Donor and transgenic plants (cv. Fielder) were grown in Sunshine^® Mix #4 Aggregate Plus potting mix with Osmocote 14-14-14 slow-release fertilizer in growth cabinets at 20°C day and 15°C night temperatures, 16/8 h day/night regime, and 600 µmol/m²s light intensity provided by fluorescent tubes. Humidity was maintained at 70%. Extra care was taken to keep donor plants healthy and reduce the risk of spreading pathogens. Powdery mildew was controlled through weekly foliar application of the plant health promoter Foliar Supreme S720 (2 mL/L, OMEX Agriculture Inc., Canada) until the heading stage. The thrip population was controlled through the weekly application of the diatomaceous earth to the heads and leaves of the donor plants.

The wheat transformation was done as described in Hayta et al. (2021). In brief, following seed sterilization, immature embryos were collected at the early milk stage (∼10–13 days post-anthesis, 1–2 mm in diameter) into 1.7-mL microcentrifuge tubes containing 1 mL of wheat inoculation medium (WIM—0.44 g/L Murashige and Skoog, 10 g/L glucose, and 0.5 g/L 2-(N-morpholino) ethanesulfonic acid (MES)) with 0.05% Silwet L-77 (∼100 embryos/tube). The WIM was replaced with the fresh media, and embryos were centrifuged for 10 min at 14,000 rpm at 4°C. The transformation was performed using the hypervirulent Agrobacterium tumefaciens strain AGL1 carrying the pSoup helper plasmid and the JD633 plasmid containing the cloned gRNA seed sequences. In some cases, the Agrobacterium cultures with two different gRNAs were mixed in equal volumes immediately before embryo inoculation for co-transformation experiments. WIM was removed, and 1 mL of Agrobacterium solution (WIM supplemented with 100 μM acetosyringone, OD₆₀₀ = 0.5) was added; tubes were inverted repeatedly for 30 s and incubated for at least 20 min. The suspension was removed with the pipette, and 25–30 embryos were transferred to fresh 90-mm-diameter plates, scutellum side up, with a wheat co-cultivation medium (WIM supplemented with 100 μM acetosyringone, 5 μM AgNO₃, 1.25 mg/L CuSO₄·5H₂O, and 8 g/L agarose). Petri plates were sealed and incubated in the dark at 24°C for 3 days, followed by excision of the embryogenic axes. The embryos, scutellum side up, were transferred to fresh wheat callus induction medium plates (WCI—4.3 g/l MS, 30 g/L maltose, 1.0 g/L casein hydrolysate, 10 mL 100× vitamin stock, 2 mg/L picloram, 0.5 mg/L 2,4-D, 1.25 mg/L CuSO₄·5H₂O, 160 mg/L timentin, and 5 g/L agarose) and incubated at 24°C in the dark for 5 days. Following the resting period, the embryos were transferred to fresh WCI plates supplemented with 15 mg/L hygromycin and kept at 24°C in the dark for 2 weeks. Eventually, the calli were split into clumps of ∼4 mm² and transferred to fresh selection plates (WCI plates) but with 30 mg/L hygromycin. The plates were moved back to 24°C in the dark for 2 weeks, then transferred to a lit culture chamber (100 μmol/m²/s) at the same temperature with a 16-h photoperiod, and covered with a single-layer paper towel for a week. The transformed calli started to turn green and produce small shoots during this period. The calli were transferred a final time to a wheat regeneration medium (WRM—4.3 g/l MS, 30 g/L maltose, 1.0 g/L casein hydrolysate, 10 mL vitamin stock, 1.0 mg/L zeatin, 1.25 mg/L CuSO₄·5H₂O, 3.5 g/L phytagel, 160 mg/L timentin, and 20 mg/L hygromycin) and kept under fluorescent lights (100 μmol/m²/s) at 24°C with a 16-h photoperiod. The regenerated shoots with prominent roots were transferred to magenta jars containing the WCI medium without growth regulators and supplemented with 15 mg/L hygromycin. The well-developed shoots with strong root systems were eventually transferred to soil and covered with a plastic dome for acclimatization.

