Acoustic detection

The imaging of microorganisms is critical to cancer foci locating and diagnosis in our project. In order to make our microorganism 'visible' under ultrasound, acoustic reporter genes(args) on pET28a vector(kind gifts from team RDFZ_China) was transferred to E.coli BL21(DE3)(Fig.1A). The bacteria were then induced by 0.5mM IPTG at 30℃ for 40hr. Gas vesicles consist of all-protein shells were formed during ARGs expression, for which they can be detected by ultrasound. The initial ultrasound imaging experiments were conducted in vitro , by mixing induced bacterial with molten 1% agarose at 50 ℃. Two sets of ARGs , ARG1 and ARG2, were used for detection.

Our experiments show that ultrasound signals from E.coli-expressed ARGs can be detected(Fig.1B). However there is a possibility that the signals were caused by other factors, so the secondary imaging experiments focused on gas vesicles collapsing and determining density gradient. Samples were prepared by similar procedures, the concentration of agarose was decreased to make collapse easier to occur. The power of the machine was turned up to 100% and sustained for 5 min to collapse gas vesicles(Fig.1C).

All of the experiments indicate that ECHO can be detected by ultrasonic methods, and signals from ECHO can be distinguished from background signals (such as body tissues) by collapsing(Fig.2).


The expression of gas vesicles in bacteria and the detection of ECHO

The collapse of signals from ECHO

Cancer targeting

When things came to cancer targeting, we turn to a kind of tumor-targeting oligo peptides. Previous research shows that these three peptides (T pep) consist of 15 amino acid residues can combine with the T-antigen that exists on the surface of colorectal cancer cells.

However, we have to find a way to bring the oligo peptides to the surface of the bacteria for effective attachment. Two kinds of outer membrane protein, OmpA and INP are chosen as the peptides carriers. The INP-peptide fusion protein and OmpA fusion protein was constructed into pCDFDuet-1 plasmid, functioning as a peptide surface display system. The system was transformed into BL21(DE3) together with eGFP as the reporter gene.

A series of experiments were performed to determine whether the cell surface display system correctly transports peptides to outer membrane of the bacterium, as well as distinguishing if the peptides are effective to build a cell-bacteria junction. We used cell-climbing cover glasses as the solid basement on which the cell-bacteria adhesion experiment was carried out. Cell line HT29(human colorectal cancer cell) was chosen for the experiment and was grew over the glasses. These cover glasses were carefully observed via the fluorescent microscope after a period of incubation with the bacteria, thus the adhesive properties between cell and bacteria could be determined (fig 4.).

Before the experiments, we certificate the HT29 expression of T antigen by FACS. A specified antibody that binds with T antigen was applied to detect this biomarker on HT29’s surface. We used FACS to analysis the antigen-antibody reaction. Those antibodies -- Peanut agglutinin(PNA) were marked by green fluorescence, called Fluorescein isothiocyanate -labeled peanut agglutin (FITC-PNA). They were mixed with both HT29 and human immortalized keratinocyte (HACAT) cell suspension in order to investigate whether there are T antigens on the HT29 surface, and the concentration of the antibody was 10ug/mL. The mixture was incubated under 37℃ for 0.5h, and was directly loaded to flow cytometer (fig 3.). In this step FITC-PNA gave the only fluorescent signal, the output that was high in green fluorescent intensity indicates that there were great amount of T antigens expressed on the cell surface. The result showed that there is a distinguishing difference between the adhesion rate of PNA-HT29 and PNA-HACAT(Fig.3). The antibody showed a higher tendency to bind to HT29 cells, proved by the 35.5% positive rate compared with 2.56% for the HACAT cell. Generally speaking, the results certificated the existence of T antigen.

