Multiplexed promoter and gene editing in wheat using a virus‐based guide RNA delivery system

Summary The low efficiency of genetic transformation and gene editing across diverse cultivars hinder the broad application of CRISPR technology for crop improvement. The development of virus‐based methods of CRISPR‐Cas system delivery into the plant cells holds great promise to overcome these limitations. Here, we perform direct inoculation of wheat leaves with the barley stripe mosaic virus (BSMV) transcripts to deliver guide RNAs (sgRNA) into the Cas9‐expressing wheat. We demonstrate that wheat inoculation with the pool of BSMV‐sgRNAs could be used to generate heritable precise deletions in the promoter region of a transcription factor and to perform multiplexed editing of agronomic genes. We transfer the high‐expressing locus of Cas9 into adapted spring and winter cultivars by marker‐assisted introgression and use of the BSMV‐sgRNAs to edit two agronomic genes. A strategy presented in our study could be applied to any adapted cultivar for creating new cis‐regulatory diversity or large‐scale editing of multiple genes in biological pathways or QTL regions, opening possibilities for the effective engineering of crop genomes, and accelerating gene discovery and trait improvement efforts.


Introduction
Crop improvement using the CRISPR-Cas-based editing relies on understanding the function of genes involved in the regulation of biological processes affecting phenotypic variation. While major advances were made towards linking genes with major agronomic phenotypes in wheat and other crops, and genome sequences facilitated inter-species extrapolation of functional information among related species, the mechanistic basis of most of the traits remains poorly characterized. The advances in gene mapping and large-scale genomic analyses helped to identify a number of quantitative trait loci (QTL) and biological pathways associated with major traits (He et al., 2022). However, the number of candidate causal genes detected in these studies still remained beyond the technical capabilities of existing genomic screens available for gene function validation in major crop species. The ability of CRISPR technology to introduce targeted changes into genomes has been broadly utilized in model systems to develop high-throughput functional screens greatly accelerating the characterization of causal genes and pathways (Shalem et al., 2015). Although, the large-scale CRISPR-Cas editing is a powerful gene discovery tool, the scope of its application in crop genetics is limited by the complexity and cost of such projects (Liu et al., 2020) and warrants the development of more effective gene editing strategies.
In many crops, including wheat, the realization of the full potential of CRISPR technology is hindered by a combination of methodological challenges. While CRISPR technologies based on the Cas9 and Cas12a editors have been successfully applied to edit single and multiple genes in the wheat genome (Gil-Humanes et al., 2017;Liang et al., 2017;S anchez-Le on et al., 2018;Wang et al., 2014;Zhang et al., 2019), they show relatively low editing efficiency and required transformation of a large number of plants or screening of large populations in the next generation of transgenic plants to recover desired mutations (Wang et al., 2018a(Wang et al., , 2021. Moreover, the genetic transformation protocols developed for wheat, and for many other crops, are restricted to few varieties showing high regenerative capabilities (Debernardi et al., 2020). This reduces the utility of CRISPR technology for the high-throughput editing of a large number of genes or the direct editing of adapted cultivars for testing the effects of novel CRISPR-induced alleles in diverse genetic backgrounds. The recent discovery of growth regulators Baby boom, Wuschel (Lowe et al., 2016), and GRF-GIF (Debernardi et al., 2020) significantly improved the regeneration efficiency and broadened the range of wheat cultivars amenable to genetic transformation. However, plant transformation remains a timeand resource-consuming effort that requires specialized expertise, limiting its application in most of the breeding or research programmes.
Recently, virus-based CRISPR delivery systems were developed and tested for several major crops, including wheat ( Cerm ak et al., 2015;Ellison et al., 2020;Gil-Humanes et al., 2017;Hu et al., 2019;Li et al., 2021). Compared with the Agrobacteriumor biolistic-based delivery of editing reagents, these systems rely on the natural ability of viruses to spread across the plant cells and allow for omitting the plant transformation and regeneration steps (Baltes et al., 2014). A DNA-based viral replicon was successfully used for delivering CRISPR, Cas9, and DNA templates for homology-directed repair into tomato plants ( Cerm ak et al., 2015). The RNA viruses used for delivering the guide RNAs could maintain the high editing efficiency in the wheat and tobacco lines expressing Cas9 enzymes (Ellison et al., 2020;Li et al., 2021). The gene editing system based on the barley stripe mosaic virus (BSMV) was capable of generating both somatic and heritable mutations in single or multiple genes in wheat plants expressing Cas9 (Hu et al., 2019;Li et al., 2021). Because of the ease of viral infection procedure, the virus-based guide RNA delivery system could potentially be adopted by any genetic research and breeding programmes and scaled up to edit many targets, overcoming the limitations of genetic transformation in wheat and other crops.
