Sensitive in vitro assays for substrate preference and site of nucleic acid binding, modification, and cleavage agents

Abstract

Methods and compositions for conducting highly sensitive in vitro assays to determine substrate preferences and off-target sites of nucleic acid binding, modification, and cleavage reagents.

Description

[0001] Federally Sponsored Research or Development

[0002] This invention was made with U.S. Government support under Grant No. HR0011-17-2-0042 awarded by the Defense Advanced Research Projects Agency (DARPA). The U.S. Government has certain rights in this invention. Technical Field

[0003] Provided herein are methods and compositions for conducting highly sensitive in vitro assays to determine substrate preferences and off-target sites of nucleic acid binding, modification, and cleavage reagents. Background Art

[0004] Off-target activity is a major challenge to the safe or effective use of proteins with customizable DNA-binding activity, including but not limited to homing endonucleases, zinc fingers, transcription activator-like effectors (TALEs), and CRISPR-Cas9 system proteins, in clinical, industrial, and research settings. Summary of the invention

[0005] Provided herein are methods and compositions for conducting highly sensitive in vitro assays to determine substrate preferences and off-target sites of nucleic acid binding, modification, and cleavage reagents.

[0006] Provided herein is a method for identifying a double-stranded DNA sequence that is cut, modified, or bound by an enzyme. The method comprises (i) providing a plurality of linear dsDNA oligonucleotides of known sequence, each having a 5' end and a 3' end, and having at least two copies of a unique identifier sequence at or near both the 3' and 5' ends of the oligonucleotide, and a common sequence present at the 5' and 3' ends of each of the plurality of oligonucleotides; (ii) incubating the plurality of oligonucleotides in the presence of an enzyme selected from a site-specific nuclease, a DNA modifying protein, and a DNA binding domain under conditions sufficient for cutting, modification, or binding to occur; (iii) selecting and optionally enriching the oligonucleotides that are cut, modified, or bound; and (iv) determining the sequence of the selected oligonucleotides that are cut, modified, or bound, thereby identifying a double-stranded DNA sequence that is cut, modified, or bound by an enzyme.

[0007] Also provided herein is a method for identifying a double-stranded DNA sequence that is cleaved, modified, or bound by an enzyme. The method comprises (i) providing an initial plurality of linear dsDNA oligonucleotides of known sequence, each oligonucleotide having a 5' end and a 3' end, and having two copies of a unique identifier sequence at or near both the 3' and 5' ends of the oligonucleotide, and a common sequence present in each of the plurality of oligonucleotides; (ii) incubating the plurality of oligonucleotides in the presence of an enzyme selected from a site-specific nuclease, a modifying protein, and a DNA binding domain under conditions sufficient for cleavage, modification, or binding to occur; (iii) selecting an oligonucleotide that is not cleaved, modified, or bound; and (iv) determining the sequence of the selected oligonucleotide that is not cleaved, modified, or bound; and (v) comparing the sequence of the selected oligonucleotide that is not cleaved, modified, or bound to the sequence of an initial plurality of pre-enriched linear dsDNA oligonucleotides of known sequence; wherein the non-selected linear dsDNA oligonucleotides in the initial plurality are identified as being cleaved, modified, or bound by an enzyme.

[0008] In addition, the present invention provides a method for identifying a double-stranded DNA sequence modified by a base editing enzyme (e.g., a cytidine deaminase that converts deoxycytidine to deoxyuridine or an adenine base editing enzyme that converts deoxyadenine to deoxyinosine). The method includes (i) providing a plurality of linear dsDNA oligonucleotides of known sequence, each oligonucleotide having a 5' end and a 3' end, and having two copies of a unique identifier sequence at or near both the 3' and 5' ends of the oligonucleotide, and a common sequence present in each of the plurality of oligonucleotides; (ii) incubating the plurality of linear dsDNA oligonucleotides in the presence of a base editing enzyme under conditions sufficient for modification to occur; (iii) during DNA synthesis, converting the edited base pairs into an equal mixture of canonical base pairs; (iv) amplifying the oligonucleotides (i.e., wherein dATP nucleotides are incorporated from opposite to dU or dCTP nucleotides are incorporated from opposite to dI) with a polymerase that converts a dU:dG base pair to an equivalent mixture of dT:dA and dC:dG base pairs, or converts a dI:dT base pair to an equivalent mixture of dA:dT and dG:dC base pairs, such that the oligonucleotides that have been modified by the base editing enzyme are amplified as a mixture of the original barcode-ligated sequence from the preprocessed library and the modified sequence containing the substitution (e.g., dC->dT or dA->dG); and (iv) determining the sequence of the amplified oligonucleotides, thereby identifying the double-stranded DNA sequence modified by the base editing enzyme.

[0009] In addition, the present invention provides a method for identifying a double-stranded DNA sequence modified by a cytidine deaminase base editing enzyme that converts cytidine to uridine and generates a nick on the opposite chain. The method includes (i) providing a plurality of linear dsDNA oligonucleotides of known sequence, each oligonucleotide having a 5' end and a 3' end, and having two copies of a unique identifier sequence at or near the 3' and 5' ends of the oligonucleotide, and a common sequence present in each of the multiple oligonucleotides; (ii) incubating the plurality of linear dsDNA oligonucleotides in the presence of a base editing enzyme under conditions sufficient for modification, and then incubating the plurality of linear dsDNA oligonucleotides in the presence of an enzyme to generate a single-strand break (nick) at a site with a uridine nucleotide, thereby generating a dsDNA oligonucleotide containing two nicks with a 5' phosphate on the opposite chain, thereby generating an overhang; (iii) using a DNA polymerase (e.g., T4 DNA polymerase or Phusion DNA polymerase or Phusion U The invention relates to a method for preparing a dsDNA oligonucleotide comprising: (i) incubating the dsDNA oligonucleotide with a double-stranded DNA polymerase; (ii) capturing the phosphorylated blunt end with a double-stranded DNA adapter comprising a primer sequence; (iii) amplifying the sequence using a primer specific for the adapter and a primer specific for the consensus sequence backbone; (iv) optionally performing additional selection by size selection of smaller cleaved fragments before or after amplification; and (iv) determining the sequence of the amplified oligonucleotide to identify the double-stranded DNA sequence modified by the base editing enzyme.

[0010] In addition, the present invention provides a method for identifying a double-stranded DNA sequence modified by an adenine base editing enzyme that converts deoxyadenine to deoxyinosine and generates a nick on the opposite strand. The method includes (i) providing a plurality of linear dsDNA oligonucleotides of known sequence, each oligonucleotide having a 5' end and a 3' end, and having two copies of a unique identifier sequence at or near both the 3' and 5' ends of the oligonucleotide, and a common sequence present in each of the plurality of oligonucleotides; (ii) incubating the plurality of linear dsDNA oligonucleotides in the presence of a base editing enzyme under conditions sufficient for modification, and then incubating the plurality of linear dsDNA oligonucleotides in the presence of an endonuclease V enzyme to generate a single-strand break (nick) at a site with an inosine nucleotide, thereby generating a dsDNA oligonucleotide containing two nicks with a 5' phosphate on the opposite strand, thereby generating an overhang; (iii) using a DNA polymerase (e.g., T4 DNA polymerase or Phusion DNA polymerase or Phusion U) that generates a 5' phosphorylated blunt end from the overhang. The invention relates to a method for incubating a dsDNA oligonucleotide with a base editing enzyme (DNA polymerase); (iv) ligating the phosphorylated blunt ends to a double-stranded DNA adaptor comprising a primer sequence; (v) amplifying the sequence using one primer specific for the adaptor and one primer specific for the consensus sequence backbone; (vi) optionally performing additional selection by size selection of smaller cleavage fragments before or after amplification; and (iv) determining the sequence of the amplified oligonucleotide; thereby identifying the double-stranded DNA sequence modified by the base editing enzyme.

[0011] A method for recognizing a double-stranded DNA sequence modified by an adenine base editor that converts deoxyadenine to deoxyinosine and creates a nick on the opposite strand or a cytidine deaminase base editor that converts cytidine to uridine and creates a nick on the opposite strand, the method comprising: (i) providing a plurality of linear dsDNA oligonucleotides of known sequence, each oligonucleotide having a 5' end and a 3' end, and having two copies of a unique recognition sequence at or near both the 3' and 5' ends of the oligonucleotide, and a common sequence present in each of the plurality of oligonucleotides; (ii) in a sequence from an extremely thermophilic archaeon (Thermococcus kodakarensis) incubate multiple linear dsDNA oligonucleotides in the presence of endonuclease MS (TkoEndoMS) to induce double-strand breaks (DSBs) at deamination sites in the substrate DNA to generate DNA fragments with single-stranded, 5 base pair overhangs centered at the deamination sites; (iii) treating the DNA fragments with uracil DNA glycosylase and endonuclease VIII to remove deoxyuracil bases from the ends of the DNA fragments; (iv) end-repairing and / or A-tailing the ends of the DNA fragments; (v) ligating adapter oligonucleotides (preferably comprising sequences for high-throughput sequencing) to the ends; and (vi) sequencing the DNA fragments.