The wheat transformation was done with the JD633 plasmid containing the TaCas9 expression cassette together with TaU6:gRNA and the GRF4–GIF1 embryogenesis chimeric gene (Debernardi et al., 2020). The plasmid was a gift from Jorge Dubcovsky (Addgene plasmid # 160393; http://n2t.net/addgene:160393; RRID: Addgene_160393). Complementary oligos (Supplementary Table S6) were annealed (2 µM each in 50 µL of water), and 1 µL of the annealed product was mixed with 100 ng of the AarI-digested vector, 1x T4 DNA ligase buffer, and 400 U of T4 DNA ligase. The ligation mix was incubated for 1 h at room temperature and transformed into E. coli commercial chemical competent cells. Cloned seeding sequences were verified by the Sanger sequencing method, and the plasmids were transformed into the A. tumefaciens strain AGL1 (pSoup) using a chemical transformation protocol (An et al., 1988). gDNA was isolated from leaves of the soil-acclimated putative transgenic plants, and the presence of the transgene was confirmed using PCR with HYG-specific primers (HYG-F and HYG-R, Supplementary Table S6). Segregation analysis for the T₁ plants was done using the same HYG-specific primers and with up to 10 plants per T₀ transgenic line.

The editing efficiency at the target sites was assessed using either the cleaved amplified polymorphic sequence (CAPS) assay, as described previously (Xu et al., 2022), or the qPCR method. In the CAPS assay, the gRNA#10 target regions were amplified using sub-genome-specific or universal primers (Supplementary Table S6) and digested with SmaI. The digested products were separated on 1% agarose gel, and the editing efficiency was calculated as a percentage of undigested band intensity measured using Image Lab software (Bio-Rad, United States).

A multiplexed probe-based qPCR method was used to assess the editing efficiency for gRNAs tested in this study. The primers were designed to amplify the editing region with the 5′-FAM (6-fluorescein)-labeled probe annealing to the cleavage sites of the gRNAs for three sub-genomes (A, B, and D, Supplementary Table S6). The PINb gene (puroindoline-b, two copies per hexaploid genome) was used as a reference (Collier et al., 2017). The reference gene probe was 5ʹ-HEX (hexachlorofluorescein)-labeled, and both gRNA and PINb gene probes contained ZEN and Iowa Black Hole Quencher 1 (Integrated DNA Technologies, Coralville, IA, United States). The 1x Luna^® Universal Probe qPCR Master Mix (cat. #M3004S; New England Biolabs, United States), 500 nM of each primer pair (for the endogenous reference gene and the transgene), and 250 nM each of gRNA and reference probes were mixed in the total volume of 20 μL. qPCR was run on the CFX96 Real-Time PCR Detection System (Bio-Rad) with 50 ng/sample of gDNA. PCR conditions consisted of one cycle of initial denaturation at 95°C (60 s), followed by 45 cycles of denaturation at 95°C (15 s) and extension at 60°C (30 s). Reactions were run in triplicate, and average ΔΔCt was calculated for transgenic and wild-type control (WT, cv. Fielder) samples for corresponding gRNAs using CFX Manager, v. 3.1 (Bio-Rad). The GE rates were calculated through the normalization of ΔΔCt values of transgenic plants to WT control (ΔΔCt_norm), and the 1/ΔΔCt_norm ratios were used for analysis.

Target regions for three selected gRNAs (gRNAs #2, 4, and 6) were PCR-amplified with either sub-genome-specific or universal primers (Supplementary Table S6) using gDNA of the T₁ plants that demonstrated edits in the qPCR assay. Fragments were analyzed on the gel and cloned into the pJET1.2 plasmid (cat. #K1231, CloneJET PCR Cloning Kit, Thermo Fisher Scientific). Sequencing was done using Primordium Labs (United States) and analyzed using Geneious Prime software.

The T-DNA copy number in T₀ plants and their progeny was measured using digital droplet PCR (ddPCR), as described by Collier et al. (2017), with some modifications. gDNA was isolated from young leaves using the method that was developed by Warner et al. (2002). The concentration was quantified using a Qubit fluorometer and the dsDNA quantification kit (cat. #Q32851), as described by the manufacturer (Thermo Fisher Scientific, Waltham, MA, United States). For every sample, 1 µg of gDNA was digested either for 1 or 12 h at 37°C using EcoRI-HF (cat. #R3101S) in 1x rCutSmart Buffer, followed by heat inactivation at 65°C for 20 min. The digested gDNA was purified using the DNA Clean-Up and Concentration Micro-Elute Kit (cat. #67200, Norgen Biotek Corp., ON, Canada) and quantified again using a Qubit fluorometer. Two different amounts of digested gDNA/reaction were tested and compared for the efficiency of transgene copy number estimation—350 ng (Collier et al., 2017) and 10 ng (Jouanin et al., 2020). The transgene-specific primers/probe mix was explicitly designed for TaCas9 (Supplementary Table S6), and normalization was done against the PINb gene. The transgene and reference gene probes were labeled with 5′-FAM (6-fluorescein) and 5ʹ-HEX (hexachlorofluorescein). All probes were double-quenched with ZEN and Iowa Black Hole Quencher 1 (Integrated DNA Technologies, Coralville, IA, United States). The ddPCR master mix for amplification contained 1x ddPCR Supermix for Probes (no dUTP; cat. # 186-3024; Bio-Rad Laboratories), 500 nM of each primer pair (for the endogenous reference gene and the transgene), and 250 nM of each probe. The total volume was adjusted to 20 µL with the ultrapure water. The droplets were produced following the manufacturer’s instructions using a droplet generator (Bio-Rad) and Bio-Rad Droplet Generation Oil (cat. # 186-3005). Droplets (40 µL) were transferred to a 96-well skirted Eppendorf PCR plate (cat. # 951020362), sealed with pierceable foil (cat. # 181-4040; Bio-Rad) and placed into the Bio-Rad thermocycler. Initial denaturation was done for 10 min at 95°C, followed by 40 cycles of 94°C (30 s) and 60°C (1 min) with a final step at 98°C for 10 min. A temperature ramp rate was set to 2°C/s, according to the manufacturer’s instructions. The samples were then transferred to a QX200 droplet reader (Bio-Rad). Data analysis was done using the Bio-Rad QuantaSoft Analysis Pro software (v1.0.596) with manual settings for threshold determination to distinguish positive and negative droplets.

