CRISPR/Cas9-mediated mutagenesis of VvMLO3 results in enhanced resistance to powdery mildew in grapevine (Vitis vinifera)

Grapevine (Vitis vinifera), one of the most economically important fruit crops in the world, suffers significant yield losses from powdery mildew, a major fungal disease caused by Erysiphe necator. In addition to suppressing host immunity, phytopathogens modulate host proteins termed susceptibility (S) factors to promote their proliferation in plants. In this study, CRISPR/Cas9 (clustered regularly interspaced short palindromic repeats/CRISPR-associated 9) technology was used to enable the targeted mutagenesis of MLO (mildew resistance Locus O) family genes that are thought to serve as S factors for powdery mildew fungi. Small deletions or insertions were induced in one or both alleles of two grapevine MLO genes, VvMLO3 and VvMLO4, in the transgenic plantlets of the powdery mildew-susceptible cultivar Thompson Seedless. The editing efficiency achieved with different CRISPR/Cas9 constructs varied from 0 to 38.5%. Among the 20 VvMLO3/4-edited lines obtained, one was homozygous for a single mutation, three harbored biallelic mutations, seven were heterozygous for the mutations, and nine were chimeric, as indicated by the presence of more than two mutated alleles in each line. Six of the 20 VvMLO3/4-edited grapevine lines showed normal growth, while the remaining lines exhibited senescence-like chlorosis and necrosis. Importantly, four VvMLO3-edited lines showed enhanced resistance to powdery mildew, which was associated with host cell death, cell wall apposition (CWA) and H2O2 accumulation. Taken together, our results demonstrate that CRISPR/Cas9 genome-editing technology can be successfully used to induce targeted mutations in genes of interest to improve traits of economic importance, such as disease resistance in grapevines.


Introduction
Grapevine is one of the most extensively cultivated and economically valuable horticultural crops in the world, with 7.6 million hectares in production and an annual value of~$3.6 billion in wine export markets alone (Organisation Internationale de la Vigne et du Vin, 2017). Unfortunately, grape production is limited by various diseases, including powdery mildew caused by the fungus Erysiphe necator. Powdery mildew can be controlled to a certain extent by fungicides, but this increases production costs and can negatively affect the environment and human health. In addition, the indiscriminate use of pesticides increases carbon emissions 1 .
A more cost-effective and environmentally friendly strategy for limiting losses due to diseases is the development of disease-resistant cultivars through traditional breeding and/or precision breeding with the aid of biotechnology. The identification and characterization of the genes and genetic pathways that confer resistance can greatly facilitate the latter approach. Advances in functional genomics are changing the preferred experimental strategy from using traditional forward genetics (i.e., from phenotype to genotype) to employing both forward genetics and reverse genetics (i.e., inferring gene function based on the phenotypes caused by mutations of a target gene) 2 . The introduction of functional mutations through targeted genome editing is a powerful reverse genetics approach for understanding gene function. CRISPR-Cas9 technology is a relatively simple yet effective technique that can be exploited for gene mutation, the repression/ activation of gene expression, subtle modification of gene function and epigenome editing 3 . In 2013, three research groups simultaneously reported the use of the CRISPR-Cas9 system for targeted genome modification in Arabidopsis thaliana (Arabidopsis), tobacco, rice and wheat for the first time [4][5][6] . This system has also been successfully applied in maize 7 , cotton 8 , and other plants. Grapevine is a woody perennial plant with a highly heterozygous genome that can be genetically transformed only through lengthy and difficult procedures. To date, there have been only a few reports on the use of CRISPR-Cas9 for genome editing in grapevine 5,[9][10][11] .
