Theses Doctoral

Microfluidic and computational technologies to improve cell therapy manufacturing

Anandakumaran, Priya Nivashini

Cell therapies are an emerging form of therapy, with the potential to treat and cure a variety of diseases. As more cell therapies become approved and commercialized, challenges remain in the manufacturing of these often single-batch products due to their complexity and patient-to-patient variability, which limit their cost-effectiveness and reproducibility. In this dissertation, we aim to improve the manufacturing of two different cell therapies, namely, organoid-based cell therapies using hydrogel scaffolds, and adoptive cell therapies using deep learning and microfluidics, to facilitate their widespread clinical use.

First, we develop new tools to manufacture organoids, which are widespread in drug-screening technologies, but have been sparingly used for cell therapy as current approaches for producing self-organized cell clusters lack scalability or reproducibility. Here, we use alginate microwell scaffolds to form pre-vascularized organoids composed of endothelial cells and mesenchymal stem cells, where the size and structure can be readily tuned by varying the cell source, ratio of cells, or size of the microwells. Furthermore, by uncrosslinking the alginate scaffold, the organoids can be harvested in a gentle manner without damaging their structure or impairing their functionality. Finally, we assess the ability of the pre-vascularized organoids to restore vascular perfusion in a mouse model of hindlimb ischemia. By making use of the dynamic nature of hydrogels, this method can offer high yields of reproducible, self-organized multicellular aggregates for use in cell therapies.

Next, we shift our focus to the identification of antigen-specific T cells, which is a critical step in the manufacturing of adoptive cell therapy. Conventional techniques for selecting antigen-specific T cells are time-consuming, making them difficult to adapt for large-scale manufacturing, and are limited to pre-defined antigenic peptide sequences. Here we train a deep learning model to rapidly classify videos of antigen-specific CD8+ T cells by distinguishing the distinct interaction dynamics (in motility and morphology) between cognate and non-cognate T cells and dendritic cells (DCs). The model is able to classify high affinity antigen-specific CD8+ T cells from OT-I mice with an area under the curve (AUC) of 0.91, and generalizes well to other types of high and low affinity CD8+ T cells. We also show that the experimental addition of anti-CD40 antibodies amplifies the differences between cognate and non-cognate T cells and DCs, thereby improving the model’s ability to discriminate between them. This workflow can be used to better understand the role of cognate T cell – DC interactions in the pathogenesis of cancer and autoimmune diseases, and can be integrated into a device to simplify and accelerate the selection of antigen-specific T cells for use in adoptive cell therapy.

Finally, we sought to develop a device to address two other issues associated with the selection of antigen-specific T cells: low-throughput screening, and the inability to assess a mixed population of T cells against a library of antigens, both of which are necessary to identify rare T cells, and improve clinical outcomes of the corresponding cell therapy. A few specialized assays exist that can assess T cells against multiple antigens, but they are often limited by an increased manufacturing burden. Here, we develop a microfluidic artificial lymph node, which is inspired by the efficient selection of antigen-specific T cells in vivo. In particular, our flow-through design consists of multiple compartments, each containing microcarrier beads coated with DCs presenting a distinct antigen, such that T cells that are flowed sequentially through each compartment can stably arrest to cognate DCs, becoming captured in the appropriate compartment. We test a single-compartment device computationally using agent-based simulations, and experimentally using a mixed population of antigen-specific and wild-type (WT) (non-specific) T cells, and in both cases we observe a preferential accumulation of cognate, antigen-specific T cells. This proof-of-concept single-compartment device can be readily scaled up to systematically test many T cells against multiple antigens.

Underlying this work is the development of technologies to enable the large-scale manufacturing of cell therapies. Cell therapies are undergoing a transformation to a new class of therapeutic modality, and there are many emerging questions, especially related to the scale-up and scale-out of production processes. Together, this work aims to engineer technologies to improve cell therapy manufacturing processes, facilitate their clinical translation, and ensure their availability to all patients who would benefit from them.


  • thumnail for Anandakumaran_columbia_0054D_16859.pdf Anandakumaran_columbia_0054D_16859.pdf application/pdf 8.45 MB Download File

More About This Work

Academic Units
Biomedical Engineering
Thesis Advisors
Sia, Samuel K.
Ph.D., Columbia University
Published Here
October 13, 2021