Here, we investigated the ability of the BSMV-based viral system to induce multiple targeted mutations and fragment deletions in the genomes of wheat cultivars carrying the introgression of a high-expressing allele of the Cas9 gene. By multiplexing BSMV-gRNA constructs, we created a series of deletions in the promoter of a transcription factor controlling domestication traits in wheat (Simons et al., 2006). According to recent studies, CRISPR/Cas9-induced cis-regulatory mutations in genes controlling productivity traits could broaden the range of phenotypic variation and help to overcome the negative impact of epistasis on major agronomic traits (Rodr ıguez-Leal et al., 2017;Soyk et al., 2017). However, the application of this strategy for mining beneficial cis-regulatory diversity remained limited in crops mostly due to the difficulties associated with the production of cis-regulatory mutants. In our study, we demonstrate that a BSMV-based viral system is an effective tool for creating new allelic variants in the regulatory regions of the wheat genome. In addition, using wheat lines with the low-and high-expressing Cas9 loci, we investigated the relationship between (i) the levels of Cas9 expression and the efficiency of BSMV-based genome editing, (ii) the frequency of somatic editing and the heritability of BSMV-Cas9-induced mutations, and (iii) the levels of BSMV-gRNA construct multiplexing and the efficiency of target editing. Our study shows that marker-assisted introgression of a highexpressing Cas9 locus provides efficient means for BSMV-sgRNA-based genome editing in any adapted cultivar and opens new possibilities for the analysis and discovery of new allelic diversity for crop improvement.

Efficiency of BSMV-based CRISPR editing depends on the Cas9 expression levels
To assess the ability of BSMV transcripts directly applied onto the wheat leaves to deliver sgRNAs into wheat cells and induce mutations, we used a highly efficient QT1 gRNA (Table S1)  previously designed to target the coding regions of the three homoeologous copies of the Q gene (TraesC-S5A02G473800, TraesCS5B02G486900, TraesCS5D02G486600), which controls major domestication traits in wheat (Simons et al., 2006). The Cas9-induced mutations at the QT1 site are expected to generate the loss-of-function q alleles, which should result in speltoid spear-shaped spikes characteristic of the wild relatives of domesticated wheat (Simons et al., 2006). The QT1-sgRNA was subcloned into pBSMVcPDS (Figures 1a and S1) to replace the TaPDS gene fragment (henceforth BSMV-QT1) ( Figure 1b). The 2nd leaf of the two-leaf seedlings from the progeny of a transgenic Bobwhite line (7438) constitutively expressing Cas9  was inoculated with BSMV-QT1 ( Figure 1c). The leaves of inoculated plants showed white streaks and spots, mosaic symptoms, and necrosis consistent with systemic viral infection ( Figure 1d). The next-generation sequencing (NGS) of the QT1 target site amplicons generated using DNA isolated from the 4th leaf of plants two weeks after inoculation (Figure 1e) was able to detect the Cas9-induced mutations at the frequency of 0.43% (Figure 1f and Table S2). The target site mutations were not detected in the next generation of these BSMV-QT1-inoculated plants. We concluded that the low expression level of Cas9 in transgenic line 7438 was one of the contributing factors to such low editing efficiency ( Figure S2).
Measuring the relative levels of Cas9 expression in multiple transgenic Bobwhite lines, we identified lines C413 and 707 (henceforth, high-Cas9 lines), both expressing Cas9 at levels~15fold higher than line 7438. Five plants derived from the high-Cas9 line C413 were inoculated with BSMV-QT1. The high levels of somatic editing efficiency, reaching 98% on average ( Figure 1g and Table S3), were detected at the QT1 site in all plants. The recovery of the nonfunctional q alleles in these M 0 plants is also supported by the speltoid spike phenotype (Figure 1h). The NGS of the M 1 generation plants confirmed the heritability of the BSMV-induced mutations ( Figure 1i). Among 49 analysed M 1 plants, 31 carried mutations in at least one genome copy and 25 plants were homozygous or heterozygous for mutant alleles in all three homoeologous copies of the Q gene ( Figure 1i). This result suggests that the high-Cas9 expression is required to support effective genome editing using sgRNAs delivered via BSMV.

