Efficient plant genome engineering using a probiotic sourced CRISPR-Cas9 system

CRISPR-Cas systems are adaptive immune mechanisms in bacteria and archaea that have been adapted for genome engineering across organisms, including plants. SpCas9 is the dominant Cas9 used in plants for diverse applications such as base editing, prime editing, transcriptional regulation, and epigenetic editing. However, SpCas9’s canonical 5'-NGG-3' PAM restricts target scope. Two strategies have addressed this: (1) discovery of Cas9 orthologs with alternative PAM requirements (e.g., SaCas9, St1Cas9, NmeCas9, Nme2Cas9, CjCas9, BlatCas9; or orthologs with shorter PAMs like ScCas9 [5'-NNG-3'] and FrCas9 [5'-NNTA-3']), though many have complex PAMs and limited activity in plants; (2) engineering SpCas9 variants with relaxed PAMs (xCas9, SpCas9-NG, SpRY), which expand target scope but often reduce editing efficiency in plants, sometimes due to self-cleavage effects. Cas12a provides an alternative with 5'-TTTV-3' PAM, multiplexing advantages, and promoter editing utility, but suffers from temperature sensitivity and lacks nickase forms, limiting some precise editing tools. Consequently, a need remains for a plant-optimized Cas9 with alternative, A/T-rich PAM preference, robust activity at lower temperatures, and high specificity. This study reports a type II-A Cas9 from probiotic Lactobacillus rhamnosus (LrCas9) recognizing a 5'-NGAAA-3' PAM, enabling efficient and versatile plant genome engineering.

Prior plant genome editing predominantly utilized SpCas9, which recognizes 5'-NGG-3' PAM, facilitating gene knockouts and precise tools (base and prime editors, CRISPRi/a). To overcome PAM constraints, researchers explored Cas9 orthologs (SaCas9 [5'-NNGRRT-3'], St1Cas9, NmeCas9/Nme2Cas9, CjCas9, BlatCas9), though many require longer or complex PAMs (6–8 bp) and show limited activity in plants. Orthologs with shorter or alternative PAMs (ScCas9 [5'-NNG-3'], FrCas9 [5'-NNTA-3']) were reported; ScCas9 worked in rice but with low efficiency. Engineered SpCas9 variants (xCas9 and SpCas9-NG for 5'-NG-3'; SpRY for near-PAMless 5'-NR/YN-3') expand targetability but often at the cost of lower efficiency and potential self-targeting in plants. CRISPR-Cas12a (LbCas12a) targets 5'-TTTV-3' PAM, excels in multiplexing and promoter editing, but is temperature sensitive and historically lacked nickase versions, impeding base/prime editing (recent dCas12a base editors were reported; potent Cas12a CRISPRa remains limited). Overall, SpCas9 remains the most versatile but with PAM limitations, motivating discovery of plant-suitable Cas9 orthologs with alternative, especially A/T-rich, PAMs and high activity.

