(L-459) A cell-based assay and image analysis program to evaluate protein binders for EV targeting

Introduction:: Extracellular vesicles (EVs) are lipid bilayer nanoparticles produced by cells that have recently garnered interest as biologic drug delivery systems. Targeting EVs to specific tissues can be achieved by loading EVs with binder proteins on their surface, such as antibodies, that target proteins on tissues of interest. However, when a new cell type or surface marker is being targeted, it is not always clear which binders will correctly bind their target when displayed on the surface of an EV. EV-cell binding assays can be used to test binders, but this workflow is often inadequate due to the slow, low throughput process of EV harvesting. What is needed is a method to rapidly quantify the capacity of membrane-bound, cell-expressed binder proteins to bind target cell types. Here, we present a new method to quickly evaluate such a capacity. Since binders on EVs are anchored in a membrane, we hypothesized that we could screen candidate binders in a cell-cell binding assay rather than an EV-cell binding assay, resulting in faster and higher throughput of candidate screening. Specifically, we developed a workflow and image analysis program that quantifies cell-cell binding of fluorescently-labeled cells displaying binder proteins with cells displaying the cognate target. To demonstrate the utility of this workflow, we recapitulated a T cell-targeting study by evaluating T cell-targeting of an αCD2 single-chain variable fragment (scFv) on 3 different membrane-anchored scaffolds. Our method presents a user-focused pipeline that should accelerate the identification of targeting molecules and improve EV-mediated therapeutics.

Materials and Methods:: To evaluate binder-target interactions, we used an ELISA-like, cell-based assay and an image analysis program (Figure 1). Briefly, HEK293FTs were transiently transfected with DNA encoding an αCD2 scFv binder tethered to different linkers and separate DNA encoding a red mKate2 protein. Jurkat T cells were dye-labeled with green DiO and natively express CD2. HEK293FTs were adhered to the plate, and Jurkats were added in suspension to the same plate to allow for cellcell interactions. The plate was then washed to remove any unbound cells, and microscopy images were taken using fluorescence channels corresponding to the fluorescence of the two cell populations. A strong binder results in more green cells (Jurkats) remaining after washing than a weak binder. The images were then fed into a custom image analysis program written in Python whose output is a pixel ratio of fluorescent pixels in each channel. Each fluorescence image was converted into greyscale where each pixel represents a single-value intensity. With input from the user, the program performed a background subtraction step and counted the number of remaining fluorescent pixels in each channel. This was repeated for both channels, and the program divided the number of pixels from the Jurkat (CD2) cells by the number of pixels from the HEK293FT (αCD2 scFv) cells. This pixel ratio was then used as a quantitative measure to evaluate the binding of binder and target proteins, allowing for down selection of the best binders in the evaluation pool.

Results, Conclusions, and Discussions:: The cell-based assay and image analysis program closely replicated the rank-order of binders in the original study mentioned in the introduction. The linkers used for αCD2 scFv were helical, hinge, and 40 glycine-serine (40 GS), and a non-targeted condition was included as a negative control for residual binding (no scFv). As seen in Figure 2, all linkers resulted in a significantly higher pixel ratio when compared to the negative control indicating that our method can distinguish between the occurrence and lack of binding. Excitingly, the helical and 40 GS performed similar to one another and outperformed the hinge linker – this rank order recapitulates the results produced when evaluating these binders on EVs. This recapitulation of EV-cell binding supports our hypothesis that the cell-cell binding assay can be used as a useful screening tool for evaluating binding and targeting of binders on the surface of EVs. Of note, a single cell-based assay is performed in 3 days while a typical EV-cell assay takes 7 days, significantly improving analysis time. This meaningful reduction in experimental time will enable more binders to be evaluated. In future studies, we wish to explore and extend this method in several ways. First, we plan to evaluate the robustness of our workflow to experimental design choices. For example, one parameter to explore is whether the cells being evaluated grow in a suspension or adherent state. Different permutations of adherence and suspension can be investigated in order to generalize the proposed methodology to different cell types depending on the user’s application. In the same approach of generalization, different methods of fluorescence labeling will be explored to understand how labeling method affects the pixel ratio metric in order to support future user needs. In conclusion, we have developed an accelerated method of quantifying cell-cell binding in order to down select candidate binder or target proteins for applications in EV targeted delivery. This method should enable more rapid screening of targeting proteins, which will offer better targeting for EV-based therapeutics.

Acknowledgements (Optional): :

References (Optional): :