, we finally selected F97TAG as our ultimate site to incorporate Tet v2.0.
Before TCO-DOX, we wanted to check the de-caging capabilities and properties of the bacteria. The detection of DOX was comparatively difficult, so after investigation, we caged the fluorescence of coumarin with TCO3, though later experiments showed the fluorescence wouldn't be caged completely. We synthesized TCO-Coumarin through a two-step reaction (Figure 19), and was verified through mass spectrometry (Figure 20):
We tested TCO-Coumarin first on molecular dimension, which is, de-caging through tetrazine rather than protein or bacteria. By adding the Tet v2.0 and TCO-Coumarin (Figure 21), we preliminarily assumed the adequate de-caging time was about 30 minutes.. Later, the result was collected and analyzed by dry lab, to model the process of reaction so that we could better control the reaction. (See more: please refer to dry lab) However, some problems still occurred in such experiments because of the limited time:
- A.We didn't precisely weigh the TCO-Coumarin we got, since it's only several milligrams, so we didn't get the precise de-caging percentage. If time's available, we will re-do the experiment to get the percentage, so that we can decide the best dosage of our design.
- When communicating with the dry lab, they expected a much more complex system to simulate the effect of the environment on the de-caging reaction. For example, we are recommended to experiment with culture medium such as DMEM or RPMI 1640. However, the TCO-Coumarin was placed at -20°C for too long, and such fluorescence detection could be disturbed by the culture medium, we might think of a better way to design the experiment.
After verifying the effect in molecular dimension, we continued our experiment on bacterial dimension. By adding TCO-Coumarin to PBS-washed bacteria and culturing for a certain time, we tested the fluorescence from the supernatant after centrifugation and OD600 normalization. (Figure 23) S89TAG and F97TAG all displayed a similar de-caging effect, while S89TAG performed a faster reaction rate, and F97TAG exhibited a better reaction percentage at 30 minutes.The problem with such an experiment is that Coumarin might also be centrifuged under the bacteria. This result partially supported our selection of F97TAG rather than S89TAG.
After verification of the de-caging effect of TCO-Coumarin, we came to the final stage of testing our design in the dimension of cells. In the selection of prodrug and cell line, we decided on both investigations and Human Practices. Finally, we set doxorubicin4 (DOX) as our prodrug to examine cell cytotoxicity.
After selecting colorectal cancer as our target, we reached out to Doctor Qi as a Human Practices, to find out the problems of our design.
Initially, we set MMAE as our drug, since it's a widely used chemotherapeutic that has already been used in fields of ADCs. However, Doctor Qi stated that for colorectal cancer, MMAE might not be a preferred option, since its effectiveness is relatively poor to colorectal cancer, he suggested doxorubicin and Irinotecan might be a better option. Doxorubicin already has references to be caged by TCO, thus we set DOX as our first candidate. It is also recommended to find a site to cage Irinotecan by TCO. We told our findings to the dry lab. However, they haven't yet found a place to cage by TCO.
For more information, please refer to Human Practices Or take a look at other details about safety.We synthesized TCO-DOX through a two-step reaction (Figure 24) and verified the molecular through mass spectrometry (Figure 25).
After obtaining TCO-DOX, we first tested it on the molecular dimension. After mixing TCO-DOX and overdose Tet v2.0, we analyzed the de-caging properties by using HPLC (Figure 26) TCO-DOX rapidly de-cage within 30 minutes, which is compatible with the result we had before on TCO-Coumarin. However, the experiment still had several drawbacks. The standard of doxorubicin hydrochloride we used could not produce a peak with a specific retention time in HPLC. So, we couldn't identify which was the de-caged DOX and how much we got. Thus, we set one peak as DOX, and set the peak area at 125 minutes as 100%, to show its de-caging property. This could not show either the de-caging percentage.
