Rice (Oryza sativa) Japonica cultivar Nipponbare were grown on ½ MS basal salt medium (Murashige and Skoog, 1962) in the dark at 28°C for 14 days. Etiolated rice leaves were used for protoplast isolation. Suspension cells derived from embryogenic calli of Hamlin sweet orange (Citrus sinensis) line H89 were maintained in liquid MT (Murashige and Tucker) media (Murashige and Tucker, 1969) (Phytotech Catalog #M5525) and subcultured every 3 weeks.

All Cas12a variants with or without nuclear localization signals (NLSs) (Behlke et al., 2018; Vakulskas et al., 2020; Zhang et al., 2020) and crRNAs with or without end modifications were provided by IDT (Integrated DNA Technologies, Inc., United States). LbCas12a crRNA scaffold sequence was used for LbCas12a genome editing, while AsCas12a crRNA scaffold sequence was used for AsCas12a. All crRNA sequences are included in Supplementary Table S1. To assemble RNP to a final concentration of 0.1 µM in the protoplast cell culture, with a 1:5 Cas12a:crRNA ratio, 20 µg Cas12a nuclease was mixed with 667 pmol crRNA in a 1 X NEBuffer™ 3.1 solution. Reduced amount of Cas12a nuclease and crRNA with the same ratio were applied to assemble RNP to lower final concentrations. The 20 µl mixture was then incubated at room temperature for 10 min. The amount of Cas12a and crRNA used for each protoplast transfection was adjusted according to the concentration and Cas12a:crRNA ratio. The 1:5 Cas12a:crRNA ratio was used for rice protoplast transfection except when testing different ratios. The 1:2 Cas12a:crRNA ratio was used for citrus protoplast transfection to ensure enough crRNAs were provided.

T-DNA vectors were constructed as previously described (Tang et al., 2017; Zhang et al., 2019b). The crRNA targeting OsEPFL9 was synthesized as duplexed oligonucleotides, which were phosphorylated, annealed, and ligated into pYPQ141-ZmUbi-RZ-As (for AsCas12a-Ultra, Addgene #86196), and pYPQ141-ZmUbi-RZ-Lb (for LbCas12a and LbCas12a-E795L, Addgene #86197) at the BsmBI site (Supplementary Table S2) (Tang et al., 2017). Mutations were introduced to pYPQ230 (Addgene #86210) and pYPQ220 (Addgene #86208) (Tang et al., 2017) to generate LbCas12a-E795L expression vector pYPQ230-E795L (Addgene #176890) and AsCas12a Ultra expression vector pYPQ220-Ultra (Addgene #176889), respectively, and using the Q5® Site-Directed Mutagenesis Kit (New England BioLabs). These three Cas12a expression vectors (pYPQ230, pYPQ230-E795L, and pYPQ220-Ultra) were assembled with their corresponding crRNA expression vectors and the destination vector pYPQ203 (Addgene #86207) (Tang et al., 2017) using the three-way Gateway assembly (Zhang et al., 2019b).

Rice protoplasts were isolated and transfected as previously described (Lowder et al., 2015). For RNP delivery, 20 µl assembled RNP mixture (with different concentrations and/or Cas12a:crRNA ratios) was mixed with 200 µl protoplast (1 × 10⁶ cells/ml) and 220 µl PEG solution (40% (w/v) PEG 4000, 0.2 M mannitol, 0.1 M CaCl₂). For plasmid delivery, 30 µl plasmid DNA (30 µg) was mixed with 200 µl protoplast (1 × 10⁶ cells/ml) and 230 µl PEG solution. After 30 min incubation at room temperature, the transfection was terminated by adding 900 μl W5 buffer. Protoplasts were collected by centrifugation and resuspended in 1.33 ml W5 buffer. The mixtures were transferred into 12-well culture plates and incubated at 32°C (the default temperature) or 25°C (only for Figure 2) in the dark for 2 days. The protoplasts were collected and lysed for target site amplification using the Phire Plant Direct PCR Kit (Thermo Scientific™).

