Heritable transgene-free genome editing in plants by grafting of wild-type shoots to transgenic donor rootstocks

CRISPR-Cas9 enables targeted genome editing through guide RNAs (gRNAs) that program Cas9 to create double-strand breaks at specific genomic loci. In plants, generating transgene-free edited lines typically requires transient delivery of Cas9/gRNA or outcrossing to eliminate transgenes, both of which are time-consuming, technically challenging, or not feasible for many species. Existing approaches such as protoplast/embryo delivery and regeneration demand specialized equipment and protocols, and many crop species are recalcitrant to transformation or have long generation times. The authors aimed to develop a method to deliver Cas9 and gRNAs from a transgenic rootstock to a non-transgenic scion via grafting, leveraging RNA mobility conferred by tRNA-like sequence (TLS) motifs known to mediate long-distance RNA transport over graft junctions or into parasitic plants. They hypothesized that fusing TLS motifs to Cas9 mRNA and gRNAs would promote root-to-shoot transport into wild-type scions, enabling genome editing in scion tissues and heritable edits in seeds without the presence of transgenes in the progeny.

The study builds on foundational CRISPR-Cas9 work in plants and recognized challenges in producing transgene-free edits. Prior methods include Agrobacterium-mediated transformation requiring segregation to remove transgenes, plant-virus-mediated delivery of editing components, and direct delivery of Cas9–gRNA ribonucleoproteins into protoplasts or embryos followed by regeneration—approaches that are time-consuming, costly, or limited in species scope and may induce genome instability. RNA mobility in plants has been documented, including long-distance transport of endogenous mRNAs and mobile RNA motifs (e.g., tRNA-related sequences and FT-derived motifs) that can license transport of heterologous RNAs. TLS motifs have previously been shown to promote systemic mRNA transport and to deliver transcripts affecting meiosis (e.g., DDMCI variants) into reproductive cells across grafts. Viral vectors with mobile sgRNAs have enabled heritable editing in some species. The present work extends these concepts by creating mobile, non-viral Cas9 and gRNA TLS fusions to achieve heritable, transgene-free editing via grafting.

Design and constructs: The authors engineered Cas9 and gRNA transcripts fused to RNA mobility motifs. Two TLS variants were used: TLS1 (tRNA^Met) and TLS2 (tRNA^Met-ADT lacking D and T loops). Cas9 (zCas9, ~4.2 kb, plant codon-optimized, NLS, C-terminal FLAG) was placed under an estradiol-inducible promoter (pMDC7 vector with rbcs-e9t terminator). Two gRNAs targeting Arabidopsis NIA1 (AT1G77760) were driven by U6-26 and U6-29 Pol III promoters to generate a ~1 kb genomic deletion; gRNAs were built with or without TLS fusions and a short poly-A tail. Alternative targets included two gRNAs (gVenus1/2) designed to delete the H2B-Venus::35S terminator region in a 35S promoter::H2B-Venus::35S terminator::BastaR transgene cassette. Co-fold RNA structure predictions indicated TLS fusions preserved proper folding of both TLS and gRNA/Cas9 sequences. Transgenesis and plant material: Constructs were introduced into Arabidopsis thaliana Col-0 via Agrobacterium (AGL1) floral dip. Lines expressing gNIA1/gNIA1-TLS1/TLS2 or gVenus-TLS1/TLS2 were crossed to lines expressing estradiol-inducible Cas9, Cas9-TLS1, or Cas9-TLS2, respectively. Segregation and antibiotic selection ensured stable lines without visible growth phenotypes. Grafting protocols: Hypocotyl grafting was performed—Arabidopsis-to-Arabidopsis homografts and Arabidopsis rootstocks to Brassica rapa scions for heterografts. Arabidopsis seedlings (6–7 DAG) were grafted using silicone micro-tubes; plants grown on 0.5× MS, with 5 µM estradiol induction post-grafting, and adventitious roots on scions were removed daily. Adult grafts were transferred to soil or carried to flowering on estradiol. For B. rapa heterografts, Arabidopsis rootstocks (2 weeks) were grafted with 1-week-old B. rapa scions under long-day conditions; estradiol was applied after graft take; plants were grown to flowering on MS in jars. Mobility assays: Three weeks after grafting (juvenile) and at flowering (adult), roots and scion tissues (rosette, cauline leaves, stem, flowers, siliques) were sampled. RT-PCR (45–50 cycles) detected Cas9, gRNA, and controls (kanamycin, hygromycin; negative mobility controls). RT-qPCR quantified Cas9-TLS transcript levels (UBQ10 reference; 2^−ΔΔCt) to estimate root-to-shoot delivery ratios. Editing detection: Visual phenotyping for nia1 (chlorosis on NH4+-deficient medium) in scions indicated functional editing. Genomic PCR detected NIA1 deletions (wild-type 1,469 bp; edited ~430 bp) and Venus deletions (wild-type 1,719 bp; edited ~250 bp), with enrichment by restriction digest (HindIII for NIA1; PstI for Venus) to suppress wild-type amplicons. Amplicons were cloned and Sanger sequenced (20 clones per graft condition) to confirm edits and characterize junctions. Heritability and frequency estimation: Seeds from wild-type scions grafted on Cas9/gRNA rootstocks were screened. For NIA1, progeny were grown on NH4+-limited medium for phenotyping and pooled genomic PCR screening (~70–100 seedlings per pool). For Venus, pooled genomic PCR used ~40 seedlings per pool and confocal microscopy assessed Venus fluorescence loss in subsets. Editing frequencies were calculated as minimum estimates per 1,000 seedlings based on pooled PCR positivity. Transgene absence (Cas9, Kan) in progeny was verified by RT-qPCR/RT-PCR. Cross-species validation: Arabidopsis rootstocks expressing Cas9-TLS2 × gNIA1-TLS2 were grafted to B. rapa scions. RT-PCR/RT-qPCR evaluated TLS-mediated movement; genomic PCR and Sanger sequencing assessed B. rapa NIA1 edits in siliques and flowers. Graft success and integrity were monitored over 10–40 days.

