The CRISPR/Cas family of proteins comprises DNA- or RNA-cleaving enzymes that can be programmed to cut specific sequences using oligonucleotide-length RNA, making them suitable candidates for genome editing but also targeted sequencing, given their high specificity.

CRISPR RNAs (crRNAs) program Cas9 to bind and cleave DNA at sites that are identical (or highly similar) to the crRNA sequence. Candidate crRNAs need to be determined by the user and can be found by bioinformatic searches of reference genomes (detailed below). When Cas9 is loaded with crRNA identical to the intended target sequence, together with trans-activating CRISPR RNA (tracrRNA - a structural RNA required for catalytic activity) it forms a ribonucleoprotein complex (RNP). This complex of RNA and protein searches the genomic DNA for the target region using the crRNA sequence. The complex attempts to seed the crRNA by melting duplex DNA, and, if the crRNA matches and base-pairs with the target sequence, Cas9 cuts both strands of the target sequence.

Cas9 requires a 20mer target site (the protospacer), plus the protospacer-adjacent motif (PAM - sequence NGG). The PAM is immediately next to the 3’ end of the target and defines the boundaries of the protospacer sequence. Cas9 cuts both strands of the target sequence 3 bp upstream of the PAM. The production of a newly-exposed end, what we call ‘deprotection’, at a defined site is the basis for the selectivity of the Cas-mediated enrichment method.

For Cas9 targeted sequencing, we are able to selectively protect and deprotect ends using CRISPR/Cas and enrich for regions of interest in a background prior to long read nanopore sequencing of the native DNA molecule.

There are several ways to design a Cas9 targeting experiment, and the method depends on the target region of interest. Multiple factors determine which method should be used:

• Size of region of interest (ROI)
• Length of input DNA
• Are the sequences either side of the region of interest known?
• How much coverage of region of interest is required?

There are three main options, detailed below.

crRNA probes are design to target regions either side of the region of interest to excise that region (ideal for particular gene targets).

This method should be used:

• When the region of interest is <20 kb (largest target region if multiple regions are being targeted)
• When both ends of the region of interest are known, to be able to design probes on either side

Figure 1. Excising a region of interest using four cuts (two each side of the ROI). Cutting once on each side results in incomplete cleavage, as shown by the high coverage depth of the region behind the ROI after the first cut (bottom panel).

We generally recommend excising a ROI by making four cuts, two upstream of the ROI targeting the (+) strand, and two downstream targeting the (-) strand. Four cuts instead of two are for redundancy, in case one or more crRNAs yield incomplete cleavage. This method should provide the highest coverage of a target region, as on-target strands will have sequencing adapters ligated to both ends. Regions larger than 20 kb can be targeted using this approach, but this is limited by the length of the input DNA and users may experience drops in coverage towards the middle of the region of interest.

A crRNA probe is designed at one end of a target region to read into the unknown (ideal for characterisation of integration sites within a genome).

This method should be used:

• When the region of interest is <20 kb (largest target region if multiple regions are being targeted)
• When only one side of the region of interest is known

Figure 2. Excising a region of interest using a single cut, when only one side of the ROI is known. The coverage depth is high both for the ROI and the region downstream (bottom panel).

This method will give lower throughput due to only ligating sequencing adapters to one end of the strand. Regions larger than 20 kb can be targeted using this approach, but this is limited by the length of the input DNA and users may experience drops in coverage towards the end of the region of interest.

Figure 3. Example coverage plots generated using the Oxford Nanopore bioinformatics tutorial “Evaluation of read-mapping characteristics from a Cas-mediated PCR-free enrichment” for the HTT gene with alternative probe design methods. An enrichment experiment run on a single MinION R9.4.1 flow cell targeting the HTT gene with a single probe (panel 1), an excision approach using one probe on either side of the region of interest (panel 2) and the recommended excision approach using 4X probes for redundancy (panel 3). Red lines on the coverage plot highlight the range in the input BED file used in the bioinformatics tutorial. This example ranges from outer probe to outer probe for the 4X excision method for all plots. The HTT gene falls between these red lines, so the average coverage of the HTT gene is the highest coverage observed on these graphs. Therefore the average coverage of the HTT gene with a single probe is around 100X, 1600X for an excision approach with 2X probes and 2100X coverage with a 4X probe excision.

crRNA probes are designed along a region of interest in 5-10 kb overlapping chunks that allows for even coverage across the region of interest.

