1 Screening of suitable hypoxia-inducible promoters

1.1 Construction of plasmids and characterization of HIPs

Escherichia coli Nissle 1917 (EcN), as a probiotic, has tumour-targeting properties to some extent. So, our aim was to enhance this tendency by specifically targeting EcN to the hypoxic tumour microenvironment (TME). Therefore, we focused on hypoxia-inducible promoters (HIPs), hoping to achieve explicit hypoxia targeting of EcN.

HIPs are regulated by the Fumarate and Nitrate Reduction regulator (FNR)$^1$. Under hypoxic conditions, FNR homodimerizes through a 4Fe–4S²⁺ cluster and binds to a specific DNA site termed ‘FNR box’. Upon its binding, the homodimer interacts with the RNA polymerase binding to TATA box at -10 region, thus activating transcription (Fig. 3). Oxygen can inhibit this process. Previous study$^2$ has shown that pPepT, as an endogenous HIP of EcN, can produce a large amount of expression under hypoxic conditions. Mengesha et al$^3$ made improvements to the pPepT from Salmonella by mutating its FNR box and TATA box, resulting in an enhanced hypoxia-responsive curve.

Thus, on the base of wild type PepT promoter, we strived to change the intensity of its expression by mutating TATA box, mutating FNR box, and changing the RBS. We used the PJUMP41-2A (sfGFP) plasmid (BBa_J428365, this plasmid referred below are abbreviated as O12, as it corresponds to the O12 well on the biobrick) as our vector This plasmid is characterized as a low-copy plasmid with spectinomycin resistance and carries the BBa_J23100 promoter along with the sfGFP sequence. Through reverse PCR and Gibson assembly (details are shown in “Engineering”), we successfully replaced the J23100 promoter and RBS with the following 10 promoters + RBS (see BBa_K4713116 - BBa_K4713125):

Table 1 | The promoter + RBS sequences used in screening.

We used sfGFP as the fluorescence output and tested the intensity of 10 variants. In order to achieve normalization of fluorescence intensity, the empty plasmid was used as a blank control, and the ratio of the intensity of sfGFP fluorescence controlled by BBa_J23100 under normoxic and anaerobic conditions was used to explain the insufficient folding and weak fluorescence of sfGFP under anaerobic conditions (coefficient R in formula)

Relative Expression level was analyzed with the following formula:

The result was shown in Fig. 2.

Our results show that FF+20 variant exhibits the strongest sfGFP fluorescence under anaerobic conditions, although it also has the highest baseline expression level. Mutating the TATA box enhances expression levels under anaerobic conditions, but it also leads to increased expression under normoxic conditions. Similar effects are observed when changing the RBS. However, altering the FNR box did not result in significant expression distinction.

Meanwhile, we acknowledged that our characterization was conducted on a macroscopic scale, and some measurement errors may be inherent. Therefore, we aimed to observe individual or small groups of cells at the microscopic level to obtain more precise data. To achieve this, we employed microfluidic technology.

1.2 Microfluidics for Precise Characterization at the Single-cell Scale

In order to measure fluorescence intensity with single cell accuracy at precise oxygen concentrations, we specially designed a microfluidic chip. With the help of Prof. Chunxiong Luo, we designed a chip that can simultaneously measure the fluorescence intensity of 4 E. coli strains under 6 oxygen concentration gradients$^4$, as shown in Fig. 3.

We have made the microfluidic chip according to the drawing, and the real picture is shown in Fig. 4.

We demonstrated the viability of the microfluidic chip by verifying the presence of an oxygen concentration gradient and the capture of E. coli by the trap chamber. As shown in Fig. 5a, when the mixture was changed from air to nitrogen air at 10 minutes, the fluorescence intensity was found to be increased, and the 6 channels were different. The following data (Fig. 5b - c, Video 1) further confirms the reliability of our chips

Unfortunately, we didn't have enough time to apply our microfluidic chip on formal measurement. However, we believe that microscopic result will be similar to the macroscopic result.