We used end-point PCR to estimate the intactness of the left and right border (LB and RB, respectively) integration events in the T₀ plants. For RB, forward primer (RB-F, Supplementary Table S6) was used in combination with either RB-R1 or RB-R2 primers, one of which anneals at RB (RB-R2, L = 355 bp), whereas another one is 122-bp 5′-upstream from RB (RB-R1, L = 233 bp). The PCR reaction was run on 1% agarose gel, and the presence of the larger band was indicative of the intactness of RB. For LB, common reverse primer (LB-R) was mixed with either of the three forward primers annealing either at LB (LB-F1, L = 500 bp), 75-bp 3′-downstream (LB-F2, L = 425 bp), or 317-bp downstream (LB-3, L = 183 bp). The presence of the 500-bp PCR fragment indicated the integrity of LB.

In experiments where two Agrobacterium cultures carrying different gRNAs were mixed, co-integration of T-DNAs was assessed using PCR with the common primer for the JD633 vector backbone (JD633-F) and reverse primer specific to the gRNA seeding sequence (Supplementary Table S6). The PCR reactions were analyzed separately on 1% agarose gel for every T₀ transgenic plant.

Total RNA was isolated from leaves of transgenic T₁ seedlings using a NucleoSpin RNA Plant Kit (cat. # 740949.50, Macherey-Nagel, Germany), according to the manufacturer’s instructions. cDNA was synthesized using a SuperScript IV First-Strand Synthesis System (cat. # 18091050, Thermo Fisher Scientific, United States), following the manufacturer’s protocol. For every ddPCR, 20 ng of cDNA/sample was used and combined with 1x ddPCR Supermix for Probes (no dUTP; cat. # 186-3024; Bio-Rad Laboratories), 500 nM of each primer pair (for the endogenous reference gene and the transgene), and 250 nM of each probe. Expression from the transgene cassette was measured using ddPCR (Bio-Rad, United States) with TaCas9 primers/probe (Supplementary Table S6). TaGA3PD (TraesCS7A02G313100, TraesCS7B02G213300, and TraesCS7D02G309500) was used as an endogenous reference (Garrido et al., 2020) (TaGA3PD, 5′-HEX-labeled, double-quenched with ZEN and Iowa Black Hole Quencher 1, Supplementary Table S6). ddPCR gene expression data were normalized as previously described (Coulter, 2018).

The chromatin marks for the wheat (cv. Chinese Spring) were extracted from the following datasets—SRP126222 (H3K27me3, H3K36me3, H3K4me3, and H3K9ac), SRP133674 (DNA methylation) (Concia et al., 2020a), GSE133885 (ATAC-Seq), and GSM3929161 (RNA polymerase II ChIP-Seq) (Jia et al., 2021). The gRNA regions, including the three-nucleotide protospacer adjacent motif (PAM), were BLAST-searched using the WheatOmics 1.0 browser (http://wheatomics.sdau.edu.cn/), and nucleotide resolution data were extracted for corresponding chromatin marks. The average value across the respective gRNA target sequence plus PAM was calculated for three sub-genomes. The data set for every chromatin mark was normalized for comparison on a scale of 0–1, with 1 indicating the highest level of that feature, and used to construct the heatmap.

The ATAC-Seq coverage data for wheat chromosomes (IWGSC RefSeq v1.0 (International Wheat Genome Sequencing, 2018)) were plotted using the R package karyoploteR (Gel and Serra, 2017). Eventually, the gRNA and centromere positions were displayed as markers.