We first tested the cell toxicity in dimension of molecular. We chose CT26 as our model of colorectal cancer. It was murine colorectal cancer cell line that we bought from Pricella. We didn't select human colorectal cancer cell line such as 4T1, because if possible, it was more convenient for us to later test our design in mouse model.
For each of the group, we added different concentrations of DOX, TCO-DOX and TCO-DOX de-cage with Tet v2.0 (Figure 28). DOX exhibited an IC50 of 0.26μM, while TCO-DOX exhibited an IC50 of 5.4μM, which was compatible with the fact that TCO caging the toxicity of DOX. After adding Tet v2.0 to each system, the cell toxicity significantly increased, the IC50 rose to concentration of 1.2μM, verifying the de-caging effect caused by Tet v2.0. We previously proved that our bacteria could display a large amount of tetrazine on the curli fiber, we are quite confident that our design could release prodrug efficiently and kill tumor cells.
We previously proved that our bacteria could display a large amount of tetrazine on the protein. So, we incubated 1:1 F97TAG with 12μM TCO-DOX and compared the cytotoxicity assay with F97TAG and the prodrug TCO-DOX itself (Figure 28). Group F97TAG with 12μM TCO-DOX showed the lowest survival rate. This indicates the successful de-caging effect by F97TAG. This gives us confidence that our design will work in vivo.
Due to the needs of molecular biology experiments in the laboratory, we require a promoter with high expression efficiency, such as the arabinose promoter. However, for subsequent animal experiments or real clinical use, we will choose a more suitable promoter for the hypoxic environment of the middle and low rectum, that is, the hypoxia-sensitive promoter—the JWW promoter.
To qualitatively and quantitatively analyze the differences between the two promoters, we designed two plasmids and introduced them into the EcN strain, using the GFP expression of bacteria under induced conditions to characterize the efficiency of the promoters.
The results are shown in the figure below, with the ratio of fluorescence value to OD600 as the Y-axis for bar graph analysis and comparison.
Using the same culture and treatment methods, but employing different induction conditions: arabinose induction or hypoxic environment induction, and reflecting promoter efficiency by measuring the ratio of fluorescence value to OD600.
We found that: compared to the arabinose promoter, the hypoxic promoter has a very low expression efficiency, which is difficult to meet our expectations for the expression quantity of the target protein by the engineered bacteria. In response to this, we have designed a new expression system, introducing the T7 promoter system, using the hypoxic promoter to express T7 polymerase, and then using the "cascade reaction" to induce GFP production. The plasmid design is as follows.
In six months of experiments, we started our experiments from constructing the bacteria, to examine the generation of curli fiber, and then the insertion of Tet v2.0 into CsgA, and finally examining the effect of our design in tumor cells. We found the comparatively better insertion site, S89TAG, and proved its effectiveness. TCO-DOX is the prodrug we use, and we examined its efficacy and properties in tumor cell line. For the next step, we plan to furthermore improve our design, finding better prodrugs, and conduct animal experiments. For now, we are quite confident that we can stand out from other bacterial therapies in treating rectal cancer.
1. Praveschotinunt, Pichet et al. “Tracking of Engineered Bacteria In Vivo Using Nonstandard Amino Acid Incorporation.” ACS synthetic biology vol. 7,6 (2018): 1640-1650.
2. Blizzard, Robert J et al. “Ideal Bioorthogonal Reactions Using A Site-Specifically Encoded Tetrazine Amino Acid.” Journal of the American Chemical Society vol. 137,32 (2015): 10044-7.
3. Fan, Xinyuan et al. “Optimized Tetrazine Derivatives for Rapid Bioorthogonal Decaging in Living Cells.” Angewandte Chemie (International ed. in English) vol. 55,45 (2016): 14046-14050.
4. Versteegen, Ron M et al. “Click to release: instantaneous DOXorubicin elimination upon tetrazine ligation.” Angewandte Chemie (International ed. in English) vol. 52,52 (2013): 14112-6.