For its non-invasive nature, good penetrability and reversibility, temperature stands out for its unique advantages in microbial therapy. We chose temperature as the trigging factor of our switch, intending to use the thermo-sensitive switch to function and solve the problems in microbial therapy.

Cold-inducible On-switch:

Our cold-inducible on-switch functions on the base of the evolved transcription factor and the evolved protease. The expression of GOI (goal of interest) was inhibited by the evolved transcription factor and the evolved protease could activate the reporter gene by cleaving the sensitive repressor. More details see PROJECT-SWITCH.

Figure 1. The cold-inducible ON-switch is encoded on two plasmids. The circuit is shown with the genetic parts and relationships among them.

We firstly used ELIASA to characterize the tendency of fluorescence change of a series of cold-inducible ON-switches using different TEVts mutants under different temperatures. As shown in Figure 2, all five switches show high fluorescence under low temperature, while their fluorescence all decreases when temperature rises. Then we chose two kinds of cold-inducible ON-switches, consisting of TEVts#6 and TEVts#18 correspondingly, to measure their temperature response curve more precisely by flow cytometer in TOP10 strain. From Figure 3 we can see that two group both show narrow transition ranges and have ∼100-fold induction, which was achieved within less than ten degrees.

Figure 2. The tendency of fluorescence change of a series of TEVts mutants under different temperatures (measured by ELIASA)
Figure 3. The induction curve of the cold-inducible ON-switches (TOP10)

For the further use of our platform to serve microbial therapeutics, we found a more suitable E.coil strain, Nissle 1917, a probiotic with more than 100 years of medical application. We measured the best one, consisting of TEVts#18, in Nissle 1917. It also shows high performance similar with it in TOP10 strain, indicating the robustness also stability of our system.

Figure 4. The induction curve of the cold-inducible ON-switches (Nissle 1917)

Thus, our cold-inducible ON-switches show high performance and versatility, which ensures the potential for basic research, as well as industrial and biomedical applications, and truly makes engineered bacteria precisely controlled.

Heat-inducible On-switch:

Our cold-inducible on-switch functions on the base of TCI transcription factor family and TlpA family, whose activity will be lost under high temperature. When it is above threshold temperature, TF will be allosteric, losing its activity. Then the inhibition of target gene is relieved while the target gene is expressed. More details see PROJECT-SWITCH.

Figure 5. The heat-inducible ON-switch is encoded on one plasmid. The circuit is shown with the genetic parts and relationships among them.

We first characterized the performance of 6 members from the two family in Top10 strain, the result shows that most of the transcription repressors show sharp thermal transitions, especially TCI and TCI42, with more than 100-fold induction within 10 degrees Celsius. Their impressive performances make them candidate parts for our further circuit design.

Figure 6. The induction curve of the heat-inducible ON-switches (TOP10)

Also, we tested these heat-inducible ON-switch in the chassis E.coli Nissle 1917, a probiotic with more than 100 years of medical application, their robustness give us more confidence in the stability and preciseness of our ark.

Figure 7. The induction curve of the heat-inducible ON-switches (Nissle 1917)

Double-status switches

The temperature-dependent on-off switch consists of both the two types of switches mentioned above. We can choose different transition temperature to meet different environments and goals. Also, by combining the heat-inducible with the cold-inducible ON-switch, we can get a double-status switch to realize more functions in one creature.

Figure 8. The double-status switch is encoded on three plasmids. The circuit is shown with the genetic parts and relationships among them.

Based on the temperature response curve of cold-inducible ON-switches and heat-inducible ON-switches, we chose the best mutant, TEVts#6 and TCI42 to build our double-status switch. But actually, users can choose other thermo-sensitive protease or transcription factors to build other more suitable double-status switches based on their needs.

AS shown in Figure 9, our cold-inducible ON-switch works when the temperature is low and then express green fluorescence. When the temperature rises, the activity of cold-inducible ON-switch is inhibited and the heat-inducible ON-switch works. Therefore, we can turn on different genes’ expression at different temperatures. Similar with our single thermo-sensitive parts, this double-status switch also shows high performance and versatility.

Figure 9. The induction curve of the double-status switches (TOP10)


Cold-inducible On-switch + Toxin

One way to address the biosafety issue is using toxin-antitoxin pairs. Toxin is a kind of protein which is the output of the genetic circuit. Under certain condition such as temperature, toxin begins to be expressed by the engineered bacteria and eventually kill those bacteria.


In our engineering bacteria, we used Doc, a toxin interferes with basic metabolism at the level of translation, to associate with our cold-inducible on-switch, so that under low temperature, the switch is turned on to express Doc protein.

Figure 10. The design of establishing our cold-inducible kill switch.