2 Success on construction of a hybrid promoter sensing both hypoxia and high lactate

In order to reduce the escape rate and improve targeting to the tumour microenvironment, our initial idea was to build an AND-gate system to sense hypoxia and high lactate simultaneously. Fig. 6a shows one of the traditional AND-gate systems$^5$. This system involves two signal inputs (Ara and aTc), which produce protein sicA and invF respectively that can bind and act as a transcriptional activator to initiate the expression of downstream rfp gene. We initially wanted to imitate this AND-gate structure and replace the downstream output with the necessary gene (asd) for EcN. The asd gene expresses aspartate semialdehyde dehydrogenase, which is required for the biosynthesis of lysine, threonine and methionine. Deletion of asd causes bacterial cell wall rupture and death. Supplementation with diaminopimelic acid (DAP) can promote the growth of asd knockout strains. Notably, DAP cannot be produced or metabolized from the host cell environment, making it an ideal strategy to reduce escape rates. High lactate is sensed through the native lldPRD lactate operon of EcN$^6$. But we are worried that this AND-gate structure may bring a greater expression burden to EcN.

However, after gaining a deeper understanding of the structure of the lactate operon, we came up with a fantastic idea. Because the principle of the lactate operon (Fig. 6b) is that when lactate is low, lldR acts as an inhibitory protein and inhibits promoter expression by binding to the O1 and O2 sites. When the lactate concentration rises, lactic acid molecules trigger an allosteric reaction by binding to the lldR protein, making the inhibitory effect disappear. Additionally, the lldR protein that binds lactic acid molecules can also bind to the O1 site and act as a transcription activator. Therefore, we wanted to replace the promoter between the O1 and O2 sites with our pPepT (Fig. 6c), allowing it to achieve an AND-gate-like structure and sense hypoxia and high lactate simultaneously. Facts have proved that we successfully constructed this type of promoter through the following experiments.

2.1 Knockout of asd gene in EcN

To enable EcN to respond to downstream output from the bybrid promoter, we deleted the asd gene of EcN. The two-plasmid system of Yang et al $^7$ $^8$ was applied for gene knockout of EcN. Plasmids of pEcCas and pTarget were the gift from Prof. Changtao Jiang. The N20 sequence was designed via chopchop (chopchop.cbu.uib.no).

Through this method, the asd gene was successfully knocked out in four EcN strains (EcN ∆asd 1 ~ 4). (Fig. 7a) By exogenously supplementing DAP at different concentrations (Fig. 7b), we found that concentrations of 50 μg/ml and above could enable the survival of asd knockout strains.

2.2 Hybrid promoter responds to hypoxia and high lactate

Before constructing the hybrid promoter, we imagined two failure situations (Fig. 8a): first, the lldR protein binding at the O1 site can cover up the FNR binding site of PepT, so that the FNR protein cannot bind to the FNR box to initiate downstream transcription due to steric hindrance (condition 1); second, as the distance between promoters O1 and O2 lengthens (pPepT is 19 bp longer than the original promoter), the active lldR protein bound to the O1 site cannot activate transcription (condition 2).

Therefore, we constructed 6 versions of the plldR operon through shortening or lengthening the distance between O1 and O2. Combined with 4 variants of pPepT (BBa_K4713111 to BBa_K4713114), a total of 24 candidates was constructed (Fig. 8b). We replaced the BBa_J23100 promoter of O12 plasmid with our candidate. In order to improve the transformation efficiency (the transformation efficiency of EcN ∆asd is about one-tenth that of EcN), we replaced the original LB medium with SOC medium and appropriately increased the supplementation amount of DAP (100 μg/ml). All promoters were successfully constructed, but only 14 candidates were successfully transformed into EcN ∆asd (the remaining 10 were also successfully transformed recently, but have not yet been characterized). Subsequently, we evaluated the promoter strength under different oxygen (normoxic or anaerobic) and lactate concentrations (0, 0.1 mM, 1 mM or 10 mM), with the results showing as follows:

It can be seen from the results that appropriately shortening the distance between O1 and O2 (version S1, S3) shows a better response. We finally selected the best-performing promoter S3-FA (BBa_K4713008), which exhibited both low expression at baseline and high expression under conditions of low oxygen and high lactate.