[0012] Additionally, provided herein are methods for identifying double-stranded DNA sequences bound by a catalytically inactive Cas9 in the presence of a selected gRNA or other DNA binding domain. The method comprises (i) providing a plurality of linear dsDNA oligonucleotides of known sequence, each oligonucleotide having a 5' end and a 3' end and having two copies of a unique recognition sequence at or near both the 3' and 5' ends of the oligonucleotide and a common sequence present in each of the plurality of oligonucleotides; (ii) incubating the plurality of oligonucleotides in the presence of a DNA binding domain attached to magnetic beads (e.g., covalently bound or bound by an affinity handle), such as a Cas9 enzyme complexed with an sgRNA or other DNA binding domain, under conditions sufficient for binding to occur; (iii) selecting and optionally enriching the bound oligonucleotides by pulling down one or more sets of magnetic beads and washing in an appropriate buffer that promotes dissociation of unbound molecules into the supernatant, and then eluting the bound DNA in an appropriate buffer that promotes dissociation of any bound DNA or in a buffer containing a protease, such as proteinase K, that degrades proteins bound to the magnetic beads and releases the bound DNA; and (iv) determining the sequence of the selected oligonucleotide that was cleaved, thereby identifying the double-stranded DNA sequence bound by the DNA binding domain.

[0013] In some embodiments, the linear dsDNA oligonucleotides used in the methods described herein comprise (i) a collection of all potential off-target sequences in a reference genome that have a certain number of mismatches or less relative to the identified on-target site (similar to a genomic DNA library); (ii) a comprehensive collection of potential off-target sites with a certain number of mismatches or less (similar to a random base substitution library); (iii) a library of potential off-target sequences present in a collection of variant genomes from a specific population (i.e., a genomic DNA library designed to reflect the DNA sequence variants present in a population of individuals); or (iv) other relevant specific collections of potential off-target sites (e.g., oncogene hotspots or sequences from tumor suppressor genes).

[0014] In some embodiments, the pre-enriched linear DNA library members are first synthesized as individual single-stranded DNA sequences, for example on a high-density oligonucleotide array; and the single-stranded DNA sequences are converted into double-stranded DNA molecules by priming the consensus sequence, optionally before or after release from the chip.

[0015] In some embodiments, the pre-enriched linear DNA library members represent 1) a collection of all potential off-target sequences in a reference genome that have a certain number of mismatches or less relative to the on-target site (similar to a genomic DNA library), 2) a comprehensive collection of potential off-target sites with a certain number of mismatches or less (similar to a random base substitution library), 3) a library of potential off-target sequences present in a collection of variant genomes from a specific population (i.e., a genomic DNA library designed to reflect the DNA sequence variants present in a population of individuals), or 4) other relevant specific collections of potential off-target sites (e.g., oncogene hotspots or sequences from tumor suppressor genes).

[0016] In some embodiments, the pre-enriched linear DNA library comprises 1,000, 2,500, 5,000, or more than 10,000 members and 10 6 , 10 7 , 10 8 , 10 9 , 10 10 , or 10 11 There are 10-100k different sequences.

[0017] In some embodiments, the pre-enriched linear DNA library members comprise sequences of 50 to 500, such as 100 to 400, such as 150 to 300 bp in length, such as 200 to 280 bp in length.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the art to which the invention belongs. Methods and materials for use in the present invention are described herein; other suitable methods and materials known in the art may also be used. The materials, methods, and examples are illustrative only and are not intended to be limiting. All publications, patent applications, patents, sequences, database entries, and other references mentioned herein are incorporated herein by reference in their entirety. In the event of a conflict, the present specification, including definitions, will prevail.

[0019] Other features and advantages of the invention will be apparent from the following detailed description and drawings, and from the claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 . Illustration of the difference in library complexity between base substitution libraries (SEQ ID NOs: 1-6, 1, and 7-11) and genomic DNA libraries (SEQ ID NOs: 1 and 12-16). For the genomic library, the target site (red box in this example) has very little similarity to the other approximately 3 billion genomic sequences. For the base substitution library, the preselected library is enriched for sites that are not necessarily present in the genome but are similar to the predetermined target site. Substitutions are represented by lowercase letters.

[0021] Figure 2 .Description overview of exemplary methods. A user-determined set of potential DNA substrates (typically but not limited to 10,000 to 100,000 sequences) is generated by synthesis on a high-density oligonucleotide array and then made into a double-stranded strand. Three potential examples of the set of sequences used include all sequences with less than six mismatches in the human genome, less than eight mismatches in the human exome, or a set of all possible DNA sequences with less than three mismatches. The double-stranded DNA library can then be used for screening of sequence modifications, screening of sequence deletions, or selection of sequence modifications and / or cutting. Black lines represent constant sequences present in each member of the library and are used for primer binding sites in amplification steps and bioinformatics processing.

[0022] Figure 3 .The target site library can be selected for cleavage (strategy 1), or the target site library can be screened to eliminate cleaved sequences (strategy 2). The black lines represent constant sequences present in each member of the library and primer binding sites used in the amplification step and bioinformatics processing. SEQ ID NO.: 17-22 are shown.

[0023] Figure 4. Enrichment of Cas9 cleavage sites by selecting from a random base substitution library. The predetermined target site (SEQ ID NO: 23) is listed below the heat map. Each black box represents the abundance of a specific nucleotide (represented on the left) at a position corresponding to the target site nucleotide listed below, with black representing the most abundant nucleotide / each position and white representing no abundance.

[0024] Figure 5 . Target sequences were identified by screening random base substitution libraries and screening genomic DNA libraries. Substrates in the substrate spectrum library and the genome-inspired library have fewer mutation numbers (represented by a smaller m number, where (m_d_i)->m=mutation number, i=insertion number, d=deletion number), where Xm0 represents m mismatches without any insertions, RNAmd represents a target site deletion of length d with m mismatches at the remaining base pairs in the site, and DNAmi represents a target site insertion of length I with m mismatches at the remaining base pairs in the site).

[0025] Figure 6 . Shows the composition of a representative genomic DNA oligonucleotide library. The number of each genomic site is listed according to the number of mismatches and bulges. Unless otherwise stated, these libraries will be used for the experiments outlined in the subsequent figures.

[0026] Figure 7 .Library characterization. After oligonucleotide synthesis, library amplification, and Illumina sequencing, the uniformity metric and drop out percentage are displayed for each library. The 90th percentile sequencing count refers to the number of sequencing reads of the library member obtained in the 90th percentile when sorted by increasing reads. The 90 / 10 ratio is the ratio of the number of sequencing reads of the 90th percentile library member divided by the 10th percentile library member, and is a measure of library uniformity. Dropout refers to the number of sequences that are not represented in the sequenced amplified library.

[0027] Figure 8 .Enrichment of known GUIDE-seq sites. Clustered scatter plots showing representative cut selections using the methods described herein (an example of which is referred to as the ONE-seq method). Each circle represents the aggregated read counts normalized to the on-target sequence for a given guide RNA selection (listed at the top) for a single library member. Black stars represent on-target library members. Solid circles represent sites identified by published GUIDE-seq experiments. There are no published RNF2 GUIDE-seq sites.

[0028] Fig. 9 .Enrichment of highly enriched CIRCLE-seq sites. Clustered scatter plots showing representative cut selections using the ONE-seq method. Each circle represents the aggregated read counts normalized to the on-target sequence for a given guide RNA selection (listed at the top) for a single library member. Black stars represent on-target library members. Solid circles represent sites with >100 read counts identified by published CIRCLE-seq experiments. There are no published RNF2 GUIDE-seq sites.

[0029] Fig.10 .Enrichment of moderately enriched CIRCLE-seq sites. Clustered scatter plots showing representative cut selections using the ONE-seq method. Each circle represents the aggregated read counts normalized to the on-target sequence for a given guide RNA selection (listed at the top) for a single library member. Black stars represent on-target library members. Solid circles represent sites with 10-99 read counts identified by published CIRCLE-seq experiments.

[0030] Fig.11 .Enrichment of low-enriched CIRCLE-seq sites. Clustered scatter plots showing representative cut selections using the ONE-seq method. Each circle represents the aggregated read counts normalized to the on-target sequence for a given guide RNA selection (listed at the top) for a single library member. Black stars represent on-target library members. Solid circles represent sites with 1-9 read counts identified by published CIRCLE-seq experiments.

[0031] Fig.12 Venn diagram showing an exemplary approach (darker circles) that identified all 60 of the 62 highly enriched (>100 reads) CIRCLE-seq sites for six SpCas9:sgRNAs using a cutoff of 1% of on-target ONE-seq aggregate read counts. The 2 sites that did not exceed the 1% ONE-seq cutoff do not necessarily represent true off-target sequences and could be false positives in the CIRCLE-seq approach. CIRCLE-seq failed to identify 478 ONE-seq identified sites for these guide RNAs.