Over the past 25 years, the plant immune system has been extensively studied, and fundamental discoveries have led to a better understanding of the molecular basis of plant disease resistance. While the development of disease-resistant crop varieties has typically relied on the introduction of dominant resistance (R) genes into elite cultivars via classical breeding, recessive mutations in host susceptibility (S) genes have also been utilized to confer resistance. R-gene-induced defense responses often result in localized cell death or a hypersensitive response (HR) 12 ; by contrast, the loss of Sgenes required for the successful invasion of a particular pathogen may not display strong defense activation. For example, the disruption of S-genes such as PMR5 13 , DMR6 14,15 , LOB1 16 , and SWEET14 17,18 can confer broadspectrum disease resistance with no or minimal activation of defenses in several economically important plant species. Sgene-mediated resistance is generally more durable than Rgene-based resistance because the pathogen must overcome dependence on a host S factor 15 . The disruption of the MLO gene in barley confers durable powdery mildew resistance, as demonstrated by the fact that no powdery mildew strains capable of overcoming mlo-mediated resistance have been found in the field for many decades 19 .
Studies suggest that the pathogenicity of adapted powdery mildew species relies on a functional host MLO gene 20,21 . MLO proteins constitute a family of highly conserved 22 , plant-specific, seven-transmembrane domain proteins that are structurally similar to G protein-coupled receptors (GPCRs) with a calmodulin-binding domain 23 . The mechanism by which MLO proteins act as powdery mildew susceptibility factors is unknown, and the biochemical function of MLO proteins in the host remains largely unexplored 24 . In barley, an mlo mutation was found to confer broad-spectrum and durable resistance to barley powdery mildew isolates of Blumeria graminis f. sp. hordei 25 , but a range of developmentally controlled pleiotropic effects occurred in the host, such as the formation of callose-containing cell wall appositions (papillae) in the absence of any pathogen and the premature onset of leaf senescence. These effects reduced the grain yield of the mlo mutants; thus, the dysfunctional MLO proteins may indeed result in a competitive disadvantage in natural populations 26 . In bread wheat, TALEN-induced mutations in all three TaMLO homologs confer heritable broad-spectrum resistance to powdery mildew 27 . In apple, the suppression of MdMLO11 and MdMLO19 through RNA interference (RNAi) reduces susceptibility to powdery mildew (Podosphaera leucotricha) 28 . In Arabidopsis, the Atmlo2 Atmlo6 Atmlo12 (Atmlo2/6/12) triple mutant is completely resistant to powdery mildew 21 . The grapevine genes VvMLO3, VvMLO4, and VvMLO17, orthologous to the AtMLO2, AtMLO6, and AtMLO12 genes, respectively 29 , are reported to exhibit rapid induction upon powdery mildew infection, suggesting that these VvMLO proteins may play a role in modulating antifungal defense responses in grapevine. One previous study used RNAi to suppress specific VvMLO genes and found that the simultaneous knockdown of VvMLO6 and VvMLO7 in grapevine reduced susceptibility to powdery mildew 30 .
The grapevine cultivar Thompson Seedless is an ideal material for studying powdery mildew resistance in grapevine because it is amenable to transformation and susceptible to powdery mildew infection. There are 17 VvMLOs in Thompson Seedless. In this study, we used CRISPR/Cas9 technology to target two grapevine MLO genes, VvMLO3 and VvMLO4, and obtained VvMLO3edited grapevine lines with enhanced resistance to powdery mildew. Our work suggests that CRISPR/Cas9-targeted mutagenesis is a valuable tool for generating disease-resistant grapevines.

Target selection within MLOs and vector construction
For each VvMLO gene, two target loci were selected (Fig.  1a). Schematic representations of the target sites of two small guide (sg) RNAs and their gene expression cassettes used for the Agrobacterium-mediated transformation of grapevine embryogenic calli is shown in Fig. 1a, b. The sgRNA intermediate vectors are based on the pUC18 backbone described previously 31 . The binary vector for plant transformation contains a plant codon-optimized Cas9 gene. In total, four DNA constructs, designated CM3G1 (Cas9/sgRNA1-MLO3-Grapevine), CM3G2, and CM4G3 (Cas9/sgRNA3-MLO4-Grapevine) and CM4G4, were produced and used for transformation (Fig. 1b, c).

