Unraveling CAR Differentiation Dynamics through in Vitro and in silico Perturbations

Gene and cellular therapies extend the hope of cure to wider range of human diseases. Durability of these therapies is correlated to persistence of the adoptively conferred cells; while mature effector functions are key to effectiveness. Physiologically, cells are rarely simultaneously endowed with indefinite self-renewal and mature effector functions. Cellular therapies must, in response to environmental signals, dynamically transition between self-renewing and mature effector states while avoiding dysfunctional states.

Chimeric Antigen Receptor (CAR) T cells have emerged as effective therapies for B cell and plasma cell malignancies. While they induce prolonged remission in most patients, fewer than half achieve durable control of their disease. Factors associated with durable remissions can be broadly categorized into CAR T cell extrinsic, encompassing malignancy type, tumor burden, antigen escape and the tumor microenvironment, and intrinsic factors, pertaining to the dynamics of transition between the persistent and mature effector states. Strategies to design of CAR T therapies to surmount intrinsic mechanisms of failure are essential.

We carried out CRISPR activation screening targeting all human transcription factors and epigenetic modifiers in anti-CD20 CAR T cells cultured in vitro under conditions designed to provoke exhaustion. We observed statistically significant depletion in sgRNA targeting the TGFb (including KLF10) and TNFa (including NFATC4) signaling pathways in CAR with immunophenotypes associated with durability and effectiveness. Model-based Analysis of Genome-Wide CRISPR-Cas9 Knockout (MAGeCK) for the central memory population revealed enrichment of MFNG, a regulator of NOTCH signaling; NFIB, a master regulator of cell differentiation; SMYD2, an epigenetic regulator controlling T_reg development; and TRIAP1, a TP53 regulated inhibitor of apoptosis. Although biologically plausible, these targets play important roles in regulating proliferation and cytotoxicity and entirely disabling these host-protective mechanisms raises important safety concerns.

Transcription factors and epigenetic modifiers have broad effects on cellular states and regulatory mechanisms. By contrast, cis-regulatory elements (CRE) directly affect only the transcription of genes in their vicinity. Perturbating CRE alters T cell regulation with a finer degree of control and is incompletely explored. Systematically screening the large number of candidate CRE is not feasible via current experimental approaches.

We developed a model, trained on single cell multi-omics data, to predict the time evolution of transcriptomes and regulomes. The introduction of chromatin state in the description of cellular dynamics, permits prediction of cellular dynamics after enhancing or disrupting specific and computes therapeutically salient metrics such as overall proliferation rate and proportions of CAR in memory and functional effector subsets.

Our model predicts that the conditional knock out of the trans-acting factor TCF7 will increase the transition rates from naïve and memory T cell subsets into effector states; a prediction that agrees with the established biological roles for TCF7. By contrast, targeting the disruption of a candidate cis-regulatory element (cCRE) in the vicinity of KLF10 (chr8:101,790,788-101,791,270) is predicted to strongly increase transition probabilities into the Naïve and Memory subsets.

Our model casts the problem of optimizing cellular therapies into the framework of dynamical systems theory and provides a rigorous approach for prioritizing the experimentally intractable number of perturbations by metrics directly relevant to persistence of efficacious adoptive cells. Future directions include the incorporation of additional single cell multiomic modalities to better characterize regulatory dynamics.