The efficiency of somatic editing correlates with the heritable mutation rate
It was successfully demonstrated that improvement in gene editing efficiency could be achieved by targeting sgRNAs to shoot apical meristem cells using mobile RNAs (Ellison et al., 2020). In these experiments, the endogenous mobile RNA encoded by the FT gene was used (Li et al., 2011). However, prior studies of mobile RNA fusions on genome editing efficiency produced contradicting results in different plant species (Ellison et al., 2020;Li et al., 2021). We decided to re-evaluate the effect of mobile RNAs on editing efficiency in our experimental system by fusing Arabidopsis FT (AtFT), wheat Vrn3 (Yan et al., 2006), and methionine and isoleucine tRNAs (tRNA met and tRNA ile ) with the sgRNA targeting the TaGW2 gene (Figures 2a and S1), which affects grain size and weight in wheat and rice (Song et al., 2007;Wang et al., 2018b). Compared with the nonfused BSMV-GW2T2 construct, which showed 77% editing efficiency in inoculated leaves (Figures 2b and S3,  Table S4), AtFT, Vrn3, and tRNA ile fusions, except tRNA met , had substantially reduced somatic editing efficiency (29%, 33%, and 6%, respectively). The tRNA met fusion resulted in a minor efficiency reduction reaching 76% compared with nonfused BSMV-GW2T2. Consistent with these results, gene editing efficiency reduction caused by mobile RNA fusion was also observed for a sgRNA targeting the PDS gene in the wheat study (Li et al., 2021).
By genotyping the M 1 generation, we show that the proportion of M 1 plants with inherited target site mutations correlated with the frequency of somatic mutations detected in the leaves of the M 0 plants. The BSMV-GW2T2 produced the highest number of mutants in M 1 generation (88%), followed by BSMV-GW2T2-tRNA met (48%) and BSMV-GW2T2-AtFT (19%). The proportion of M 1 mutants created using BSMV-GW2T2 with all six copies of TaGW2 mutated (Figure 2d) reached 61%. Consistent with our previous report (Wang et al., 2018c), the TaGW2 mutants had significantly increased grain weight (13%) and width (6%) compared with wild-type plants (Figure 2e), though we did not detect significant changes in grain length. We assessed the frequency of somatic mutations induced by BSMV-GW2T2 and BSMV-GW2T2-tRNA met using DNA isolated from the 2nd (inoculated leaf), 4th and 6th leaf. The frequency of somatic editing positively correlated with the frequency of M 1 mutants in the 6th leaf, while a less obvious correlation was observed in the 2nd and 4th leaves (Figures 2f and S4).

Application of pooled BSMV-sgRNAs for editing multiple genes and generating promoter deletions
Previously, it was demonstrated that the BSMV-based system could be used to simultaneously edit multiple loci in wheat (Li et al., 2021). However, it remained unclear whether the efficiency of editing is affected by the BSMV-sgRNA pooling compared with the individual BSMV-sgRNA constructs. We performed multiplex editing by inoculating the high-Cas9 lines with the pool of RNAs synthesized from the BSMV-sgRNA constructs targeting the regions of TaGW2, TaUPL3 and TaGW7 conserved across all three wheat genomes (henceforth, the BSMV-GUG pool) ( Figure 3a). Compared with plants inoculated with the BSMV-sgRNAs targeting a single gene, the BSMV-GUG pool showed lower somatic editing efficiency for target sites GW2T2 and GW7T6, while the efficiency of UPL3T11 editing remained the same (Figures 3b and S5, Table S5). The number of mutants identified in the M 1 generation of plants inoculated with the pooled and individual BSMV-gRNAs showed similar trends (Figure 3c,d). The proportion of plants with mutations at the GW2T2, UPL3T11, and GW7T6 target sites dropped from 96% to 2.3%, from 27% to 19%, and from 51% to 21%, respectively. Only 5.7% of M 1 plants carried mutations at two target sites, and no mutants with all three targets edited were recovered ( Figure 3c). These results suggest that with an increase in the multiplexing level performed by simple pooling of BSMV-sgRNAs, we should expect a decrease in the efficiency of editing of individual targets and the rate of multiplex gene editing.