Targeted mutagenesis in regenerated transgenic grapevines
In total, 210 plantlets were regenerated from the grapevine anther filament-induced embryogenic calli subjected to coincubation with Agrobacterium cells (Fig. 2). To verify the presence of the CRISPR-Cas9 sequence, genomic DNA from each regenerated plantlet was analyzed by PCR. As shown in Fig. 3a, a fragment of 508 bp was amplified from Cas9-positive plantlets, whereas Cas9-negative and nontransgenic control plantlets showed no amplification. A total of 125 plantlets were identified as Cas9 positive for any of the four targets of the two VvMLO genes, and these plantlets were further analyzed for the presence of the intended sequence variation. A total of 68 independent PCR-positive plantlets for VvMLO3-sgRNA1, 39 for VvMLO3-sgRNA2, 13 for VvMLO4-sgRNA3, and 5 for VvMLO4-sgRNA4 were analyzed by Sanger sequencing ( Table 1). The corresponding DNA target sites of sgRNA1, sgRNA2, or sgRNA3 contained mutations ( Table 1), indicating that three of the four CRISPR/Cas9-DNA constructs (i.e., CM3G1, CM3G2, CM4G3) enabled successful gene editing in the regenerated grapevine plantlets. Different target sites were mutated with different editing efficiencies, varying from 0 to 38.5% (Table 1). Singlenucleotide insertions and nucleotide deletions of various lengths were identified among these edited lines (Fig. 3b). The mutations detected in the sgRNA1 lines were mainly small deletions and insertions, while those of the sgRNA2 lines were mainly small insertions, and those of the sgRNA3 lines were mainly small insertions ( Fig. 3b and Table S1). Most of the insertions consisted of only 1 bp, and the majority of deletions were short in the sgRNA1 and sgRNA2 lines. Double-strand breaks (DSBs) usually occur at a position three base pairs upstream of the PAM sequence 32 . Among the 20 VvMLO-edited lines, 16 showed mutation sites exactly at the 4th base from the PAM site ( Fig. S1 and Table S1).
Four different types of mutation sequence chromatograms are shown in Fig. 3c. These 20 edited lines were detected among the 125 transgenic plantlets that were obtained and analyzed (Table 1); thus, the overall geneediting efficiency was 16.5%. It was interesting that no edited lines were detected among any of the five transgenic lines containing the CM4G4 construct. Among the 20 edited lines, one (5%) was homozygous for a 1-bp insertion, three (15%) were biallelic for small deletions and insertions (Indels), seven (35%) were heterozygous (i.e., only one allele contained a mutation), and nine (45%) were chimeric (i.e., more than two mutated alleles were detected in one line) (Fig. 3d). Detailed information about the mutation found in one edited line, CM3G2-12, is provided as an example in Fig. 3e; both alleles of VvMLO3 in this line contained a single-nucleotide (A) insertion in the target site, resulting in the formation of a premature stop codon, which in turn was predicted to produce a VvMLO3 mutant with a C-terminal truncation.

Phenotypic characterization of genome-edited lines
The vast majority (93%) of the regenerants obtained after Agrobacterium infection in this study showed abnormal growth and development (not shown). When grown under axenic conditions, 14 of 20 3-month-old VvMLO3-or VvMLO4-edited plantlets (e.g., CM3G1-65 and CM3G2-25; Fig. 4a) showed similar phenotypes, which were characterized by senescence-like chlorosis and necrosis. Among these 14 plantlets, CM3G2-12, homozygous for a 1-bp insertion, died in the medium when it was two months old. However, the nonedited plants displayed normal growth and development during the same period.
