When colorectal cancer is caught early, around nine out of ten people survive the next five years after diagnosis. However, the 5-year survival rate drops to almost one out of ten people when it is discovered in late stages [1]. This highlights the importance of early diagnosis of this disease. Unfortunately, the current most used screening method (FIT) is not optimal. It is only performed every two years in people over the age of 55 and users have to take their own stool sample, which they have to send to a laboratory for further analysis. Additionally, the false positive rate is high (30%), meaning that lots of colonoscopies are performed unnecessarily [2]. These two factors create a big burden on people and the healthcare system.
Colourectal aims to solve these problems by designing an early diagnostic tool with a higher specificity than the current method. This is in the form of a self-test, which can be entirely performed at home, relieving the pressure on the user and the healthcare system. Thanks to these characteristics, the test could be performed more frequently and starting from a younger age.
Here, we aim to prove step-by-step that the different stages explained in project implementation could work in a real-life setting. We successfully tested most of the components of the system in the final organism: Escherichia coli Nissle 1917 (EcN).
Prediction of behaviour of the living diagnostic tool in the gut
The diagnosis process with Colourectal takes place over the period of one week. In order to know the dose needed to keep our living diagnostic in the colon for a week and to ensure a sufficient concentration to perform its function, we used the model of the colon developed in our project. Different situations were simulated, and ultimately the model showed that one dose per day during the first three days of the process was the most adequate strategy for our objectives (Figure 1). For more information on how this simulation was performed in our colon model, visit our modelling page.
Figure 1: Simulation performed in our colon model showing that three doses, once per day, provide a strong population of our living diagnostic in the proximal mucus for around 6 days.
Thanks to this information, we established a four-pill system, as is also explained in the implementation section.
Check; was our living diagnostic tool to do its job?
The first three pills contain our living diagnostic and should be taken on days one, two and three. The third pill contains, in addition to the bacteria, a set concentration of rhamnose. This is used to induce a positive control mechanism based on the expression of a pink chromoprotein that will be visible in the stool, serving as verification of the proper functioning of our test. We demonstrated that the pink chromoprotein anm2CP can be successfully produced in Escherichia coli DH5a (E. coli DH5a) (Figure 2), although secretion was not yet achieved. For more information on the structure and production of this chromoprotein check out the signalling results section.
Figure 2: Luria Bertani agar plate supplemented with Kanamycin with E. coli Nissle 1917 showing pink colouration due to anm2CP production.
This positive control will be induced by the three-step inducible system. We designed this system to be able to express three separate genetic circuits at three different concentrations of the same inducer metabolite. This was needed because of the lack of inducers that are suitable for the whole Colourectal rationale. The three-step system works by tuning the gene expression using CRISPRi by driving our dSpyCas9 with the inducible expression system RhaRS/PrhaB and by playing with the length of our CRISPR spacers (Figure 3). To test this system, we needed a titratable inducer system. For this, we selected the inducible expression system RhaRS/PrhaB, induced by rhamnose, based on a paper by Hjelm et al. [3]. To do this, we knocked out two genes in EcN preventing the cells from transporting and catabolizing rhamnose. We then tested this strain by growing it solely on rhamnose as C source and observed a decline in growth (Figure 3). If you want to know more about inducer systems go to the corresponding results page.
Figure 3: Growth experiment with the EcN WT strain and the EcN knockout strain on M9 with 25 mM of rhamnose. Additionally, both strains were grown in M9 with 50 mM of glucose and no rhamnose.
To test the three-step inducible system, an experiment was performed using three different length spacers with known binding capacities (Figure 4). From this graph it can be observed that using the spacers of 9 and 20 nucleotides, we were able to silence the targeted genetic circuit and stop the silencing by increasing the amount of inducer metabolite. In this experiment the system reacts with both spacers at 250 μM at the same fluorescence. Our future goal is to separate these peaks and create two distinguished steps for our system. If you would like to have more insight into this system and how we want to improve it in the future, go to our three-step inducible system result page.
Figure 4: Three separate knock-out Escherichia coli Nissle 1917 strains were created, each containing the double repression plasmid and a plasmid containing monomeric Red Fluorescent Protein and a different length of spacer P1. These three were all put on rhamnose concentrations up to 1000 μM. Furthermore, a positive control, containing both plasmids, but with a non-targeting spacer, was added.