• In silico discovery: Analyzed 33,825 proteomes (32,023 bacteria; 1,832 archaea) using HMMER with TIGRFAMs and Pfam seed models to identify 257,745 putative Cas proteins and 30,495 CRISPR clusters (29,586 bacterial; 909 archaeal). Identified 4,963 clusters containing Type II Cas9. Predicted crRNA/tracrRNA features; typical CRISPR loci ~5–10 kb (mostly ~7 kb); Cas9 proteins ~1,000–1,400 aa (mostly ~1,400 aa); tracrRNA 100–500 nt; crRNA ~36 nt. Mapped anti-spacers in phage/viral genomes via BLAST, extracted flanks to infer PAMs (WebLogo). After de-redundancy and scoring, selected 42 CRISPR-Cas9 candidates (mostly Type II-A, some II-C). Aligned crRNAs revealed conserved 5'-GUUUU-3' and 3'-AAAAC-3' motifs.
• Candidate selection and initial testing: Selected four Cas9s for empirical testing, including BAI42646.1 from Lactobacillus rhamnosus GG (LrCas9). Built all-in-one vectors expressing crRNA, tracrRNA, and Cas9; transformed rice protoplasts. LrCas9 edited three sites with 4.3–64.2% mutation rates; others were inactive.
• PAM validation: Performed an E. coli double-antibiotic PAM depletion assay (LrCas9 expressed under J23100; crRNA/tracrRNA under J23119; selection on Kan/Amp) showing strong depletion at 5'-NGAAA-3' PAM, consistent with prediction.
• sgRNA engineering: Combined crRNA/tracrRNA into sgRNA scaffolds (V1.0–V3.0 with varying tetraloops). In rice protoplasts, V2.0 and V3.0 gave ~20–25% editing at two sites; optimized spacer lengths at Os-AG04 across 14–22 nt.
• Activity profiling: Tested 19 endogenous rice targets; LrCas9 cuts 3 bp upstream of PAM with predominant 1 bp deletions; insertions generally exceeded deletions. Temperature assays indicated robust activity at 28–32 °C, reduced at 22 °C.
• Benchmarking: Compared LrCas9 to LbCas12a variants (LbCas12a, ttLbCas12a, LbCas12a-RRV) at four rice loci; and to SpCas9-NG and SpRY at seven 5'-NGA-3' sites using rice protoplast assays.
• Cross-species testing: Assessed LrCas9 editing in wheat, tomato, and Larix protoplasts across 7, 7, and 6 targets respectively.
• Stable rice transformation: Generated T0 lines with single sgRNAs targeting OsPDS, OSDEP1, OSBADH2, Os03g0568400, Os03g0603100; assessed editing by SSCP and Sanger sequencing. Built tRNA-based multiplex sgRNA arrays to edit OsPDS and OSDEP1; created dual-OsPDS constructs (Os-TG01/Os-TG02) to induce large deletions.
• Promoter engineering: Designed multiplex sgRNAs to delete ~2.1 kb (KR-A to KR-F) in OsWx promoter; screened T0 lines by PCR and Sanger; obtained homozygous deletion; measured OsWx expression (qPCR), amylose and total starch content, and seed phenotypes (iodine staining, SEM).
• Specificity assessment: (1) Systematic single-nucleotide mismatches across a 20-nt spacer; measured editing via amplicon deep sequencing. (2) GUIDE-seq in rice protoplasts with 10 or 100 pmol dsODN GUIDE; NGS to quantify on- and off-target insertions; analyzed top sites. (3) Validated predicted off-targets in edited T0 plants (Os-AG04, Os-TG02) by Sanger sequencing.
• Base editor construction: Built LrCas9-D13A nickase fusions with PmCDA1-UGI (CBE) for rice and wheat; tested at multiple loci in protoplasts and in stable rice T0 lines. Constructed ABE V1.0 (ecTadA-7.10) and ABE V2.0 (ecTadA-8e) fusions with LrCas9-D13A; tested editing in rice.
• CRISPRi/CRISPRa systems: Built dLrCas9 (D13A, H858A) fused to KRAB-SRDX (dLrCas9-KS) for repression; targeted multiple genes around TSS. Built dLrCas9-TV12 (dLrCas9-TV) and a nuclease-active LrCas9-TV system using truncated (14-nt) spacers to abolish cleavage; measured activation by qPCR.
• Data analysis: Deep sequencing via Novaseq; mutation analysis with CRISPRMatch; off-target nomination with GUIDE-seq scripts and CRISPR-GE; statistics with GraphPad Prism.

• Discovery and PAM: Identified LrCas9 (Type II-A) from Lactobacillus rhamnosus GG recognizing a unique A/T-rich 5'-NGAAA-3' PAM, validated by E. coli depletion assay.
• Editing efficiency: In rice protoplasts across 19 sites, LrCas9 showed robust activity, cutting 3 bp upstream of PAM with mainly 1 bp deletions and overall higher insertion frequency than deletion.
• Temperature: LrCas9 maintained comparable activity at 28–32 °C and reduced at 22 °C, suggesting lower temperature sensitivity than SpCas9.
• Benchmarking: At four rice loci, LrCas9 performed similarly to or better than LbCas12a variants, significantly outperforming LbCas12a at three sites; overall potency comparable to ttLbCas12a and higher than LbCas12a. Against SpCas9-NG and SpRY at seven 5'-NGA-3' sites, LrCas9 had significantly higher editing at 5/7 sites; two sites were refractory to all three.
• Cross-species robustness: Demonstrated editing at all tested targets in wheat, tomato, and Larix protoplasts, averaging ~5% mutation rates, comparable to SpCas9 benchmarks for these systems.
• Stable rice editing: Single sgRNAs yielded T0 editing efficiencies of 16.7–85% with 10–40% biallelic edits. Multiplexing OsPDS/OSDEP1 showed T0 editing of 30.4% (OsPDS; 13% biallelic) and 52.2% (OSDEP1; 47.8% biallelic). Dual OsPDS targeting achieved 85.2% (81.5% biallelic) at Os-TG01 and 70.4% (29.6% biallelic) at Os-TG02, with 14.8% large deletions recovered.
• Promoter engineering (OsWx): Designed ~2.1 kb promoter deletion encompassing KR-A to KR-F; obtained a homozygous deletion line (232-6). OsWx expression reduced to ~40% of wild type; amylose content reduced by 39.3%, with only 8.1% reduction in total starch. Seeds showed distinct iodine staining and altered starch granule morphology (smaller granules, occasional holes) by SEM.
• Specificity: Single-nucleotide mismatches across the spacer largely abolished LrCas9 editing, including at 5' positions. GUIDE-seq showed 87.8–98.8% GUIDE integration at on-target sites across four spacers; captured off-target sites had ≥6 mismatches and were not validated by Sanger; no off-target edits detected at GUIDE-seq/CRISPR-GE nominated sites in T0 plants, indicating high specificity.
• Base editors: LrCas9-CBE (PmCDA1-UGI) achieved up to ~35% C-to-T editing in rice protoplasts and up to ~10% in wheat, with preferred editing toward the 5' spacer region; minimal indels. In stable rice, CBE yielded 90.5% (OsDEP1) and 57.1% (OsPDS) mutation rates with successful base conversions. LrCas9-ABE V2.0 (ecTadA-8e) produced A-to-G edits in T0 rice, while ABE V1.0 (ecTadA-7.10) did not.
• CRISPRi/a: dLrCas9-KS repressed target gene expression to ~20% or lower in rice protoplasts with sgRNAs near or around TSS. dLrCas9-TV activated targets 2–3-fold. A nuclease-active LrCas9-TV with 14-nt spacers (abolishing cleavage) achieved strong activation: 19-fold (OsmiR528) and 150-fold (OsWx), representing ~11x and ~60x improvements over dLrCas9-TV.
• Additional: LrCas9 exhibited variable activity at non-canonical 5'-NGAA-3' PAMs. A preliminary LrCas9-based PE2 prime editor produced detectable but low (~0.1%) editing in rice protoplasts.

