Fine-tuning CRISPR/Cas9 gene editing in common bean (Phaseolus vulgaris L.) using a hairy root transformation system and in silico prediction models

Abstract

The stable transformation of common bean is a challenging and time-consuming process. Although CRISPR/Cas9 has revolutionized gene editing with its high efficiency and specificity, the performance of the system can be affected by multiple factors, such as sgRNA specificity and effectiveness, and the choice of promoter used to drive Cas9 expression. The use of a hairy root transformation system to initially check the efficiency of sgRNAs and the impact of different promoters could speed up this process and increase the chances of success. We initially tested three different transformation methods to induce hairy roots and selected a preferred method suitable for a variety of different common bean genotypes. This method involved inoculating a severed radicle with Rhizobium rhizogenes K599 and was fast, had a high transformation frequency of 42-48%, and resulted in numerous hairy roots. This method was further used for the transformation of explants using R. rhizogenes harboring different CRISPR/Cas9 constructs and evaluated the on-target activity of sgRNAs targeting raffinose family oligosaccharides biosynthetic genes and the impact of different promoters driving Cas9 on the gene editing efficiency. Additionally, we evaluated the reliability of the in silico tools, CRISPOR, CRISPR RGEN, and inDelphi to predict the sgRNA efficiencies and resulting mutations. Our results showed that the hairy root transformation system allows for rapid evaluation of multiple sgRNAs and promoters. We also identified several highly efficient sgRNAs that induced frameshift mutations at rates of up to 70% when a parsley ubiquitin promoter was driving Cas9 expression, providing valuable information for the selection of the most effective sgRNAs and promoters for future transformation experiments. Although most of the computational models used to predict the sgRNA efficiency did not match the in planta results, the Lindel model proved to be the most reliable for P. vulgaris, accurately predicting the sgRNA efficiency and the type of induced mutation in most hairy roots. Furthermore, the inDelphi algorithm could correctly predict deletions and single nucleotide insertions resulting from DNA double-strand breaks in common bean. These results offer promising implications for enhancing precise editing in plants because they provide the possibility of predicting repair outcomes.