New Frontiers in Genotoxicity Detection by Multiparametric Analyses for Gene Therapy

In hematopoietic stem cell gene therapy, integrating vectors insert therapeutic transgenes into semi-random genomic sites within hematopoietic stem and progenitor cells to correct genetic defects. However, vector integration can be genotoxic, leading to proto-oncogene activation or tumor suppressor inactivation triggering clonal expansions or oncogenesis in clinical trials with c-retroviral vectors (c-RVs) and even Self-Inactivating Lentiviral Vectors (SIN-LV). To minimize the risk of insertional mutagenesis, vector designs with reduced risk should be developed and tested using sensitive methods to assess genotoxicity, in preclinical studies and follow-up in clinical settings. Traditional genotoxicity assessments rely on vector integration sites (IS) analysis, focusing on the identification of common insertion sites and/or clonal dominance. While informative, these readouts offer limited resolution. To enhance genotoxicity detection, we developed a strategy incorporating three additional readouts: (i) Abnormal Exon Targeting Frequency, where exon-targeting frequencies for each gene were compared to expected values, using chi-square test based on exon/gene length ratios (‡ 3 ISs); (ii) Percentileranked IS Abundance, where the distribution of the clonal rank abundances of IS targeting the same gene is compared to the global IS distribution of all genes, enabling comparison across samples with different clonal complexity and sequencing depths; (iii) Integration Orientation Bias, where ISs occur preferentially on the same or opposite transcriptional orientation relative to the targeted gene, suggesting promoter insertion or enhancer-mediated activation. To evaluate this strategy, we analyzed IS datasets from clinical trials where clonal expansions and leukemogenesis triggered by insertional mutagenesis were well described. These include the LV-based SCID-X1 trial (n=19, 3.9 million ISs) where the vector carried a cryptic splice in the long terminal repeat (LTR) increasing gene deregulation risk and two trials using c-RV for ADASCID (n=24 62,653 IS) and Wiskott-Aldrich Syndrome (WAS) (n=2). These IS datasets enabled benchmarking of our readouts across high-risk clinical profiles. In total, 1800 genes were flagged as possible culprits of insertional mutagenesis by at least one read out in the different clinical trials. In SCID-X1 patients we identified 1755 putative culprits, in WAS and ADA patients treated with a c-RV showed 28 and 75 culprits respectively. Enrichment analysis showed a significant increase in cancer gene (up to 14% in our selected culprit Vs 6% expected). Recurrently altered genes across patients were identified highlighting biological reproducibility. Exonic insertions favored the last exons and 3¢ untranslated regions (3¢ UTR), significantly enriched in microRNA targets involved in cancer, whose disruption may cause oncogene activation. Moreover, the vector type and the regulatory elements dictated predominant genotoxicity mechanisms and influenced their frequency. Unlike earlier studies identifying few recurrent cancer-related genes (e.g., LMO2, HMGA2, UBR2), our pipeline uncovered a broader range, including already known culprits as well as hundreds of novel protein coding genes and, intriguingly, several long noncoding RNAs. Our multiparametric pipeline enhances genotoxicity detection, outperforms existing methods by several orders of magnitude, supporting vector selection to improve gene therapy safety and study of vector-host genome interactions.