We found that when the VvMLO-edited plantlets were subcultured to induce proliferation, their rooting potential was very low, even in the presence of IBA at a concentration as high as 1.0 mg·L −1 . Only a few lines (i.e., CM3G2-30 and CM3G1-51) showed normal rooting (Fig. 4b). Compared with the nontransgenic control, the growth of the VvMLO-edited plantlets was poor, and the lower part of the plantlets showed accelerated defoliation.
Powdery mildew-triggered mesophyll cell death, cell wall apposition and H 2 O 2 production are associated with mlomediated resistance To evaluate the effect of targeted mutations in the two VvMLO genes on resistance to powdery mildew, two VvMLO3-edited lines, CM3G2-30 and CM3G1-51 (sequencing chromatograms are shown in Fig. 5a), and one nontransgenic control plant were transplanted to pots (Fig. 5b), and detached leaves of these plants were inoculated with the powdery mildew isolate En. NAFU1 33 15 days after transplantation. Leaves inoculated with En. NAFU1 were collected at 24 and 72 hpi and stained  with trypan blue for the visualization of fungal structures by microscopy. At 24 hpi, both the VvMLO3-edited and nontransgenic lines produced appressoria with no apparent differences. However, at 72 hpi, the edited line showed much less fungal growth compared with the control line (Fig. 5c), with only~1/4 of the total hyphal length per colony being observed on CM3G2-30 and~1/ 10 on CM3G1-51 compared with that in the control line (P < 0.01) (Fig. 5d). In addition, there were fewer secondary hyphae on the two VvMLO3-edited lines compared with the nontransgenic control (not shown), indicating that both edited lines exhibited enhanced resistance to powdery mildew. Unfortunately, all five VvMLO4-edited lines produced deformed leaves and died. Thus, it remains to be determined whether VvMLO4 plays a role in powdery mildew pathogenesis. The potted VvMLO3-edited plants were transferred to a greenhouse for further growth and evaluation (Fig. 6a). In the edited  lines harboring heterozygous mutations in VvMLO3, clear powdery mildew-induced cell death was observed (Fig. 6b). Careful examination of the trypan blue-stained leaf sections revealed that in many cases, the mesophyll cells underneath the infected epidermal cells were stained blue (Fig. 6c), suggesting that signals from mildewinvaded epidermal cells trigger the collapse of neighboring mesophyll cells as a result of the mutation of VvMLO3 in these heterozygous lines.
In a previous report, increased cell wall apposition (CWA) or papilla formation was observed beneath the site of attempted penetration in mlo mutant plants 21 . We used aniline blue staining to check for the presence of callose in the CWAs after the inoculation of powdery mildew onto the detached leaves of VvMLO3-edited lines. At 7 dpi, clear CWAs were observed in the heterozygous mutant line CM3G2-40 but not in nontransgenic control plants (Fig. 6d). In addition, H 2 O 2 production and accumulation visualized as brownish precipitates upon 3,3′-diaminobenzidine (DAB) staining were more pronounced in mildew-infected cells of the heterozygous mutant lines CM3G1-91 and CM3G2-40 at 7 dpi ( Fig. 6b and Fig. S3a). The quantification of H 2 O 2 accumulation using ZEN 2012 software showed that~22% and~42% of the infected leaf epidermal cells of the two VvMLO-edited plants exhibited H 2 O 2 accumulation, which was three to six times higher than that (6.5%) in the infected leaf epidermal cells of the wild-type plants (Fig. S2a, b). Taken together, our data suggest that Vvmlo3-mediated resistance to powdery mildew is associated with callose deposition and H 2 O 2 accumulation in powdery mildewinfected epidermal cells and possibly the demise of mesophyll cells underneath.