This method should be used:

• When the region of interest is >20 kb (especially when limited by the length of the input DNA)

Figure 4. Probe design for a “bricklayer” tiling approach. Two sets of probes are designed covering a large region of interest (>20 kb). This breaks up the ROI into overlapping sections, ensuring good coverage of the whole length of the ROI.

This method involves designing two pools of probes which should be prepared as two separate experiments and pooled in the final step (before the final clean up) and loaded together onto the same flow cell. There are two options for designing the probes. For most use cases, we recommend a “bricklayer” approach, where each pool contains (+) and (–) probes that are designed to overlap each other. The second “highway” option is to have one pool of probes cutting in one direction (+) and one pool of probes cutting in another (–) (probe directionality is explained below). The “bricklayer” approach has shown to give a higher coverage of the target region.

Figure 5. Decision tree for designing the Cas9 targeting approach based on the size of the target region.

Independent of which Cas9 cut method is used, the probe design process is the same and the same features that make a good probe apply to all. The diagram below shows what the RNP complex looks like bound to the target sequence (in black and PAM in purple, Figure 6 (1) and (2)). Upon RNP binding, the duplex is locally melted, and the crRNA hybridises to the non-target DNA (bottom strand in Figure 6 (2) which is complementary to the target and crRNA sequence). This means that the crRNA sequence is same sequence as the ROI (Figure 6 (2), shown in red). Therefore, when designing a crRNA probe, the sequence should be the same as the region of interest.

Figure 6. (1) the RNP complex bound to the region of interest. (2) The RNP melts the DNA duplex upon hybridisation with the strand complementary to the target region. The crRNA sequence is identical to the target sequence upstream of the PAM.

The orientation of the probes is critical to designing a Cas9 cleavage experiment, the orientation being defined by the PAM and target sequence. Cas9 cuts at the PAM-proximal end of the protospacer 3 bp upstream of the PAM.

Figure 7 (3) shows details of the Cas9 cleavage reaction. The protection of the PAM-distal end by Cas9, which remains bound, provides the directionality for the strand but this process is imperfect. Sometimes Cas9 can release the DNA, but our internal data suggests that the reads towards the PAM site outnumber the reads away by a factor of at least 3:1, up to at least 10:1 (shown by the grey arrows in Figure 7 (3) and the small arrows in Figure 8).

Figure 7. (3) The Region of Interest is cleaved at the 5’ side, and the PAM-distal site is protected by the bound Cas9. (4) The cleavage and protection happens on both sides of the ROI. (5) The directionality of the read is determined by the protection of the PAM-distal site by Cas9. In most cases, the ROI is enriched for, and read by the nanopore. In a minority of cases, the DNA is read in the other direction, away from the ROI.

Figure 8. The directionality of the sequenced read when the Region of Interest is cleaved by the excision approach (twice on each side).

To design a crRNA probe, you initially must search the region upstream, and downstream if doing an excision approach, of the region of interest for a PAM sequence (NGG). We recommend a minimum of 1 kb flanking region between the region of interest and probe target sequence. This is very important when looking at repeat expansions and regions of low complexity to help with scaling of signal during basecalling, and this flanking region can be expanded if the region of interest is very small and could be lost in subsequent clean-ups (<3 kb). All crRNA probe sequences should be checked against the whole target genome to see if there are off-target alignments, which would result in cuts elsewhere in the genome. Assess if there are any known SNPs in the crRNA probe target sequence, as these SNP mismatches will reduce the efficiency of a probe, potentially causing the absence of a cut at the region of interest in non-reference samples.

Cut kinetics are acceptable for panels of up to ~100 crRNAs. Single or multiple gene fragments may be targeted in an experiment. To excise a ROI, at least four crRNAs are required, but up to a hundred or more may be used. For simplicity, we advise users to pre-mix crRNAs at equimolar concentration and form all ribonucleoprotein complexes (RNPs) simultaneously in a single tube (unless performing a tiling experiment where each pool should be kept separate). The greater the number of crRNAs in a single cut reaction, the lower the concentration of each probe, and therefore the slower the cut kinetics. The reaction is performed at excess RNPs with respect to target. In-house data at Oxford Nanopore Technologies shows that the kinetic efficiency of the cut reaction is complete within ~15 min and acceptable for panels of up to ~40 crRNAs. Larger panels may require longer cut times.