The study addresses limitations of existing plant genome editing platforms by introducing LrCas9, a probiotic-derived Cas9 with an A/T-rich 5'-NGAAA-3' PAM. This PAM complements SpCas9’s 5'-NGG-3' and overlaps with Cas12a’s 5'-TTTV-3' (reverse complement), enhancing access to promoter-rich, A/T-biased regions. LrCas9 demonstrated robust nuclease activity across monocots (rice, wheat) and dicots (tomato), and even conifer cells (Larix), with tolerance to standard plant temperatures, making it broadly applicable. When matched on the same targets, LrCas9 outperformed or matched leading alternatives (LbCas12a variants, SpCas9-NG, SpRY), highlighting its potency. High specificity was supported by mismatch intolerance assays, GUIDE-seq showing predominant on-target integration, and absence of validated off-target edits in plants. Beyond knockouts, LrCas9 enabled efficient promoter deletions to modulate quantitative traits (e.g., reducing OsWx expression and amylose content with minimal impact on total starch), and supported potent CRISPRi and enhanced CRISPRa using a nuclease-active activation strategy with truncated spacers. Development of LrCas9-derived CBEs and ABEs expands precise editing to A/T-rich PAM contexts, facilitating allele-specific and regulatory edits not readily accessible to SpCas9. The probiotic origin may also assist public acceptance of genome-edited crops. Collectively, LrCas9 enriches the plant genome engineering toolbox with complementary targeting space, high efficacy, and precision.

This work establishes LrCas9 from Lactobacillus rhamnosus GG as a highly efficient and specific Cas9 for plant genome engineering with a unique 5'-NGAAA-3' PAM. LrCas9 outperforms or matches LbCas12a variants and engineered SpCas9 variants on shared targets, functions across diverse plant species, and supports multiplexed genome editing, large promoter deletions for trait modulation, and highly effective CRISPRi/a. LrCas9-enabled CBE and ABE expand base editing into A/T-rich contexts. Future directions include: improving LrCas9-based prime editing (e.g., optimizing sgRNA scaffolds and pegRNA designs), engineering LrCas9 protein and guide RNAs to broaden PAM compatibility (including non-canonical NGAA variants), integrating simultaneous editing and transcriptional activation for complex trait engineering, and validating performance across additional crops and field-relevant conditions.

• PAM complexity: The 5'-NGAAA-3' PAM is more restrictive than 5'-NGG-3' (SpCas9), potentially reducing the number of candidate sites per gene, though complementary to existing tools.
• Temperature sensitivity: While less sensitive than SpCas9 in some assays, LrCas9 activity decreased at 22 °C, indicating residual temperature effects.
• Prime editing: Initial LrCas9 PE2 showed low efficiency (~0.1%) in rice protoplasts, possibly due to the longer sgRNA scaffold (139 nt) and secondary structure constraints; further optimization is required.
• Non-canonical PAMs: Variable performance at shorter/non-canonical 5'-NGAA-3' PAMs suggests room for engineering to reliably expand PAM scope.
• Generalizability: Most benchmarks were performed in protoplasts and select loci; broader locus- and species-wide validation, as well as field conditions, remain to be assessed.