To assess the impact of the VvMLO3 gene mutation on whole-plant resistance to powdery mildew, two independent experiments were carried out. Potted transgenic or nontransgenic control plants were inoculated with fresh conidia of powdery mildew En. NAFU1. This strain was able to infect both the WT control and heterozygous mutant lines CM3G1-91 and CM3G2-40 (Fig. 7a). However, it was noted that the leaves of the edited lines, but not those of the control line, were covered with abundant white velvety fungus at 15 dpi, which seemed to correlate with the presence of necrotic lesions in the infected leaf tissue in the edited lines but not in the control line at 15 dpi (Fig. 7b). The quantification of the number of spores per mg of fresh infected leaf tissue at 20 dpi revealed that there was an~2-fold reduction in fungal sporulation in the edited line CM3G1-91 in comparison with the WT line, while a small, yet still statistically significant reduction was detected in the other edited line, CM3G2-40 (Fig. 7c).

Discussion
Since the first report on the application of CRISPR/Cas9 technology for the editing of grapevine genes in 2016 34 , two more recent studies have explored the use of this new genome-editing technology in grapevine using protoplasts 11 or embryogenic callus cells 9 . Two other examples of successful targeted mutagenesis using the CRISPR/ Cas9 system in grapevine resulted from studies conducted for the functional characterization of the WRKY52 gene 10 and the VvCCD8 gene 35 . In this study, two grapevine MLO genes, VvMLO3 and VvMLO4, were edited by using CRISPR/Cas9 technology, and the reduction of VvMLO3 expression by mutating one allele significantly improved the resistance of a susceptible grapevine cultivar against the powdery mildew isolate En. NAFU1.
Charrier et al. performed targeted mutagenesis in apple and pear using the CRISPR-Cas9 system, and they classified the obtained mutants as homozygous, heterozygous, biallelic, and chimeric 36 . According to their classification, in a "homozygous" line, both alleles of the target gene contain the same mutation; in a "heterozygous" line, only one allele is mutated; in a "biallelic" line, both alleles are mutated, but the mutations are not identical; and finally, in a "chimeric" line, more than two alleles are detected, which is indicative of the presence of cell lineages derived from different edited cells during somatic embryogenesis. In this study, all four abovementioned types of grapevine mutants were detected, and the "chimeric" type accounted for 45% of the mutants obtained, suggesting that CRISPR/ Cas9 is enzymatically active during the early stage of embryogenesis initiated from a single transformed somatic cell.
As a woody perennial fruit crop, grapevine is known to be recalcitrant to genetic transformation. To increase the chance of success, we designed two sgRNAs for each of the two MLO target genes, one of which targeted a site near the initiation codon, while the other targeted a site in the middle of the coding sequence for each gene (Fig. 1a).
This strategy was found to be effective for VvMLO3, as mutations were identified in two target sites. For VvMLO4, sgRNA3 (CM4G3) achieved an~23% editing  Black arrows indicate leaf cell necrotic lesions, while red arrows denote clustered dead or dying mesophyll cells stained by trypan blue. c Quantification of the number of spores per mg of fresh leaf tissue in WT and VvMLO3-edited grapevines at 20 dpi. Each data point is the mean of three independent experiments (±standard deviation); asterisks indicate significant differences in comparison with the WT as determined by Student's t test (*P < 0.05, **P < 0.01).
efficiency, but sgRNA4 (CM4G4) failed to produce a single mutant line ( Table 1). The latter situation may simply be due to the insufficient number (eight) of transgenic lines generated or to the presence of unusual local chromatin structures that are less accessible by CRISPR-Cas9 37 . The DNA mutations induced by CRISPR/Cas9 in plants mainly consist of short indels and replacements that occur during the process of DSB repair via nonhomologous end joining (NHEJ) 38 . Such diverse mutations are expected to occur during NHEJ-mediated DNA repair. This mutation pattern was also observed in this study, as most of the mutations induced in VvMLO3 were indels. Among the 20 edited lines obtained, only one was homozygous for a single 1-bp insertion, while all other lines showed a combination of different types of mutations (Fig. 3c).