We recommend that even users familiar with existing CRISPR design criteria read the section below carefully, because there are rules specific to nanopore target enrichment. Oxford Nanopore Technologies recommends free crRNA probe design tool CHOPCHOP (chopchop.cbu.uib.no) and ordering IDT Alt-R™ Cas9 probes: idtdna.com/crispr-cas9.

CHOPCHOP (chopchop.cbu.uib.no) identifies all possible target sites, based on the availability of PAM sequences, and evaluate each site for the efficiency of cutting and possible off-target effects (based on experimental data). A CHOPCHOP search would normally involve:

Input parameters (user-defined): Figure 9. The CHOPCHOP homepage, with the fields selected for CRISPR/Cas-based nanopore enrichment.

• A FASTA sequence or specific genomic coordinates. Remember to include 1 kb flanking regions, and a separate search is required for probes upstream and probes downstream of the region of interest (‘Target’ box on CHOPCHOP homepage)
• The target genome to search for off-target sites (‘In’ box on CHOPCHOP homepage)
• Which CRISPR/Cas system to use (‘Using’ and ‘For’ box on CHOPCHOP homepage. Select ‘CRISPR/Cas9’ and ‘nanopore enrichment’)
• Manual parameters, e.g. the PAM, protospacer length, and the algorithms used to score the efficiencies of probes (select ‘Options’ and change Efficiency score to ‘Doench et al. 2014 – only for NGG PAM’)

Figure 10. Recommended parameters in CHOPCHOP for probe design.

The search tool, performed within CHOPCHOP:

• Identifies all possible protospacer candidates based on the locations of PAM sequences
• Scores and ranks them based on:
  • Predicted efficiency
  • Self-complementarity (how likely the crRNA is to form unwanted secondary structure)
  • Number of off-target sites in the genome bearing 0, 1, 2 or 3 mismatches relative to the candidate protospacer sequence, as judged by an alignment of each candidate protospacer sequence against the entire genome

Filtering the search results, user-performed: to remove candidates with significant secondary structure, off-target effects, or low efficiency.

CHOPCHOP will only allow for a search within 20 kb at a time. For loci >20 kb, multiple searches should be performed, but the results from multiple searches can be pooled. This is required if using a tiling approach.

CHOPCHOP returns the target sequence, i.e. the protospacer + PAM sequence. For S. pyogenes Cas9, this will be a 20mer sequence + NGG. Cas9 will not cleave the target if the PAM is included in the crRNA sequence.

Example for Cas9: if target = GTTAGTGTCCCCATACAACGGGG, the 20mer crRNA = GTTAGTGTCCCCATACAACG.

Oxford Nanopore Technologies recommends ordering IDT Alt-R™ Cas9 probes: idtdna.com/crispr-cas9. Alt-R™ probes contain proprietary modifications that increase the stability of the crRNA and its resistance to nuclease-mediated degradation.

The crRNA requires a short sequence at the 3’ end of the protospacer sequence that hybridises to tracrRNA; however, this sequence is automatically applied to the crRNA sequence at the time of ordering; only the protospacer sequence (no PAM, as DNA provided by CHOPCHOP) is required. As an RNP complex is required to perform a Cas9 experiment, users must ensure that they also order tracrRNA (IDT Alt-R™, Cat # 1072532, 1072533 or 1072534) and S. pyogenes Cas9 (Alt-R® S. pyogenes HiFi Cas9 nuclease V3 IDT Cat #1081060 or 1081061).

1. A 5' G is needed to make the in vitro transcript, which limits the candidates by 4-fold. Adding one where it does not already exist causes issues.
2. Transcribing RNA in vitro can yield highly variable results.

If you would still like to try, skip the annealing step and add 500 nM final sgRNA during the RNP formation step.

The crRNA sequences for each of these probes are:

• HTT_2561 = TTTGCCCATTGGTTAGAAGC
• HTT_2662 = TCTTATGAGTCTGCCCACTG
• HTT_7412 = GGACAAAGTTAGGTACTCAG
• HTT_9569 = CTAGACTCTTAACTCGCTTG

These can be ordered from IDT and used in the initial step of the sample prep protocol as a control experiment or as an in-run control with other targets.

