Review our results below!

Adhesion Results

Figure 1. Optical density measurements of GFP and anti-GFP cultures. The figure above displays OD600 values of GFP and anti-GFP cultures, with and without Atc induction (first, third, and fifth columns are not induced, while second, fourth, and sixth are). The optical density was measured at three different intervals to measure the changes in cell density within the supernatant. The error bars represent the standard deviation.

Figure 2. Optical density measurements of C-tag and anti-C-tag cultures. The figure above displays OD600 values of C-tag and anti-C-tag cultures, with and without Ara induction (first, third, and fifth columns are not induced, while second, fourth, and sixth are). The optical density was measured at three different intervals to measure the changes in cell density within the supernatant. The error bars represent the standard deviation.

Figure 3. Optical density measurements of GFP and anti-GFP cultures. The figure above displays OD600 values of GFP and anti-GFP cultures, with and without Atc induction (blue columns are not induced while red are). The optical density was measured at three different intervals to measure the changes in cell density within the supernatant.

There is strong adhesion between the C-tag antigen and nanobody cells as seen by the large clumps of aggregated bacteria in Figure 4 and decreasing optical density of the corresponding supernatant, as indicated in Figure 2. However, the GFP antigen and nanobody cells display no adhesion, as indicated by Figure 1 with increasing optical density measurements of the corresponding supernatants. We also observe little fluorescence between the mixed cultures of the GFP antigen and nanobodies, as seen in Figure 5. We were led to hypothesize that the GFP antigen we utilized was faulty as the GFP nanobody was able to bind to purified GFP, as showcased in Figure 5. We switched out the avGFP antigen for EGFP and tested its efficiency through our aggregation assay. Unfortunately though, as demonstrated in Figure 3, the GFP nanobody and EGFP antigen cells show no adhesion, with optical density measurements of each supernatant increasing over time.

Secretion Results

Over the course of the summer, a series of Western Blot experiments were conducted, yielding four successful outcomes after addressing issues related to experimental Western Blot procedures and secretion assays. The data from all four blots consistently demonstrated high GFP expression within the cell pellet and minimal GFP expression in the supernatant. These preliminary findings raised the possibility that all five secretion constructs encountered challenges in releasing the GFP molecule, initially suggesting a concern with our plasmid constructs. However, before definitively ruling out the viability of these constructs, we embarked on an investigation to assess their ability to release the EGF ligand.

To investigate this, we collected supernatants from bacteria expressing proteins induced for 24 hours and plated them on ERK-activated HEK29T cells, equipped with a BioSensor.This biosensor came from the findings of the following paper Regot et, al (2014) High-sensitivity measurements of multiple kinase activities in live single cells. ERK activation is a downstream event in the EGF receptor pathway; thus, successful ERK activation implies proper functioning of the EGF pathway. Activated of Erk is shown in these biosensors when ERK-KTR is localized in the nucleus upon ERK activation for a period of time before eventually heading back tot he cytoplasm. Visually this results in the appearance of a darker nucleus. On the other hand if ERK expression is not found to be active, no ERK-KTR localization will occur- (the cell nucleus will not change).Through a time series experiment involving the plating of 100 µL of protein-secreted supernatant onto the cells, we observed induced ERK activation in supernatants originating from specific protein-secreted plasmids

Note on Controls: The EGF Control refers to the sample in which 50 nM concentration of pure EGF ligand was added onto ERK biosensor Mammalian cells. Because these biosensors have the ability to detect ERK activation, with the addition of the pure EGF ligand to the cells, we expected to see ERK-KTR activation. Thus this control was used as a relative measure of comparison to see what ERK activation would look like on our HEK-29T Mammalian Cells. The Optimem control was used to see whether the media the cells were plated on would activate ERK expression. This was a necessary control to verify the the ERK biosensors were solely dependent on EGF ligand binding on not other experimental factors.

As observed by the plasmids in comparison to the Opti Mem and EGF control plasmids, plasmids containing Signal Peptide 6 and 7, the O33434 signal peptide and the Amylase Signal Peptide respectively were able to successfully secrete the EGF ligand, triggering ERK activation when the protein supernatant was added onto our Mammalian Cells. Plasmid 4, containing the Cholera Signal Peptide Toxin also showed some extremely interesting results. When the protein supernatant was plated on top of the ERK activated Mammalian Cells, immediate apoptosis was seen. This is characterized by the interesting blebbing of the cells that can distinctively be seen at the 22.5 time after supernatant addition on the figured above. The reasons behind the activation of apoptosis for the Cholera Toxin derived signal peptide will need further research/experimental insight to account for it. But it is suspected that the EGF ligand interacted with the signal peptide in someway to program apoptosis of the cell.

