RNA-guided genome editing with the CRISPR-Cas9 system has great potential for basic and clinical research, but the determinants of targeting specificity and the extent of off-target cleavage remain insufficiently understood. Using chromatin immunoprecipitation and high-throughput sequencing (ChIP-seq), we mapped genome-wide binding sites of catalytically inactive Cas9 (dCas9) in HEK293T cells, in combination with 12 different single guide RNAs (sgRNAs). The number of off-target sites bound by dCas9 varied from ∼10 to >1,000 depending on the sgRNA. Analysis of off-target binding sites showed the importance of the PAM-proximal region of the sgRNA guiding sequence and that dCas9 binding sites are enriched in open chromatin regions. When targeted with catalytically active Cas9, some off-target binding sites had indels above background levels in a region around the ChIP-seq peak, but generally at lower rates than the on-target sites. Our results elucidate major determinants of Cas9 targeting, and we show that ChIP-seq allows unbiased detection of Cas9 binding sites genome-wide.

Supplementary data are available at Bioinformatics online.

Imaging chromatin dynamics is crucial to understand genome organization and its role in transcriptional regulation. Recently, the RNA-guidable feature of CRISPR-Cas9 has been utilized for imaging of chromatin within live cells. However, these methods are mostly applicable to highly repetitive regions, whereas imaging regions with low or no repeats remains as a challenge. To address this challenge, we design single-guide RNAs (sgRNAs) integrated with up to 16 MS2 binding motifs to enable robust fluorescent signal amplification. These engineered sgRNAs enable multicolour labelling of low-repeat-containing regions using a single sgRNA and of non-repetitive regions with as few as four unique sgRNAs. We achieve tracking of native chromatin loci throughout the cell cycle and determine differential positioning of transcriptionally active and inactive regions in the nucleus. These results demonstrate the feasibility of our approach to monitor the position and dynamics of both repetitive and non-repetitive genomic regions in live cells.

The CRISPR system has become a powerful biological tool with a wide range of applications. However, improving targeting specificity and accurately predicting potential off-targets remains a significant goal. Here, we introduce a web-based CRISPR/Cas9 Off-target Prediction and Identification Tool (CROP-IT) that performs improved off-target binding and cleavage site predictions. Unlike existing prediction programs that solely use DNA sequence information; CROP-IT integrates whole genome level biological information from existing Cas9 binding and cleavage data sets. Utilizing whole-genome chromatin state information from 125 human cell types further enhances its computational prediction power. Comparative analyses on experimentally validated datasets show that CROP-IT outperforms existing computational algorithms in predicting both Cas9 binding as well as cleavage sites. With a user-friendly web-interface, CROP-IT outputs scored and ranked list of potential off-targets that enables improved guide RNA design and more accurate prediction of Cas9 binding or cleavage sites.

Genes or their encoded products are not expected to mingle with each other unless in some disease situations. In cancer, a frequent mechanism that can produce gene fusions is chromosomal rearrangement. However, recent discoveries of RNA trans-splicing and cis-splicing between adjacent genes (cis-SAGe) support for other mechanisms in generating fusion RNAs. In our transcriptome analyses of 28 prostate normal and cancer samples, 30% fusion RNAs on average are the transcripts that contain exons belonging to same-strand neighboring genes. These fusion RNAs may be the products of cis-SAGe, which was previously thought to be rare. To validate this finding and to better understand the phenomenon, we used LNCaP, a prostate cell line as a model, and identified 16 additional cis-SAGe events by silencing transcription factor CTCF and paired-end RNA sequencing. About half of the fusions are expressed at a significant level compared to their parental genes. Silencing one of the in-frame fusions resulted in reduced cell motility. Most out-of-frame fusions are likely to function as non-coding RNAs. The majority of the 16 fusions are also detected in other prostate cell lines, as well as in the 14 clinical prostate normal and cancer pairs. By studying the features associated with these fusions, we developed a set of rules: 1) the parental genes are same-strand-neighboring genes; 2) the distance between the genes is within 30kb; 3) the 5′ genes are actively transcribing; and 4) the chimeras tend to have the second-to-last exon in the 5′ genes joined to the second exon in the 3′ genes. We then randomly selected 20 neighboring genes in the genome, and detected four fusion events using these rules in prostate cancer and non-cancerous cells. These results suggest that splicing between neighboring gene transcripts is a rather frequent phenomenon, and it is not a feature unique to cancer cells.

The past decade has seen a revolution in genomic technologies that enabled a flood of genome-wide profiling of chromatin marks. Recent literature tried to understand gene regulation by predicting gene expression from large-scale chromatin measurements. Two fundamental challenges exist for such learning tasks: (1) genome-wide chromatin signals are spatially structured, high-dimensional and highly modular; and (2) the core aim is to understand what the relevant factors are and how they work together. Previous studies either failed to model complex dependencies among input signals or relied on separate feature analysis to explain the decisions. This paper presents an attention-based deep learning approach, AttentiveChrome, that uses a unified architecture to model and to interpret dependencies among chromatin factors for controlling gene regulation. AttentiveChrome uses a hierarchy of multiple Long Short-Term Memory (LSTM) modules to encode the input signals and to model how various chromatin marks cooperate automatically. AttentiveChrome trains two levels of attention jointly with the target prediction, enabling it to attend differentially to relevant marks and to locate important positions per mark. We evaluate the model across 56 different cell types (tasks) in humans. Not only is the proposed architecture more accurate, but its attention scores provide a better interpretation than state-of-the-art feature visualization methods such as saliency maps.1

SUMMARY The MLL gene is a common target of chromosomal translocations found in human leukemia. MLL-fusion leukemia has a consistently poor outcome. One of the most common translocation partners is AF9 (MLLT3). MLL-AF9 recruits DOT1L, a histone 3 lysine 79 methyltransferase (H3K79me1/me2/me3), leading to aberrant gene transcription. We show that DOT1L has three AF9 binding sites, and present the NMR solution structure of a DOT1L-AF9 complex. We generate structure-guided point mutations and find they have graded effects on recruitment of DOT1L to MLL-AF9. ChIP-Seq analyses of H3K79me2 and H3K79me3 show that graded reduction of the DOT1L interaction with MLL-AF9 results in differential loss of H3K79me2 and me3 at MLL-AF9 target genes. Furthermore, the degree of DOT1L recruitment is linked to the level of MLL-AF9 hematopoietic transformation.

Deep neural network (DNN) models have recently obtained state-of-the-art prediction accuracy for the transcription factor binding (TFBS) site classification task. However, it remains unclear how these approaches identify meaningful DNA sequence signals and give insights as to why TFs bind to certain locations. In this paper, we propose a toolkit called the Deep Motif Dashboard (DeMo Dashboard) which provides a suite of visualization strategies to extract motifs, or sequence patterns from deep neural network models for TFBS classification. We demonstrate how to visualize and understand three important DNN models: convolutional, recurrent, and convolutional-recurrent networks. Our first visualization method is finding a test sequence’s saliency map which uses first-order derivatives to describe the importance of each nucleotide in making the final prediction. Second, considering recurrent models make predictions in a temporal manner (from one end of a TFBS sequence to the other), we introduce temporal output scores, indicating the prediction score of a model over time for a sequential input. Lastly, a class-specific visualization strategy finds the optimal input sequence for a given TFBS positive class via stochastic gradient optimization. Our experimental results indicate that a convolutional-recurrent architecture performs the best among the three architectures. The visualization techniques indicate that CNN-RNN makes predictions by modeling both motifs as well as dependencies among them.