BLISS is a versatile and quantitative method for genome-wide profiling of DNA double-strand breaks

Abstract

Precisely measuring the location and frequency of DNA double-strand breaks (DSBs) along the genome is instrumental to understanding genomic fragility, but current methods are limited in versatility, sensitivity or practicality. Here we present Breaks Labeling In Situ and Sequencing (BLISS), featuring the following: (1) direct labelling of DSBs in fixed cells or tissue sections on a solid surface; (2) low-input requirement by linear amplification of tagged DSBs by in vitro transcription; (3) quantification of DSBs through unique molecular identifiers; and (4) easy scalability and multiplexing. We apply BLISS to profile endogenous and exogenous DSBs in low-input samples of cancer cells, embryonic stem cells and liver tissue. We demonstrate the sensitivity of BLISS by assessing the genome-wide off-target activity of two CRISPR-associated RNA-guided endonucleases, Cas9 and Cpf1, observing that Cpf1 has higher specificity than Cas9. Our results establish BLISS as a versatile, sensitive and efficient method for genome-wide DSB mapping in many applications.

Implications of human genetic variation in CRISPR-based therapeutic genome editing

“…For example, for treatment of common diseases, more guide-enzyme combinations will need to be developed given the breadth of the natural genetic variation in the patient population. Fourth, in silico screening and empirical assays 9,19,20,21,27,28,29 to assess target efficacy and genome-wide specificity should be used to identify the optimal RNA guide-enzyme combinations from the pool of selected RNA guides. In the event that no high efficacy guides are found, the number of RNA guide-enzyme combinations should be increased and tailored to the presence of multiple independent high frequency haplotypes.…”

Predicting double-strand DNA breaks using epigenome marks or DNA at kilobase resolution

“…For KBM7 cells, we used DNase-seq, CTCF, H3K4me1/me3, and H3K9me3 for prediction and BLISS for validation [40]. The model identified 163 113 bins with a high DSB score (predicted DSBs) and 115 204 bins with a low DSB score (predicted controls).…”

“…We used double-strand DNA breaks mapped by BLISS in KBM7 cells (human myeloid leukemia) from NCBI Sequence Read Archive at SRP099132 [40]. We also used double-strand DNA breaks mapped by END-seq in untreated and etoposide-treated MCF-7 cells (human breast cancer) from GSE99197 [35].…”

“…For this, we predicted four DNA shape features using simulations: minor groove width (MGW), propeller twist (ProT), roll (Roll), and helix twist (HelT) of DSB sites at base resolution. From each feature, we computed 12 predictors including quantiles (0,10,20,30,40,50,60,70,80,90, and 100%) and the variance to describe the distribution of the feature within a DSB site. We used the resulting 48 variables combined with motif occurrences to predict DSBs with random forests and obtained better accuracy (AUROC = 0.838 and AUPR = 0.915; Fig.…”

“…For DNA shape, we computed four features including MGW, ProT, Roll, and HelT of DSB sites at base resolution. For each DNA shape feature, we then computed 12 predictors, including quantiles (0, 10,20,30,40,50,60,70,80,90, and 100%) and the variance to describe the distribution of the feature within a DSB site. The DSB data were next split into two sets: the training set used for learning the model and a test set used for assessing prediction accuracy.…”