[0032] Fig.13. Validation results for three FANCF off-target sites identified by ONE-seq but not GUIDE-seq or CIRCLE-seq. Targeted amplicon sequencing was performed on three of the five most highly enriched novel off-target candidates identified by ONE-seq from HEK293T cells, sorting HEK293T cells for the top decile of expression of the SpCas9:FANCF sgRNA construct. On the left, the total number of sequence reads containing indels (edited) and the total number of reference reads are shown, and the percentage of edited and unedited candidate off-target sequences are shown. Separate data from three separate sorting and control (untreated) experiments are shown on the right. (SEQ ID NO:24-39 appear in order)

[0033] Fig.14 . Reproducibility of enrichment scores in variant libraries. ONE-seq selections were performed on the EMX1 genomic off-target library and the EMX1 genomic variant off-target library. Enrichment scores (relative to the on-target sequence) for library members shared by both libraries are shown. Overlapping lines correspond to equal enrichment scores from both selections.

[0034] Fig.15 .ONE-seq identifies candidate off-target sites that are present in the population but not in the reference genome. Normalized aggregated read counts (where 1.0 is the on-target site) for off-target candidates identified from the reference genome and paired off-target candidates containing SNPs found in the 1000 genome population are shown. Variants present in >40% of the population are indicated with solid circles. Overlapping lines correspond to equal enrichment scores from two paired library members.

[0035] Fig.16 .Base editor screening strategy. A random base substitution library designed for the EMX1 target site was incubated with BE1 in vitro and amplified by PCR with Kapa HiFi Uracil+ DNA polymerase, which converts U:G base pairs to an equal mixture of T:A and C:G base pairs during DNA synthesis (incorporating dATP nucleotides opposite dU). Therefore, upon sequencing, any library member that can be modified by BE1 will be sequenced as a mixture of the original barcode-ligated sequence from the pre-processed library and the modified sequence containing C->T substitutions (and other rare substitutions).

[0036] Fig.17 .Base editor screening shows enrichment for sites containing NGG and displays high specificity at the PAM proximal end and low specificity at the PAM distal end of the target site (SEQ ID NO: 23). Heatmap with Figure 4 Explained the same way.

[0037] Fig.18 BE3 selection strategy. In this strategy, the target site library is exposed to the BE3 enzyme and modified members are enriched by double-strand breaks formed at sites with uridine nucleotides (by USER) and nicks (by BE3).

[0038] Fig.19 .Enrichment of BE3 off-target sites by ONE-seq. Aggregated read counts normalized to eight ONE-seq selections on a genomic DNA-stimulated library are shown (where 1.0 corresponds to an on-target site). Only sites with a score of 0.01 or above (1% on-target enrichment) are shown. Black stars indicate on-target library members. Solid black circles indicate newly validated off-target sites compared to Digenome-seq (except ABE site 18, which was not tested by Digenome-seq). Hollow black circles indicate Digenome-seq candidate sites.

[0039] Fig. 20 Newly identified and validated BE3 off-target sites. Targeted amplicon sequencing data of genomic DNA from HEK293T cells expressing the indicated BE3:sgRNA complexes are shown compared to untreated controls. Experiments were performed in triplicate. Only 28 newly identified and validated BE3 off-target sites are shown compared to Digenome-seq.

[0040] Fig.21 .ABE selection strategy. In this example, an adenine base editor (ABE) was used. ABE produces an A->I change in DNA. The method we used here to determine the off-target sites of ABE is similar to the method used in Example 3, except that a different nuclease, endonuclease V, is used to produce the nick at the deoxyinosine site in DNA.

[0041] Fig. 22 . ABE selection on base substitution library. Heat map with Figure 4 The data from the selection indicated an enrichment of NGG PAMs and also indicated that relative to Figure 4 SpCas9 cleavage selection and Figure 8 BE3 selection in the enrichment of sequences containing A at position 5, indicating that A is required for activity in certain editing windows of ABE. (SEQ ID NO:23)

[0042] Fig.23.Enrichment of ABE7.10 off-target sites by ONE-seq. Normalized aggregate read counts for eight ONE-seq selections in the genomic DNA stimulation library are shown (where 1.0 corresponds to an on-target site). Black stars indicate on-target library members. Solid circles indicate validated off-target sites. Hollow black circles indicate off-target candidates sequenced in the validation study.

[0043] Fig.24 . Validated ABE off-target sites. Data from targeted amplicon sequencing of genomic DNA from HEK293T cells expressing the indicated ABEmax:sgRNA complexes are shown compared to untreated controls. Experiments were performed in triplicate.

[0044] Fig.25 Capillary electrophoresis data from experiments demonstrating the specificity of TkoEndoMS' endonuclease activity for G:U DNA mismatches in vitro. 800 base pair PCR amplicons were incubated with purified BE protein and variable sgRNA for 2 hours to induce site-specific deamination. After purification, the deaminated PCR amplicons were incubated with purified TkoEndoMS protein for 7 minutes to induce double-strand breaks at G:U mismatches. DNA was then size-separated by capillary electrophoresis and imaged.

[0045] Fig.26 .Overview of enriching binding sites by pull-down. In this method, dCas9-coated magnetic beads are incubated with a library of potential off-target sites. Unbound library members are washed into the supernatant, and bound library members are eluted by digesting the magnetic bead-bound proteins with proteinase K. The resulting eluted library can be amplified and subjected to additional rounds of pull-downs or analysis by high-throughput sequencing.

[0046] Fig. 27 .Pull-down conditions can distinguish between on-target and off-target sites. A mixture of three double-stranded DNAs of different lengths was subjected to binding site pull-down using magnetic beads coated with dCas9:EMX1 sgRNA. The on-target site was present on 280 base pairs of DNA, and one of the two off-target sites (OT2 or OT4) was present on 220 base pairs of DNA. A third 200bp ("random") DNA that contained neither the on-target site nor the off-target site was also present in the mixture. The enriched DNA was run on a QIAxcel. Lane A3 represents a size ladder. Lane A1 shows selective pull-down of the 280 base pair on-target site, but no pull-down of the OT2 site or the 200bp DNA. Lane A2 shows that the OT4 site can still be bound in the method, indicating that the conditions used for pull-down can enrich for off-target sequences that can be bound by the dCas9:EMX1 sgRNA. (SEQ ID NO: 23, 40-41 are shown in sequence)

[0047] Fig.28 . Enrichment of genomic DNA primed libraries by pulldown. Pulldowns performed in the presence of 50 ug / ml heparin in the FANCF library resulted in enrichment of the most abundant post-pulldown library members for the on-target loci (black stars).

[0048] Fig.29 .Library composition after selection of I-PpoI. Sequence logo showing sequences with normalized read counts of at least 1% of the expected I-PpoI target site. The positions in the site from the 5' to the 3' end are shown on the horizontal axis, and the height of the stack of letters represents the information content of each position (in bits). The height of each individual nucleotide emphasizes the relative contribution of the nucleotide to the information content of the position. Positions 2, 13, and 14 are the most highly described (with the most information content). Position 15 is the least described (with the lowest information content). DETAILED DESCRIPTION

[0049] In vitro / biochemical strategies to understand the on-target and off-target activities of DNA binding domains are generally divided into two types: Figure 1 ): In the first type, off-target cleavage events are searched for a collection of DNA sequences in a specific system of interest (e.g., the human genome). Examples of methods utilizing this strategy include CIRCLE-seq (Tsai et al., Nat Meth. 14: 607 (2017)], SITE-seq (Cameron et al. Nat Meth. 14: 600 (2017)), and Digenome-seq (Kim et al. Nat Meth. 12: 237 (2015)). The scope of these whole genome approaches is limited to off-target sites present in the specific genomic DNA used for the study. In contrast, in the second type of strategy, the substrate preference of DNA binding / modifying proteins is analyzed in a more unbiased manner by comprehensively searching for a library of binding sites in which certain base positions are randomly replaced by all potential optional bases (rather than being limited to a limited set of base replacements as in the case of the first strategy). Various nucleases (ZFN (Pattanayak et al. Nat Meth.8:765(2011)], TALEN (Guilinger et al. Nat Meth.11:429(2014)), and CRISPR-Cas9 (Pattanayak et al. Nat Biotech.31:839(2013))) in vitro "genomic DNA" and "random base substitution library" approaches and provide insights into the biochemical functions and specificities of these nucleases.