A tutorial delivering tools for reviewing the performance of a Cas9 targeted sequencing experiment is provided by Oxford Nanopore Technologies. The tutorial guides the assignment of sequences that can be defined as on-target, off-target and background reads and presents both tabular data and graphical plots that can be used to assess the performance of an enrichment study.

This analysis tutorial can be run with two different methods, but both will generate similar statistics, graphs and outputs. The recommended method is using the EPI2ME Labs platform.

The tutorial requires a FASTQ format sequence file generated by MinKNOW and a BED file for target coordinates as input, along with pointers to download locations for the reference genome to be used. The EPI2ME Labs tutorial has an embedded example dataset to demonstrate the type of files needed and an example of the output.

The tutorial aids with the quantification of the non-target depletion and provides information on mapping characteristics that highlight the protocol's performance. The executive summary generated at the end of the EPI2ME Labs tutorial provides the key metrics used to establish whether an experiment has been successful.

Single HTT gene target in human:

10 gene targets (5-15kb) in human:

The most important value to focus on is the “Target Coverage”, which will be an average value if multiple targets are being assessed in the same experiment. For a well-designed Cas9 targeted experiment, we would expect >200x coverage for a single target. The coverage of the region of interest is mostly dependent on probe design and input DNA quality. To assess coverage in more detail, the coverage plots show overlapping directional coverage for the region of interest. For multiple targets in the same experiment, when the headline value provides an average coverage, these plots are useful to look at individual coverage of the regions. If the coverage is lower than 200x for a region of interest, the coverage plot can show where a crRNA probe may not be performing well. Firstly, ensure that additional probes have been added to either side of the region of interest for redundancy as explained in the Probe Design section of this document, as this could significantly boost coverage of the region of interest if a probe is not performing well. For probes to be working equally well in both directions (+/-), a similar coverage plot will be observed for both directions as show in the HTT below. In the example of the SCA17 gene below, the + probe has performed better as it has generated a higher coverage than the probe in the – direction. In this example, it would be recommended to review the probes in the – direction to boost overall coverage.

The other figures on the EPI2ME Labs tutorial relate to background reads and off-target regions. Background reads, off-target effects and explanation for the % of reads on-target are described elsewhere in this document, and all of these statistics can be found in the executive summary on the EPI2ME Labs tutorial. If the total “Throughput” of the run is high (similar to that observed with a non-targeted sequencing run), it is likely that the dephosphorylation step has not worked in the preparation, and new reagents are recommended. Please note that the % of reads on-target is dependent on the size of the region of interest compared to the size of the genome. As well as the headline stats, the EPI2ME Labs tutorial provides an insight into possible off-target regions. These regions are described as areas of the genome that are outside of the region of interest but have a higher coverage than the background coverage of the genome. These regions arise from crRNA probes causing Cas9 to cut at other sequences very similar to the sequences being cut around the region of interest. To reduce the number of off-target sites, and in turn boost the coverage of the region of interest by reducing the activity of Cas9 in other regions and focusing more on the region of interest, we recommend reviewing the the crRNA probe design.

Output from the EPI2ME Labs tutorial can be assessed in IGV for a more detailed visualisation of the region of interest. The output can also be further processed by other bioinformatic tools to assess SNPs, SV and repeat counting. Refer to Bioinformatics Resource page in the Nanopore Community for more information.

As the target region is often a small part of the genome of interest, the overall throughput will be lower than a standard Ligation Sequencing Kit (SQK-LSK109) run, but more time is spent sequencing target DNA. Cas9 targeted sequencing experiments will therefore boost the coverage of the region of interest several hundred-fold, and users will see a reduction in coverage of the rest of the genome compared to a whole genome sequencing experiment.

Figure 14. Relative coverage of the whole genome and the Region of Interest with and without Cas9 targeting.

• Overall throughput
• % of reads on target
• Coverage of region of interest
• Depletion of the non-target DNA

These key metrics can easy be determined using the Oxford Nanopore Technologies Cas9 enrichment-specific data analysis tutorial, which will also generate coverage plots for the region(s) of interest and provide information about specific off-target cuts (details below).

