Mismatch tolerance of a gRNA for CRISPR-based gene activation confers broad activity critical for cell reprogramming

CRISPR activation and interference systems (CRISPRa/i) are widely used for programmable transcriptional control without introducing DNA edits. Although these technologies are capable of highly specific single-gene activity, some applications of transcriptional network reprogramming require broad, genome-wide effects. In this study, we identify a CRISPRa gRNA that is highly effective at reprogramming astrocyte transcriptional state. However, closer analysis revealed that this gRNA achieves this with widespread off-target activity capable of inducing expression changes in thousands of genes, in contrast to gRNAs targeting neighboring sites and regulating the same target gene. We use this promiscuous gRNA to further study the determinants of gRNA-driven off-target dCas9 binding in the context of transcriptional reprogramming. Employing a combination of ChIP-seq, high-throughput in vitro protein-binding microarrays, and high-throughput screening of gRNA-variant libraries in cells, we demonstrate that critical PAM-proximal bases within the gRNA seed sequence determine genomic binding, that mismatch tolerance varies by gRNA and base residue, and that mutating the PAM-proximal sequence of promiscuous gRNAs can modulate specificity. Finally, we find that CRISPRa-driven phenotypes can result from simultaneous contributions of active off-target effects and dose-dependent on-target activity. These findings highlight the potentially widespread impacts of CRISPRa off-target activity by gRNAs, underscore the need to account for cryptic effects when selecting and evaluating gRNAs for programming cell phenotypes, and demonstrate that multi-site binding in CRISPRa systems, typically viewed as a limitation, can also be exploited as a feature for network-level perturbations in cell reprogramming.