Statistical significance between groups for correlation analysis was calculated using Pearson’s correlation test, with p < 0.05 considered significant.

The Agrobacterium-mediated wheat transformation was done following a recently published protocol (Hayta et al., 2021) with the co-delivery of the wheat transcription factor GRF4 and its cofactor GIF1 (Debernardi et al., 2020). The reported transformation efficiency with the improved protocol and the GRF4–GIF1 construct co-delivery was 77.5% in cv. Fielder (number of transgenic plants per cultured embryos). Similarly, our transformation rate with the JD633 plasmid carrying the GRF4–GIF1 and Cas9/gRNA expression cassette ranged from 22% to 68% with an average of 44.1% (Table 1, total transformation rate, including both clonal and non-clonal events). Many calli following transfer to the WCI media and later to the WRM generated multiple green shoots (Figure 1) that we treated as single transgenic events. However, we did not exclude the possibility of some of them being clones if they originate from a single callus (not tested in this study). Healthy donor plants and good embryo quality were critical for the efficient regeneration of transgenic plants. It is worth noting that following the co-culturing of embryos on WIM, the embryos are transferred to WCI with timentin to control the Agrobacterium overgrowth. The WCI media must be prepared fresh, and the culture plates can be stored in the fridge for up to 2 weeks to prevent timentin from losing its activity.

The transgene copy number was measured using the ddPCR assay as described previously (Collier et al., 2017) with some modifications. Based on the wheat genome size (17.33 pg for 1C plant DNA), the original protocol calls for 350 ng of EcoRI-digested wheat gDNA/sample to quantify the transgene copy number (Collier et al., 2017). It was also suggested to perform the digestion of gDNA for 12 h. At the same time, another study on the optimization of the α-gliadin gene copy number estimation identified the optimal amount to be 10 ng of digested gDNA per 20 µL of ddPCR mixture(Jouanin et al., 2020). Therefore, we compared the T-DNA copy number values between the two different gDNA amounts for 11 T₀ plants and observed no significant difference in the copy number values between corresponding samples (Supplementary Table S1, Pearson’s correlation R = 0.99, p = 1.24e-08). Additionally, a prolonged digestion period (more than 1 h) of the total gDNA before the ddPCR analysis was not required for reproducible estimation of the transgene copy number (Supplementary Table S2, Pearson’s correlation R = 0.97, p = 6.98e-08). To our surprise, many T₀ plants selected for the study (10 out of 17) contained multiple T-DNA copies ranging from 3 to as high as 17 (Figure 2A). These data were further confirmed by estimating the transgene copy number in the T₁ population and segregation analysis in the T₂ progeny (Figures 2B, C). Consistent with the recent report (Luo et al., 2021), we did not observe a correlation between the Cas9/gRNA cassette copy number and GE rate in the T₀ plants and their progeny (Supplementary Table S3, Pearson’s correlation, R = 0.31, p = 0.33 and R = 0.24, p = 0.27 for T₀ and T₁ plants, respectively). At the same time, we observed a moderate but significant positive correlation between the transgene copy number and gene expression in respective samples (Supplementary Figure S1, Pearson’s correlation, R = 0.55, p = 0.0049).

The GE efficiency for the high-throughput functional genomics studies could be increased by co-delivering multiple gRNAs targeting the same gene (Luo et al., 2021). gRNAs could be coded on the single expression plasmid (e.g., tRNA or ribozyme expression cassettes) or the separate plasmids (Abdelrahman et al., 2021). We cloned 10 different seeding sequences into the JD633 plasmid, transformed them separately into Agrobacterium, and performed five individual co-transformation experiments by mixing two Agrobacterium isolates before the embryo inoculation. We screened 10 T₀ plants, each regenerated from five independent co-transformation events, with approximately half of them (48% ± 17.8%) being PCR-positive for both gRNAs co-delivered (Supplementary Figure S2).

During the Agrobacterium-initiated T-DNA delivery, the right border is protected from the exonuclease cleavage by covalently linked VirD2 protein, whereas the left edge is prone to truncations (Gelvin, 2017). For this reason, plant selection cassettes are usually placed closer to the left border to select for transgenic events with intact left and right border regions. We tested the integrity of the left and right T-DNA borders in 10 randomly selected T₀ transgenic plants transformed with the JD633::gRNA#10 plasmid using primers placed at different distances from the borders. Three separate forward primers (LB-F1, F2, and F3) and one standard reverse primer (LB-R) were used for the left border. For the right border, we used one common forward primer (RB-F) and two reverse (RB-R1 and R2) primers (Figure 3A). Seven out of ten plants tested had an intact left border, whereas only three plants had a complete right border (Figure 3B). At the same time, all but one plant (T₀ plant #22) had an entire gRNA cassette at the right border, as seen from the presence of the RB-1 PCR fragment. Overall, we observed truncations at both borders. Nevertheless, these deletions were at a distance from the essential components of the cassettes.

