← Back to Home

Where Did CRISPR Make Edits, and How Did Cells Respond? Superb-seq Connects Editing with Gene Activity

Identifying off-target DNA changes alone is not enough to assess their consequences. A new method from the Salk Institute links editing sites with gene expression in individual cells, adding a layer of functional evidence to gene-editing research, though it remains some distance from determining clinical safety.

By SURL BioNews

The precision of molecular scissors cannot be judged solely by whether they cut the intended target. If they also affect other DNA sites, researchers must ask further questions: Which cells are affected? Does gene activity within those cells change as well? Superb-seq, developed by the Salk Institute in the United States, aims to connect the traces left by editing with molecular responses in the same cell.

The study was published in *Nature Methods* on November 16, 2026. The team integrated CRISPR–Cas9 editing, short DNA sequences used to mark editing sites, and single-cell RNA sequencing. The markers produce RNA that reports editing locations and is read alongside the cell’s own RNA, allowing researchers to compare where DNA was altered with which genes showed differences in expression levels.

In a group of approximately 10,000 laboratory-grown human K562 leukemia cells, the team used seven guide RNAs to edit four genes involved in chromatin regulation. Superb-seq identified 5,761 cells carrying edits and detected 43 editing sites, 36 of which were outside the intended targets. These figures describe specific cells and experimental conditions and cannot be directly treated as off-target rates for other CRISPR tools or therapies.

The methodological key is to distinguish “what instructions were delivered” from “what edits actually occurred.” Perturb-seq, used for comparison, reads guide RNAs and gene expression but does not directly read the final editing outcomes. By adding this information, Superb-seq improved researchers’ estimates of downstream effects on gene expression. The paper also reported that 19 off-target edits were associated with differences in gene expression, nine of which showed cell-type specificity, suggesting that the same off-target event may have different effects depending on the cellular context.

This also matters for basic research. When scientists disrupt a gene to infer its function, unintended edits may become mixed into the observed results, making phenomena originally attributed to the target gene difficult to interpret. Pairing editing sites with cellular responses helps clarify these confounding effects. In the development of gene-editing therapies, it may also supplement early safety analyses of candidate approaches.

The team also proposed that these paired data could be used to train machine-learning models to help predict how cells respond to genetic perturbations. The specific value is that models would have access to actual editing outcomes, rather than only the instructions researchers originally designed. However, this remains a proposed use; the study did not demonstrate that AI models can already improve clinical decisions. Salk also stated that Graham McVicker, Brad Balderson, and Michael Lorenzini have filed a patent application for Superb-seq.

Superb-seq remains a research method, rather than a clinically validated treatment or safety certification tool. A change in gene expression does not mean that disease has occurred, and the absence of a measurable, clear change cannot rule out long-term risks. It offers a more detailed path for investigation: first identify which cells carry edits, then examine their molecular responses. Whether these changes affect human health still requires further validation.

References

  1. Salk Institute
  2. Phys.org (provided by Salk Institute)