At the same time, due to the low expression of endogenous lldR protein, the promoter's response to lactate concentration is weak, resulting in a certain escape expression at the baseline level. In order to further reduce this expression, we also transferred the lldR overexpression plasmid into EcN ∆asd, and inserted the T5 promoter and lldR sequence into the vector pET28a through gibson assembly, so that it can induce the controllable expression of lldR through IPTG. The EcN ∆asd co-transformed with pET28a_T5-lldR and O12-plldR-J23117 (this plasmid replaces the BBa_J23100 promoter with the BBa_K1847008, a lactate-responsive promoter, whose promoter sequence between O1 and O2 is BBa_J23117) significantly increased the level of lldR protein after induction of expression with IPTG, thereby inhibiting the expression of the plldR promoter, manifested by a significant decrease in OD_600nm (Fig. 10).

Unfortunately, due to time limits, we did not experiment with the effect of expressing lldR on our hybrid promoter, however, we believe that such expression could significantly reduce baseline levels of escape expression.

3 Expressing vesicle proteins enables precise spatial positioning

After the construction of hybrid promoter, our engineered bacteria had the ability to colonize specifically in the tumour tissues. According to our plan, the next step we need to find a method that can detect the location of the presence of bacteria. Bourdeau et al$^9$ genetically modified the vesicular protein from a photosynthetic autotrophic bacterium cyanobacterium Anabaena flos-aquae and transformed it into E. coli BL21(AI) to successfully express the vesicular protein, which could be detected by medical ultrasound probes. However, this article still has the shortcoming that the bacteria can only reach the tumour by in situ injection. The ability of engineered bacteria to self-localize tumours is necessary if this method is to be applied to clinical treatment.

Fortunately, the scheme proposed by the Peking iGEM team to target the tumour microenvironment by engineering bacteria can make up for this deficiency. We hypothesized that ARG could be introduced into the engineered bacteria containing and induce its expression vesicles. Since only the engineered bacteria colonized in the tumour tissue could survive, the signal would only be detected in the tumour tissue when detected by a medical ultrasound probe.

To prove this idea, pET28a_T7-ARG1 plasmid (addgene #106473, purchased from Miaoling Biology) was transformed into BL21(DE3) and BL21(AI). Vesicle protein expression was induced according to the method of Bourdeau et al$^9$, and a range of IPTG concentrations (0.1 mM, 0.4 mM, 1 mM, 10 mM) were used. As a negative control, BL21(DE3) transduced with pET28a_T7-eGFP (gift from A.P. Qingsong Wang) was characterized in the same way, which could only express GFP after induction.

After centrifuging for 2h at 300g, the results are shown as Fig. 11. The results for other induction concentrations are similar and will not be shown here.

Then, we used a transmission electron microscope (TME) (JEOL, JEM-F200) to observe the sample, and the results are as follows:

After confirming that the pET28a_T7-ARG1 plasmid was available, we managed to express vesicle proteins in EcN. Considering the lack of T7RNA polymerase in EcN, we replaced the T7 promoter with the T5 promoter through enzyme digestion and ligation, and transformed it (pET28a_T5-ARG1) into EcN through through electroporation. Induction was characterized in a similar way, except that we used 5 μM IPTG for induction and grew 150 ml cultures. The results are as follows:

Based on the above presentation, we successfully expressed vesicular proteins in EcN that ultrasound could detect. The pET28a_T5-ARG plasmid, combined with the hybrid promoter, is expected to obtain engineered bacteria that can specifically colonize tumours and be detected by ultrasound. The engineered bacteria are expected to be used as an alternative to PET-CT in the future to provide help for the systemic detection of metastatic lesions in advanced pancreatic cancer.