• Overall throughput = >1 Gb
• % of reads on target = 5-10%
• Average coverage of regions of interest = >100x
• Depletion of non-target DNA = ~3000-fold depletion

For larger or smaller genomes and regions of interest, coverage will decrease or increase respectively. Ploidy will also impact coverage, e.g. a haploid cell will have fewer copies per cell than a diploid or triploid cell. Users should calculate the levels of enrichment expected based on their target size, copy number, genome size, etc.

The % of reads that are on-target in a run (and the target coverage) is governed by the set-up of the experiment. Targeting multiple regions in a single sample will give a similar coverage for each target region (compared to sequencing just a single target region) but increase in % of reads on target. This is because we are increasing the proportion of the genome being targeted and therefore increasing the % of input that corresponds to the target. If a user wants to look at a single gene target but in multiple samples, which might be barcoded, then the overall yield of the experiment will be higher but the proportion of reads on-target will be like a single gene, single sample experiment. Coverage per target will also be lower for the multiple sample options.

Table 1. Calculations of enrichment data with variation of several key input parameters. Example cases based on typical sequencing runs of control human reference samples.

Non-target DNA observed during sequencing can be split into two categories: off-target and background. Each crRNA in a panel should allow Cas9 to cut genomic DNA at the site that perfectly matches its sequence, but may also cut at sites bearing multiple mismatches, leading to adapter ligation at those sites and a reduced proportion of on-target reads. This “off-target” activity can be mitigated by the careful design of crRNAs to have a minimum number of mismatches while maintaining cut efficiency. Background DNA can come from efficiency in the dephosphorylation step or ligation of sequencing adapters to non-cut DNA.

To analyse enrichment data and assess the level of off-target and background DNA sequenced, please refer to the Evaluation of read-mapping characteristics from a Cas-mediated PCR-free enrichment bioinformatics tutorial.

For optimal target coverage, Cas9 targeted sequencing experiments require at least ~5 pg of target material in the input. For example, enrichment of a 5 kb human gene would require 5 µg of input DNA.

Table 2. Experimental data from an DNA input titration from a human 10 gene-panel using the Ca9 targeted sequencing protocol. The proportion of reads or bases on target remains constant, but target coverage is roughly proportional to input amount, for 500 ng to 5 µg total input.

For an efficient enrichment experiment, high molecular weight DNA is required. The median length of molecules in a genomic DNA sample can greatly impact upon the efficiency of the enrichment in two main ways:

• The shorter the median DNA fragment length, the greater the concentration of DNA ends that must be protected from ligation (assuming constant mass of DNA). Thus, in general, the shorter the DNA, the greater the background
• If pairs of crRNAs designed to excise a ROI are placed more than one median fragment length apart, coverage drop between the cut sites will be significant.

For these reasons, we recommend purifying genomic DNA to the highest possible length and quality and matching the spacing of crRNAs to the expected median fragment length.

The table below highlights the key features and requirements of Cas9 targeted sequencing using S. pyogenes Cas9. Although other CRISPR/Cas enzymes, such as Cas12a, are available, we recommend that new users begin developing their own targeted sequencing assay using S. pyogenes Cas9 for the following reasons:

• Cas9 enzymes, including a wide variety of high-fidelity mutants, are more widely available from third parties;
• A far wider library of verified probe designs is already available for Cas9 via the scientific literature;
• The rules for predicting the efficiencies (i.e. the proportion of target predicted to be cut) and off-target effects of Cas9 crRNAs are better understood than for other enzyme crRNAs, i.e., the design criteria are better-defined;
• The Cas9 cleavage reaction yields a blunt-ended double-stranded break, so the biochemistry required to generate an end to which an adapter can be ligated is simpler.

Table footnote:

1. Streptococcus pyogenes species
2. Based on pricing of Alt-R™ synthetic crRNA at the lowest synthesis scale in vials, provided by Integrated DNA Technologies (IDT). 2 nmol is sufficient for ~200 reactions. Workflows involving multiple crRNAs require preparing an equimolar mix of crRNAs. crRNAs may be ordered in bulk in multi-well format at lower cost. Pricing correct as of September 2018.
3. Based on pricing of Alt-R™ synthetic tracrRNA from IDT. One-time purchase: the same tracrRNA is used for all Cas9 cleavage reactions. 5 nmol is sufficient for ~500 reactions. Pricing correct as of September 2018.