Previously, others and we have demonstrated that biolistic-mediated transformation can effectively generate edited wheat plants (Li et al., 2021; Xu et al., 2022). Here, we tested the efficiency of GE with the single-guide RNA (gRNA#10) targeting six copies of the wheat grain width and weight2 gene (TaGW2) (Wang et al., 2018). The gene encodes a RING-type E3 ubiquitin ligase, found initially to regulate rice grain weight by increasing the cell number of spikelet hulls (Song et al., 2007). We analyzed 46 T₀ lines for the presence of edits at the target sites using the cleaved amplified polymorphic sequence assay (CAPS), with 27 of the lines (59%) demonstrating various levels of edits at three sub-genomes (Supplementary Figure S3). All but one edited line were heterozygous/chimeric for the presence of edits. We randomly selected 12 lines and propagated them into the T₁ generation to test for the transgenerational inheritance of the edits. The 24 lines (two T₁s per parental line) were screened for the presence of the transgene cassette and the edits. Fifty percent of the screened progeny contained different levels of edits at the target site, with one plant (T₁ plant# 9-2, Supplementary Figure S4) having disrupted all six copies of the gene.

Although the CAPS assay remains a highly robust method for detecting the GE events in polyploid crops, it relies on a restriction site at the gRNA cut site, which is not always the case. We decided to compare the GE rate detection between the CAPS assay and the probe-based qPCR method for the T₀ transgenics generated with JD633::gRNA#10. The qPCR assay is less labor-intensive and allows higher-throughput analysis than the CAPS assay. The GE events were quantified for three sub-genomes in 13 plants for the CAPS assay, and an average GE rate per T₀ transgenic plant was calculated (Figure 4A). The same 13 plants were analyzed using the qPCR-based method with the universal primers designed to span the target regions for all three sub-genomes with the probe annealing to the gRNA cut site. Eventually, the GE rates for the qPCR assay were calculated through the normalization of ΔΔCt values of transgenic plants to WT control (ΔΔCt_norm, cv. Fielder), and the 1/ΔΔCt_norm ratios were used for plotting. We observed a significant positive correlation between the two quantification assays (Pearson’s correlation, R = 0.81, p = 0.0008, Figure 4B). Therefore, we decided to use the qPCR-based assay to quantify GE rates.

We generated 10 constructs with gRNA seeding sequences cloned into the JD633 plasmid, targeting seven genes (Table 2). Every gRNA was designed to mutagenize three gene homeologs per haploid genome with 30 regions (Figure 5A). We generated 65 T₀ plants with at least three independent plants per gRNA.

Mean normalized GE rates for T₀ plants were quantified using the probe-based qPCR method varied from 1.29 to 35.15, with gRNA#9 showing the highest activity (Figure 6A, Supplementary Table S4). We further confirmed the presence of edits at the target site for the gRNAs #2, 4, and 6 using Primordium sequencing (Figure 6B). We detected single-nucleotide insertions, deletions, substitutions, and small and large deletions (from 5 to 78 nucleotides) at the gRNA cut sites. Overall, we observed no correlation between activity scores (Table 2, using models devised by Doench et al., (2016)) and the editing rate of gRNAs tested in our study (Pearson’s R = −0.22, p = 0.55).

Editing efficiency in plants can vary depending on gRNA used, with both sequence-dependency and epigenetic profile suggested as the reasons for inconsistent GE rates among different loci (Naim et al., 2020; Weiss et al., 2022). Therefore, we decided to test whether the epigenetic profile at the target loci can affect the mutagenesis rates observed in our study. The nucleotide resolution data for six epigenetic profiles (DNA methylation, H3K4me3, H3K9ac, H3K27me3, H3K36me3, and ATAC-Seq) and RNA polymerase II-ChIP (Concia et al., 2020a; Jia et al., 2021) for corresponding gRNAs were extracted from available datasets for wheat seedlings (cv. Chinese Spring) and correlated with the mean GE rates observed in Figure 6A. We observed a significant positive correlation between the average mutagenesis rates for 10 gRNAs and ATAC-Seq data (Pearson’s correlation, R = 0.68, p = 0.03, Figures 5B, C). At the same time, no correlation was observed for the rest of the epigenetic marks analyzed (Supplementary Table S5).