4 Construction and debugging of dynamic encapsulation circuits

Though tumour-targeting and ultrasound-sensing has been achieved, controlled immunogenicity during in vivo delivery is still a problem. In order to avoid viremia as much as possible, we expect to mask epitopes of EcN before ultrasound imaging and achieve rapid clearance after diagnosis. Thus, we focused on engineering the capsular polysaccharide (CAP). CAP is a loose mucous substance located on the surface of the cell wall, protecting bacteria from environmental influences. It is worth mentioning that the thickness of CAP changes dynamically, and if not generated endogenously, its thickness will naturally become thinner. Harimoto et al $^{11}$ have designed an inducible capsular polysaccharide (iCAP) system. They used the kfiC gene to regulate the expression of CAP. As an endogenous gene in EcN, kfiC encodes a glycotransferase essential in CAP biosynthesis, thus its expression determines the rate of CAP production, and then impacts the thickness of cellular outer layer. Based on the previous work, we designed the following circuit (Fig. 16a): kfiC under the control of ParaB (an arabinose-inducible promoter) would be transformed into EcN ∆asd ∆kfic. Prior to delivery, the bacteria would be induced by arabinose to start kfiC expression, and then encapsulation from CAP production, preparing it for host immune evasion and tumour focus localization. As CAP gradually degrades and cellular outer layer becomes thinner over time, the remaining bacteria in blood would be rapidly cleared, thus reducing long-term inflammatory response and toxicity risks (Fig. 16b) $^{11}$.

We tried to knockout the kfiC gene from the genome of EcN ∆asd in the same way as the asd gene (see Result 2.1), yet without success. A possible cause may be that both asd and kfiC genes are essential to bacterial cell wall synthesis, thus the absence of both would be fatal to EcN. We propose to change the output necessary gene of hybrid promoter to ThyA$^{12}$. ThyA encodes thymidylate synthase in E. coli, which is required in nucleotide biosynthesis pathways, and does not interfere with cell wall synthesis. Nevertheless, we demonstrated the feasibility of our iCAP system through modelling. By changing the elimination rate $k_e$ in our MC model (see Dry Lab - Model), we were able to simulate an enrichment to the tumour compartment as well as rapid clearance in other compartments with suitable elimination rates determined by CAP thickness. This design can be utilized to optimize therapeutic bacteria administration to maximize curative effects and minimize side effects.

Reference

1. Wing, H. J., Williams, S. M. & Busby, S. J. Spacing requirements for transcription activation by Escherichia coli FNR protein. Journal of Bacteriology 177, 6704–6710 (1995).
2. Chien, T. et al. Enhancing the tropism of bacteria via genetically programmed biosensors. Nat Biomed Eng 6, 94–104 (2022).
3. Mengesha, A. et al. Development of a flexible and potent hypoxia-inducible promoter for tumour-targeted gene expression in attenuated salmonella. Cancer Biology & Therapy 5, 1120–1128 (2006).
4. Sun, Y. et al. Two-Layered Microfluidic Devices for High-Throughput Dynamic Analysis of Synthetic Gene Circuits in E. coli. ACS Synth. Biol. 11, 3954–3965 (2022).
5. Moon, T. S., Lou, C., Tamsir, A., Stanton, B. C. & Voigt, C. A. Genetic programs constructed from layered logic gates in single cells. Nature 491, 249–253 (2012).
6. Aguilera, L. et al. Dual Role of LldR in Regulation of the lldPRD Operon, Involved in l-Lactate Metabolism in Escherichia coli. Journal of Bacteriology 190, 2997–3005 (2008).
7. Jiang, Y. et al. Multigene Editing in the Escherichia coli Genome via the CRISPR-Cas9 System. Appl Environ Microbiol 81, 2506–2514 (2015).
8. Li, Q. I. et al. A modified pCas/pTargetF system for CRISPR-Cas9-assisted genome editing in Escherichia coli. ABBS 53, 620–627 (2021).
9. Bourdeau, R. W. et al. Acoustic reporter genes for noninvasive imaging of microorganisms in mammalian hosts. Nature 553, 86–90 (2018).
10. Solé, R. & Conde-Pueyo, N. Ultrasound approach tracks gut microbes. Nature 553, 36–37 (2018).
11. Harimoto, T. et al. A programmable encapsulation system improves delivery of therapeutic bacteria in mice. Nat Biotechnol 40, 1259–1269 (2022).
12. Fu, X. & Xu, J.-G. Development of a chromosome-plasmid balanced lethal system for Lactobacillus acidophilus with thyA gene as selective marker. Microbiology and immunology 44, 551–556 (2000).