[0050] Both strategies for studying off-target activity have limitations, affecting their ability to identify true off-target sites. In whole genome selection, tens to hundreds of off-target sites that are cut must be enriched from the background of billions of other uncut sites (the length of the human genome is about 3 billion base pairs, so it contains about 6 billion sites to be analyzed). For example, due to the noise in the enrichment method and sequencing results, the CIRCLE-seq method is limited to the detection of sites with no more than six mismatches relative to the target site, which only represents about 0.002% of the genomic material present in the analysis. Although some methods, such as Digenome-seq, rely on a large amount of excessive sequencing of DNA libraries treated with nucleases, methods such as CIRCLE-seq and GUIDE-seq are generally incorporated into the enrichment step of the edited sequence. This enrichment step can be carried out in cells (GUIDE-seq) or in vitro (CIRCLE-seq). Although the CIRCLE-seq method is substantially more sensitive than other methods for off-target screening, the CIRCLE-seq method requires a very large input of genomic DNA (25 μg) for each experimental sample.

[0051] In vitro selection in unbiased base substitution libraries is limited by library size (the set of sequences that can be practically tested). For example, the SpCas9 target site contains 22 potentially specified base pairs (20 from hybridization with the guide RNA and two from the PAM sequence). To analyze all potential target sites with all possible combinations of base substitutions at all positions, it would be necessary to generate and search at least 4 22 ~10 13 unique DNA molecules, neither of which is possible using current technology. For example, library construction methods are currently limited to generating 10 11 -10 12 Furthermore, even if library construction methods are improved, 12It is not feasible to sequence a single DNA molecule. To overcome this limitation, doped oligonucleotide synthesis is traditionally used to generate libraries with sites of base substitutions that follow a binomial distribution, so that the on-target site is present in more copies than each variant site in a library with a single mutation, each of which is present in more copies than each variant site in a library with double variant sites, and so on. Therefore, selections made by these random base substitution libraries are limited by the following facts: 1) it is impossible to generate completely unbiased libraries (i.e., they are heavily biased toward the expected on-target site sequences) and 2) it is impossible to generate libraries that consistently represent the potential sequence space. In addition, using the output from a specific library experiment to predict or identify off-target sites in a genomic sequence generally requires extrapolation (Sander et al. Nucleic Acids Res. 41: e181 (2013)), because there is no guarantee that all relevant genomic sequences are included in the pre-selection (limited to 10 12 sequences, which correspond to six or seven substitutions) or post-selection libraries (limited by sequencing capacity to 10 7-8 sequence).

[0052] Methods for identifying DNA binding, modification or cleavage sites

[0053] In this paper, we provide an improved method ( Figure 2 ), which can identify on-target and off-target binding, modification, or cleavage sites of DNA-modifying proteins / protein complexes (including but not limited to: dCas9 fused to an effector domain, a Cas9-based base editor, or an active Cas9 protein), and it overcomes the drawbacks of both the "genomic DNA" and "random base substitution library" methods. Using this method, a pre-enriched library of linear DNA consisting of a specific user-specified sequence can be generated by high-density oligonucleotide synthesis, and then those sequences that can be bound, modified, or cut by a sequence-specific protein or protein complex are searched. Minimally, this method allows the identification of sequences that are potential substrates for any agent whose modification can result in sequence modification, binding, or cutting of nucleic acids.

[0054] The pre-enriched linear DNA library members are initially synthesized on a high-density oligonucleotide array as individual single-stranded DNA sequences, each with a unique identifier / barcode that is present / replicated on both sides of the oligonucleotide ( Figure 2). The synthesized oligonucleotides are released from the chip and converted to double-stranded DNA molecules by priming against a common sequence present in all DNA molecules synthesized on the chip. The combined library is then incubated with the site-specific nuclease, modified protein, or DNA binding domain of the target and enriched for cleaved, modified, or bound sequences (see Examples 1, 3, and 4) or screened for modifications (see Example 2) in a selective manner. The DNA sequence of the cleavage site can then be reconstructed from one of the identical barcodes that were originally located on either side of these sites and are now separated into two molecules.

[0055] Synthetic molecules can be specified to represent: 1) a collection of all potential off-target sequences in a reference genome with a certain number of mismatches relative to the on-target site (similar to a genomic DNA library), 2) a comprehensive collection of potential off-target sites with a certain number of mismatches (similar to a random base substitution library), 3) a library of potential off-target sequences present in a collection of variant genomes from a specific population (i.e., a genomic DNA library intended to reflect DNA sequence variants present in an individual population), or 4) other relevant specific collections of potential off-target sites (e.g., oncogene hotspots or sequences from tumor suppressor genes). This strategy has important advantages for constructing these libraries. For random base substitution libraries, all sequences within a determined number of substitutions can be represented equally and can be easily sampled using current next-generation sequencing methods. For genomic or exome DNA libraries, only the most likely related sites are included (e.g., all potential off-target sites with six or less substitutions), which eliminates the noise caused by about 99.998% of sites that are not substrates. Importantly, since this method results in the generation of a double-stranded DNA (or RNA) library that is an enriched collection of potential off-target sites for DNA (or RNA) binding proteins, it can be used to determine the specificity and off-target sites of not only nucleases, but also other proteins that bind or modify nucleic acids, including but not limited to customizable base editing enzymes (Komor et al. Nature 533:420, 2016, Gaudelli et al. Nature. 551:464 (2017)), transcription activators (Mali et al., Nat Biotech. 31:833 (2013), Chavez et al. Nat Meth. 12:326 (2015)), transcription repressors (Bikard et al., Nucleic Acids Res, 41:7429 (2013), Thakore et al. Nat Meth. 12:1143 (2015)), and epigenome editing enzymes (Zentner and Henikoff. Nat Biotech. 33:606 (2015)).

[0056] The ability to determine the same barcode on both sides of a specific recognition site represents a major advance beyond previous in vitro analysis methods (US Patent 9,322,006, US Patent 9,163,284), because the sequence of the library member is encoded in at least three positions of each individual member of the DNA pool. When trying to determine the DNA modification activity (such as base editing) in which the target sequence is modified, the redundancy of this information is particularly advantageous. Even if the actual DNA sequence of the library member itself is modified, the original sequence information can be obtained from the information content contained in the barcodes on both sides. The redundancy of the information in the two barcodes and the recognition site also allows endonuclease cutting selection (or paired base modification + cutting selection) to be performed on the potential cutting site where each library member exists in a single copy, contrary to multiple copies (US Patent 9,322,006, US Patent 9,163,284). Without the current barcode strategy, the library member sequence cut in the recognition site cannot be recombined because the cutting separates the two sides of the cutting site in space (top figure, bottom right, blue area).

[0057] Example

[0058] The present invention is further described in the following examples, which do not limit the scope of the invention described in the claims.

[0059] Target sites used in the following examples:

[0060]

[0061] Example 1: DNA cleavage selection by SpCas9 and SpCas9-HF1

[0062] In this example, a random base substitution library designed for SpCas9 nuclease engineered with a guide RNA (gRNA) designed for the on-target site of the human EMX1 gene (hereinafter referred to as EMX1 gRNA and EMX1 target site) and a library of potential EMX1 gRNA off-target sites from the human reference genome were selected for cleavage by SpCas9 or SpCas9-HF1.

[0063] In this embodiment ( Figure 3 ), both selection (Strategy 1) and screening (Strategy 2) can be employed. In Strategy 1, the pooled library contains library members of approximately 50,000 barcodes (either a random base substitution library containing all possible sequences within three mismatches of the target site in EMX1 SpCas9, or a genomic DNA-stimulated library containing all possible sequences from the hg19 human reference genome within six mismatches of the target site in EMX1 SpCas9—see the Methods section for details of the libraries.)

[0064] Selections using strategy 1 in a 1:1:1 ratio of SpCas9:sgRNA:DNA library (EMX1 target site) showed enrichment of sequences that could be cleaved ( Figure 4 ). The position in the target site is on the horizontal axis (with the on-target bases listed below). The possible bases in the library (either substitutions or on-target) are represented on the vertical axis. The data are combined and summarized in a heat map, where the darker black rectangles represent a larger proportion of sites in the library after selection containing the corresponding base from the vertical axis. As a proof of principle, this heat map is consistent with previous studies that show that the N of the NGG PAM sequence is not specified and its specificity at the PAM distal end of the target site is lower than its specificity at the PAM proximal end.

[0065] Screening of the random base substitution library using strategy 2 yielded similar results ( Figure 5 ). The substrates in the substrate repertoire library and the genomic stimulation library have fewer mutation numbers (indicated by a smaller m number, where (m_d_i)->m=mutation number, i=insertion number, d=deletion number), where Xm0 represents m mismatches without any insertions, RNAmd represents a target site deletion of length d with m mismatches at the remaining base pairs in the site, and DNAmi represents a target site insertion of length i with m mismatches at the remaining base pairs in the site).