Suspension citrus cells derived from embryogenic calli of Hamlin 89 (H89) sweet orange were used for protoplast isolation and the following RNP delivery. To obtain suspension citrus cell culture, approximately 5 g of citrus embryogenic calli maintained on solid MT50 medium (MT (Murashige and Tucker, 1969) (Phytotech Catalog #M5525) (supplemented with 50 g/L sucrose, 50 µM 6-Benzylaminopurine and 8 g/L agar at pH 5.8) were crushed using a spatula and resuspended in 50 ml liquid MT50 media (MT50 medium without any agar). Suspended cells were maintained in liquid MT50 media in the dark at room temperature on an orbital shaker (120 rpm) and subsequently subcultured every 3 weeks. The cells were drained and digested in a filter-sterilized enzyme solution (0.6 M mannitol, 10 mM CaCl₂, 10 mM MES buffer, and 0.75% (w/v) cellulase Onozuka RS (Yakult Pharmaceutical IND. CO., Tokyo, Japan), 0.75% (w/v) Macerozyme R-10 (Yakult Pharmaceutical IND. CO., Tokyo, Japan), 0.1% BSA, and pH 5.6) in the dark at room temperature on an orbital shaker (80 rpm) for approximately 16 h. The digested protoplasts were purified and transfected using the same method as for rice protoplasts, except that the transfected protoplasts were incubated at 28°C in the dark for 2 days. For RNP delivery in citrus, LbCas12a were used and different concentrations (0.1, 0.01, and 0.001 µM) of RNP were tested. Protoplasts transfected with water were used as the mock control. The protoplasts were collected and lysed for target site amplification using the Phire Plant Direct PCR Kit (Thermo Scientific™).

To measure editing efficiencies of RNP-mediated and plasmid-mediated targeted mutagenesis, the restriction fragment length polymorphism (RFLP) method was first used. Amplicons harboring target sites were digested and ran through 2% TAE agarose gels. Amplicons with edits are expected to lose restriction enzyme cutting sites, resulting in digestion-resistant bands. Editing efficiencies were quantified using ImageJ (https://imagej.nih.gov/ij/). Editing efficiency = undigested band intensity/(undigested band intensity + digested band intensity) × 100%. To measure editing efficiencies using amplicon deep sequencing, target sites were PCR amplified using the Phire Plant Direct PCR Kit (Thermo Scientific™). Each sample was barcoded and sequenced using the Illumina sequencing platform. Demultiplexed NGS data was analyzed for genome editing with CRISPAltRations v1.0.0 using default parameters and the recommended window size (9 bp) for detecting Cas12a editing (Kurgan et al., 2020). All primers used for RFLP assay and amplicon deep sequencing are listed in Supplementary Table S2.

Pairwise comparisons were conducted using Student’s t-test. Significant differences were indicated using one asterisk (p < 0.05), two asterisks (p < 0.01) or three asterisks (p < 0.001). Multiple comparisons are conducted using Tukey’s Honest Significant Difference (HSD) test in R. Treatments with the same letter are not significantly different when α = 0.05.

To investigate whether engineered Cas12a variants can produce higher editing efficiencies, four Cas12a variants, including LbCas12a, LbCas12a-E795L, AsCas12a, and AsCas12a Ultra (Supplementary Table S3) were used to target four genes (OsPDS, OsROC5, OsmiR528, and OsEPFL9) in rice protoplasts individually. Cas12a and its corresponding crRNAs were delivered as RNPs. Genome editing efficiencies were evaluated using the RFLP assay (Figure 1). When Cas12a nucleases were delivered without crRNAs, no editing activity was observed. RNP delivery resulted in highly efficient genome editing at all four target sites. At the OsPDS and OsEPFL9 sites, LbCas12a, LbCas12a-E795L, and AsCas12a Ultra showed high editing efficiencies, 93.6–98.8%, which outperformed AsCas12a at 68.5–77.6% (Figures 1A,D). At the OsROC5 site, 92.5–94.6% editing efficiencies were obtained with all four variants, showing no significant differences among these variants (Figure 1B). Slightly lower editing efficiencies were observed at the OsmiR528 site, due to that EarI based RFLP can only detect edits at the last three nucleotides of the target sequence (Figure 1C). In this case, DNA with edits at other regions of the target sequence can still be digested like unedited DNA. This data suggests that Cas12a RNPs can be efficiently delivered into rice cells, resulting in >90% editing efficiencies. Engineered AsCas12a Ultra has enhanced editing activity over AsCas12a and is on par with LbCas12a and LbCas12a-E795L. The data correlates with human cells where AsCas12a Ultra showed superior editing activity (Zhang L. et al., 2021). The reason that AsCas12a Ultra did not outperform LbCas12a and LbCas12a-E795L as in human cells (Zhang L. et al., 2021) could be because a 32°C temperature, not 37°C, was used for rice cell transfection.

