Wet lab
Dry lab
The wet lab team was able to engineer our intended end chassis, the probiotic E. coli Nissle 1917, with an optimized nitric oxide sensing genetic circuit and combine it with the secretion of anti-TNFα nanobodies. Our live therapeutic is able to sense nitric oxide levels and responds by secreting anti-inflammatory nanobodies which locally reduce inflammation. With IBD NanoBiotics, we combat the lack of specificity that limits current therapy options and propose a new therapeutic strategy that acts locally and is less invasive (see Implementation for future applications of IBD NanoBiotics).
All of our plasmid maps are available here. A detailed methods and protocols section can be found here.
Aim: Our aim was to create a biosensor for the inflammation marker nitric oxide (NO).
Our NO sensing system inspired by Xiaoyu J. Chen et al. 1 consists of a promoter, sfGFP, and NorR. We were able to prove that our promoter (pNorVβ) indeed responded to different NO concentrations (Figure 1b). We decided to compare it to the NO-sensitive promoter used by the 2016 iGEM team from the ETH Zurich and observed that pNorVβ; seems to induce higher responses to NO than the ETH promoter (Figure 1c).
All GFP measurements were normalised to the OD. To account for background fluorescence, we compared our constructs to a negative control plasmid that does not contain any promoter (Figure 1a). We did all our experiments in E. coli Nissle 1917 as this was our prefered chassis. We chose E. coli Nissle because it is a probiotic strain and already used as an IBD therapy. 4
Some of our graphs show a spike in the beginning that is not present in the raw data, which we concluded is a normalisation issue due to the low OD at the beginning of the experiment.
Some graphs also show a slight dip at the beginning, especially at higher concentrations. This might be because DETA/NO is slightly toxic, and might kill some of the bacteria at higher concentrations. Dead cells would still contribute to the OD value but would not be able to produce GFP, warping the results.
At this point, our sensing range was above the IBD relevant range and did not reflect the range reached by our reference paper 1 although our constructs were very similar. The only difference found between our gene circuits was the number of ribosome binding sites (RBSs) in front of the sfGFP, so we decided to vary the number of RBSs in our construct. We assembled 3 plasmids with 1, 2 and 3 RBSs respectively. Although this did not shift the sensing range of our promoter, it did increase the GFP expression at the same NO concentrations, especially with 2 RBSs.
From Figure 2a, it also seems that the construct with 2 ribosome binding sites is more leaky than the others, showing higher GFP expression, also in the absence of induction ([NO]=0).
Conclusion: We used the gene circuit with 2 RBSs for further testing because this construct proved to have the highest GFP expression. To further investigate leakiness and behavior of single cells, we decided to run a flow-cytometry assay (see below).
Aim: Since the sensitivity of our construct still did not reach our goal range, we decided to see whether the removal of NorR from our circuit could improve the output. NorR can act as a repressor when not bound to NO, and high release of NorR would therefore result in a lower function of our constructs. To see if endogenous NorR improves the sensitivity of the promoter to NO, we tested the construct in the absence of the positive NorR feedback-loop.
In order to do this, we removed NorR from our construct by Gibson assembly and re-ran the same plate-reader experiment. However, this did not improve our range and reduced the GFP expression.
Due to time constrains, we were not able to further try to adjust the sensing range and decided to continue with our previous construct (our genetic circuit with 2 RBSs), which gave us the highest GFP expression.
Conclusion: Removing the codon-optimized transcriptional regulator NorR did not improve the NO-sensing range. Instead it reduced the GFP expression at the same NO concentration as in previous experiments.
Aim: For proper characterisation of our gene circuit we calculated the EC50, the concentration at which half of the maximal response is induced.
We measured the response of our construct with 2 RBSs at different concentrations of NO and reported the slope of the linear section of the graph for each concentration. Plotting these values against the concentration enables readout of the EC50 value.
Conclusion: From the curve, we estimated an EC50 of ca. 335µM DETA/NO. However, as IBD patients with active inflammation exhibit NO-concentrations of around 15µM 1, our sensing range is currently still to high. We hope to improve the sensing range with further modifications to the promoter and sensing circuit.
All our experiments were conducted under aerobic conditions, however the anaerobic environment of the gut might improve sensitivity of pNorVβ for nitric oxide.
Aim: To further investigate the behavior of single cells following induction, we decided to perform a flow cytometry assay.
Altogether, our earlier findings were supported by flow cytometry analysis. Again, we could show that pNorVβ is slightly more sensitive than pNorV (ETH promoter) and that higher amounts of ribosome binding sites in front of sfGFP improved the expression upon DETA/NO induction, our construct with 2 ribosome binding sites yielding the strongest response.
However, the leakiness of pNorVβ + 2RBS was confirmed, as shown in the violin plots.
Conclusion: Our genetic circuit with 2 ribosome binding sites yields the highest responses to nitric oxide induction.
Aim: To avoid horizontal gene transfer and to reduce metabolic burden, we planned to integrate our designed genes into E. coli Nissle 1917, using the clonetegration method 2.
First, we encountered issues trying to transform our bacteria with the integration plasmid pOSIP-KO from the iGEM kit 2022. PCR amplification also did not work and gel electrophoresis revealed impurities in the product. Neither pOSIP-CH nor pOSIP-TT from the kit could be successfully transformed.
Trying the same procedure with pOSIP-KO ordered from Addgene, we were able to clone the plasmid and transform our bacteria with the assembled integration vector.
Conclusion: To screen for integrants, we did colony PCR. Unfortunately, we could not prove successful integration because we later found out that the integration vector alone also would have yielded a band of the same size as expected from successful integration.
