Graphical AbstractHighlights d SpliceAI, a 32-layer deep neural network, predicts splicing from a pre-mRNA sequence d 75% of predicted cryptic splice variants validate on RNA-seq d Cryptic splicing may yield 10% of pathogenic variants in neurodevelopmental disorders d Cryptic splice variants frequently give rise to alternative splicing A deep neural network precisely models mRNA splicing from a genomic sequence and accurately predicts noncoding cryptic splice mutations in patients with rare genetic diseases. SUMMARYThe splicing of pre-mRNAs into mature transcripts is remarkable for its precision, but the mechanisms by which the cellular machinery achieves such specificity are incompletely understood. Here, we describe a deep neural network that accurately predicts splice junctions from an arbitrary pre-mRNA transcript sequence, enabling precise prediction of noncoding genetic variants that cause cryptic splicing. Synonymous and intronic mutations with predicted splice-altering consequence validate at a high rate on RNA-seq and are strongly deleterious in the human population. De novo mutations with predicted splice-altering consequence are significantly enriched in patients with autism and intellectual disability compared to healthy controls and validate against RNA-seq in 21 out of 28 of these patients. We estimate that 9%-11% of pathogenic mutations in patients with rare genetic disorders are caused by this previously underappreciated class of disease variation.(legend continued on next page) (F) Relationship between exon-intron length and the strength of the adjoining splice sites, as predicted by SpliceAI-80 nt (local motif score) and SpliceAI-10k. The genome-wide distributions of exon length (yellow) and intron length (pink) are shown in the background. The x axis is in log-scale. (G) A pair of splice acceptor and donor motifs, placed 150 nt apart, are walked along the HMGCR gene. Shown are, at each position, K562 nucleosome signal and the likelihood of the pair forming an exon at that position, as predicted by SpliceAI-10k. The genome-wide Spearman correlation between the two tracks is shown. (H) Average K562 and GM12878 nucleosome signal near private mutations that are predicted by the SpliceAI-10k model to create novel exons in the GTEx cohort.

No abstract

No abstract

The development of clinically effective CAR-T cell products for solid tumors will require substantial cell engineering to confer sufficient specificity, potency, and persistence. Advances in genome engineering and synthetic biology have provided an increasingly complex set of features that can be introduced into CAR-T cells to augment their function. However, combining multiple features may result in unpredictable negative interactions between components. Here, we report the use of high-throughput screening to optimize the design of a highly-engineered Integrated Circuit T Cell (ICT) product for the treatment of clear cell renal cell carcinoma (ccRCC). ICT cells are CAR-T cells that contain an AND logic gate requiring two antigens to be present to trigger tumor cell killing together with multiple enhancement modules. First, to create the logic gate we generated hundreds of novel scFv and VH/VHH binders targeting PSMA (as a priming target) and CA9 (as a cytolytic target) via two parallel de novo binder discovery efforts: 1) transgenic mice immunizations and 2) internally-developed phage display panning campaigns. Two independent arrayed screens with 500 PSMA prime receptors (PrimeRTM) and 750 CA9 CARs were conducted to find PrimeRs with high inducibility and CARs with strong on-target potency. From these screens, the top 25 PSMA PrimeRs and 20 CA9 CARs were combined with an shRNA cassette for targeted knockdowns along with two variations of a persistence module. We used a fully-automated workcell to perform end-to-end arrayed screening of the resulting 1,000 member library in T cells engineered from four human donors. Non-viral editing techniques were used to electroporate primary CD4/CD8 cells and robotic handlers were used to set up co-cultures. Circuit specificity and potency were assessed by flow and cytokine secretion and resistance to exhaustion was assessed in a seven day killing assay. Although the library was built from components that functioned well independently, we found that when combined, many of the circuits displayed suboptimal function. Integrated screening identified 20 variants that each far exceeded the performance of a small set of initial prototypes built from “best-guess” selections of individual components. The final candidates are significantly superior to constitutive CAR-T cells in a long term killing assay, show potent cytotoxicity of low expressing antigen lines, and display background levels of cytotoxicity against single antigen targets. Engineering multiple features into T cell products is limited by unpredictable negative interactions between components. We have overcome this limitation by using high-throughput screening which generated development-ready candidates for ccRCC with finely tuned desirability criteria in <18 months. Citation Format: Nishant Mehta, Jamie Thomas, Edward Yashin, Andrea Fua, Jonathan Li, Jonathan Chen, Laura Lim, Je Chua, Andrew Cardozo, Marian Sandoval, Duy P. Nguyen, Ziyan Hong, Jimmy Wu, Catherine Sue, Gustavo Guzman, Li Wang, Sofia K. Panagiotopoulou, Sophie Xu, Angela C. Boroughs, W. Nicholas Haining. High-throughput arrayed screening of logic-gated CARs enables the selection of candidates for ccRCC with optimal potency and fidelity traits [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 1783.