[0066] The genomic library typically consisted of all potential off-target sites in the hg19 reference human genome with zero to six mismatches relative to the on-target sequence, four or fewer mismatches combined with a DNA bulge of one or two nucleotides, and four or fewer mismatches combined with an RNA bulge of one nucleotide, and three or fewer mismatches combined with an RNA bulge of two nucleotides. Figure 6 ). Sequence the pre-selected library to evaluate quality metrics ( Figure 7 ), indicating low deletion rates (below 0.20%) and high concordance (90 / 10 ratio > -2). To our knowledge, these metrics have not been calculated for other specificity methods and are therefore not directly comparable.

[0067] Six non-promiscuous guide RNAs (HBB, RNF2, HEK2, HEK3, FANCF, and EMX1) with relatively few expected off-target sequences were selected using strategy 1 (called ONE-seq) and a genomic DNA-stimulated library. Figure 8 , black stars) are the most abundant or top 3 abundant library members among the thousands of six non-scrambled guide RNAs tested. Summarizing the six non-scrambled guide RNAs, ONE-seq enriched all 163 off-target sites identified by GUIDE-seq ( Figure 8, filled circles), with post-selection read counts ranging from 11% to 120% of the on-target sequence. This approach also enriched for highly enriched CIRCLE-seq sites (defined here as those with >100 sequence reads, Fig. 9 ), and moderately enriched to a lesser extent CIRCLE-seq sites with moderate enrichment (10-99 reads, Fig.10 ), or low enrichment (1-9 reads, Fig.11 ). If the cutoff for on-target enrichment in the ONE-seq method described here is 1%, ONE-seq identifies 60 of the 62 highly enriched CIRCLE-seq sites ( Fig.12 ), while CIRCLE-seq failed to identify 478 highly enriched ONE-seq candidates. Notably, the two highly enriched CIRCLE-seq sites that were not highly enriched by ONE-seq were likely false positives of the CIRCLE-seq approach. Validation of ONE-seq sites that were novel and not identified by GUIDE-seq or CIRCLE-seq was demonstrated by sorting HEK293T cells in the top decile of SpCas9:FANCF sgRNA expression ( Fig.13 ). These results suggest that the method described here is at least as sensitive as existing methods and may be more sensitive.

[0068] In addition, the method can be extended to any library / specific collection of nucleic acid sequences. For example, using public data from the 1000 Genomes Project, ONE-seq selection was performed on the EMX1 genomic off-target site library, which accounts for naturally occurring sequence variation on a population scale. In this exemplary library, all sequences from the reference hg19 human genome collection that were in the original EMX1 library ( Figure 6 ) and contains the SNP in the 1000 Genomes database. In addition, sequences containing the SNP were included as additional library members to account for the possibility that individuals may have off-target sequences that are not contained in the reference genome. ONE-seq cut selection performed on this EMX1 library containing the SNP provides a reproducible enrichment of off-target candidates present in the reference hg19 genome ( Fig.14 ONE-seq cut selection in variant libraries demonstrated that evaluation of thousands of variants in a population of candidate sets present at off-target sites of the EMX1 guide RNA identified several that were differentially enriched ( Fig.15 , black circles). Example 2: Screening of base editors with BE1

[0069] In this embodiment ( Fig.16), a screening strategy was used to identify base modifications produced by the BE1 enzyme (Komor et al. Nature 533:420, 2016), which typically produce C->U changes in a specific window of DNA.

[0070] Base editor screening with BE1 following the above protocol was applied to the EMX1 target site and the substrate repertoire library yielded an enrichment of the expected repertoire of tolerated off-target sites ( Fig.17 ).

[0071] Example 3: Base editor selection with BE3

[0072] In this embodiment ( Fig.18 ), a selection strategy is used to enrich for sites modified by the BE3 enzyme. Library members that can be recognized by the BE3 enzyme (Komor et al. Nature 533:420, 2016) should show both C->U modifications and nicks on the opposite strand. USER enzyme (NEB) is used to achieve double-stranded cleavage of library members that are BE3 substrates by replacing dU nucleotides with nicks. Therefore, the resulting modified library members will contain two nicks with 5' phosphates on the opposite strands and are incubated with a DNA polymerase that can blunt these DNA overhangs (e.g., T4 DNA polymerase or Phusion DNA polymerase). The resulting phosphorylated blunt ends are captured with double-stranded DNA adapters prior to amplification / selection using a primer specific for the adapter and a primer specific for the library backbone (as in Example 1). Additional selection stringency can be obtained by size selection of smaller cleavage fragments before or after amplification.

[0073] Using this approach, we examined BE3 targeting with genomic DNA-stimulated libraries for eight target sites, including all seven BE3 targets previously tested by Digenome-seq (Kim et al. Nat. Biotech. 35:475, 2017). ONE-seq selection results showed enrichment of the expected target sites for the top 13 of the tens of thousands of library members for all eight selections ( Fig.19 , black stars). In three of the eight selections, the expected on-target site was the most enriched site. All 42 off-target sites previously validated by Digenome-seq were present in the enriched post-selection libraries ( Fig.19 , open black circles), and 40 of 42 were among the top 61 sites for each selection. To further validate our ONE-seq results, we amplified and sequenced approximately 20-40 high-ranking sites from each selection from human HEK293T cells. Our results showed that there were 28 validated BE3 off-target sites that were not identified as candidates by Digenome-seq ( Fig.19, solid black circles). Six of the 28 new sites had an editing percentage greater than 1% in cells ( Fig. 20 ), up to 23.9%, indicating a higher level of sensitivity of ONE-seq in detecting not only weak off-target sites but also high-frequency off-target sites.

[0074] Example 4: Selection of base editors with ABEs

[0075] In this embodiment ( Fig.21 ), we used adenine base editors (ABE; Gaudelli et al. Nature. 551: 464 (2017)) with EMX1 gRNA and base substitution repertoire libraries and genomic DNA libraries for selection. ABE is a sgRNA-guided Cas9 nickase fused to a protein domain that catalyzes the conversion of deoxyadenosine to deoxyinosine. In this example, double-stranded cleavage of the pre-selected library was completed in two steps ( Fig.21 ). First, incubation with the ABE enzyme and guide RNA results in nicking of the strand of the identified library member that hybridizes to the guide RNA. Second, subsequent incubation with an enzyme that nicks the 3' of deoxyinosine in the library member that may be caused by ABE activity, i.e., endonuclease V, results in nicking on the non-hybridized DNA strand, resulting in a double-strand break with an overhang. Subsequent filling of the double-strand break with a DNA polymerase results in the formation of a blunt end, which can be selected according to the description of the Cas9 nuclease in Example 1.

[0076] Selection of the base substitution library showed enrichment of substrates with NGG. Furthermore, as expected, this experiment ( Fig. 22 ) shows the enrichment of substrates with A at position 5 of the target site (where 1 is the base pair farthest from the PAM), reflecting the preference of ABE for modification with A farther from the PAM than is present in the typical EMX1 target site. Notably, 95 of the 100 most abundant sequences in the post-selection library have A at position 5. These results indicate that our strategy is effective in enriching and identifying off-target sites of ABE.

[0077] Table 1. Enrichment of sequences with A at position 5 in ABE selections.

[0078]

[0079] We also performed the above selection on the EMX1 genomic DNA library (Table 2), which shows the enrichment of EMX1 on-target sites (highlighted; the 96th most abundant post-selection library sequence) and EMX1 off-target sites with the highest off-target recognition (bold and asterisks; the 9th most abundant post-selection library sequence).

[0080] Table 2. Top 96 most abundant sites in the library after selection for ABE selection in genomic DNA library of potential EMX1 off-target sites.

[0081]

[0082]

[0083] The two sequences highlighted are the most active off-target site for cleavage (chr5: 45359060) and the on-target site (chr2: 73160981) marked with an asterisk. Due to the presence of A at a more favorable position in the editing window, it can be expected that the off-target site will be more abundant in the selection.

[0084] We also performed additional selections on a genomic DNA library designed to identify off-target sequences of the six guide RNAs ( Fig.23 ). Applying the modified ONE-seq selection scheme to six ABE targets showed that for the five non-promiscuous guides tested, the expected on-target sites were enriched in the top three of the library after selection (HEK4 is a known promiscuous guide RNA). This was verified by amplicon sequencing of DNA from human HEK293T cells, which were transfected with ABE7.10:sgRNA pairs (about 20 each in each selection) of the appropriate top candidate sites, thereby identifying a total of 12 confirmed cellular off-target sites in the six target sites. The set includes three validated off-target sites identified by two guide RNAs tested by ONE-seq and EndoV-seq (Liang et al. Nature Communications. 10: 67 (2019)) or Digenome-seq (Kim et al. Nature Biotechnology. 37: 430 (2019)), as well as nine newly validated off-target sites that were not identified as potential candidates by any of these methods. Nine of the 12 sites had off-target modification rates below 1% or showed evidence of only a single nucleotide substitution, either of which could be due to sequencing error, despite stringent quality filtering of the sequenced reads (all positions in paired reads had to have a quality score > Phred 30) and validation in triplicate. To increase our confidence that these sites were true off-target sites, we performed a second round of validation experiments with cells transfected with plasmids expressing ABEmax, a codon-optimized version of ABE7.10, and GFP, and sorted the cells to enrich for the top decile of GFP and, therefore, ABEs expressing ABE ( Fig.24). Genomic DNA extraction was performed immediately after sorting without further expansion. In the sorted validation set, the frequency of on-target modification ranged from 61%-94% compared to 31%-56% in the unsorted validation set. All 12 off-targets in the unsorted validation set were modified at a higher frequency in the sorted validation set, confirming that they were true off-targets, and five additional off-target sites were identified at a modification frequency of less than 1%. An ABE_14 off-target site containing a single mismatch relative to the on-target site had 85% of the DNA modified in the sorted validation (18% in the unsorted validation), indicating that some ABE off-target sites can be modified at a high frequency.