MLO homologs have been subjected to targeted mutagenesis in many plants for the engineering of powdery mildew resistance 20,21,27,39,40 . The loss of AtMLO2, but not AtMLO6 or AtMLO12, in Arabidopsis thaliana enhances the resistance of the mutant plants to powdery mildew, while the loss of all three genes results in complete resistance 21 . These findings indicate that while AtMLO2 is the major host S protein required for powdery mildew infection, AtMLO6 and AtMLO12 possess functions that overlap with those of AtMLO2. In this study, VvMLO3 and VvMLO4, from the clade corresponding to AtMLO2, 6, and 12 in the phylogenetic tree 29 (Fig. S3), were selected for CRISPR/Cas9-mediated mutagenesis. Unfortunately, the only Vvmlo3 homozygous mutant plantlet in which VvMLO3 was knocked out died of massive leaf necrosis in the tissue culture medium (data not shown). This implies that either the loss of VvMLO3 or a genetic mutation due to the T-DNA insertion might have triggered lethal cell death in this Vvmlo3 mutant. Nevertheless, although multiple VvMLO3-edited lines were heterozygous lines, qRT-PCR revealed that there was an~4-8-fold reduction in the VvMLO3 expression level in the edited lines CM3G1-91 and CM3G2-40 in comparison with the WT line (Fig. S4), which indicated that the VvMLO3 expression level was indeed negatively impacted by such editing. Significantly enhanced mildew resistance was observed in multiple VvMLO3 heterozygous lines in comparison with the WT plants from 2 to 7 dpi (Figs. 5 and 6), suggesting that VvMLO3 is one of the functional homologs of AtMLO2 in grapevine. Taken together, our results suggest that VvMLO3 is an important (though not the only) host susceptibility gene required for powdery mildew infection. It should note that the nomenclature for grapevine MLO genes differed in two previous studies by Feechan 29 and Winterhagen 41 that were published around the same time. In the present study, the naming of VvMLO3 and VvMLO4 followed the Feechan nomenclature 29 , whereas these genes were named VvMLO11(W) and VvMLO13(W), respectively, according to Winterhagen 41 (for clarity, when we describe MLO genes using the Winterhagen nomenclature, we use VvMLO6 (W) and VvMLO7 (W)) (Fig. S3). Previous studies showed that VvMLO6 (W) and VvMLO7 (W), rather than VvMLO11 (W) or VvMLO13 (W) (i.e., VvMLO3 and VvMLO4 according to the Feechan nomenclature), may negatively regulate disease resistance 29,42 , and the silencing of the VvMLO6 (W) and VvMLO7 (W) genes resulted in enhanced resistance to powdery mildew 30 . Hence, it is likely that VvMLO6/7 (W) may also be a functional homolog of AtMLO2. Simultaneous targeted mutagenesis of all three grapevine MLO genes may therefore be necessary to develop complete mildew resistance in grapevine in the future.
In addition to powdery mildew resistance, mlo mutant plants display premature leaf chlorosis and a reduced grain yield in barley 19 , increased callose deposition in the cell wall of leaves in pathogen-free plants and slower growth and development in A. thaliana 21,43,44 , and reduced plant size in pepper 45 . Not surprisingly, similar growth phenotypes, such as early leaf senescence, were observed in our VvMLO3 heterozygous mutant grapevine plantlets (Fig.  4a). Interestingly, mildew-induced mesophyll cell death was observed in the VvMLO3 heterozygous mutant lines (Fig. 6), which, along with H 2 O 2 accumulation and callose deposition in mildew-infected epidermal cells, seemed to be associated with the increased resistance of these mutant lines to powdery mildew (Fig. 6). In this regard, it is worth pointing out that no obvious pleiotropic phenotypes were observed when VvMLO6, 7, 11, or 13 (W) was knocked down by RNAi 30 . This implies that VvMLO6, 7, and 13 (W) may be functionally distinct from VvMLO3 and/or that these four genes are functionally redundant or were not adequately silenced.