Inducible kill switch to terminate the test
The fourth pill must be taken on day seven of the diagnostic procedure, to induce the kill switch that will terminate the self-test. This pill contains rhamnose at the corresponding concentration to the third level of the three-step inducible system. Even though the engineered EcN population will most likely have disappeared by day seven, as can be seen in the model and Figure 1, we wanted to ensure with this mechanism that our living diagnostic was completely removed from the body once the test finishes. In our project we demonstrate that a highly genotoxic CRISPR-Cas Type-IC mechanism can be induced with rhamnose, effectively killing the EcN cells at concentrations above 3 mM (Figure 5). If you want to know how this kill switch works, have a look at the results page.
Figure 5: Plate reader growth experiment conducted during 24h. Escherichia coli Nissle 1917 (EcN) containing a plasmid with this part was tested in M9 medium with 50 mM of glucose and different rhamnose concentrations (0, 0.5, 1, 3 and 5 mM). The positive control corresponds to EcN wildtype in M9 medium supplemented with 50 mM of glucose.
Confine the living diagnostic tool to the colon environment
Our living diagnostic performs its function in the colon, so we have successfully implemented two biocontainment mechanisms to prevent its survival in other body parts or outside the human body.
To ensure our living diagnostic is viable only within the colon environment, we have made it dependent on mucin, which is the most abundant glycoprotein lining the inner surface of this organ [4]. For this purpose, we have engineered a two-component chimeric system which would only drive the expression of an essential gene upon mucin sensing. Figure 6 shows that in normal colon concentrations of mucin (5-10 %), gene expression dramatically increases in E. coli K12, which is similar to EcN, providing a proof of concept for implementation in our bacteria [5]. For more information in the chimeric system, go see the biosafety results page!
Figure 6: High levels of fluorescence are achieved at standard colon concentrations of mucin in cell sample containing the chimeric two component system.
The second biocontainment mechanism prevents our living diagnostic from surviving outside the human body, i.e., at temperatures below 37ºC. We have effectively implemented a temperature-dependent toxin/antitoxin kill switch system (pcspA) in EcN that does not allow cell growth when the temperature decreases as it can be seen in Figure 7. If you want to see how this mechanism works, go to the biosafety results page.
Figure 7: Cells containing the pcspA system, compared to the control cells in which no kill switch system is present, do not show growth at temperatures below 37°C.
During its time in the colon, our living diagnostic is constantly sensing for high lactate concentrations. We successfully engineered and tested a lactate sensor in EcN that is suitable for the colon environment, meaning that it works in presence of glucose and microoxic conditions. A reporter gene (sfgfp) was placed after the lactate inducible promoter (ALPaGA) and tested at different lactate concentrations, showing an increase of fluorescence with increasing lactate concentrations (Figure 8). However, this system was not able to significantly differentiate healthy from cancerous levels of lactate in the colon.
Figure 8: The response of ALPaGA, expressed as OD600 corrected fluorescence output, at different L-lactate concentrations. The green and red areas signify the L-lactate concentrations in the microenvironment of healthy and cancerous colon cells, respectively.
As a consequence, we developed a model of the lactate threshold system, and a sensitivity analysis was performed to find out which parameter should be modified to refine the dose-response so that it does not recognize healthy-state lactate concentrations but only recognizes cancer-state levels. As a result, we predicted that reducing the intracellular concentrations of L-lactate in our living diagnostic, would allow us to control the operational range of the dose-response (Figure 9). This objective, we believe, could be achieved by changing the expression levels of lactate dehydrogenase. To learn more about how we found out about this, please visit the Lactate DBTL page.
Figure 9: Sensitivity results that show what happens to the dose-response curve when the rate of LldR-lactate complex formation, which is tied to the intracellular lactate concentration, is increased/decreased.
If, regrettably, a lactate concentration corresponding to the cancer-state is detected within the colon by our living diagnostic tool, it will produce a blue chromoprotein capable of colouring the stool. We successfully expressed two different blue chromoproteins, amilCP and Ultramarine, in E. coli Nissle 1917 (Figure 10). However, we did not manage for the chromoproteins to be secreted outside the Colourectal cells yet.
Although we expect the blue colour to be easily visible to the naked eye in the stool, we decided to reduce subjective interpretations as much as possible. To this end, with the help of Tim van der Vooren, we developed a mobile app able to discern normal from blue coloured faeces. To know more about the app, check out our implementation page.
