Free-Base Editing Delivers Cleaner DNA Changes in Crops

A new strategy for precision genome editing replaces the usual single-protein base editor with two freely expressed components: a CRISPR-Cas9 nickase and the cytidine deaminase CDA1. Rather than permanently joining these proteins, the system—dubbed Free-BE—allows the deaminase to encounter DNA exposed by the nickase and convert selected cytosines into uracil, which DNA repair can then convert into thymine. The modular design was tested in the moss Physcomitrium patens, in transiently transformed maize protoplasts and in stably transformed wheat. It produced fewer unwanted insertions and deletions, fewer simultaneous edits within a target window and fewer changes outside the predicted activity window than a conventional fused editor, although it did so at a substantial cost to editing efficiency. The results suggest that separating the editing machinery may offer a useful way to control unwanted DNA changes, particularly in the development of crop varieties with precise single-letter mutations.

Most plant cytidine base editors are built as one continuous protein. A catalytically weakened version of Cas9, known as SpCas9 nickase, is fused to a cytidine deaminase enzyme. This arrangement keeps the enzyme close to the single-stranded DNA exposed when the guide RNA directs the nickase to its target. The strategy is generally powerful, but prolonged contact can allow the deaminase to modify several nearby bases, create unwanted DNA-repair outcomes or contribute to small insertions and deletions, or indels. Testing a new deaminase, linker or Cas9 variant can also require rebuilding and reconstituting the entire fusion protein, a laborious process in plants because transformation is slow.

Free-BE takes a deliberately simple approach to this design problem. The nickase and CDA1 are encoded on separate plasmids and expressed independently, without an intein, engineered protein interaction or RNA-based recruitment system. The central hypothesis was that the two proteins could still work together if they encountered the same target temporarily. If the deaminase were not tethered continuously to the nickase, it might spend less time exposed to the target DNA. That shorter exposure could preserve useful C-to-T conversions while reducing collateral editing. Researchers compared this modular configuration, called Fused-BE, with a conventional nCas9–CDA1 fusion in the moss APT reporter system, where mutations that inactivate the APT gene confer resistance to the toxic adenine analogue 2-fluoroadenine.

In moss, both systems edited APT, but the fused version was clearly more productive. With one guide RNA, Fused-BE inactivated APT in 7.1% of regenerating plants, compared with 2.9% for Free-BE; with the second guide, the rates were 5.6% and 1.9%, respectively. Expression of the nickase alone produced negligible mutation frequencies of 0.004% and 0.008%, showing that most of the activity attributed to Free-BE came from its deaminase component. The lower yield was therefore a real limitation, but it did not indicate that independent expression prevents the enzyme from functioning at a guide-defined site.

The quality of the resulting edits differed as strikingly as their frequency. With the first guide, 52.8% of the mutations produced by Fused-BE contained indels, versus 7.1% with Free-BE—a roughly sevenfold reduction that was not merely a consequence of the system’s lower overall efficiency. Free-BE also produced a higher proportion of plants carrying only one edited cytidine: 55% with the first guide and 82% with the second, compared with 6% and 49% for the fused editor. The two architectures did not generate identical molecular outcomes, however. Free-BE produced fewer canonical C-to-T transitions and relatively more transversions, especially C-to-G, suggesting that changes in the timing and proximity of the deaminase may influence how the resulting U:G mismatch is repaired.

Targeted safety checks did not reveal unintended mutations in the candidate off-target sites examined in moss. The researchers identified four potential off-target loci for one guide using a sequence-similarity score and sequenced them in plants edited with either system. No mutations were detected at these sites in the tested lines. That finding is encouraging but limited: the analysis covered only a few predicted genomic locations for a single guide and did not measure genome-wide off-target activity, RNA editing or off-target effects in maize and wheat. The result therefore supports the need for broader specificity studies rather than establishing genome-wide specificity for Free-BE.

Free-BE then moved into a crop-based reporter test in maize. Researchers engineered a nonfunctional blue fluorescent protein so that one C-to-T edit would restore a functional green fluorescent protein. Maize protoplasts receiving the fused editor converted 1.5% of reporter molecules, while those receiving Free-BE achieved 0.6% conversion. Negative controls produced almost no GFP-positive cells. The experiment demonstrated that unlinked components can generate targeted base editing in maize cells, but because the assay measured correction of a transient reporter rather than editing of an endogenous maize gene, stable maize experiments are still needed to establish how the system performs in the crop genome.

Wheat provided the most demanding validation, involving Agrobacterium-mediated transformation of the Fielder cultivar and two guide RNAs directed at TaGASR7 and TaLOX2. Each guide perfectly matched three homoeologous copies, giving six possible endogenous targets for multiplex editing. Of 291 transformed T0 plantlets carrying Fused-BE, 235 contained an edited cytidine at one or more targets, an editing rate of 80.8%. Free-BE edited at least one target in 33 of 77 transformants, or 42.9%. Both systems could multiplex several loci: 9.1% of edited Free-BE plants carried changes in five of the six target genes, while 11.5% of edited Fused-BE plants showed changes in all six. The conventional editor therefore remained more efficient, but the modular system retained considerable activity and generated combinations of edits across multiple wheat homoeologs.

The wheat sequence data also reproduced Free-BE’s cleaner editing profile. At the TaLOX2 target, which contains seven cytidines in the guide’s spacer, 50% of Free-BE-edited plants carried one edited cytidine, compared with 21.7% for Fused-BE. Dual and triple edits were correspondingly less frequent with Free-BE, occurring in 10.0% and 3.8% of plants rather than 22.6% and 12.6%. Fused-BE also generated occasional edits near the guide’s PAM-proximal edge and outside the spacer region, whereas Free-BE did not show those events in the analyzed wheat plants. Unlike in moss, Free-BE did not consistently increase C-to-G or C-to-A transversions in wheat, indicating that the effect of separating the editor components may depend on the plant species, target sequence and biological context.

Overall, the study presents Free-BE as a precision-first alternative to the speed and efficiency of fused base editing. Its principal advantages are fewer indels, a greater likelihood of isolated single-cytosine changes, no detected edits beyond the predicted window in the wheat dataset, and a modular design that allows deaminases, Cas9 variants and expression strategies to be tested without repeatedly constructing a complete fusion. Its disadvantages are equally clear: editing frequencies were consistently lower, and the spectrum of nucleotide substitutions could be less predictable in moss. Optimizing the expression levels and timing of the two components may improve performance, while stable maize tests and comprehensive DNA and RNA off-target analyses will be needed to define the system’s full utility. For precision breeding, the results point to a promising bargain: a lower chance of producing the desired allele may be acceptable if it comes with a lower chance of altering the genetic neighborhood around it.

Subject of Research: Evaluation of a modular cytidine base-editing strategy in moss, maize, and wheat

Article Title: Validation of a Modular Base Editing Strategy, Free‐BE, in Physcomitrium patens, Maize, and Wheat

Article References: Guyon‐Debast, A., Javelle, M., Perroud, P.-F., Pichon, J.-P., Paul, W., & Nogué, F. (2026). Validation of a Modular Base Editing Strategy, Free‐BE, in Physcomitrium patens , Maize, and Wheat. Plant Direct, 10(10), Article e70196. https://doi.org/10.1002/pld3.70196

Image Credits: AI Generated

DOI: 10.1002/pld3.70196

Keywords: base editing, CRISPR, genome editing, cytidine deaminase, wheat, maize, moss, plant breeding, DNA repair, precision breeding, Validation, Modular