To explore the possibility of escape for evaluating the performance of our kill switch, we measured the growth of the bacteria with DOC gene in LB solid medium at different temperatures and dilutions. The results showed that the strain cultured at 25℃ grew much worse compared with the strain grown at 37℃, and the 20h average solid escape frequency is 2.318×10^(-2) , the lowest escape rate can be limited to 10^(-3) level.

Figure 11. The escape rate of bacteria on solid LB plate after 20h. The escape rate is calculated using the formula 𝒑𝒆𝒓 𝑬𝒔𝒄𝒂𝒑𝒆 𝑭𝒓𝒆𝒒𝒖𝒆𝒏𝒄𝒚 =(𝑪𝒐𝒍𝒐𝒏𝒊𝒆𝒔 𝒐𝒏 𝒏𝒐𝒏𝒑𝒆𝒓𝒎𝒊𝒔𝒔𝒊𝒗𝒆 𝒑𝒍𝒂𝒕𝒆 × 𝒅𝒊𝒍𝒖𝒕𝒊𝒐𝒏)/(𝑪𝒐𝒍𝒐𝒏𝒊𝒆𝒔 𝒐𝒏 𝒑𝒆𝒓𝒎𝒊𝒔𝒔𝒊𝒗𝒆 𝒑𝒍𝒂𝒕𝒆 × 𝒅𝒊𝒍𝒖𝒕𝒊𝒐𝒏)[6] , the escape rate of 3 groups are (1×10^7)/(2×10^8 )=5×10^(-2) , (11×10^7)/(6×10^10 )=1.83×10^(-2) and (2×10^7)/(16×10^9 )=1.25×10^(-3) respectively(from above to below). And the 𝑬𝒔𝒄𝒂𝒑𝒆 𝑭𝒓𝒆𝒒𝒖𝒆𝒏𝒄𝒚 = 𝑨𝒗𝒆𝒓𝒂𝒈𝒆 𝑬𝒔𝒄𝒂𝒑𝒆 𝑭𝒓𝒆𝒒𝒖𝒆𝒏𝒄𝒚 ± 𝒔𝒕𝒂𝒏𝒅𝒂𝒓𝒅 𝒅𝒆𝒗𝒊𝒂𝒕𝒊𝒐𝒏 [6]=2.318 ×10^(−2) ± 0.122.

Moreover, we measured the growth curve of the strains with DOC gene at different temperatures to characterize their growth in liquid medium, and took the strains without doc gene as the control to verify the effectiveness of the toxin system. The results showed that the growth of the two strains was almost the same at 37℃, indicating relatively low leakage of the system and low while the growth of the strains with doc gene was worse at 25℃, indicating that the system with DOC as toxin was effective.

Figure 12. The growth curve of bacteria on liquid LB. Incubated in 4 mL volume in 24-deep-well plate.

Based on the design of our cold-inducible switch and integrating toxin Doc, we developed a ‘kill switch’ which sensitively and accurately responds to the natural signal of human body and the environment – temperature. We successfully limit the escape rate to 10^(-2) to 10^(-3). Considering the improvement of the robustness and performance of the whole system, also to ensure the ability of our ark to adapt to various situation with as little risk for both human and nature as possible, we design another strategy as follow.

Synthetic Auxotrophy

We assigned the TAG stop codon to incorporate a non-canonical amino acid and tried to redesign an essential gene to develop corresponding dependent bacteria.

Figure 13. The circuit for establishing our ‘synthetic auxotrophy’[1].

Getting a high-performance aminoacyl-tRNA synthetase (aaRS) system from our PI’s laboratory, we chose 3,5-dichloro-L-tyrosine (Cl2Y) as the non-canonical amino acid.[2] First, we proved the concept of our design using GFP as the target gene. We constructed mutant GFP with 9TAG and 17TAG as the potential incorporating site. When additionally providing Cl2Y, we could see a prominent increase(~100 folds) of the fluorescence, indicating the high efficiency of incorporation.

Figure 14. When provided Cl2Y, the fluorescence of GFP can be increased almost 100 folds, indicated the high efficiency of incorporation.

MS result further proved that Cl2Y was incorporated into the target site with high selectivity and efficiency.

Figure 15. Mass spectrum result of 9Cl2Y GFP. 27824 indicates the molecular weight of interested protein.

Figure 16. Mass spectrum result of 17Cl2Y GFP. 27794 indicates the molecular weight of interested protein.

Heat-inducible On-switch + ncAA

For our therapeutic bacteria, exogenous source of non-canonical amino acid may constrain its function. So, we designed a thermo-sensitive non-canonical amino acid producing part, by combined our heat-inducible on-switch with the synthetase of our targeting non-canonical amino acid, we can strictly support the growth of our bacteria only inside the human body.

We considered to use tyrosine hydroxylase to synthesize L-dopa as the non-canonical amino acid.[3] When TCI induces the endogenous supplement of L-dopa under 37℃ circumstance, proper translation, folding and function of biotin synthase BioB will be achieved, our ark can navigate safely to arrive its destination.