[0085] Example 5: Base Editor Selection with ABE or BE3 Using Enzymes That Produce Double Bond Breaks at Modified Positions

[0086] In this example, modified library members containing deoxyinosine can be made to have flat, double-stranded ends through the action of the TkoEndoMS protein (Ishino et al., Nucleic Acids Res. 44: 2977 (2016)). TkoEndoMS can be used to generate double-strand breaks at dI: dT base pairs caused by dA->dI editing of ABE. If a base editing enzyme without nicking activity is used, the DNA with double-strand breaks is subjected to the same downstream steps as in Example 1, and the adapter is connected to the phosphorylated flat-end DNA. If a base editing enzyme with nicking activity is used, end modifications performed by a DNA polymerase that produces flat ends (e.g., T4 or Phusion), such as Examples 4 and 5, are used to allow enrichment of both sides of the cleaved library members.

[0087] We have shown that TkoEndoMS can also generate double-strand breaks at base pairs with dG:dU mismatches caused by dC->dU editing (in this example by BE1), indicating its additional applicability to BE1, BE3 and other enzymes that cause dC->dU changes upon DNA binding (see USSN 62 / 571,222 and Fig.25 ). This strongly suggests that we can use TkoEndoMS on our synthetic DNA locus library to identify off-target base edits caused by various base editors that induce dC to dU edits, with or without the presence of Cas9-induced nicking.

[0088] Example 6: Enrichment of DNA binding sites from pull-down

[0089] SELEX (selective evolution of ligands by exponential enrichment) has been used to determine the DNA binding specificity of DNA binding domains (originally by Oliphant et al., Mol Cell Biol. 9: 2944, 1989). In the SELEX method, a library of randomized DNA sequences is pulled down for multiple rounds and enriched with the immobilized DNA binding domain of the target to identify sequences in the initial pool that can bind to the target DNA. The SELEX method has been applied to the zinc finger and TALE portions of ZFN (Perez et al., Nat Biotech. 26: 808 (2008)) and TALEN (Miller et al. Nat Biotech. 29: 143 (2011)), however, there are no reports on SELEX studies of Cas9 proteins. We speculate that SELEX studies of Cas proteins are difficult because it is necessary to selectively enrich for 22 base pair target sites from a large library, which must contain > 10 13 unique molecules or at least 10 12 molecules, corresponding to a 20 base pair target site if the NGG PAM is fixed.

[0090] In this example, we take advantage of our pre-selected libraries to pre-enrich the sites most likely to be bound by a given Cas9:sgRNA complex (or other DNA binding domains with predictable binding motifs). We evaluate the binding preference and specificity of Cas9 DNA by performing sequential rounds of DNA pull-down experiments on the pre-enriched libraries ( Fig.26 ). This is achieved by tethering an inactivated Cas9 (dCas9) to magnetic beads. To chemically bind dCas9 to the magnetic beads, we employed a Cas9 protein with a so-called SNAP tag. Proteins with a SNAP tag can be covalently bound to substrate molecules such as magnetic beads that are loaded with benzylguanine. We envision incubating any type of oligonucleotide library with magnetic bead-bound SNAP-tagged dCas9 and enriching for DNA substrates with high binding affinity for Cas9 by capturing bound sequences by the magnetic beads and washing away unbound sequences. This process can be repeated in multiple cycles by amplifying the eluted library members and using the resulting enriched DNA library as the starting library for magnetic bead-based selection. Using this approach, in a single cycle, we have demonstrated conditions that result in selective pull-down of on-target sites by dCas9:EMX1 sgRNA compared to off-target sites ( Fig. 27 Furthermore, pull-down of the FANCF genomic DNA-stimulated library resulted in the greatest enrichment of the on-target site relative to other sites ( Fig.28). Detailed knowledge of the on-target sites bound by Cas9 is particularly valuable for mechanistic studies to improve the performance of genetically engineered Cas9 variants such as high-fidelity Cas9. Importantly, the off-target pattern of Cas9 fusion proteins (or fusion proteins with other DNA binding domains) with limited interdependence of effector domains and DNA binding domains can be determined primarily by the DNA binding properties of the fusion protein. Therefore, DNA pull-down experiments on pre-enriched oligonucleotide libraries can contribute to our understanding of the off-target distribution of fusion proteins. In addition, by conducting DNA binding studies of Cas9 (or other DNA binding protein domains) in libraries with limited complexity, high-quality binding data with little background noise can be obtained, which can then be used to infer and predict the binding of more complex libraries such as the genome of a cell.

[0091] Example 7: Homing endonuclease selection

[0092] Homing endonucleases, such as I-PpoI, represent a group of naturally occurring nucleases with base recognition motifs longer than most restriction enzymes. Although homing endonucleases (also known as meganucleases) do not have specificity that can be easily reengineered, they may have research, commercial or clinical uses if they target the genomic sequence of the target. Here, we show that we can adapt our in vitro selection to analyze the specificity spectrum of I-PpoI homing endonucleases. We generated an unbiased library of potential I-PpoI off-targets, including all sites with less than 3 mispairings and single DNA / RNA protrusions. The I-PpoI library contains 15,533 members. I-PpoI selection enriched 501 of the 15,533 library members (Table 3), while the expected on-target sites ranked close to the top of the selection (28 of the 15,533). Sequences with one mispairing or one insertion were the most enriched library members. Analysis of the mismatch positions in the top-scoring I-PpoI off-target candidates indicated that certain positions in the recognition motif are more important for I-PpoI cleavage than others ( Fig.29 ). In particular, positions 2, 13, and 14 appear to be highly conserved and are most important for I-PpoI-mediated DNA cleavage. The adaptability of in vitro selection to homing endonucleases suggests that these selections can be widely used to analyze the off-target spectra of a variety of nucleases, including those that are cut to show sticky ends (such as I-PpoI and Cas12a). I-PpoI leaves a 4bp 3' overhang, which is a DNA end structure that is known to reduce the efficiency of off-target detection performed by existing methods such as GUIDE-seq or CIRCLE-seq. Therefore, we confirm that in vitro selection can be used to analyze nucleases that induce staggered DNA breaks.

[0093] Table 3. Top 30 most enriched sites in the I-PpoI selected library versus the unbiased DNA library.

[0094]

[0095] #,SEQ ID NO:

[0096] The most closely matched off-target candidates are enriched to the top of the selection. However, the selection indicates that I-PpoI off-target candidates are abundant.

[0097] Method: Library Generation

[0098] Oligonucleotide library synthesis on high-density chip arrays was purchased from Agilent.

[0099] Substrate library:

[0100] 1) Development of an oligonucleotide backbone with 50% GC content and no potential canonical PAM sequence (NGG for S. pyogenes Cas9).

[0101] 2) Generate 13-14 base pair barcodes that differ from all other barcodes by at least two substitutions, have 40-60% GC, and do not contain any typical PAM sequences for a minimal unbiased library:

[0102] 3) Generate potential off-target sites for all possible combinations of substitutions, insertions and deletions of the SpCas9 target site (this can be variable):

[0103] Displacement Single base pair deletion Single base pair insertion <＝3 0 0 <＝1 1 0 0 2 0 <＝1 0 1

[0104] 4) Incorporate barcodes / potential off-target sites for all i off-target sites (i is about 50,000) into the backbone:

[0105]

[0106] Genome-stimulated library:

[0107] 1) Generate potential off-target sites by CasOffFinder according to the table below (these parameters can vary) and add 20-113 bp (this can be variable) of genomic flanking sequence

[0108] Displacement Single base pair deletion Single base pair insertion <＝6 0 0 <＝4 <＝2 0 <＝3 0 <＝2 4 0 1

[0109] For the EMX1 locus, here is an example of the number of sequences that exhibit the above parameters given.

[0110]

[0111]

[0112] 2) Incorporate barcodes / potential off-target sites for all i off-target sites (i is approximately 50,000) into the backbone, for the smallest unbiased library with the largest genomic flanking context:

[0113]

[0114] Constant backbone sequences can be added as the flanking genomic environment changes.