Intriguingly, the supernatants associated with ERK activation showed no apparent correlation with GFP secretion on the Western Blots. (GFP secretion could not be detected via Western Blot in supernatant samples of our constructs)

Similar phenomena, where fluorescent protein detection was absent while other proteins were detected in fusion protein samples, have been documented in various protein fusion experiments. Researchers noted that when GFP was fused to a signal sequence optimized for the secretion of the malE (maltose-binding protein gene), fluorescence was not detectable1). However, when the signal sequence was omitted, fluorescence was observed. This observation suggests that challenges may arise when GFP is fused with a signal protein, further indicating that GFP encounters folding difficulties during secretion encountering for lack of observed fluorescence).

Further investigation into literature revealed several potential reasons for these difficulties. Firstly, GFP, when secreted has been found to exhibit a preference for secretion through the TAT pathway rather than the SEC pathway.(3) Studies have shown that GFP tends to fold properly within the cytoplasm, rather than refolding in the periplasm. Although there is emerging evidence that certain GFP variants can be secreted via the TAT pathway, most literature has shown that GFP folding only occurs via the SEC pathway (3). Additionally, the accumulation of protein aggregates in the GFPD2 channel, the main transport machinery in the T11SS that moves proteins from the periplasm into the extracellular space, has been linked to challenges in proper protein folding(2). Given that EGF is secreted prior to GFP, it is plausible that an excessive accumulation of EGF in the channel may impede the successful transport and detection of GFP.

While we successfully demonstrated our ability to design bacterial constructs capable of secreting the EGF ligand by showcasing that proteins in the bacterial supernatant (secreted into the extracellular matrix) were sufficient in activating ERK expression, we sought to determine whether these constructs could secrete a sufficient quantity of protein to induce ERK expression without undergoing lysis. To address this question, once we had established the successful development of the constructs, we conducted a microscopy experiment using live bacterial cell cultures. This experiment proved that our engineered bacterial constructs secrete enough protein to affect ERK activation in Mammalian Cells.

Integration Results:

Bacterial construct phenotype

Mammalian cell line phenotype

Mammalian cell line phenotype

Surface-expressed anti-GFP

Surface-expressed GFP

Consistent adhesion

Surface-expressed eGFP (new GFP construct)

Surface-expressed anti-GFP

Inconsistent adhesion; issues related to GFP folding, intimin protein, bacterial lysis

Surface-expressed EGFR nanobodies (EgB4 and 7D12)

HEK293T (blank)

No adhesion

Surface-expressed C-tag

Surface-expressed anti C-tag

Not verified

Surface-expressed anti C-tag

Surface-expressed C-tag

Not verified

Microscopy showed that bacterial adhesion was not random and that the nanobody-antigen adhesion approach resulted in specific binding to cells. Furthermore, more binding was observed on cells with higher surface expression levels.

GFP-displaying bacteria (green) binds to membrane-expressed anti-GFP on HEK293T cells (red membrane).
Anti-GFP-displaying bacteria (red) binds to membrane-expressed GFP on HEK293T cells (red and green membrane).

Control: Blank bacteria (red) does not bind membrane-expressed GFP on HEK293T cells (red and green membrane).

We corroborated our endocytosis hypothesis with a DAPI test, staining the bacteria after an overnight culture. As expected, bacteria that were not adhered to the cells absorbed the stain, while bacteria that were adhered did not (indicating that they were within the mammalian cell membrane).
DAPI staining to confirm bacterial endocytosis.

We attempted to cotransform both the secretion and adhesion constructs along with a labeling plasmid into our E.Coli cells and co-culture this with mammalian cells. To do this, the pPF058 ERK biosensing mammalian cells were transfected with a plasmid containing surface expressed GFP (pBR20). This mammalian cell line was then sorted and co-cultured with the triple-transformed bacterial constructs and examined via confocal microscopy through a time series. However, due to issues with culturing our mammalian cells (did not grow properly), we were not able to evaluate this “triple transformed'' construct with a high amount of certainty.

Modeling Results

Based on our modeling work, we derived the following equation to model the efficacy of combined bacterial adhesion and ligand secretion versus secretion alone:

Ls = concentration of secreted ligand from bacteria that do adhere to HEK293Ts

Lns = concentration of secreted ligand from bacteria that do not adhere

R = mammalian cell radius

K = rate of ligand degradation

D = rate of ligand diffusion

Ls/Lns is directly related to the size of the cell (cell radius/3) and the rate of ligand degradation (sqrt(K)), and inversely related to the rate of ligand diffusion (sqrt(D)), since the EGF ligand is subject to degradation once it has been secreted.