Figure 17. The design of heat-inducible on-switch combined with ncAA system for our ‘synthetic auxotrophy’, providing a solution for the current obstacles of dependence on exogenous source of noncanonical amino acid.

We successfully synthesized L-dopa in the proper situation, this part of result can be seen in PROJECT-PARKINSON'S DISEASE. With a developed DHPRS and tRNA¬CUA system[9], we can incorporate L-dopa into the intended site, as you can see in the mass spectrum below.

Figure 18. Mass spectrum result of 17L-Dopa GFP. 27726 indicates the molecular weight of interested protein.

However, due to the limitation of the system itself, the selectivity and efficiency of the L-dopa incorporation were not as good as our expectation. We could predominantly reduce the background by adding TAG codons.

Figure 19. Relative Fluorescence of 17+3TAG GFP with/without L-Dopa. The background was predominant controlled by assigning another TAG stop codon into the interested protein.


Microbial therapeutics has unique advantages on this severe problem by providing a sustained, stable and trace supply of levodopa.

Figure 20. TyrH catalyzes tyrosin to levodopa

Figure 21. Gene of TyrH under a constitutive promotera

Firstly, we transferred the tyrosine hydroxylase(TyrH) gene into E.coli, so that the engineering bacteria can synthesize dopa in the presence of tyrosine. After that, DOPA will enter the brain area through the blood and maintain normal dopamine concentration.

To characterize the efficiency of TyrH, we conducted the fermentation experiment of levodopa, and the result, as shown in Figure 22, shows that the maximum concentration of dopa can reach more than 130 mg/L after 8 to 12 hours of fermentation. By the way, the subsequent decline maybe due to oxidation of dopa during fermentation, but this hardly occurs in the gut.

Figure 22. The fermentation curve of levodopa

The comparison of taking drugs by pill and microbial therapeutics

We employed a model to explain the advantage in dose controlling the drug.

We abstract the body into three compartments, absorption compartment –the intestinal tract, central compartment –blood circulation, and effect compartment, which is related to clinical effect. The rate of drug delivery is in direct proportion of the drug amount.

Then we consider two ways of drug delivery—by pills and by bacteria. Their difference origins in the absorption compartment. If by pills, some drugs are released initially, and then no drug is produced. If by bacteria, as we built a model to simulate the distribution of bacteria in the intestinal tract and found the distribution tends to be even, thus the drugs are produced and released continuously.

Then we made a comparison between bacterial delivery and oral delivery, by comparing the clinical effect versus time, and by comparing the drug concentration in plasma, which is related to possible side effects.

Figure 23. Drug Concentration Versus Time by Different Methods, for Fluctuate Patients

Figure 24. Clinical Effect Versus Time by Different Methods, for Fluctuate Patients

(Note: Here we only showed the result of the fluctuate patients).

As shown in Figures 23 & 24, obviously, compared to oral delivery, the bacterial delivery generally has a high-and-stable clinical effect, while having a much lower drug concentration in plasma, thus will bring much fewer side effects. So the bacterial delivery can perfectly reduce the side effects while maintaining a good clinical effect, and with such method, patients no longer need to take pills frequently.

See more details about this model here.

Finally, combining biocontainment module of our platform with the circuit for releasing TyrOH can make this microbial therapeutic for Parkinson's disease safer and meet the requirements of clinical and commercial applications.


Considering the unique advantages of microbiological therapy in treating cancer, we selected cancer as one of our application and built a double-status switch in our platform to solve the problem of biocontainment and biosafety, which is used to lower the off-target effect simultaneously. [4]

In our project, we used CD47 nanobody as a drug molecule to block CD47 on the surface of cancer cells, so that they could be killed by microphages.[5][6] More details about CD47 nanobody see PROJECT-CANCER.

Because the oxygen concentration around the cancer tissue is generally low, some anaerobes, like Bifidobacterium, tend to colonize tumor tissue.[4] Based on this character, we connected the gene of CD47 nanobody behind the anaerobic induced promoter, PfnrS, to ensure its expression is only near the cancer tissue.

Figure 25. CD47 nanobody under the anaerobic induced promoter, PfnrS

To achieve lower off-target effect, we used our thermo-sensitive switches to construct a biosafety module besides original biocontainment module. Combining both of them together, we built a double-status switch.More details see PROJECT-CANCER.

Figure 26. The combination of double-status switch and the circuit of microbial therapeutics

Combining this double-status switch with the circuit of microbial therapeutics, we can lower the off-target effect significantly and make full use of unique advantage of microbial therapeutics.

Incubated in the anaerobic environment in vitro, our engineered bacteria successfully expressed CD47 nanobody, indicating its potential for precisely controlled expression near the cancer tissues.

