This week we profile a recent publication in Nature Biotechnology from Dr. Daniel Kirouac (pictured) and a team in Dr. Peter Zandstra’s lab at Notch Therapeutics and UBC.
Can you provide a brief overview of your lab’s current research focus?
Notch Therapeutics is working to maximize the benefit of T cell therapies. Current T cell therapies (e.g., Chimeric Antigen Receptor T cells (CAR-Ts)) require patient-specific manufacturing, resulting in significant batch variability and extensive ‘vein-to-vein’ wait times, negatively impacting patient outcomes. Our iPSC-based technology platform enables precision control of Notch signaling, a pathway required to induce T cell development from stem cells. This proprietary T cell-production platform combines sophisticated product design with commercial-compatible manufacturing, delivering the ability to design and manufacture a uniform and unlimited supply of therapeutic T cells. Our growing team combines expertise in stem cell biology, immunology, bioengineering, cell manufacturing, genomics, and systems biology. The systems biology group uses advanced data science (mathematical modelling, bioinformatics, and machine learning) to support research and development projects. One fundamental question we are tackling is, what are the design specifications for an optimal T cell therapeutic? In this paper, we attempt to answer this question using publicly available clinical data.
What is the significance of the findings in this publication?
Genetically engineered T cells hold immense promise for the treatment of cancers and other diseases. However, clinical experience with CAR-Ts has revealed that they make unwieldy therapeutics. These ‘living drugs’ proliferate, differentiate, actively traffic between tissues, and interact with patient immune systems in complex and poorly understood ways. The resultant pharmacology is highly variable between patients, obscuring the relationship between administered dose, exposure, and clinical outcome.
We hypothesized that the principles governing T cell dynamics during infection also govern the pharmacology of CAR-Ts. We tested this theory using a mathematical description of T cell regulatory control, wherein transitions between T cell states (memory, effector, and exhausted) are coordinated by tumor antigen engagement. The model was trained on clinical data from CAR-T products in different hematological malignancies, yielding biological insights as well as the ability to make quantitative clinical predictions about patient outcomes.
Our model predicts that poor clinical responses, characterized by minimal expansion of CAR-T cells in circulation and marginal tumor reduction, are caused by cell-intrinsic deficits in the infused CAR-T product. Specifically, deficits in the proliferative capacity of memory cells and the cytotoxic potency of effectors are the primary determinants of response. We confirmed this prediction using single-cell transcriptome data from pre-infusion CD19 CAR-T products matched with clinical outcomes in different hematological malignancies. Memory cell sub-populations from poor responders display transcriptional features indicative of diminished function. Using a machine-learning workflow we demonstrate that product-intrinsic differences can accurately predict patient outcomes based on pre-infusion transcriptomes, and do so with better accuracy than standard immunophenotyping, while additional pharmacological variance arises from cellular interactions with patient tumors. We believe these insights and the computational tools we have made available will enable a new phase of predictive CAR-T product development.
What are the next steps for this research?
As we work to create T cell therapies with enhanced efficacy, safety, and batch-to-batch consistency, robust predictive models are required to guide product design and treatment. Transcriptome profiling combined with our machine learning workflow can be used to evaluate the proliferative capacity and potency of novel cell products. We are using this internally at Notch Therapeutics for product development and optimization, and we believe others could do so as well, both as a screening assay during development and potentially for release testing. Our mathematical model can be used to simulate T cell pharmacokinetics and test alternate dosing protocols in silico, optimizing treatment regimens which maximize anti-tumor response while minimizing excess toxicity. We are applying this to our pre-clinical pharmacology studies, and plan to extend to clinical trial design and analyses once we get there.
Our goal is to develop cell therapies that maximally benefit patients, and this work represents progress towards determining the design specifications of such products. We’ve developed computational methods to sift through the biological complexity and quantify features that are predictive of positive patient outcomes. Further improvements on these methods require additional clinical data for model training, testing and refinement. However, the availability of such data remains limiting. Hundreds of CAR-T clinical trials are in progress – if the underlying data from these were to be made publicly available, the scientific and medical community could fully leverage the power of computational models to design and deliver the next generation of T cell therapies.