Conclusion
CRISPR/Cas9-enabled targeted mutagenesis suggested that VvMLO3 is an S-gene required for the powdery mildew infection of grapevine and that the targeted mutation of VvMLO3 in a grapevine cultivar results in enhanced resistance to powdery mildew. This demonstrates that CRISPR/Cas9-targeted mutagenesis can be used to develop disease-resistant cultivars and facilitate the functional characterization of genes of interest in grapevine.

Cloning of grapevine MLO genes
Total RNA was extracted from Thompson Seedless leaves using the E.Z.N.A. Plant RNA Kit (Omega, Guangzhou, China) according to the manufacturer's instructions. First-strand cDNA was generated from 1.5 μg of total RNA using PrimeScript RTase (Takara Bio Inc., Dalian, China). The synthesized cDNA product was diluted 10-fold and used as a template for the cloning of MLO genes. Two VvMLO genes, VvMLO3 and VvMLO4, identified in a previous study 29 were selectively amplified using oligonucleotide primers derived from VvMLO gene sequences available in the National Center for Biotechnology Information (NCBI) database (https://blast. ncbi.nlm.nih.gov/Blast.cgi). PrimeSTAR HS DNA Polymerase was used for amplification. The PCR products were cloned into the pMD18-T vector (Takara Bio Inc., Dalian, China) and sequenced at the Beijing AuGCT Biotech, Yangling, Sequencing Department. The sequenced DNAs were aligned with the genome of Vitis vinifera PN40024.

CRISPR-Cas9 binary construct design
Target sequences were selected within two VvMLO exons, and four sgRNAs, designated sgRNA1 to sgRNA4, were designed within the two coding regions (Table S3). To increase the editing success rate for each gene, we selected two target loci in each MLO gene, seeking loci with a moderate GC content. We screened the target sequences to minimize potential off-target alignments using the online software Grape-CRISPR (http://biodb. sdau.edu.cn/gc/) and CRISPR-P (http://crispr.hzau.edu. cn/CRISPR2/) 47 . For each gene, we located one target site near the start codon and another in a region corresponding to the functional domain. Next, we performed a BLAST-P search based on the National Center for Biotechnology Information (NCBI) database (https://blast. ncbi.nlm.nih.gov/Blast.cgi) of the target sequences (including PAM) against the grapevine genome sequence to ensure that the target sequence was as specific as possible. We avoided sequences containing more than four consecutive Ts to prevent RNA Pol III from using such sites as a transcription termination signal. (The selected target sequences are shown in Fig. 1a.) Target adapter oligos were synthesized for each target locus using Ao ke ding sheng (http://www.augct.com/? ClassID=2) in the Yangling department (Table S2).
The binary pYLCRIPSR/Cas9-N 48 vector and two sgRNA cassettes driven by the AtU3b and AtU6-1 promoters were used to generate vectors expressing CRISPR/ Cas9-MLOs. For each sgRNA, a pair of DNA oligonucleotides was synthesized by Beijing AuGCT Biotech and annealed to generate dimers, which were subsequently integrated upstream of the sgRNA scaffolds in the plasmid vector. Two sequential rounds of overlapping PCR were carried out to insert double-stranded, gRNA spacer DNA sequences between double Bsa I restriction enzyme sites to generate intermediate vectors with one or two target sites. Intermediate sgRNA expression cassettes with target sequences were generated by overlapping PCR (Figs. S5 and S6). We used a binary CRISPR/Cas9 vector, pYL-CRISPR/Cas9P35S-N containing the CaMV 35S promoter and the NPTII gene as a selectable marker. The Cas9 vector was built in the backbone of pCAMBIA1300. The final purified PCR products were assembled into the binary pYLCRIPSR/Cas9 vector containing the NPTII gene and 2 × 35 s promoter by Golden Gate ligation 49 . The ligated products with one or two sgRNA expression cassettes were directly used to transform DH5α E. coli competent cells.