Previously, the multiplexed genome editing strategy was used to generate variation in the promoters of genes controlling productivity traits (Rodr ıguez-Leal et al., 2017). Here, we tested the ability of the pooled BSMV-sgRNAs to create deletions in the promoter of the Q gene allele on chromosome 5A (Figure 4a and  Table S1), which controls a number of domestication traits in wheat (Debernardi et al., 2017;Simons et al., 2006). Its expression levels positively correlate with compact spike morphology, free-threshing habit, and rachis fragility. In addition to domestication traits, the Q allele was linked with increased spike length, number of spikelets per spike, and number of grains per spikelet and spike . By editing the Q gene's promoter region, we might remove regulatory elements that have positive or negative effects on gene expression levels or regulate distinct biological pathways controlled by the Q gene. The Q gene promoter region on chromosome 5A was highly divergent from the promoter regions on homoeologous chromosomes 5B and 5D ( Figure S6). The Sanger sequencing of the 3 kb region upstream of the Q gene region revealed that cultivar Bobwhite has one SNP, one 160 bp deletion, and one 160 bp insertion compared with the reference genome of cultivar Chinese Spring (The International Wheat Genome Sequencing Consortium (IWGSC), 2018) ( Figure S7). The Q gene promoter was edited by inoculating the high-Cas9-expressing line C413 with a pool of five BSMV-sgRNA transcripts (henceforth, BSMV-pQT).
The means of somatic editing efficiency for each of the five target sites, including pQT17, pQT18, pQT23, pQT25, and pQT26, were 37%, 28%, 25%, 27%, and 5%, respectively ( Figure 4b, Table S6). The proportions of M 1 plants carrying heritable mutations at each target site were 15%, 11%, 6%, 6%, and 0%, respectively (Figure 4c), and highly correlated with somatic editing efficiency (Figure 4d). About 32% of M 1 plants carried heritable mutations in at least one target site. There were no M 1 plants carrying heritable mutations at three or more targets. A total of 5% of M 1 plants had heritable mutations at two target sites, which is nearly 5 times lower than the proportion of M 1 plants with mutations at a single target site (27%) (Figure 4c). Further, we screened the edited plants for deletions spanning the regions between the pairs of the BSMV-sgRNA target sites. First, we tested M 0 plants for deletions between target sites pQT17 and pQT18 by performing PCR using a pair of flanking primers 17F and 18R (Figure 4a). The NGS of amplicons showed that 5.2%-19.9% of reads carried the 71-bp deletion expected upon successful targeting of both the pQT17 and pQT18 sites ( Figure 4a, Table S7).
The Q gene promoter deletion screening in M 1 plants was performed using 11 pairs of PCR primers (Table S1). We identified five M 1 plants that had deletions in the homozygous or heterozygous state or presented as mosaic somatic mutations (Figure 4a,e). Two of these plants, pQT-C413-1-12-28-7 and pQT-C413-1-12-28-21, were homozygous for deletions between the pQT17 and pQT18 target sites and between the pQT18 and pQT23 target sites, respectively (Figure 4a).
Compared with wild-type plants, these two deletion mutants did not exhibit visible differences in development or spike morphology ( Figure S9). The plants with deletions in heterozygous state or present as nonfixed somatic mutations also did not show visible phenotypic effects on spike morphology. The quantitative RT-PCR analysis of Q gene expression in the 3rd and 6th leaves at the 6-leaf developmental stage and in the spikes of M 1 plants also did not detect significant changes compared with the wild-type plants ( Figures S8-S10). The lack of expression changes in mutants could be linked to the lack of overlap between the induced promoter mutations and regulatory regions controlling Q gene expression. Alternatively, promoter mutations led to changes in the Q gene expression domain without affecting absolute Q expression levels. It is also possible that the phenotypic effects of promoter mutations are small and require larger population sizes to detect subtle changes in spike morphology. Further studies aimed at creating additional deletions within the upstream regulatory region of the Q gene and more extensive phenotypic evaluation of created mutants are planned in the future.

BSMV-sgRNA-based gene editing in the winter and spring wheat cultivars
Even though the BSMV-based method of sgRNA delivery method bypasses the wheat transformation and regeneration steps (Li et al., 2021), it still depends on the availability of a wheat line expressing the Cas9 gene. Despite recent advances in the wheat regeneration methodology (Debernardi et al., 2020), the development of transgenic wheat cultivars remains a time-consuming process and has varying levels of success. To broaden the range of wheat cultivars amenable to editing using the BSMV-based sgRNA delivery system, we transferred the highly expressed allele of Cas9 from line 707 into elite spring wheat line 3 613 474 from CIMMYT and winter wheat line KS080093K-18 from Kansas wheat breeding programme (Figure 5a,b). The backcrossed progeny of both lines, now carrying the highly expressed allele of Cas9, was inoculated with the RNA transcripts synthesized from the BSMV-sgRNA constructs targeting the Q and TaGW7 genes. Similar to the results obtained by inoculating the high-Cas9 C413 line with BSMV-QT1 (Figure 1), the BC 2 F 3 plants from the 3 613 474 families inoculated with BSMV-QT1 also exhibited spear-shaped speltoid spikes suggestive of the presence of the loss-of-function mutations in the Q gene. Consistent with this result, a high level of somatic editing (97%) was observed for the QT1 target in these plants (Figures 5c and S11, Table S8). In the M 1 generation, 61% of plants carried mutations in at least one homoeologous copy of the Q gene with nearly 30% of plants carrying mutations in all three homoeologs of the Q gene ( Figure 5d). The plants from the 3 613 474 and KS080093K-18 backcross families inoculated with BSMV-GW7T6 showed 59% and 64% somatic editing efficiency, respectively (Figure 5c). In the M 1 generation, 11% and 17% of 3 613 474-and KS080093K-18-derived plants inherited mutations in the TaGW7 gene, respectively (Figure 5d).