After IPTG induction, the bacteria would express INP-peptide fusion protein or OmpA-peptide fusion protein, and is ready for properties examination. We first build the co-incubation system by culturing HT29 cells on the cell climbing cover glasses in a 24-well plate. We then added suspension of bacteria with or without peptides into the well after HT29 had covered 70%~80% of the cover glasses, and placed on a shaking table for incubation. Controls were applied during our incubation-in-well experiment as well. We used two kinds of bacteria, one expressed the INP-tPep surface display system while the other one didn’t. Both kinds of E.coli were marked by EGFP, and they should perform different efficiency of cell adhesion. We have tried sets of temperatures and time conditions during the experiments. We first tried incubating at 4℃ overnight but then we realize under this incubation condtion it is easy for cells to detach from the cover glass during washing procedure, since the retraction of cell pseudopodium after the low temperature treatment. Then we tried 37℃ incubation with different incubating time. Finally we came to the conclusion that the best incubating temperature is 37℃ with the incubation last for 0.5h(Fig.4A). This condition is just adequate for the cell-bacteria adhesion, any incubation longer than 0.5h would lead to the increase of the non-specific adhesion inferred from the enlargement of the number of peptide-free bacteria adhesion in the control group. After the appropriate condition was solidified, we further tested the function of OMPA surface display system with peptide-1 and peptide-2 using the same procedure.

After the incubation, we loaded the cover glasses to the microscope slides and added 4% PFA and DAPI for fixation and staining(Fig.4A). The slides were later observed under a fluorescent microscope, and the cell-binding ability of the bacteria could be revealed intuitively by the pictures. The pictures were composed of shots from Alex Fluorescence 488 and DAPI channels, and the result shows that, under the condition of roughly equivalent cell amounts, bacteria with peptide expressed outnumbered the control bacteria in terms of the adhesion to cells(Fig.4B,4C). All three kinds of peptides showed the same result but differed only in the improvement degrees. .

Being Expecting to demonstrate the effect of the surface display system more accurately and quantitively,we’ve also applied FACS to HT29 cell suspension incubated with peptide-expressing BL21. In this case, BL21 with peptide displayed on the outer membrane should attach to the HT29 cells in the suspension. The reaction system was treated with 7AAD by which the dead cells can be differentiated from living ones in a low dye concentration after the co-incubation (Zembruski et al. 2012). And since BL21 we used expressed EGFP, it could be easily detected.

To identify the positive signal, that is, the bacteria adhered to the HT29 cell, from the cell-bacteria suspension, we decided to screen out the free BL21 before we test the fluorescent intensity. The signal that shows a similar scale of cells as well as strong fluorescence intensity is inferred as the positive signal after free bacteria are excluded. In order to draw a statistically meaningful conclusion, each group had four parallel repeats. As the results shown, positive rate in INP-tPep + EGFP (BL21 with INP-tPep peptide carrier and EGFP expressed), is 46.4% in average, being substantially higher than 11.9% in average of EGFP(BL21 with only EGFP expressed) group(Fig.5).

In conclusion, we have certified the effectiveness of INP-tPep as the joint component of our E.coli detector.

Micro-fluidic chip application in the test of cell-bacteria adhesion :

Micro-fluidic Chip (MFC) was also introduced to our cell-bacteria adhesion experiments to a further understanding of the effectiveness of surface displayed peptides. This part of experiments was implemented under the guide of a doctoral student (seen in the Attribution). In this experiment, we used a grate-like MFC to screen cells and free bacteria from their suspension. The grate was constructed in an ‘U’-like pocket which receives suspension from the open end. With liquid flowing from the open end to the bottom end, the chip grated the suspended particles by their scale and discharges all particles whose diameter is smaller than the gate space of the pocket. Treated by this procedure, the cells could be separated from bacteria except for those binding to the cell by peptides and T antigens. Thus, we could observe this process under a fluorescent microscope (Fig.6). To carry out the experiment, we also made some cell suspension co-incubated with INP-tPep E.coli marked by EGFP under the condition of 37℃ and 0.5h on shacking table. 200ul of the suspension was later injected into the chip by a specified pump. The gate space of this chip was 10um, while the average diameter of our HT29 cell is roughly 13um. The result showed that, most cells remained in the pocket, and correspondingly the E.coli whose diameter was no more than 5um was almost totally pump out of the grate. The only EGFP signals we could detect were those who bond to the cell. We further recorded a video of how the cell moved with the E.coli attached. The video is available below.

