Colon cancer is the second leading cause of cancer death in the United States. Each year, almost 140,000 people are diagnosed with colon cancer, and 50,000 people die from the disease. Colon cancer is the name we give to the uncontrolled proliferation of cells the gut or rectum. These tumor cells can metastasize into other parts of the body, leading to life-threatening effects. Colon cancers generally begin as benign polyps in the lining of the large intestine. As cancer cells grow and multiply, they invade the surrounding colon tissue and grow beyond the wall of the colon7.
Meanwhile, Escherichia coli naturally colonize the colon as a part of the gut microflora that serve protective, structural, and metabolic roles. However, E. coli is also the most common bacterial cause of intestinal disease. Some strains of pathogenic E. coli cause diarrheagenic disease while others cause urinary tract infections (UTIs)8.
We wanted to take the E. coli that naturally colonize the gut and engineer them to specifically bind to cancer cells. Our strategy was to co-opt the natural binding system found in some E. coli strains, both pathogenic and commensal, called Type 1 pili. These hairlike appendages have a protein called FimH on the tip that bind to the sugar mannose found on human epithelial cells. The adhesion of pathogenic E. coli to the urinary tract in UTIs is mediated by the FimH adhesin9, and the structure is adapted to withstand high shear forces present in the gut and urinary tract.
As we began thinking about how we would design a system which could bind specifically to cancer cells, we realized we were facing a two part problem. As we describe above, fimH adhesin naturally binds to mannose sugar, which present outside of all eukaryotic cells. This means that even with the addition of a cancer-specific binding peptide, fimH would bind via mannose to cancerous and healthy cells alike. Without careful design consideration, it would be impossible for us to create a targeted therapy. This is why we decided to focus our engineering efforts on first knocking-out nonspecific mannose binding and then reintroducing specific binding via a tumor binding peptide.
To achieve the first goal, we combed the literature for mutations to fimH which would knock out mannose binding. We were able to locate two that had been shown to disrupt mannose binding via standard assays, one a 1 amino acid substitution at site 4910 and another 2 amino acid substitution at site 13611, from here on out to be called mannose KO 49 and mannose KO 136, respectively. We introduced these modifications separately to our fimH-HisTag plasmid via site directed mutagenesis, including the His Tag for reliable measurement. This set us up to to test our hypothesis that these mutations would disrupt mannose-mediated agglutination of yeast cells.
Meanwhile, we searched the literature for a binding peptide that we could insert into the FimH protein so that we could make pili bind specifically to cancer cells. What we discovered was a cancer-binding peptide called RPMrel from the TumorHoPe database. This peptide was used by the 2012 HKUST Hong Kong iGEM Team in their project on colon cancer. RPMrel was discovered using a technique called phage display and tested in vitro in a variety of cancer cell lines. It was tested with a payload of a mitochondrial toxin that selectively killed cancer cells12. We determined that this was the most promising tumor-specific binding peptide for our purposes and inserted it via site directed mutagenesis into N-terminal domain of our mannose KO fimH. We tested the binding of our engineered pili to Caco-2 colon cancer cells, a widely used cell line in laboratories.
Pili Purifications were performed to validate pili expression of these mannose KO constructs. This was especially important because we needed to make sure that a negative agglutination result wasn’t because of some problem with assembling the pili structure caused by our modifications. We needed to show that our pili were fully assembled, but not mannose-binding.
Cultures were grown to a standard OD, induced with either rhamnose, arabinose, both or neither, and pili-purified. Samples were run on an SDS-PAGE gel and stained with Coomassie G250. Images below:
We see a strong band at fimA’s molecular weight for all 16.5 for all arabinose induced cultures of both mannose KOs, indicating the presence of pili. This suggests that even if we can’t detect binding activity of fimH, the mannose-KO constructs are successfully producing pili.
Following the protocol, we mixed OD standardized cultures of our induced and uninduced mannose KO constructs with S. cervisiae yeast along with controls. We found that our mannose KOs did not agglutinate yeast when induced while an identical plasmid without our binding KO mutations under the same conditions did agglutinate. This shows that the pili we determined were being expressed by our mannose KO’s above had little or no mannose binding capability, as determined by a standard assay for mannose binding. This assay was performed in biological triplicates with a similar result. Image below:
Having shown that we could control off-target binding, we were then concerned whether our cancer-mannose KO fusion fimH could express Type 1 Pili. A pili purification was performed.
Cultures were grown to a standard OD, induced with either rhamnose, arabinose, both or neither, and pili-purified. Samples were run on an SDS-PAGE gel and stained with Coomassie G250. Image below:
As with the mannose KOs, we see a band of fim A in each arabinose induced culture containing the cancer binding mannose KO fimH. This indicates that we are assembling pili when we induce the operon, and once again suggests we have a leaky rhamnose promoter.
At this point we had a fimH construct that we believed could bind specifically to colon cancer. To test for specific binding, we developed a dot blot protocol, which would test whether our recombinant cancer-mannose KO could bind specifically to Caco-2 colorectal cancer cells. We obtained our cancer cells from Trevor Nash in the Joshi Lab. We prepared our cells by washing with PBS, adding trypsin, incubating, and adding media. We took out 10 ml to use for our dot blot assay.
At this point the Caco-2 cells were then dotted in 10 ul drops on nitrocellulose paper and blocked in TTBS milk. The blocked blots were bathed in induced and uninduced E. coli containing the cancer-binding mannose KO fimH and operon plasmids, with negative controls, and transferred to a TTBS wash to remove unbound bacteria. Finally, the blots were stained with α-LPS primary antibodies and horseradish peroxidase conjugated secondary antibodies to visualize the E. coli still bound to the Caco-2 dots. This assay was performed with the rpmREL cancer binding peptide fused to both mannose KO 49 and mannose KO 136, because we were unsure if either mannose knockout modification would disrupt rpmREL’s function. The assay was repeated in biological triplicates and we saw some variability between whether the mannose KO 136 +rpmREL or mannose KO 49 + rpmREL had more affinity. The following results are representative of the pattern:
Based on these preliminary results, we were able to control specific adhesion to cancer cells with BactoGrip! This opens up the potential to harness the toolbox of synthetic biology to develop a bacterial therapeutic.
Although we spent most of our time in lab this summer tackling the problem of controlling bacterial adhesion, we were always thinking about the problem that originally provoked our project-- colon cancer. Our wet lab achievements show we were able to achieve controllable, specific adhesion to colon cancer cells. We think this is the biggest and most important step in creating a viable biotherapy to treat colon cancer. Indeed, in our ongoing exploration of the real world impact of our work, we spoke to experts who saw the potential of a targeted biotherapy. We also thought about how our bacteria could act as a therapeutic, including producing a cytotoxic compound in low levels that would be toxic to the cancer cells in close contact with BactoGrip bacteria who targeted them, but not toxic to healthy cells which BactoGrip would not interact with. There are some obvious safety concerns with this approach, which we explored with experts. This led us to think about ways in which we could “trigger” the adhesion and release of BactoGrip to cancer cells, which led us to collaborate with Worcester Polytechnic Institute who worked this summer on a novel method for disrupting bacterial adhesion in biofilms by using antifreeze proteins. In the future, we are most interested in developing a binding system built on BactoGrip which could not only bind specifically, but also transmit a signal to the cell in response to a binding event. We are also concerned with the future safety concerns involved with our project, and incorporated feedback from our safety experts by changing our chassis strain to E. coli Nissle.