Figure 27. The SDS-PAGE analysis result of CD47 nanobody purification product. An around 14kDa protein can be observed, which is corresponding to the theoretical molecular weight of CD47 nanobody (17.42kDa) our bacteria are supposed to express. The bacteria were incubated in a 500mL conical flask in the anaerobic environment.

To control the unintended proliferation of our engineered bacteria into healthy tissues, also prevent the possible side effect to healthy tissues caused by CD47nb, our design of combining TCI42 with a lysis gene, ΦX174, precisely restricted the bacteria lysis and the drug release in the cancer tissues specifically heated by our hardware. We tested the efficiency of bacteria lysis under 42℃. As we expected, we could observe a prominent decrease of OD600 when our engineered bacteria were incubated under a circumstance of 42 ℃, where the controlled group bacteria were observed to grow in a steady increasing trend. What’s more, in the environment of 37 ℃, the growth of our bacteria was similar to the controlled group bacteria, indicating rather low leakage or burden of the whole system.

Figure 28. The curve of OD600 change of experiment/ controlled group under 37/42℃ condition.



We developed an innovative electronic capsule, cPlus, which can perfectly assist diagnosis and treatment of intestinal diseases. cPlus consists of a heating system, a sampling device, a clamping device, a positioning device and a remote-control system. With the accurate positioning(<1cm) and heating (<1000s to reach the target temperature within 1cm range) control, cPlus can be precisely located and heat the specific lesion to assist the targeted drug release from the engineered therapeutic bacteria. Not only for treatment, but cPlus contains a sampling device and can be combined with current imaging system for diagnosis. In the present research of intestinal flora, original technology of sampling from the feces is difficult to meet the requirements of sampling intestinal flora at specific location (including long-term fixed flora), cPlus fills a vacancy of intestinal flora accurate sampling. We believe cPlus will bring an evolution in the industry of medical microdevice.

Structure of cPLUS

Our capsule consists of 6 large parts, including heating system, sampling device, clamp device, positioning system, energy and remote control system. More details see PROJECT-HARDWARE.

Figure 29. Overview of our cPLUS


1. To select the most suitable PTC element for our capsule, we carried out heating experiments on pig’s small intestine using 2 different kinds of PTC elements.

● This is the first set of data that we measured

Figure 30. The PTC element is connected to a 6V voltage to heat the pig's small intestine at 5V 40℃

Data4: less than 1cm away from the heat source, it can reach over 40 ℃after 500s heating, and basically stable at 43-44 ℃ after 1500s

Data2:1-1.5cm away from the heat source, it can reach more than 40 ℃after 1500s, and basically stable

Data1&Data3: 1.5cm away from the heat source, heating to more than 40 ℃may take longer

● This is the second set of data that we measured

Figure 31. The PTC element is connected to a 6V voltage to heat the pig's small intestine at 5V 50℃

Data4: less than 1cm away from the heat source, the heating temperature continues to rise to more than 45℃ and will be higher, so it is not suitable

Data1 &2&3: 1.5cm away from the heat source, it may take longer to heat above 40 ℃

Of course, there are many other sets of test data, and here is just the conclusive data we found after comparing them

The conclusion was that we found the right PTC element and the right heating voltage to meet the requirements of our experiment -- that is, the PTC element was connected to the 6V voltage at 5V40℃

2. After building the positioning device, we did the following experiment to determine its location precision.

● We randomly selected 13 points for magnetic induction positioning test, and finally found that the average error was less than 1cm, which could well meet our requirements.

Figure 32. The circle represents the position error and the possible range of the actual position


[1] Mandell, D. J. et al. Biocontainment of genetically modified organisms by synthetic protein design. Nature 518, 55-60, (2015).

[2] Liu, X. et al. Significant expansion of fluorescent protein sensing ability through the genetic incorporation of superior photo-induced electron-transfer quenchers. J Am Chem Soc 136, 13094-13097, (2014).

[3] Kappock TJ, Caradonna JP. 1996. Pterin-Dependent Amino Acid Hydroxylases. Chemical Reviews 96:2659-2756.

[4]Zu C , Wang J . Tumor-colonizing bacteria: A potential tumor targeting therapy[J]. Critical Reviews in Microbiology, 2014, 40(3):225-235.

[5]Chowdhury Sreyan,Castro Samuel,Coker Courtney,Hinchliffe Taylor E,Arpaia Nicholas,Danino Tal. Programmable bacteria induce durable tumor regression and systemic antitumor immunity.[J]. Nature medicine,2019,25(7).

[6]Willingham S B , Volkmer J P , Gentles A J , et al. The CD47-signal regulatory protein alpha (SIRPa) interaction is a therapeutic target for human solid tumors[J]. Proceedings of the National Academy of Sciences of the United States of America, 2012, 109(17):6662-6667.