Measuring Growth and Gene Expression Dynamics of Tumor-Targeted S. Typhimurium Bacteria

Using this procedure, we are able to generate time courses for the growth and gene expression dynamics of bacteria colonizing tumors. While these measurements are routinely performed in vitro in batch culture or microfluidics devices, they are much more difficult to perform in vivo.

There are several modifications that can be applied to these methods. While we used the OVCAR-8 cell line to generate our mouse models, a number of other cell lines may be used equivalently. For example, luciferized cell lines may be used that allow for 3D colocalization of tumor and bacterial luciferase channels. Additionally, while we used a bilateral hind flank model, a different number and locations of injection sites may be used to generate models to study different phenomena. For example, if tumor size is an important variable, serial dilutions of cells can be prepared and injected to different sites simultaneously, generating smaller- and larger-sized tumors. Finally, while we used S. typhimurium strain ELH430, other strains of S. typhimurium or other bacteria such as C. novyi, E. coli, V. cholorae, B. longum may be used equivalently with these methods, though the bacterial counts for injection may need to be adjusted.

The most critical steps in this protocol involve the creation of the mouse cancer model, and the preparation and injection of bacteria. For the creation of the cancer model, precision in the cell count and concentration, injection volume, and injection site placement will greatly enhance the consistency of the tumors generated. We have found that equivalently sized, shaped, and placed tumors create the most quantitatively comparable data. For the preparation of bacteria, it is critical to completely wash them prior to injection and to control for the bacteria concentration, volume, and injection site carefully. These steps are required to avoid excessive immunological impact of the injection, and to maintain quantitative comparison between separate mouse models. Since the number of bacteria that enter the tumor is a small fraction of the injected dose, variations in the number of injected bacteria between mice can result in dramatic differences on colony progression. For additional information, see several additional references on bacterial preparation for quantitative in vivo studies ^{3, 4},

While these techniques are highly adaptable to a variety of specific studies, some limitations exist for extending to certain applications. For example, intravenous delivery of high bacterial counts may not be desirable for certain bacteria and/or mouse models due to immunological impact. One alternative route of administration is oral delivery (gavage), a technique which is described on JoVE²³. Additionally, time-lapse imaging can be quite labor-intensive depending on the specific experiment length, time-resolution, and equipment available. Automating the imaging process would alleviate many of these burdens, but also introduces new technical challenges. Mouse vitals such as heart rate, breath rate, and temperature must be continuously monitored by additional equipment. Additionally, for longer-term anesthesia, eye ointments must be used to maintain adequate moisture. Finally, the level of anesthetic delivered to the mouse must be continuously adjusted to maintain the proper depth of anesthesia. A closed-loop feedback control system that integrates mouse vital measurements with anesthetic delivery would greatly facilitate this process.

While synthetic biology has achieved much success^{17, 19, 20}, applying engineered gene circuits to in vivo applications will require quantitative, in vivo time courses to inform design principles. S. typhimurium is an excellent strain for clinical synthetic biology for cancer therapy since it is similar to E. coli, specifically colonizes tumors, and has been shown safe in human clinical trials^{6, 12, 14, 22}. Additionally, recent studies have found that many published circuits function in S. typhimurium without modification²⁴. Using the experimental platform described here, we recently described how dynamic in vivo expression profiles can be programmed using the inherent instability of plasmid vectors²⁵. In addition, recent research has developed novel methods for stably maintaining plasmid vectors in vivo²³.

While luciferase is a common reporter used for in vivo imaging for its sensitivity⁵, there are examples of excellent studies on quantitative in vivo gene expression using GFP²⁶. Using GFP would allow the researcher to study the spatial-temporal dynamics of individual bacteria in vivo. Additionally, while we used a subcutaneous model, extending these studies toward more relevant models of cancer would further increase their predictive power⁷. Future studies like this may enable a next generation of in vivo synthetic biology for clinical applications.

Developing both experimental and computational techniques in concert will be critical to engineering in vivo genetic circuits. Computational modeling can rapidly probe system parameters to explore potential outputs but must remain closely tied to experimental results to remain relevant. Thus far, these results have been difficult to obtain quantitatively, due in part to a lack of methods for studying genetic circuits in vivo.

Building on this platform, future applications will include engineered gene circuits that further extend the range of expression dynamics, sensing tumor-specific stimuli and self-regulating cargo production as needed. As the sophistication of these in vivo circuits increases, the requirements on quantitative data required to tune them will also increase. We believe that the methods described here, along with new innovations in long-term mouse imaging, will make this process possible.