- TLS-dependent RNA mobility: Without TLS, neither Cas9 mRNA nor gRNA moved from rootstocks to wild-type scions; with TLS1 or TLS2, both Cas9-TLS and gRNA-TLS were detected in scion tissues (juvenile and adult) by RT-PCR/RT-qPCR. Antibiotic resistance transcripts (Kan, Hyg) were absent in scions, ruling out contamination and indicating specificity. - Functional editing in scions: nia1 chlorotic phenotype appeared in grafted scion leaves on NH4+-deficient medium in 20/28 (Cas9-TLS1 × gNIA1-TLS1) and 26/30 (Cas9-TLS2 × gNIA1-TLS2) plants; 0/20 in non-TLS controls. Genomic PCR detected NIA1 deletions (~430 bp) in all tested TLS scions and in rootstocks; Sanger sequencing confirmed precise deletions between the two gRNA sites. - Adult tissue mobility and activity: In adult Arabidopsis scions, Cas9-TLS transcripts were detected by RT-qPCR in siliques, flowers, stems, cauline and rosette leaves, with an estimated delivery ratio of ~0.09% (≈1/1,000 root-produced transcripts) to scions; both TLS1 and TLS2 behaved similarly. NIA1 deletions were confirmed in siliques and flowers. - Heritable edits (Arabidopsis NIA1): Offspring from wild-type scions grafted on TLS rootstocks showed heritable edits in 11/15 grafts (TLS1) and 17/22 grafts (TLS2); 0/11 in non-TLS controls. Minimal editing frequencies: 5.7 edits/1,000 seedlings (TLS1) and 5.0 edits/1,000 (TLS2). Homozygous nia1 phenotypes were observed at ~1.17 (TLS1) and ~1.41 (TLS2) per 1,000 seedlings. - Alternative target (Venus transgene): Editing detected in progeny of 19/20 (TLS1) and 17/18 (TLS2) grafts; 0/19 in controls. Minimal editing frequencies: 14.9 (TLS1) and 15.9 (TLS2) edits/1,000 seedlings. Loss of Venus fluorescence indicating homozygous deletion occurred in 7/1,557 seedlings (~0.45%). Planned BASTA resistance was not achieved because gRNA1 cut upstream in the 35S promoter, yielding a truncated promoter and no BastaR expression. - Cross-species mobility and editing (Arabidopsis→Brassica rapa): Cas9-TLS2 and gNIA1-TLS2 transcripts were detected in B. rapa scion siliques, flowers, stems, and leaves; none detected without TLS. RT-qPCR indicated a higher scion/root delivery ratio in heterografts (~0.468%, ≈1/250 transcripts) compared to Arabidopsis homografts (~0.091%). NIA1 deletions were found in 4/6 siliques and 4/6 flowers of B. rapa scions; controls showed no edits. - Scale and novelty: The ~4.2 kb Cas9-TLS represents one of the largest mobile mRNAs reported in Arabidopsis. The approach produced transgene-free edited seeds in one generation without tissue culture, segregation, or viral vectors.