The high editing efficiency observed with RNP delivery in rice protoplasts suggests that RNP CRISPR-Cas12a delivery could result in improved editing efficiencies over traditional plasmid delivery methods. To compare editing efficiencies between RNP delivery and plasmid delivery, three of the highest performing Cas12a variants, LbCas12a, LbCas12a-E795L, and AsCas12a Ultra, were used to target the OsEPFL9 site at two temperatures, 32°C and 25°C, which are relevant temperatures for plant genome editing and tissue culture. Both RFLP (Figures 2A,B) and next generation sequencing (NGS) of PCR amplicons (Figures 2C,D) methods were used to detect editing efficiencies. The results from these two methods were highly consistent with each other, indicating RFLP is a reliable method to detect editing efficiencies, although it is much cheaper and faster than NGS. RFLP revealed slightly higher editing efficiencies than NGS, possibly due to the different sensitivity of these two methods. RFLP assays showed significantly higher editing efficiencies were obtained using RNP delivery than plasmid delivery at both temperatures (Figures 2A–C). Notably, RNP delivery resulted in two-fold or higher editing efficiencies over plasmid delivery at both temperatures, based on the RFLP assay (Figure 2B) and deep sequencing of PCR amplicons (Figure 2C). RNP delivery of LbCas12a-E795L generated the highest editing frequency at 32°C, 69.4% as quantified by amplicon deep sequencing (Figure 2C). In all conditions except plasmid delivery of LbCas12a-E795L, the Cas12a nucleases performed better at 32°C than at 25°C (Figures 2B,C). When Cas12a nucleases and crRNAs were delivered as RNPs, LbCas12a-E795L showed significantly higher editing efficiency than AsCas12a Ultra at both temperatures (Figures 2B,C). When CRISPR reagents were delivered as plasmids, LbCas12a showed significantly higher editing efficiency than AsCas12a Ultra at 32°C, while LbCas12a and LbCas12a-E795L both showed significantly higher editing efficiency than AsCas12a Ultra at 25°C. Overall, Cas12a nucleases showed higher activity at 32°C than 25°C, except when LbCas12a-E795L was delivered as plasmids. AsCas12a Ultra is more sensitive to lower temperatures than LbCas12a and LbCas12a-E795L. These results are consistent with our previous observation on Cas12a temperature sensitivity and hyper temperature sensitivity of AsCas12a in plants (Malzahn et al., 2019). Notably, RNP delivery even at the lower temperature (25°C) outperformed plasmid delivery at the higher temperature (32°C). The data hence support that RNP delivery of CRISPR-Cas12a can drastically improve editing efficiencies at low temperatures.

We further analyzed the editing profiles using deep sequencing data (Figure 2D). The majority of the edits were deletions, as reported previously (Tang et al., 2017). No obvious differences were observed among the Cas12a variants. Similar editing patterns were obtained using different delivery methods at both temperatures, supporting that Cas12a nuclease activity, not DNA repair, and is impacted by temperature (Malzahn et al., 2019).

The high editing efficiencies observed with RNP delivery of CRISPR-Cas12a suggest that the RNP dosage used could be near saturation. To investigate the effects of dosage on genome editing using RNP delivery, we edited the OsEPFL9 site using different concentrations of RNPs. At high concentrations (0.01, 0.025, 0.05, and 0.1 µM), LbCas12a and LbCas12a-E795L maintained approximately 100% editing efficiency assessed by RFLP (Figure 3A). At 0.01 µM, AsCas12a Ultra dropped to 81.8%, while AsCas12a dropped to 19.9% (Figure 3B). These results indicate that AsCas12a Ultra is more efficient than AsCas12a in rice cells, as observed in human cells (Zhang L. et al., 2021). At low concentrations (0.0001, 0.0005, 0.001, and 0.005 µM), editing efficiencies of LbCas12a and LbCas12a-E795L dropped along with the dosage titration (Figure 3C). AsCas12a Ultra and AsCas12a showed much lower activity than at high concentrations (Figure 3C). The comparison at low RNP concentrations revealed that LbCas12a-E795 is slightly more robust than LbCas12a (Figure 3C). These results suggest that 0.01 µM RNPs of LbCas12a or LbCas12a-E795L are sufficient to obtain highly efficient genome editing in rice cells.