Chimeric antigen receptor T cell (CAR T) therapy has demonstrated unprecedented therapeutic activity in hematologic malignancies. However, generating potent clinical responses against solid tumors remains a challenge for CAR T therapy. As the field strives to improve the therapeutic efficacy of CAR T cells with novel target antigens and enhanced potency, the risks of on-target toxicity pose a major barrier to progress. To address these challenges, we have developed engineered CAR T cells to target solid tumors through AND logic gates, where CAR expression is conditionally induced by a transcription factor released from a priming receptor (PrimeRTM) upon binding to the PrimeR antigen. The AND gate limits off-tumor toxicity as it requires both CAR and PrimeR antigen expression in the tumor microenvironment. To ensure PrimeR expression and signal transduction upon antigen binding, while minimizing residual ‘‘leaky’’ CAR induction in the absence of PrimeR antigen, we screened hundreds of PrimeR binders using both arrayed and pooled strategies. In an arrayed strategy, we engineered T cells from four donors in multiwell plates using CRISPR-mediated, non-viral, site-specific integration of logic gates bearing a variable PrimeR binder and a fixed MSLN CAR. In addition, we employed a pooled screening strategy, where we engineered T cells from two independent donors with a pool containing a subset of >300 of the same logic gates. Engineered T cells from both strategies were co-cultured with cell lines bearing either both CAR and PrimeR antigens or a single antigen, in order to evaluate fidelity and on-target functionality. In the arrayed setting, on-target functionality was quantified based on the levels of CAR induction, cytokine secretion, T cell activation, and target cell killing in the presence of both antigens, while fidelity was assessed based on the absence of these activity signals in the presence of a single antigen. In the pooled setting, sorting based on functional markers was performed and sequencing was used to quantify the relative abundance of cells with each logic gate in different sorted populations. On-target activity and circuit fidelity were then quantified based on enrichments in different sorted populations. Results from the pooled and arrayed screens were highly concordant. We combined the screen readouts to nominate a small set of PrimeR binders that exhibited both high fidelity and on-target functionality. We confirmed the desired characteristics of these binders with targeted arrayed screens in additional conditions as well as in in-vivo models. We have applied both screen strategies to select a small set of leads from hundreds of candidate PrimeR binders in the context of a logic-gated MSLN CAR. As pooled and arrayed screens come with different sets of limitations and advantages, both serve as important tools for the effective selection of receptors in the development of novel cell therapies. Citation Format: Li Wang, Sofia Kyriazopoulou Panagiotopoulou, Rona Harari-Steinfeld, Dasmanthie De Silva, Michelle Tan, Laura Lim, Angela Boroughs, Cate Sue, Jon Chen, Jamie Thomas, Mary Chua, Ed Yashin, Christine Shieh, Ryan Fong, Sophie Xu, Grace Zheng, Brendan Galvin, Aaron Cooper, Tarjei Mikkelsen, Nicholas Haining. High throughput screening strategies in the development of logic gated cell therapies. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5329.