Detecting off-targets in the lab: GUIDE-seq, CIRCLE-seq, and Digenome-seq compared

This is Part 14 of the CRISPR from Bench to Analysis series.

The Off-Target Problem

We like to think of CRISPR/Cas9 as molecular scissors that cut only at our target genomic address. In reality, Cas9 is slightly more forgiving. It tolerates small base-pair mismatches between the guide RNA (gRNA) and the genomic DNA, especially if they occur far from the PAM site (the "seed" region).

This means Cas9 can cleave off-target loci, potentially leading to unintended gene disruption, chromosome rearrangements, or activation of oncogenes.

While computer algorithms can predict potential off-target sites based on sequence similarity (we'll cover these in post 15), biology is more complex. Access to off-target sites is heavily influenced by chromatin structures, DNA methylation patterns, and spatial organization. A site that is easily cut in a tube of purified DNA might be tightly wrapped around histones and completely protected in a living cell.

To know where Cas9 actually cuts in your system, you have to measure it experimentally. Let's look at the four primary laboratory methods for finding off-targets: GUIDE-seq, CIRCLE-seq, Digenome-seq, and DISCOVER-seq.

Cell-Based vs. In Vitro: The Great Divide

Experimental off-target detection methods are split into two categories:

1. In Vitro Methods (in a test tube): Purify genomic DNA from cells, clear it of all proteins (including histones), and incubate it with recombinant Cas9 RNP. Because there is no chromatin blocking access, these methods identify every possible target that Cas9 can cut. They are highly sensitive but include false positives (sites cut in the tube that would never be cut in a living cell).
2. Cell-Based Methods (in living cells): Introduce Cas9/gRNA into living cells and capture the double-strand breaks (DSBs) as they are created or repaired. These methods capture physiological cuts but are limited by cell transfection efficiency and sequencing coverage.

1. GUIDE-seq (Cell-Based)

Developed by the Joung Lab at Harvard, GUIDE-seq (Genome-wide Unbiased Identification of DSBs Enabled by Sequencing) is the most widely-used cell-based off-target detection assay.

How it works

1. You co-transfect your cells with the Cas9/gRNA plasmid and a small, chemically modified double-stranded oligodeoxynucleotide (dsODN) tag.
2. When Cas9 cuts the genome, the cell's Non-Homologous End Joining (NHEJ) repair machinery accidentally integrates this dsODN tag directly into the double-strand break.
3. You harvest the genomic DNA, shear it, and use PCR to amplify only the regions flanking the integrated tag.
4. You sequence the PCR products. Every read containing the tag reveals a site where Cas9 cut inside the living cell.

Pros & Cons

• Pros: Measures editing inside living cells (physiologically relevant); relatively simple workflow.
• Cons: Requires transfecting cells with foreign DNA (can be toxic in primary cells or stem cells); relies on active DNA repair (NHEJ) to integrate the tag.

2. CIRCLE-seq (In Vitro)

Also developed by the Joung Lab, CIRCLE-seq (Circularization for In vitro Reporting of Cleavage Effects by Sequencing) is an extremely sensitive in vitro assay.

How it works

1. You extract genomic DNA from cells and shear it into fragments (~300 bp).
2. You circularize these fragments using ligase and degrade any remaining linear DNA with exonucleases.
3. You incubate the circularized DNA with your Cas9 RNP.
4. Cas9 cleaves the circular DNA at its target/off-target sites, converting them back into linear molecules. Fragments without cut sites remain circular.
5. You add sequencing adapters only to the linear molecules, PCR amplify, and sequence.
6. The reads represent the exact cleavage sites.

Pros & Cons

• Pros: Extremely sensitive (can detect off-target events with frequencies $< 0.1\%$); does not require transfecting living cells.
• Cons: Captures off-targets that may be biologically blocked by chromatin in vivo (high false-positive rate); requires a large amount of purified genomic DNA.

3. Digenome-seq (In Vitro)

Developed by the Kim Lab at Seoul National University, Digenome-seq (Digested Genome Sequencing) is a straightforward in vitro approach using whole-genome sequencing (WGS).

How it works

1. You digest purified genomic DNA in vitro with Cas9 RNP.
2. You sequence the digested genome using standard whole-genome sequencing (WGS) at high depth (typically 30x to 40x).
3. In a normal WGS run, reads map randomly across the genome, creating ragged, uneven alignment boundaries. But at a Cas9 cleavage site, all digested molecules start and end at the exact same nucleotide.
4. Computational analysis searches for these sharp, vertical "cliffs" in read alignments to map target and off-target cleavage sites.

Pros & Cons

• Pros: No PCR amplification required (no amplification bias); highly unbiased.
• Cons: Extremely expensive (requires whole-genome sequencing at high depth); requires sophisticated bioinformatics to separate Cas9 cliffs from random genomic shearing cliffs.

4. DISCOVER-seq (Cell-Based)

Developed by the Corn Lab at ETH Zurich, DISCOVER-seq (Double-Strand Breaks Identified by Chromatin Immunoprecipitation) is a tag-free, cell-based method.

How it works

1. When a double-strand break occurs inside a cell, repair proteins immediately rush to the site. One of the first responders is a protein complex containing MRE11.
2. You express Cas9/gRNA in cells, fix them with formaldehyde, and use an antibody to pull down the MRE11 protein (a process called Chromatin Immunoprecipitation, or ChIP).
3. You sequence the DNA bound to MRE11 (ChIP-seq).
4. Peak alignments identify the exact location of DSBs in vivo without inserting foreign DNA tags.

Pros & Cons

• Pros: Tag-free (can be used in primary cells, organoids, and even in vivo animal models); measures cutting in native chromatin.
• Cons: ChIP-seq protocols are notoriously difficult and sensitive; lower sensitivity compared to tag-based methods.

Head-to-Head Comparison

Choosing the Right Tool

Which method should you use for your project?

• If you are working with standard, easy-to-transfect cell lines (e.g. HEK293T, HeLa): Use GUIDE-seq. It is the standard, is widely supported by analysis pipelines, and tells you what actually cuts in a cellular environment.
• If you are designing a therapeutic gRNA and need to identify every possible off-target risk: Run CIRCLE-seq or Digenome-seq first to create a comprehensive list of candidate sites, then validate those specific sites in your target cells using targeted amplicon NGS (which we learned in post 13).
• If you are working with primary cells, stem cells, or animal tissues that cannot tolerate dsODN transfections: Use DISCOVER-seq.

What's next?

Now that we understand the experimental methods to discover off-targets, how do we use computational tools to predict them before we even start our experiment?

In the next post, we will look at the bioinformatics of off-target prediction: Cas-OFFinder, CRISPR-P, and how scoring matrices work.