As the results showed,We hold the belief that our surface display system worked and E.coli has the ability to attach to colorectal cancer cells by a mechanism similar to antigen-antibody binding reaction.


Construct the surface display system. The system we constructed is useful for other researchers whoever willing to make bacteria target cancer cells. We believe it is helpful and of importance in cancer research and bacterial application in cancer. The surface display system can also be used for other purpose.

Display short peptides on the surface of the bacteria and achieve ‘ECHO’-cancer attachment.

Mice experiments

Based on solid validations in vitro, we conducted animal experiments to further confirm the in vivo performance of engineered E.coli in mice colon.

Characterization of AOM/DSS-induced CRC in mice:

The azoxymethane (AOM) and dextran sodium sulfate (DSS) induced colitis murine model is commonly used to study carcinogenesis and of colorectal cancer [1][2], and to test the specificity and sensitivity of diagnostic tools, such as metagenome analysis based CRC risk screening [3]. C57BL/6 mice were divided into groups of three to four per time point and injected with 1mg/kg AOM, followed by 2-2.5% (w/v) DSS in drinking water for 7 days, with 14 days for recovery, and repeat for 3 cycles in modeling group (Fig.7A). The endpoints of 50 days, 60 days and 70 days were chosen for morphological observation and animal experiments. Clinical symptoms were observed during the progress of carcinogenesis, such as the loss of body weight, watery diarrhea and fecal blood, consistent with the previous report. Body weight loss was observed from day 9 in C57BL/6 mice (Fig.7B), with a decrease of 6±0.2% weight loss per day in the first DSS administration cycle, and turned to mild variation in the next two cycles, which was consistent with previous studies (). Multiple polypoid masses, and occult bleeding of colon could be observed in dissected CRC modeling mice, compared to control (Fig.7C). For gross examination, 2-3 polyps around 2cm diameters could be observed from day 50 mice (Fig.7D). Correspondingly, signs of surface epithelial regeneration, moderate infiltration of inflammatory cells to the mucosa, unusual distribution of adenocarcinomatous glands pronounced the clinical adenocarcinoma status of mice (Fig.7E) moderately differentiated.

Characterization of T antigen in CRC tissue

Recently, molecular mechanism laying behind the binding behavior of specific bacteria to CRC, such as Fusobacterium nucleatum was demonstrated [Ref04]. In which Thomsen-Friedenreich (TF) antigen (Galβ1, 3GalNacα-O-Ser/Thr) was characterized as typical features expressed on colorectal adenocarcinoma[Ref05-06], giving us the clue to design corresponding structure that could bind to this antigen. We firstly confirmed the expression of TF antigen on colorectums of CRC modeling mice. Fluorescein isothiocyanate (FITC)-labeled peanut agglutin (PNA), a Gal-GalNAc [Gal-β(1-3) GalNAc] lectin was specifically bound to TF antigen. We assessed the level of TF in tissue by applying this label onto colonic mice colorectum and control mice colorectum, followed with flow cytometry analysis and fluorescence microscope observation. FITC fluorescence signals were significantly higher in adenocarcinomas compared with control tissue, with PNA-FITC positive ratio of 29.3%±6.2% in CRC and 1.03%±0.36% in control, p=0.045 (Fig.8B and D). During the progression of colorectal carcinogenesis, both of day 50 and day 70 dissected CRC mice exhibited an increase of TF antigen in colorectal tissue (P<0.05) (Fig.8C). Interestingly, we observed a trend of decrease in Day 70 mice, suggesting a higher possibility of detecting early colorectal carcinoma through peptide-TF antigen binding.