Agrobacterium-mediated transformation of grapevine
Transformation was carried out via the Agrobacteriummediated transformation system as previously described 50 with minor modifications. Agrobacterium tumefaciens strain GV 3101 was used for transformation. Transformation was carried out in several batches to avoid failure. The bacterial solution (OD600, 0.6-0.8) was poured into a sterile flask, and the grapevine tissue was soaked in the flask for 15-20 min with gentle shaking. The tissue was then placed on two layers of sterile filter paper to remove excess bacterial culture and transferred to 2 ml of liquid MS medium (containing 20 mg·L −1 AS and 3% sucrose) for cocultivation. After 2 days of cocultivation at 25°C in the dark, the material was further maintained for 1 month without any antibiotics before screening to allow the transformed material to recover. The tissue was then washed three times in sterile water and transferred to KBN delayed-screening medium (

Genotyping and mutant verification
The genotyping of transgenic lines was carried out essentially as described previously 31 . The genomic DNA of each line was isolated from 0.5 to 1.0 g of regenerated plantlet tissue using a CTAB-based method 51 . To verify the presence of the engineered sequences, specific oligonucleotide primers (SP-L-35S and SP-R) were used to amplify the CRISPR/Cas9 sequence (Table S2). The PCR products (ca. 400-600 bp) were sequenced directly using internal specific primers, for which the ideal binding positions are~150-250 bp upstream of the target sites (Beijing AuGCT Biotech Yangling Sequencing Department). The detection of target gene editing requires the sequencing of multiple clones of the PCR amplicons derived from the same genomic DNA sample. The PCR amplicons were cloned into a pMD18-T plasmid vector, and five to ten randomly selected E. coli colonies were used for plasmid preparation and insert sequencing. SnapGene Viewer software was used for mutation analysis. The sequence results are shown in Table S1.

Pathogen inoculation and evaluation of resistance to powdery mildew
Four MLO-edited lines, CM3G1-51 and -91 and CM3G2-30 and -40, were assessed for resistance to powdery mildew. The transgenic mutant and nontransgenic control lines were inoculated with the powdery mildew isolate En NAFU1 as described previously 33 . For detached leaf inoculation, CM3G2-30, CM3G1-51, and wild-type (WT) plants were used at 15 days after transplantation from the subculture medium. Fully expanded leaves at the third position from the shoot tip were chosen for inoculation. Representative micrographs and hyphal development were observed at 24 and 72 hpi, respectively.
For plant leaf inoculation, CM3G2-40, CM3G1-91, and wild-type (WT) plants were grown under phytotron conditions for 6 months for propagation and then subjected to inoculation. DAB and trypan blue-stained epidermal cells of WT, CM3G2-40 and CM3G1-91 plants were observed at 5 or 7 dpi. To assess the impact of the VvMLO3 gene mutation on whole-plant resistance to powdery mildew, plants of the transgenic lines CM3G2-40 and CM3G1-91 or nontransgenic control plants that had been grown in pots for six months were inoculated with fresh conidia of En. NAFU1 powdery mildew, and the detection of cell death was performed by trypan blue staining in the leaves of CM3G2-40 or CM3G1-91 plants at 15 dpi. Three individuals from each CM3G2-40 and CM3G1-91 line were evaluated. Symptoms of powdery mildew infection were evaluated after inoculation using an Olympus BX51 microscope under visible light. DAB staining, trypan blue staining and aniline blue staining were carried out as previously described to detect the accumulation of H 2 O 2 , fungal structures, cell death and CWAs 52 .

Statistical analysis
Disease severity data were analyzed with ZEN 2012 and SigmaPlot 10.0 software. All data were analyzed using paired Student's t-tests (http://www.physics.csbsju.edu/ stats/). The mean values ± standard deviation of the mean (SD) were calculated based on the results of at least three replicates, and significant differences compared with controls are represented by *p < 0.05 and **p < 0.01.