Assessing Off-Target Effects in Epigenetic Editing

"Off-target" covers two mechanistically different problems, and conflating them leads to under-powered specificity claims:

1. Guide-directed off-targets. The editor is recruited to a sequence resembling the intended site, and the mark is written there instead of (or in addition to) the target. This is a sequence-similarity problem, inherited from CRISPR.
2. Effector-directed off-targets. The fusion protein modifies chromatin beyond the intended locus — through spreading from the target, through untethered effector activity, or through global changes in enzyme abundance and localization.

A targeted amplicon panel detects the first and misses the second. A global methylation profile detects the second and may miss focal events below its detection threshold. A credible specificity package needs both.

Candidate off-target sites are predicted from sequence similarity and PAM proximity, and the biology adds constraints worth knowing:

• • dCas9 binding does not require cutting, so off-target binding is not limited by nuclease activity constraints. Occupancy-based off-target effects can occur at sites with no detectable editing in a nuclease experiment.
• • Effector recruitment amplifies weak binding. A site bound at low frequency may accumulate a mark over time, which is why marks can appear at sites that show no functional consequence.
• • Predicted sites are a floor, not a ceiling. Unbiased detection (below) is the only way to catch unanticipated sites.

Practical candidate set: the top predicted sites by sequence similarity, any site within the same locus, and any site within a gene whose expression you measure.

Three mechanisms produce changes that no guide prediction captures:

• • Spreading. Repressive effectors can modify chromatin over roughly a kilobase from the recruitment site; methylation-based editors can extend further. The consequence: a guide placed at one element may partially silence its neighbour, which is a confound in element-resolution studies.
• • Untethered activity. Some effector configurations show activity independent of dCas9 binding, particularly with high effector expression; the mark changes without recruitment.
• • Global shifts. Overexpression of a chromatin enzyme can alter the mark globally, changing the baseline against which your locus effect is measured.

Detection requires genome-wide or region-wide data: methylation by WGBS or EM-seq, histone marks by ChIP-seq or CUT&RUN with spike-in normalization.

A defensible minimum for a publication-grade claim: targeted per-CpG data at the intended site + a global methylation profile + expression of neighbouring genes. Add mark-level ChIP/CUT&RUN when the effector writes a histone modification rather than methylation.

Reporting "no off-target effects were observed" is not a specificity claim unless it comes with a detection threshold.

What to report:

• • Number of candidate sites examined and how they were chosen.
• • Detection limit of the assay at the depth used — for WGBS, the minimum methylation difference detectable at your coverage.
• • Focal changes, with magnitude and position, including in the target region itself.
• • Global summary statistics (mean CpG methylation, mark density) compared with matched controls.
• • Functional correlates where available: expression of the nearest genes to any changed region.

A useful convention: state the specificity claim as a bounded statement — "no change exceeding 15 percentage points was detected at any of 24 predicted sites, and global methylation differed by less than 2% from control".

• • Non-targeting guide with the same effector. Separates effector biology from guide-directed effects.
• • Catalytically dead effector with the same guide. Separates recruitment from catalysis.
• • Two guides at the target. Distinguishes guide-specific artifacts from locus effects.
• • Matched clones from the same campaign. Controls for clonal and selection effects.
• • Untreated parental cells. Baseline for global comparisons.

The first two are the ones that convert a descriptive survey into an interpretable result, because they define what "background" means for that effector.

Three situations deserve a change of claim rather than a supplementary figure:

• • A methylation change at a neighbouring promoter that could explain the expression phenotype. If the neighbour regulates the same pathway, the mechanism is no longer locus-specific.
• • Global hypomethylation or hypermethylation relative to control, which weakens any claim that a specific locus drives the phenotype.
• • Mark spreading across an element boundary, which matters most in element-resolution studies where the neighbour is a distinct regulatory unit.

In each case, the honest response is to narrow the claim: from "this element's methylation controls the gene" to "the edited region, including adjacent CpGs, is associated with the change".

EDITGENE delivers epigenome-editing lines with the specificity package built in: per-CpG target mapping, predicted-site panels, global methylation or mark profiling, matched non-targeting and catalytically dead controls, and neighbour-gene expression.

Tell us the locus and the claim. Contact EDITGENE to request a project consultation or quote.

Off-target rates and spreading distances vary by effector, expression level, cell type, and locus; figures here describe detection logic rather than expected rates.

A specificity claim is bounded by the assay's detection limit and by the sites examined, and should be reported with both.

• • Tycko J, et al. Mitigation of off-target toxicity in CRISPR-Cas9 screens for essential non-coding elements. Nature Communications. 2019;10:4063.
• • Galonska C, et al. Genome-wide tracking of dCas9-methyltransferase footprints. Nature Communications. 2018;9:597.
• • Lin L, et al. Genome-wide determination of on-target and off-target characteristics for RNA-guided DNA methylation by dCas9 methyltransferases. GigaScience. 2018;7:giy011.
• • Nuñez JK, et al. Genome-wide programmable transcriptional memory by CRISPR-based epigenome editing. Cell. 2021;184:2503–2519.
• • Pulecio J, et al. CRISPR/Cas9-based engineering of the epigenome. Cell Stem Cell. 2017;21:431–447.