Specific binding of tPep expressed E.coli to tumor tissue

To confirm that engineered E.coli attachment to CRC is peptide-TF antigen binding mediated, we applied ice nucleation protein (INP) expressed EGFP E.coli and EGFP E.coli to CRC tissue and normal mice (control) tissue, co-cultivated for 30 min at 37℃, and embedded those tissues in frozen medium, made to frozen slice, and sealed for microscopy observation (Fig.9A). Obviously, on-tumor sites of CRC tissue, which exhibited abnormal hyperplasia in H&E stain slice and higher signal of T antigen (Fig.8F), bound higher amount of INP-tPep and EGFP co-expressed E.coli (47%±3%), compared to off-tumor site (11.5%±1.5%), p<0.05 (Fig.9 B and C). Besides, an increase of INP-tPep and EGFP co-expressed E.coli was observed in the apical colon, when comparing CRC tissue with the control, illustrating the binding specificity of T peptide to T antigen. Notably, certain signals (highlighted in yellow triangles) could be observed in the basolateral side of colon, which were recognized as background because we applied E.coli onto the intact colon which prevented the in-depth invasion of bacteria to colon (Fig.9E). Therefore, we chose the apical side of the colon as the region of interest (ROI), to calculate the number of attached E.coli (Fig.9D). The overall number of E.coli was higher in INP-tPep and EGFP co-expressed E.coli treated CRC compared to that treated control, with significantly difference (p<0.0001). Besides, the number of E.coli was higher in INP-tPep and EGFP co-expressed E.coli treated CRC compared to EGFP expressed E.coli treated CRC (p<0.001).

Ultrasonic imaging of bound engineered E.coli in colon

Before observing the gas vesicle signal in colon, we firstly built and testified the robustness of our detection system. We positioned the mice colon in 0.2% agarose gel, filled with corresponding bacteria that solidified in the 0.2% agarose, to facilitate the non-invasive ultrasonic observation of the organ. We collected the ultrasonic video and summed each frame in one figure, calculated the gray value of the image based on representative ROI. Signals of gas vesicle ARG expressed E.coli could be observed in mice colon both in transverse view (18.73±5.87) and longitudinal view (40.24±8.50), compared to nonengineered ones, p<0.05 (8.09±5.81 and 7.62±1.72 respectively) (Fig.10A, B and C).


tPep expressed E.coli successfully bind to CRC tumor tissues thus realize cancer targeting of "ECHO".

Successfully detect the signals from ‘ECHO’ by ultrasound, meaning that "ECHO" is effective in CRC diagnosis.

Use AOM/DSS to induce CRC in mice successfully.

NO sensing

Given that higher NO level is a marker of inflammation and early cancer, we decided to use promoters sensing NO to control the transcription of azurin gene(BBa_K250001) which induce apoptosis in mammalian cells. PyeaR (from BBa_K381001)and PnorV was chosen at last. Different amounts of Sodium Nitroprusside(produce NO) or nitrate(turn to NO in bacteria) were added into E.coli DH5α cultures that had been transferred into BBa_K381001(GFP as reporter gene)to test whether PyeaR is sensitive to NO. The results are shown below.

So it is believed that PyeaR can function as a NO sensor. We are terribly sorry that the results of PnorV activity at different NO concentration is missing. Further, we constructed PyeaR-azurin-histag and PnorV-azurin-histag-norR(the gene coding repressor protein of PnorV) into pCDFDuet-1 plasmid. The expression of azurin in both case after SNP induction was confirmed by Western Blotting. The azurin gene maybe changed into other genes that have the same function.


Construct PyeaR/PnorV azurin circuit into pCDFDuet-1 (so the surface display system and the treatment circuit are on the same vector.)

Realize NO sensing and the NO-controlled expression of azurin.