Our model finds a strong relationship between cell size and Ls. In other words, for large mammalian cells, adherent bacteria will be more efficient at secreting higher concentrations of bacteria than non-adhered bacteria. Furthermore, for secreted ligands which degrade quickly or are slow to diffuse, adhered bacteria secrete a higher concentration of ligand to the target cell compared to non-adherent bacteria. Varying size has the greatest impact on the efficacy of “stick-and-secrete” versus secretion alone, as degradation and diffusion rates are subject to square roots.

Thus, we hypothesize that in large-cavity cells like tumor cells, our “stick-and-secrete” is preferable against bacteria which only secrete and do not bind to cells. Our system is also preferable for cases in which secretion ligands degrade rapidly or have slow diffusion rates.

Future Work

Secretion Future Work

To address the challenges related to GFP detection observed in the Western Blots, the following future experiments are proposed. Firstly, it would be valuable to create a control construct without GFP fused to the signal peptide construct. This will help establish whether the challenges in GFP detection are indeed related to its fusion with the signal peptide or if there are other underlying factors affecting its detection.

Additionally, conducting Western blot analysis using an EGF antibody for detection, rather than the GFP antibody, could provide insights into the expression and secretion of EGF, which is known to precede GFP in these experiments. This alternative detection method might reveal any discrepancies in EGF expression that could be linked to the observed difficulties with GFP detection. It also circumventing the need to use GFP at all, completing cutting out the issues of proper GFP folding.

Another promising approach would be to employ a different detection method such as the use of FLAG tags over GFP.. They were similarly to our use of GFP in our methodology in that they attach to a protein and are detected via antibody recognition. However because they are smaller and easily detected, they offer greater benefits that GFP (we likely won’t run into a protein folding issue with FLAG tags because of their size)

Interkingdom Future Work

EGFR (epidermal growth factor receptor) is a receptor naturally found on human cells. However, our EGFR-nanobody displaying bacteria (verified via aggregation assay) did not bind to the human cells. We hypothesize that our HEK293T strain does not express enough EGFR on its surface and that there were not enough places for the E. coli to bind. In future experiments, we would use HEK293Ts which overexpress EGFR to test these E. coli constructs.

Additionally, we would like to verify the C-tag/anti C-tag adhesion pairs–the mammalian constructs for these pairings were transfected and sorted, but limited lab time kept us from testing them against their complementary bacterial constructs.

We would also like to repeat these microscopy experiments with a new fluorescent label (such as BFP) on the bacteria in order to be able to better distinguish bacteria from mammalian cells.

To ensure reproducibility, we would also like to more thoroughly quantify our integration experimental results, either by using macros for image fluorescence analysis or by other methods. Specifically, we would like to quantify (1) expression levels of receptors on HEK293T cells, (2) expression levels of nanobodies on E. coli, and (3) binding efficacy of E. coli to HEK293T cells in order to evaluate the validity of our “stick-and-secrete” system.

With regard to moving away from a proof-of-concept stage project, we would also repeat these trials in bacterial strains that are more likely to be used in clinical applications, such as the probiotic Nissle E. coli or S. Typhimurium (Huang et al, 2021).

Eventually, we hope to design the a singular plasmid with controlled capability of adhesion and secretion together with a bluorescent biomarker to avoid the need of double and triple transformations. Not only would this unify the while project but also contribute to the ease of ease of experimentation and collaboration.


Dammeyer, T., & Tinnefeld, P. (2012). Engineered fluorescent proteins illuminate the bacterial periplasm. Computational and Structural Biotechnology Journal, 3(4). (3)

Feilmeier BJ, Iseminger G, Schroeder D, Webber H, Phillips GJ. Green fluorescent protein functions as a reporter for protein localization in E. coli. J Bacteriol. 2000 Jul;182(14):4068-76. doi: 10.1128/JB.182.14.4068-4076.2000. PMID: 10869087; PMCID: PMC94594.(1)

Huang, X., Pan, J., Xu, F., Shao, B., Wang, Y., Guo, X., & Zhou, S. (2021). Bacteria-Based Cancer Immunotherapy. Advanced Science, 8(7), 2003572.

Wang, H., & Chong, S. (2003). Visualization of coupled protein folding and binding in bacteria and purification of the heterodimeric complex. Proceedings of the National Academy of Sciences, 100(2), 478–483. (2)