[0115] For example, with 10 bp of genomic flanking sequence on both sides:

[0116]

[0117] Other library generation strategies:

[0118] -Incorporation of population-based SNPs into genomic sequences

[0119] -Generate libraries based only on coding DNA sequences

[0120] -Generate libraries of oncogene hot spots or tumor suppressor gene loci

[0121] The following is an example of a method using an off-target library constructed using the principles described above.

[0122] Methods for in vitro selection of library members for cleavage

[0123] 1. Library Amplification

[0124] We amplified the oligonucleotide library using primers that bind to constant flanking regions found in all library members. These primers contain 5' end overhangs that introduce additional length and unique molecular identifiers. The library was amplified using the following protocol using 2 μl of 5 nM input library.

[0125]

[0126] SV - Sample Volume

[0127] RV - reaction volume

[0128] *KP_Extended_New_FW, Primer sequence:

[0129]

[0130] *KP_Extension_New_rev, primer sequence:

[0131]

[0132] 2. DNA purification:

[0133] DNA purification was performed with AMPure magnetic beads at a 0.9X sample:bead ratio according to the manufacturer's protocol.

[0134] 3. Enzyme incubation:

[0135] Incubate 300 ng of the chip-synthesized library with the target protein at different enzyme concentrations and incubation times. In most cases (Cas9, Cas9HF, BE3, ABE), 1-2 h of enzyme incubation in activity buffer with 300 ng of oligonucleotide library at a molar ratio of 10:1:1 for protein, sgRNA, and DNA substrate, respectively, is sufficient. Depending on the specific protein function, these parameters may need to be optimized.

[0136] 4. Optional DNA nicking:

[0137] Depending on the protein analyzed, the enzyme incubation does not result in the generation of DNA double-strand breaks (DSBs). In the case of BE3 and ABE, both enzymes only nick one strand of DNA and base edit the other. This DNA nick can be converted into a staggered DSB by using USER enzyme or endonuclease V for BE3 and ABE, respectively (see Figure 8 and 9 To achieve this, incubate the magnetic bead-purified DNA from step 4 with USER enzyme or Endonuclease V in their respective activity buffers at 37°C.

[0138] 5. DNA purification:

[0139] DNA purification was performed with AMPure magnetic beads at a 1.5X sample:bead ratio according to the manufacturer's protocol.

[0140] 6. Optional DNA passivation:

[0141] If an additional nicking step is required (5), staggered DSBs are blunted by incubation with Phusion polymerase at 72°C for 20 min followed by cooling to 4°C.

[0142] 7. DNA purification:

[0143] DNA purification was performed with AMPure magnetic beads at a 1.5X sample:bead ratio according to the manufacturer's protocol.

[0144] 8. Adapter ligation:

[0145] Semi-functional Y-shaped adapters were then ligated to the blunted DNA from step 7. To achieve this, we provided a 10-fold molar excess of adapters over library fragments and performed the ligation using the NEB Quick Ligation Kit, incubating the reaction at 25°C for 10 minutes.

[0146] 9. Gel purification:

[0147] We then gel-purified the ligation reaction by employing 2.5% agarose gel. Electrophoresis was performed for 1 hour at 120 volts. After 1 hour, samples containing lanes around the 180 bp fragment size were cut out and DNA was extracted using the Qiagen gel extraction kit according to the manufacturer's protocol.

[0148] 10.PCR-amplification:

[0149] The eluate from step 9 is then used as input for two PCR reactions that amplify the protospacer-adjacent site and the PAM-adjacent site of the excised library member. The primers used for this PCR contain 5' overhangs that can be used to subsequently attach an Illumina sequencing barcode. Optionally, QPCR can be performed to determine the minimum number of PCR cycles required. PCR is performed using the following parameters:

[0150]

[0151] 11. DNA purification:

[0152] DNA purification was performed with AMPure magnetic beads at a 1.5X sample:bead ratio according to the manufacturer's protocol.

[0153] 12. Quality control using capillary electrophoresis:

[0154] Quality control was performed by checking the PCR products by capillary electrophoresis.

[0155] 13. PCR-based NGS library preparation:

[0156] Append sequencing adapters to the PCR products from step 12 by performing PCR with primers containing Illumina sequencing adapters. Perform PCR using the following parameters:

[0157]

[0158] 14. DNA purification:

[0159] DNA purification was performed with AMPure magnetic beads at a 1.5X sample:bead ratio according to the manufacturer's protocol.

[0160] 15. Next-generation sequencing on Illumina sequencers:

[0161] The DNA library from step 14 was quantified by digital droplet PCR and sequenced on an Illumina sequencer according to the manufacturer's protocol.

[0162] Method for the enrichment of DNA binding sites by pull-down

[0163] 1) Resuspend Snap Capture Beads (NEB)

[0164] 2) Pipette 80uL of magnetic beads into a new 1.5mL Eppendorf tube

[0165] 3) Place the tube in a magnetic particle separator and discard the supernatant

[0166] 4) Add 1 mL of fixation buffer (20 mM HEPES, 150 mM NaCl, 0.5% Tween 20, 1 mM DTT, pH 6.5) and vortex slowly.

[0167] 5) Place the tube in a magnetic particle separator and discard the supernatant

[0168] 6) Prepare protein: Add Engen Spy dCas9 (SNAP-tag) (NEB) (4.5uL of 20uM per pull-down reaction) to 500uL of fixation buffer

[0169] 7) Add the diluted protein to the magnetic beads and mix well by pipetting

[0170] 8) Incubate at room temperature with shaking for 1 hour

[0171] 9) Place the tube in a magnetic particle separator and discard the supernatant

[0172] 10) Wash the beads. Add 1 mL of fixation buffer, pipette to mix well, then place the tube in a magnetic particle separator and discard the supernatant.

[0173] 11) Repeat step 10 twice, for a total of 3 washes. Perform the final wash with fixation buffer containing 10 ug / mL of heparin

[0174] 12) Each time you pull down, resuspend the beads in 45uL of fixation buffer

[0175] 13) Mix the following ingredients:

[0176]

[0177] 14) Incubate at 25°C for 10 minutes

[0178] 15) Add 0.9 pmol of library

[0179] 16) Incubate at 37°C for 30 minutes

[0180] 17) Place the tube on a magnetic bead separator and discard the supernatant.

[0181] 18) Wash the magnetic beads 5 times with 200uL of fixation buffer containing 10ug / mL heparin

[0182] 19) Add 50uL of water and 2uL of proteinase K and incubate at room temperature for 10 minutes while shaking.

[0183] 20) Clean up the pull-down product with DNA purification magnetic beads (e.g. Ampure) and elute with 10uL of 0.1X buffer EB (QIAgen).

[0184] Other Implementations

[0185] It should be understood that although the invention has been described in conjunction with the detailed description of the invention, the foregoing description is intended to illustrate rather than limit the scope of the invention, which is defined by the scope of the appended claims. Other aspects, advantages and modifications are within the scope of the appended claims.

Claims

1. An in vitro method for identifying double-stranded DNA sequences cleaved, modified or bound by an enzyme, the method comprising: (i) providing a library of linear dsDNA oligonucleotides of known sequence, each library member having a known sequence comprising, in order: a first sequence common to each oligonucleotide of the library; A first barcode sequence unique to the library member; potential DNA substrate sequences for nucleases; a second copy of the barcode sequence; and a second sequence common to each oligonucleotide of the library; (ii) incubating the library in the presence of an enzyme selected from the group consisting of a site-specific nuclease, a DNA modifying protein, and a DNA binding domain under conditions sufficient for cleavage, modification, or binding to occur; (iii) selecting the oligonucleotide to be cleaved, modified or conjugated; and (iv) determining the sequence or sequences of the selected oligonucleotides that were cleaved, modified or bound, thereby identifying the double-stranded DNA sequence that was cleaved, modified or bound by the enzyme.

2. The method of claim 1, further comprising enriching for cleaved, modified or conjugated oligonucleotides.

3. An in vitro method for identifying double-stranded DNA sequences cleaved, modified or bound by an enzyme, the method comprising: (i) providing an initial library of linear dsDNA oligonucleotides of known sequence, each library member having a known sequence comprising, in order: a first sequence common to each oligonucleotide of the library; A first barcode sequence unique to the library member; potential DNA substrate sequences for nucleases; a second copy of the barcode sequence; and a second sequence common to each oligonucleotide of the library; (ii) incubating the library in the presence of an enzyme selected from the group consisting of a site-specific nuclease, a modifying protein, and a DNA binding domain under conditions sufficient for cleavage, modification, or binding to occur; (iii) selecting oligonucleotides that have not been cleaved, modified or conjugated; (iv) determining one or more sequences of the selected oligonucleotide that have not been cleaved, modified or conjugated; and (v) comparing the sequence of the selected oligonucleotide that was not cleaved, modified or bound to the sequence of an initial library of linear dsDNA oligonucleotides of known sequence; wherein the unselected linear dsDNA oligonucleotides in the initial library are identified as being cleaved, modified or bound by the enzyme.