Discussion
Our study shows that the efficiency of the BSMV-based editing and the chance of obtaining heritable mutations depend on the sgRNA design and the levels of Cas9 expression. This dependency holds for both single BSMV-sgRNA and multiplexed BSMV-sgRNA constructs, indicating that the development of the transgenic wheat line with adequate levels of Cas9 expression is critical for successful BSMV-sgRNA-based editing. Once a high-Cas9 line is developed, it can be used either directly in gene editing experiments or as a donor of the high-expressing Cas9 allele for introgression into other cultivars.
We showed that somatic editing frequency in the M 0 plants inoculated with the BSMV-sgRNA transcripts correlates with the frequency of heritable mutations in the M 1 generation. It appears that in the BSMV-sgRNA-based gene editing experiments we can  (Wang et al., 2019), and instead directly assess the sgRNA editing efficiency by sequencing the DNA extracted from the newly grown seedling leaves of the BSMV-sgRNA inoculated plants. Such early assessment of somatic editing efficiency can help to quickly adjust ongoing experiments. For example, in case of discovering that a critical sgRNA has low editing efficiency, the number of the BSMV-sgRNA inoculated plants could be increased to increase the heritable mutation recovery rate or sgRNA could be re-designed to improve its efficiency in the follow-up experiments.
A previous study (Li et al., 2021) show that the multiplex wheat genome editing is possible with the pooled BSMV-sgRNAs. Our results indicated that the level of multiplexing for achieving simultaneous editing of all targets could be limited. The likelihood of recovering mutations at multiple sites across a genome appears to be the product of the editing efficiency of individual targets. Probably this happens because the chance for the same plant cell to be infected by multiple BSMV particles with distinct guides is reduced with an increase in the level of multiplexing. Superinfection exclusion could be another factor contributing to the reduced efficiency of multiplexed editing by preventing secondary infection of plant cells already infected by one BSMV particle, thereby reducing the probability of getting a genome of the same cell edited by multiple gRNAs. For effective editing of multiple targets using the BSMV-sgRNA-based method, we should explore alternative strategies. One of the possible approaches is to create BSMV-sgRNA constructs expressing an array of multiple guide RNAs simultaneously. However, the level of multiplexing in this case will likely be limited by the restrictions on the size of a genome that could be packed into a viral particle. This could partially be overcome by arraying sgRNAs in a single viral vector because prior studies suggest that no processing signals are required to separate sgRNA units in the viral transcripts (Cody et al., 2017;Ellison et al., 2020). The usage of both the b and c chains of the BSMV genome for subcloning the sgRNAs or development of the BSMV-sgRNA system based on the Cas12a editors (Wang et al., 2021;Zetsche et al., 2015), which utilize much shorter guides, could also be used for expanding the multiplexing capacity of the BSMV-sgRNA-based editing system. The utility of CRISPR-induced cis-regulatory diversity for improving productivity traits was clearly demonstrated in tomato (Rodr ıguez-Leal et al., 2017). In our study, we showed that the multiplexed BSMV-sgRNA-based system could be an effective tool for creating deletions in the upstream regulatory region of the Q gene. The approach was highly effective and allowed us to quickly recover deletion variants in the next generation of the BSMV-sgRNA-inoculated wheat lines. The distribution of editable sites that could be targeted by Cas9 in the Q gene promoter did not allow us to create deletions near the transcription start site and recovered mutants did not show significant changes in the gene expression level or phenotype. Nevertheless, our results demonstrate that this editing strategy opens opportunities for exploring novel cis-regulatory diversity in the wheat genome and modulating the regulatory effects of transcription factors controlling agronomic traits. One of the shortcomings of pooling multiple BSMV-sgRNAs was the reduction in the efficiency of individual target editing, which resulted in the relatively low frequency of any given deletion variant. To increase the chance of recovering specific deletions using the BSMV-sgRNA-based system, the reduced levels of multiplexing should be considered (e.g. pooling only two BSMV-sgRNAs targeting sites flanking the region of interest).
