Machine learning reveals sequence and methylation determinants of SaCas9-PAM interactions in bacteria

Area of Science:

• Microbiology
• Molecular Biology
• Bioinformatics

Background:

• Cas9 nucleases, like Staphylococcus aureus Cas9 (SaCas9), are bacterial defense systems against foreign DNA.
• SaCas9, guided by single guide RNAs (sgRNAs), has potential as an antimicrobial and genome-editing tool.
• Understanding SaCas9-target DNA interactions is crucial for optimizing its bacterial applications.

Purpose of the Study:

• To develop a machine learning model for predicting SaCas9 activity in bacteria.
• To identify sequence features, particularly downstream of the protospacer adjacent motif (PAM), that influence SaCas9 activity.
• To investigate the impact of DNA adenine methylation on SaCas9 function.

Main Methods:

• Generated large-scale SaCas9/sgRNA activity datasets in bacterial systems.
• Trained a machine learning model (crispr macHine trAnsfer Learning) to predict SaCas9 activity.
• Performed plasmid cleavage assays in DNA adenine methyltransferase (DAM)-deficient E. coli to assess methylation effects.

Main Results:

• Predictive model performance improved by including sequences flanking the NNGRRN PAM, with T-rich dinucleotides at positions [+1] and [+2] enhancing activity.
• SaCas9 exhibited significantly reduced activity (~10-fold) at sites with a 5'-NNGGAT[C]-3' PAM sequence.
• Adenine methylation at GATC motifs within PAM sequences was shown to inhibit SaCas9 activity, with its removal enhancing and its introduction reducing cleavage.

Conclusions:

• Machine learning can effectively identify key biological determinants of Cas9 nuclease activity.
• DNA adenine methylation directly inhibits SaCas9 function, suggesting a mechanism for self/non-self discrimination or counteracting foreign DNA.
• Findings provide insights for designing more effective SaCas9-based genome-editing and antimicrobial strategies.