The study demonstrates that fusing TLS motifs to Cas9 mRNA and gRNAs enables their long-distance movement from transgenic rootstocks into non-transgenic scions, where they are translated and enact genome editing. This directly addresses the bottleneck of removing transgenes after editing by producing edited, transgene-free progeny in a single generation via grafting. The observed mobility across diverse scion tissues, including reproductive organs, supports delivery to germline progenitors, consistent with prior observations of graft-transmissible RNAs and siRNAs to meiotic precursors. The efficacy across species boundaries (Arabidopsis-to-Brassica rapa) highlights the potential for using easily transformable donor rootstocks to edit graft-compatible, less-transformable crops. Quantitatively, the delivery efficiency (~0.09% in Arabidopsis and ~0.47% in B. rapa) was sufficient to achieve detectable somatic and heritable edits, with minimal per-seed editing frequencies ranging from ~0.5–1.6% for pooled heterozygous events (depending on target and estimation method) and ~0.1–0.45% for homozygous edits, acknowledging underestimation due to pooled sampling and reliance on dual-cut deletions. Compared to protoplast-based RNP delivery or viral editing, the grafting approach avoids tissue culture regeneration, outcrossing, or viral vectors, while offering similar overall timelines and the ability to rapidly screen large pools by multiplex PCR. Mechanistically, while TLS-driven mobility pathways require further elucidation, the results suggest robust, systemic delivery sufficient for germline editing, and indicate that mRNA size is not an insurmountable barrier. The method also enables creation of useful chimeric mosaics to analyze otherwise lethal mutations.

The authors present a graft-based, TLS-mediated RNA mobility system that delivers Cas9 and gRNAs from transgenic rootstocks to non-transgenic scions, producing heritable, transgene-free genome edits in Arabidopsis and Brassica rapa within a single generation. The approach is non-viral, bypasses tissue culture and transgene segregation, and uses grafting—a widely practiced agricultural technique—suggesting broad applicability across graft-compatible crops. Future directions include optimizing gRNA design for diverse species and targets, exploring additional or multiple RNA mobility motifs to enhance delivery, extending to monocot crops where grafting techniques are emerging, and applying the method to multiplex editing and trait stacking in breeding programs. Establishing donor rootstock libraries tailored to conserved targets across crop families could further expand the platform’s utility.

- Editing frequency, while practical, is modest (approximately 5–16 edits per 1,000 seedlings for pooled detection; ~0.1–0.45% for homozygotes depending on target), and estimates likely undercount single-site edits due to reliance on dual-cut deletions and pooled assays. - Requires generation and maintenance of transgenic donor rootstocks expressing Cas9-TLS and gRNA-TLS constructs, adding an initial time and resource investment. - Grafting compatibility and efficiency vary across species; cross-family grafts may be challenging, and mobility efficiency may differ. - gRNA design is critical; unexpected cut sites (e.g., gVenus1 in the 35S promoter) can lead to unintended outcomes and failure to activate selectable markers. - Mosaicism in scion tissues is inherent; while sometimes advantageous for studying lethal mutations, it can complicate phenotyping and selection. - Mechanisms of TLS-mediated mobility and factors influencing delivery efficiency are not fully understood, and transcript abundance in specific tissues (e.g., siliques) may not correlate linearly with editing outcomes. - Target sequence conservation across donor and recipient species must be sufficient; distant species may require species-specific gRNAs.