Alternatively, as was discussed above, constructs incorporating multiple guides into a single viral transcript need to be designed (Cody et al., 2017;Ellison et al., 2020;Wang et al., 2021;Zetsche et al., 2015).
We showed that introgression of the high-expressing Cas9 locus into a target cultivar can maintain the effective BSMV-sgRNA-induced genome editing, removing the cultivar-specific constraints on targeted genetic modification of the wheat genome associated with transformation-based approaches. Introgression of Cas9 locus and recovery of up to 98% recurrent parent genome in crosses could be quickly accomplished in three rounds of backcrossing using locus-specific and genome-wide markers (Hospital, 2003). As soon as the donors of highexpressing Cas9 lines and a virus-based CRISPR delivery system are available (Ellison et al., 2020;Li et al., 2021;Yin et al., 2015), a similar gene editing approach could be used for modifying the genomes of other crops. Contrary to approaches that are based on random genome-wide chemical mutagenesis of thousands of lines (Henry et al., 2014;Krasileva et al., 2017), these functional screens can be used to analyse the effects of mutations without the confounding impact of mutations in other parts of the genome. Considering the ease with which the BSMV-sgRNA constructs can be designed to target hundreds of targets and the simplicity of BSMV-sgRNA inoculation procedure, lines with the Cas9 introgression provide an opportunity for establishing the high-throughput functional screens for genetic studies aimed at identifying causal mutations or novel variants affecting agronomic traits.

Plasmid construction
The previously reported plasmids (Scofield et al., 2005) with the a, b, and c chains of the ND18 strain of barley stripe mosaic virus (BSMV), referred to as pBSMVa, pBSMVb and pBSMVc, were used in our study (Figure 1a). The virus-induced gene silencing (VIGS) constructs pBSMVcPDS previously developed for targeting the barley phytoene desaturase gene (HvPDS) was used in our study to silence TaPDS in wheat (Holzberg et al., 2002;Scofield et al., 2005) (Figure 1a). For delivering sgRNAs into plant cells using BSMV, the sgRNA sequences were inserted into the pBSMVcPDS plasmid after removing the PDS fragment by PacI and NotI digestion (New England BioLabs, Ipswich, MA, catalogue numbers R0547 and R0189). The sequences targeted in our study are listed in Table S2. All the primers, DNA oligos and doublestranded DNA fragments are synthesized by Integrated DNA Technologies (Coralville, IA) (Table S1). PCR reactions were performed using NEBNextâ High-Fidelity 2X PCR Master Mix (New England BioLabs, catalogue number M0541) following the manufacturer's protocol unless specified differently. For example, to insert the sgRNA for target QT1 into pBSMVcPDS, the sgRNA was amplified from the pU6sg-QT1 construct  using a primer pair targetVIGSsF and targetVIGSsR. The PCR fragments and the plasmid pBSMVcPDS were digested using PacI and NotI; the PCR fragment and the plasmid fragment without PDS gene were isolated from agarose gel using QIAquick Gel Extraction Kit (Qiagen, catalogue number 28706). The digested PCR products and the pBSMVc plasmid were ligated using T4 ligase (New England BioLabs, catalogue number M0202). This plasmid was named pBSMVcQT1 (Figure 1b). All inserted DNA fragments were validated by Sanger sequencing. All the Sanger sequencing in this study was done using BigDye TM Terminator To test the effect of mobile RNAs on BSMV-sgRNA-induced gene editing, the fragments including GW2T2-sgRNA and mobile RNAs were synthesized at Integrated DNA Technologies (Table S1). The mobile RNAs included the coding sequences of FT gene from Arabidopsis (Ellison et al., 2020) and its wheat ortholog Vrn3 (Yan et al., 2006), and wheat methionine and isoleucine tRNAs tRNA Met and tRNA Ile (Figure 2a). The GW2T2-sgRNA without mobile elements were amplified from the previously developed pBUN421-GW2T2 plasmid (Wang et al., 2018a) using the primer pair VIGS-GW2T2-sgRNA-F and VIGS-sgRNA-R. Both the synthesized DNA fragment and the PCR products were subcloned into PacI-and NotI-digested pBSMVcPDS using NEBuilderâ HiFi DNA Assembly Master Mix (New England BioLabs, catalogue number E2621) following the manufacturer's protocol.