Aim: Express anti-TNFα nanobodies in E. coli MC1061, then extract and purify them to show their ability to bind to TNFα
3 monovalent and 6 bivalent nanobodies were cloned into the pSB_init expression vector, then extracted through periplasmic extraction in the case of monovalent nanobodies and whole cell lysis for bivalent constructs. These were then purified by immobilized metal anion chromatography (IMAC) and run on a gel to see if we received the correct bands.
Afterwards, we tested the binding ability to TNFα of the purified nanobodies with an ELISA.
Conclusion: Monovalent and bivalent nanobody constructs targeting TNFα can be expressed and extracted from E. coli MC1061. Additionally these nanobodies effectively bind TNFα as shown with an ELISA.
Aim: Showing that inhibiting TNFα-actions by nanobody binding indeed influences the immune response of monocytes to the inflammatory actions induced by the cytokine.
We used the human monocytic cell line THP-1 to show that inhibiting TNFα-actions influences the immune response of human monocytes. For this, we incubated first the monocytes with different nanobody constructs and then stimulated the cells for 24 hours with different concentrations of human recombinant TNFα. We then measured the inflammatory response of the cells indirectly by comparing the IL-1β expression to the housekeeping gene GAPDH. IL-1β is an important inflammation mediator and therefore a good marker to prove functional TNFα-inhibition. Adalimumab is a commercially available anti-TNFα monoclonal antibody already used clinically to treat IBD patients. It served as a positive control in our assays.
Conclusion: Our analysis showed that TNFα alone induces a significant expression of IL-1β and all nanobodies, monovalent and bivalent constructs, were able to reduce the inflammatory response by up to 4-fold. Additionally, most nanobodies performed as good or even better than the monoclonal antibody adalimumab.
Aim: Express and secrete anti-TNFα nanobodies via the hemolysin A secretion system in E. coli MC1061.
E. coli MC1061 was transformed with a plasmid containing the parts necessary for the hemolysin A secretion system and another containing a nanobody with a myc-tag and a HlyA-tag attached. Nanobody production was induced using arabinose. A Western blot using the supernatant was performed to confirm the presence of the targeted proteins. To prove that the addition of the C-terminal tags and the secretion process still allow the nanobodies to bind to TNFα, an ELISA was done using the secreted nanobodies from the bacterial supernatant.
Conclusion: Monovalent and bivalent nanobodies can be produced and secreted in E. coli using the hemolysin A secretion system. The secreted nanobodies also retain their ability to effectively bind TNFα.
Aim: Demonstrate that anti-TNFα nanobodies can be expressed and secreted in an alternative E. coli strain.
The probiotic E. coli Nissle 1917 was transformed with the secretion system and select nanobodies tested in E. coli MC1061. The supernatant was once again used for a western blot and an ELISA to show protein presence and retained binding ability.
Conclusion: Monovalent and bivalent nanobodies can be produced and secreted in the probiotic E. coli strain Nissle 1917. The secreted nanobodies also retain their ability to effectively bind TNFα, with the bivalent construct showing increased affinity compared to that secreted in E. coli MC1061. The clear bands in the Western blot show that certain constructs are quite pure, but others secreted in E. coli MC1061 show artefacts that match in size to cleaved HlyA-tags that may have dissociated from the nanobody while retaining the myc-tag, allowing for their detection in the Wwstern blot.
Aim: Use nitric oxide (NO) to trigger the production and secretion of anti-TNFα nanobodies in E. coli Nissle 1917.
The nanobody constructs were transferred to the plasmid containing the nitric oxide sensor pNorVβ previously described. Only a single construct was viable for a double transformation in the probiotic E. coli strain. These colonies were induced using DETA/NO, and the supernatant was once again taken to perform a western blot and ELISA to show the presence of protein as well as their binding abilities.
Conclusion: The monovalent nanobody construct VHH#2B could be produced and secreted in E. coli Nissle 1917 through nitric oxide induction. However, we realised that there is significant nanobody production happening in the samples that did not get induced by nitric oxide, suggesting that the promoter might be leaky.
Aim: Previous experiments showed the expression of nanobodies even without the additional nitric oxide source. We analysed the obtained Western blot bands and used flow cytometry to investigate this issue further.
We used imageJ to measure the intensity of the Western blot bands seen in figure 13A:
For each condition a numerical average was calculated. On average the bands from the 2mM DETA/NO are 41% more intense than the control indicating an increased protein secretion upon DETA/NO induction. However, the non-induced expression appears to be quite significant. The follow-up experiment using flow cytometry to analyse liquid bacteria cultures that express GFP upon nitric oxide induction has already been described above.
Conclusion: While the construct with 2 RBS has higher overall GFP expression values, it is also leakier than the constructs with one or three ribosomal binding sites. If high GFP expression is required, but some leakiness does not matter much, we recommend choosing BBa_K4387006 with two RBS. If lower leakiness is essential, but GFP expression does not need to be very high, we recommend using parts BBa_K4387005 or BBa_K4387007 instead with one or three RBS respectively.
We simulated the gut environment and all relevant particles. The backbone of the model is an emission-diffusion system that propagates particles over space and time. We can manufacture a wide range of scenarios by implementing various parameters such as decay times, diffusion constants, particle concentrations, and severeness of inflammation. The model lets us navigate the mystifying disease that is IBD, but may also offer insights into other systems and diseases. Keeping the model general and source code available allows for adaptation to different illnesses, and certain aspects like diffusion decay and emission could be used in various models in the future.
When applying the model to the scope of IBD, we can see in figure 13 that no matter the severity of inflammation, the nanobodies will lead to an exponential decay of TNFα and eventually converge.
More surprising is that the results of figure 15 seem to be invariant with the production of nanobodies from a single bacteria, enabling future research to focus on the viability of the bacteria. Furthermore, we can tip the trade-off between competitiveness and nanobody production further toward the survival of the bacteria.