4. An in vitro method for identifying a double-stranded DNA sequence modified by a base editing enzyme, the method comprising: (i) providing a library of linear dsDNA oligonucleotides of known sequence, each library member having a known sequence comprising, in order: a first sequence common to each oligonucleotide of the library; A first barcode sequence unique to the library member; potential DNA substrate sequences for nucleases; a second copy of the barcode sequence; and a second sequence common to each oligonucleotide of the library; (ii) incubating the library in the presence of a base editing enzyme under conditions sufficient for modification to occur; (iii) during DNA synthesis, amplifying the oligonucleotide by a polymerase that converts the edited base pairs to an equivalent mixture of canonical base pairs; and (iv) determining the sequence of one or more of the amplified oligonucleotides, Thereby recognizing the double-stranded DNA sequence modified by the base editing enzyme.

5. The method of claim 4, wherein the base editing enzyme is a cytidine deaminase that converts deoxycytidine to deoxyuridine or an adenine base editing enzyme that converts deoxyadenine to deoxyinosine.

6. The method of claim 4, wherein the polymerase is a uracil-tolerant polymerase that converts dU:dG base pairs into an equal mixture of dT:dA and dC:dG base pairs, or converts dI:dT base pairs into an equal mixture of dA:dT and dG:dC base pairs.

7. An in vitro method for identifying a double-stranded DNA sequence modified by a cytidine deaminase base editor enzyme that converts cytidine to uridine and creates a nick on the opposite strand, the method comprising: (i) providing a library of linear dsDNA oligonucleotides of known sequence, each library member having a known sequence comprising, in order: a first sequence common to each oligonucleotide of the library; A first barcode sequence unique to the library member; potential DNA substrate sequences for nucleases; a second copy of the barcode sequence; and a second sequence common to each oligonucleotide of the library; (ii) incubating the library in the presence of a base editing enzyme under conditions sufficient for modification to occur, and then incubating the library in the presence of an enzyme to produce a nick at a site with a uridine nucleotide, thereby producing a dsDNA oligonucleotide containing two nicks with a 5' phosphate on opposite strands, thereby producing an overhang; (iii) incubating the dsDNA oligonucleotide with a DNA polymerase that generates 5' phosphorylated blunt ends from the overhangs; (iv) capturing the phosphorylated blunt end using a double-stranded DNA adaptor comprising a primer sequence; (v) amplifying the sequence using a primer specific for the adaptor and a primer specific for the consensus sequence backbone; and (vi) determining the sequence of the amplified oligonucleotide; This allows the recognition of double-stranded DNA sequences modified by base editing enzymes.

8. The method of claim 7, further comprising performing additional selection by size selection of smaller cleavage fragments before or after amplification.

9. The method according to claim 7, wherein the DNA polymerase is T4 DNA polymerase or Phusion DNA polymerase or Phusion U DNA polymerase.

10. An in vitro method for identifying a double-stranded DNA sequence modified by an adenine base editing enzyme that converts deoxyadenine to deoxyinosine and creates a nick on the opposite strand, the method comprising: (i) providing a library of linear dsDNA oligonucleotides of known sequence, each library member having a known sequence comprising, in order: a first sequence common to each oligonucleotide of the library; A first barcode sequence unique to the library member; potential DNA substrate sequences for nucleases; a second copy of the barcode sequence; and a second sequence common to each oligonucleotide of the library; (ii) incubating the library in the presence of a base editing enzyme under conditions sufficient for modification to occur, and then incubating the library in the presence of an endonuclease V enzyme to produce a nick at a site with an inosine nucleotide, thereby producing a dsDNA oligonucleotide containing two nicks with a 5' phosphate on opposite strands, thereby producing an overhang; (iii) incubating the dsDNA oligonucleotide with a DNA polymerase that generates 5' phosphorylated blunt ends from the overhangs; (iv) connecting the phosphorylated blunt end to a double-stranded DNA adapter comprising a primer sequence; (v) amplifying the sequence using a primer specific for the adaptor and a primer specific for the consensus sequence backbone; and (vi) determining one or more sequences of the amplified oligonucleotides; Thereby recognizing the double-stranded DNA sequence modified by the base editing enzyme.

11. The method of claim 10, further comprising performing additional selection by size selection of smaller cleavage fragments before or after amplification.

12. The method according to claim 10, wherein the DNA polymerase is T4 DNA polymerase or Phusion DNA polymerase or Phusion U DNA polymerase.

13. An in vitro method for identifying a double-stranded DNA sequence modified by an adenine base editor enzyme that converts deoxyadenine to deoxyinosine and creates a nick on the opposite strand, or a cytidine deaminase base editor enzyme that converts cytidine to uridine and creates a nick on the opposite strand, the method comprising: (i) providing a library of linear dsDNA oligonucleotides of known sequence, each library member having a known sequence comprising, in order: a first sequence common to each oligonucleotide of the library; A first barcode sequence unique to the library member; potential DNA substrate sequences for nucleases; a second copy of the barcode sequence; and a second sequence common to each oligonucleotide of the library; (ii) incubating the library in the presence of an endonuclease MS from Thermococcus kodakarensis to induce a double-strand break (DSB) at a deamination site in the substrate DNA to generate a DNA fragment having a single-stranded, 5-base pair overhang centered at the deamination site; (iii) treating the DNA fragments with uracil DNA glycosylase and endonuclease VIII to remove deoxyuracil bases from the ends of the DNA fragments; (iv) end repair and / or A-tailing of the ends of the DNA fragments; (v) ligating adaptor oligonucleotides to the termini; and (vi) sequencing the DNA fragment.

14. The method of claim 13, wherein the adaptor oligonucleotide comprises a sequence for high throughput sequencing.

15. An in vitro method for identifying a double-stranded DNA sequence bound by a catalytically inactive Cas9 in the presence of a selected gRNA or other DNA binding domain, the method comprising: (i) providing a library of linear dsDNA oligonucleotides of known sequence, each library member having a known sequence comprising, in order: a first sequence common to each oligonucleotide of the library; A first barcode sequence unique to the library member; potential DNA substrate sequences for nucleases; a second copy of the barcode sequence; and a second sequence common to each oligonucleotide of the library; (ii) incubating the library in the presence of a DNA binding domain attached to magnetic beads under conditions sufficient for binding to occur; (iii) selecting the bound oligonucleotides by pulling down one or more pools of magnetic beads and washing in an appropriate buffer that promotes dissociation of unbound molecules into the supernatant, and then eluting the bound DNA in an appropriate buffer that promotes dissociation of any bound DNA or in a buffer containing a protease that degrades proteins bound to the magnetic beads and releases the bound DNA; and (iv) determining the sequence or sequences of the selected oligonucleotide that was cleaved, thereby identifying the double-stranded DNA sequence bound by the DNA binding domain.

16. The method of claim 15, further comprising enriching the bound oligonucleotides.

17. The method of claim 15, wherein the DNA binding domain is a Cas9 enzyme complexed with an sgRNA or other DNA binding domain.

18. The method of claim 15, wherein the DNA binding domain is attached to the magnetic beads by covalent binding or binding via an affinity handle.

19. The method of claim 15, wherein the protease is proteinase K.

20. The method according to any one of claims 1 to 19, wherein the library comprises: (i) a collection of all potential off-target sequences in the reference genome that have a certain number of mismatches or less relative to the identified on-target sites; (ii) a comprehensive set of potential off-target sites with a certain number of mismatches or less; (iii) a library of potential off-target sequences present in a collection of variant genomes from a particular population; or (iv) Other relevant specific sets of potential off-target sites.

21. The method of claim 20, wherein the potential off-target site is an oncogene hotspot or a sequence from a tumor suppressor gene.

22. The method according to any one of claims 1 to 19, wherein: The library members are first synthesized as individual single-stranded DNA sequences; and The single-stranded DNA sequences are converted into double-stranded DNA molecules by priming the consensus sequence, optionally before or after release from the chip.

23. The method of claim 22, wherein the library members are first synthesized as individual single-stranded DNA sequences on a high-density oligonucleotide array.

24. A method according to any one of claims 1 to 19, wherein the library members represent 1) a collection of all potential off-target sequences in a reference genome with a certain number of mismatches or less relative to the on-target site, 2) a comprehensive collection of potential off-target sites with a certain number of mismatches or less, 3) a library of potential off-target sequences present in a collection of variant genomes from a specific population, or 4) other related specific collections of potential off-target sites.

25. The method of claim 24, wherein the potential off-target site is an oncogene hotspot or a sequence from a tumor suppressor gene.

26. The method according to any one of claims 1 to 19, wherein the library comprises 1,000 to 10 11 Different sequences.

27. The method according to any one of claims 1 to 19, wherein the library comprises sequences with a length of 50 to 500 bp.

Images