The UPL3T11-sgRNA with flanking sequences overlapping with the PacI-and NotI-digested pBSMVcPDS (Table S1) were synthesized at Integrated DNA Technologies. The sgRNAs with flanking sequences for GW7T6, pQT17, pQT18, pQT23, pQT25, and pQT26 were obtained by amplifying the sgRNA scaffold of the pBUN421 plasmid (Xing et al., 2014) using the forward primers paired with reverse primer VIGS-sgRNA-R (Table S1). Both synthesized DNA fragments and PCR products were subcloned into PacI-and NotI-digested pBSMVcPDS using NEBuilderâ HiFi DNA Assembly Master Mix.

Sequencing of the Q gene promoter in cv. Bobwhite
Three pairs of primers (Table S1) targeting overlapping genomic regions were used to amplify the Q gene promoter in cv. Bobwhite. The PCR fragments of primer pair pQT4MiF -pQT17MiR and primer pair pQT17MiF -pQT25MiR were subcloned into pCRblunt using Zero Blunt TM TOPO TM PCR Cloning Kit (ThermoFisher Scientific, catalogue number 450245) following the manufacturer's protocol. For each subcloned fragment, three colonies were sequenced. The PCR product of primer pair pQT25checkF -Q5endR3 was sequenced directly without subcloning.

Plants and growth conditions
A transgenic line 7438 expressing previously reported wheatcodon-optimized Cas9 ( Figure S12) in the background of cv. Bobwhite  was utilized for the initial testing of BSMV-sgRNA-based gene editing. Later, to improve the BSMVbased gene editing efficiency, additional transgenic lines in the background of cv. Bobwhite was screened to identify plants with high levels of Cas9 expression (high-Cas9). The 707 and C413 lines expressing maize-codon-optimized Cas9 ( Figure S12) at a high level were used for further research ( Figure S12). The 707 line was used as a donor of the high-Cas9 allele for introgression into CIMMYT spring wheat 3 613 474 and Kansas winter wheat KS080093K-18. These two lines were crossed with 707, and then, Cas9-positive F 1 plants were selected by PCR using the primer pair zCas9-F and zCas9seq1 (Table S1) and backcrossed one and two times to KS080093K-18 and 3 613 474, respectively. The Cas9-positive lines from the backcrossed plants were used for BSMV-sgRNA inoculation.
All plants were grown in ¼ litter square pots filled with the Berger BM1 growing mix (Hummert, catalogue number 10120000) in a growth chamber at 24°C day/20°C night conditions with 16 h supplemental light. The plants were grown in ¼ litter square pots for two weeks after inoculation and then transplanted into 1 L square pots. The M 1 generation plants were grown in the 128-well plastic trays filled with the Berger BM1 nutrient retention soil.

In vitro transcription of BSMV-sgRNA constructs and plant inoculation
The a, b and c virus chains were transcribed in vitro from plasmids pBSMVa, pBSMVb, and pBSMVc linearized by digesting with MluI (New England BioLabs, catalogue number R3198S), SpeI (New England BioLabs, catalogue number R0133S), and MluI enzymes, respectively. The pBSMVcPDS plasmid and all plasmids expressing BSMV c chain carrying sgRNAs were linearized with BssHII (New England BioLabs, catalogue number R0199S). In vitro transcription was performed in 20 lL reaction volume using HiScribe TM T7 High Yield RNA Synthesis Kit (New England BioLabs, catalogue number E2040S) following the Capped RNA Synthesis protocol provided by the manufacturer. The m7G(5 0 )ppp(5 0 )G RNA Cap (New England BioLabs, catalogue number S1404L) was used as Cap Analog with 4:1 of Cap Analog:GTP ratio. The quality and concentration of RNA transcripts (usually within 2-2.5 lg/lL range) were assessed on the agarose gel.
The second leaf of wheat seedlings at the two-leaf stage was inoculated with the FES buffer, BSMV, BSMV-PDS, or BSMV-sgRNAs. Each plant was inoculated using the mixture of 60 lL FES buffer and transcription products of BSMV a, b and c chain (2.5 lL each). The FES buffer includes 7.51 g/L glycine, 10.45 g/L K 2 HPO 4 dibasic, 10 g/L sodium pyrophosphate decahydrate, 10 g/L bentonite, and 10 g/L celite. The inoculation was performed by hand-rubbing the 2nd leaf from the base to the tip while wearing clean nitrile gloves. The procedure was repeated 3 times for each plant, each time applying 20 lL of the mixture. Immediately after inoculation, the plants were covered by plastic bags to keep the moisture. Plastic bags were removed 5-7 days after inoculation. In each experiment, three groups of controls including FES buffer, wild-type BSMV, and BSMV-PDS were used. The virus infection symptoms were usually observed 7 days after inoculation, while the PDS RNAi phenotype appeared 12 days after the inoculation.