For RNP delivery, it is important to know the optimal molar ratio of Cas12a:crRNA. Theoretically, a 1:1 ratio is expected to configure the RNP complex with minimal waste of either component. However, more crRNA is often used in practice, due to concerns associated with crRNA degradation (Zhang et al., 2021b). To investigate the effects of Cas12a:crRNA ratio on genome editing using RNP delivery, we edited the OsEPFL9 site using four Cas12a:crRNA ratios, 1:1, 1:2, 1:3, and 1:5, at RNP concentrations of 0.1, 0.01 and 0.001 µM. We compared LbCas12a-E795 and AsCas12a Ultra in this experiment and assessed editing efficiency by RFLP assay. Regardless of RNP dosage, no significant differences were observed when using different Cas12a:crRNA ratios for LbCas12a-E795L and AsCas12a Ultra (Figure 4). Interestingly, sequential reductions of RNP concentration results in ∼50% reductions of editing frequencies in each concentration (Figure 4). At 0.001 µM, editing activity by AsCas12a Ultra was undetectable regardless of Cas12a:crRNA ratios (Figure 4C), which is consistent with the earlier data (Figure 3C). Our results suggest that the 1:1 Cas12a:crRNA ratio is optimal for efficient genome editing in rice cells.

RNA modification could potentially enhance their stability in cells (Hendel et al., 2015). We tested one type of crRNA modification for LbCas12a-E795L and four types of crRNA modification for AsCas12a Ultra at the OsEPFL9 site, using RNP concentrations of 0.01 and 0.001 µM. All of these modifications will presumably protect crRNAs from degradation, before and after transfection. Strikingly, no significant differences were observed with or without end modification, or between different types of modification at either RNP concentration (Figure 5). These results suggest that crRNA modifications have minimal effects on genome editing efficiency in rice cells. We speculate that the RNP assembly is so efficient that crRNAs are minimally exposed to RNases before and after transfection. On the contrary, DNA delivery of CRISRP-Cas12a requires individual expression of Cas12a and crRNA before they form RNPs within cells, thereby providing many opportunities for crRNA degradation. Hence, our data support that RNPs of CRISPR-Cas12a are highly stable in rice cells.

NLS is important for Cas12a to enter the nucleus for genome editing. It is possible CRISPR-Cas12a genome editing efficiency may be improved through NLS optimization. To this end, we tested four types of NLSs (Supplementary Table S4) for LbCas12a-E795L at the OsEPFL9 and OsROC5 sites with RNP concentrations of 0.1 and 0.01 µM. No significant differences for editing efficiency were observed using different types of NLSs as measured by RFLP, except at the OsROC5 site with 0.01 µM RNPs (Figure 6). Without NLS, editing efficiencies at the four target sites ranged from none (at OsROC5) (Figure 6B) to 75.4% (at OsEPFL9) (Figure 6C). The relatively high editing efficiency by LbCas12a-E795L at OsEPFL9 even without NLS is likely attributed to the high efficiency of the crRNA for this site and the high RNP concentration (0.1 µM) (Figure 6C). At 0.01 µM of RNPs, the editing activity by LbCas12a-E795L significantly dropped without an NLS at this same target site (Figure 6D). These data suggest that nuclear entry is a limiting step for Cas12a:crRNA RNPs and NLS is critical for efficient Cas12a mediated genome editing.

Although our initial RNP CRISPR-Cas12a delivery data was derived from rice protoplasts, the protocols developed in principle should be widely applicable to protoplasts systems derived from other plant species. To expand the application of Cas12a-mediated genome editing using RNP delivery, we also conducted genome editing using LbCas12a in the protoplasts derived from embryogenic calli of Hamlin sweet orange, a dicot fruit tree species. We first established a pipeline that yielded high quality protoplasts from suspension culture cells derived from an embryogenic sweet orange callus line Hamlin 89 (Figure 7A). With these protoplasts, three RNP concentrations, 0.1, 0.01, and 0.001 µM, were used to assess genome editing at the CsPH5 gene. The RFLP assay showed 90.8% editing efficiency at 0.1 µM, 40.1% editing efficiency at 0.01 µM, and 35.3% editing efficiency at 0.001 µM (Figure 7B). Although these editing efficiencies were slightly overestimated due to residual uncut bands by RFLP in the mock samples, the results clearly indicate that RNP Cas12a delivery provides as a robust means for highly efficient genome editing in citrus cells. Notably, this experiment was conducted at 28°C. Whether this genome editing efficiency in citrus can be reproduced or improved at other temperatures needs to be further investigated. By contrast, our attempts of plasmid delivery of CRISPR-Cas12a yielded nearly undetectable editing activity in citrus protoplasts (data not shown). Since the protoplasts used were derived from embryogenic calli of citrus that have high potential for plant regeneration, high efficiency genome editing by RNP delivery of CRISPR-Cas12a suggests a promising avenue to engineering transgene-free genome-edited citrus varieties that are potentially resistant to devastating diseases such as citrus greening disease. A protocol that renders successful regeneration from such protoplasts to embryogenic calli and subsequently to whole plants will need to be established. It is also envisioned that absence of selection as a nature of RNP delivery should not be a problem due to the high editing efficiency.