Abstracts

Rationale: Tuberous sclerosis (TS) affects 1 in 6,000 individuals born in the United States and is a multisystem mTORopathy characterized by benign tumors, drug-resistant epilepsy, and other cognitive manifestations. This condition is caused by loss-of-function (LOF) genetic variants in the TSC1/2 complex which results in increased mTOR signaling. As a result, clinical trials are ongoing to test mTOR inhibitors as a precision therapy for individuals with TS. However, access to precise treatment requires a precise genetic diagnosis. In fact, in TSC2 alone, more than 2,220 variants of uncertain significance (VUSs) have been documented in ClinVar—and this number is likely to increase as next-generation sequencing becomes increasingly affordable and accessible. Resolving TSC2 VUS to either likely benign or likely pathogenic has the potential to improve patient care, inform whether additional genetic testing is necessary, and may enable access to precision medicine treatments or clinical trials. To address the growing challenge presented by VUSs, there is a critical need to develop tools to resolve their functional impact. Our long-term aims are to (1) develop a TSC2-specific machine learning (ML) algorithm for variant pathogenicity prediction, and (2) establish a high-throughput functional assay for TSC2 VUS resolution.

Methods: We have developed a ML model which utilizes ~20 features associated with variants in TSC2, including features related to evolutionary conservation and protein structure. In order to develop an assay for TSC2 VUS resolution, we used CRISPR/Cas9 genome editing to either knockout TSC2 (TSC2^KO) in HAP1 cells or edit the cells to contain a known pathogenic missense variant, TSC2 p.Arg611Gln. We then pooled gene edited cells (TSC2^KO or TSC^Arg611Gln) and TSC2^WT cells and tested whether fluorescence-activated cell sorting (FACS) sorting based on P-S6 would be sufficient to enrich for cells with TSC2 LOF.

Results: We observed that cells with high S6 phosphorylation were enriched for TSC2^KO alleles. Similarly, cells with constitutive activation of mTORC1 (high P-S6) were enriched for TSC2 p.Arg611Gln relative to unsorted cells or cells with low P-S6. Taken together, these results suggest our functional assay readily distinguishes pathogenic TSC2 alleles from the reference allele.

Conclusions: Based on this and our previous work on mTORopathy-associated variants in SZT2, we conclude that sorting based on P-S6 labeling distinguishes LOF variants from benign variants. We will adapt this approach to incorporate prime editing-mediated saturation genome editing, increasing throughput to thousands of TSC2 variants. This data will be used to test the validity of our ML pathogenicity predictions and to refine the performance of this classifier. This gene-specific workflow for improving the rate of VUS resolution can be adapted to perform in other mTORopathy genes, such as NPRL2, MTOR, and DEPDC5.

Funding: AES Junior Investigator Award