Genotyping of BSMV inoculated plants and screening of mutants in M1 generation
To detect target editing events, a 2 cm-long segment of the 4th leaf was sampled about two weeks after inoculation when BSMV-PDS-inoculated control plants showed photobleached leaf symptoms. DNA was isolated using the TPS buffer (100 mM Tris-HCl, 10 mM EDTA, 1 M KCl, pH8.0) as described (Wang et al., 2021). The gene editing efficiency in the M 0 and M 1 plants was assessed for each subgenome separately using the next-generation sequencing (NGS) approach, as previously described (Wang et al., 2018a). We used the following NGS read coverage thresholds to call the genotypes at the edited gene loci in M 1 plants. If the proportion of NGS reads with edited alleles was between 30% and 70%, the locus was defined as heterozygous. The homozygous genotype calls had more than 90% of NGS reads carrying either wild-type or mutated variants. Any M 1 plant was considered as edited if it had at least one subgenome carrying heterozygous or homozygous mutations.
The deletions in the Q gene promoter region were detected using various combinations of PCR primers (Figure 4a,e, showing evidence of deletions in the Q gene promoter are shown in Figure 4.

RNA isolation and qPCR
RNA isolation was performed using the TRIZOL reagent (Thermo-Fisher Scientific, catalogue number 15596026) following the manufacturer's protocol. The cDNA was obtained by reverse transcription using SuperScript TM III First-Strand Synthesis Super-Mix for qRT-PCR (ThermoFisher Scientific, catalogue number 11752050). The Cas9 expression in the transgenic plants was measured by qRT-PCR using the primer pair zCas9-F and zCas9seq1 and RNA isolated from the 2nd leaf of two-weekold seedlings. The Q gene expression in the M 1 plants was measured by isolating RNA from the 3rd and 6th leaf at the 6-leaf development stage. The samples were frozen in liquid nitrogen immediately and then stored under -80°C until RNA was isolated. The qPCR reaction was performed using the PowerUP SYBR Green Master Mix (ThermoFisher Scientific, catalogue number A25741) following the manufacturer's protocol. NEBNextâ High-Fidelity 2X PCR Master Mix (New England BioLabs, catalogue number M0541) and SybrGreen were used to assess the expression level of the Q gene on chromosome 5A. The specificity of the Q gene primers, Q5ArtF4 and Q5ArtR4 (Table S1), was validated using nullisomic-tetrasomic genetic stocks ( Figure S10). The TaActin gene expression was used as a reference.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Figure S1 The sequences of barley stripe mosaic virus (BSMV) plasmids used in the study. Figure S2 The expression level of Cas9 in transgenic plants. Figure S3 Somatic editing induced by the BSMV-sgRNA constructs with and without mobile RNA in the noninoculated 4th leaf of the C413 line. Figure S4 Relationship between the mutagenesis ratio in the M1 progeny of plants inoculated by BSMV-GW2T2 (a, b, and d) or BSMV-GW2T2 (c and e) and the somatic editing efficiency evaluated in the 6th (a), 4th (b and c), and 2nd (d and e) leaf. Figure S5 The somatic editing efficiency of each target site in the 4th leaf (developed after inoculation) of plants inoculated by multiplex editing pool (BSMV-GUG) or BSMV carrying single guide RNAs. Figure S6 The alignment of the Q gene promoter region from the A, B, and D genomes along with the CRISPR-Cas9 targets. Figure S7 Alignment of the Q gene (chromosome 5A) promoter from cultivars Chinese Spring and Bobwhite. Figure S8 Expression of the Q gene's A genome homoeolog in the M1 plants at the 6-leaf stage. Figure S9 The Q-5A gene expression level in spikes of M2 plants with edits in the Q-5A promoter. Figure S10 Validation of the A genome-specific primers for RT-PCR of the Q gene. Figure S11 The somatic editing efficiency of BSMV-QT1 and BSMV-GW7T6 in line 3 613 474 and the somatic editing efficiency of BSMV-GW7T6 in KS080093K-18. Figure S12 The Cas9 plasmids used for creating transgenic plants. Table S1 The primers, oligos, and synthesized double-strand DNA used in this study. Table S2 The target sites selected for BSMV-sgRNA-based editing in this study. Table S3 The efficiency of the Q gene (chromosome 5A) editing based on the BSMV-sgRNA delivery system in transgenic wheat lines with the low (7438) and high (C413) levels of Cas9 expression. Table S4 The efficiency of somatic editing in the wheat leaves inoculated with the BSMV transcripts carrying the TaGW2 gRNA with and without mobile elements.