Bacteria killing

Azurin must be released from bacteria cells to function, as well as biosafety should be taken into consideration to prevent ‘ECHO’ from releasing into the environment. Thus, bacterial lysis in vivo should be well controlled in a harmless and easy way. Fortunately, arabinose is a widely used, and more importantly, it is edible and won’t be digested and absorbed by human digestive system, which make arabinose an ideal inducer. Thus, bacteria lysis can be controlled by taking arabinose orally. We chose to use gene ΦX174E(BBa_K2500006) for cell lysis with an arabinose induced promoter PBAD before it to control bacterial lysis. Considering ΦX174E gene and azurin were used previously by other researchers, we have faith in these genes and their production.


The controlled lysis of ‘ECHO’, making it more effective and safe.

Further explorations: We measured the T2 relaxation of ARG E.coli, control blank bac., and agarose gel using magnetic resonance imaging. The Outcome showed a significantly different T2 relaxation curve between ARG E.coli and blank bac. with p-value < 0.005. The T2 relxation time of agarose gel, blank bac., ARG E.coli, are around 140ms, 125ms, 120ms, respectively. It is indicated that the presence of gas vasicles within ARG E.coli produced a T2 time drop of 5ms compared with that of blank bac.


[1] Neufert, C., Becker, C., & Neurath, M. F. (2007). An inducible mouse model of colon carcinogenesis for the analysis of sporadic and inflammation-driven tumor progression. Nature Protocols, 2, 1998–2004.

[2] Suzuki, R., Kohno, H., Sugie, S., & Tanaka, T. (2004). Sequential observations on the occurrence of preneoplastic and neoplastic lesions in mouse colon treated with azoxymethane and dextran sodium sulfate. Cancer science, 95(9), 721-727.

[3] Sakuma, S., Yu, J. Y., Quang, T., Hiwatari, K. I., Kumagai, H., Kao, S., ... & Kitamura, T. (2015). Fluorescence‐based endoscopic imaging of T homsen–F riedenreich antigen to improve early detection of colorectal cancer. International journal of cancer, 136(5), 1095-1103.

[4] El-Sayed, I. H., Lotfy, M., & Moawad, M. (2011). Immunodiagnostic potential of mucin (MUC2) and Thomsen-Friedenreich (TF) antigens in Egyptian patients with colorectal cancer. Eur Rev Med Pharmacol Sci, 15(1), 91-97.

[5] Liu, R., Li, X., Xiao, W., & Lam, K. S. (2017). Tumor-targeting peptides from combinatorial libraries. Advanced drug delivery reviews, 110, 13-37.

[6] Abed, J., Emgård, J. E. M., Zamir, G., Faroja, M., Almogy, G., Grenov, A., . . . Bachrach, G. (2016). Fap2 Mediates Fusobacterium nucleatum Colorectal Adenocarcinoma Enrichment by Binding to Tumor-Expressed Gal-GalNAc. Cell Host & Microbe, 20, 215–225.

[7] Wong, S. H., Kwong, T. N., Chow, T. C., Luk, A. K., Dai, R. Z., Nakatsu, G., ... & Ng, S. S. (2017). Quantitation of faecal Fusobacterium improves faecal immunochemical test in detecting advanced colorectal neoplasia. Gut, 66(8), 1441-1448.

[8] Bourdeau, R. W., Lee-Gosselin, A., Lakshmanan, A., Farhadi, A., Kumar, S. R., Nety, S. P., & Shapiro, M. G. (2018). Acoustic reporter genes for noninvasive imaging of microorganisms in mammalian hosts. Nature, 553(7686), 86.

[9] Archer, E. J., Robinson, A. B., & Süel, G. M. (2012). Engineered E. coli that detect and respond to gut inflammation through nitric oxide sensing. ACS synthetic biology, 1(10), 451-457.

[10] Din, M. O., Danino, T., Prindle, A., Skalak, M., Selimkhanov, J., Allen, K., ... & Hasty, J. (2016). Synchronized cycles of bacterial lysis for in vivo delivery. Nature, 536(7614), 81.

[11] Zembruski, N. C., Stache, V., Haefeli, W. E., & Weiss, J. (2012). 7-Aminoactinomycin D for apoptosis staining in flow cytometry. Analytical biochemistry, 429(1), 79-81