Optimizing the CRISPR/Cas9 system for genome editing in grape by using grape promoters

The efficacy of the CRISPR/Cas9 system in grapevine (Vitis vinifera L.) has been documented, but the optimization of this system, as well as CRISPR/Cas9-mediated multiplex genome editing, has not been explored in this species. Herein, we identified four VvU3 and VvU6 promoters and two ubiquitin (UBQ) promoters in grapevine and demonstrated that the use of the identified VvU3/U6 and UBQ2 promoters could significantly increase the editing efficiency in grape by improving the expression of sgRNA and Cas9, respectively. Furthermore, we conducted multiplex genome editing using the optimized CRISPR/Cas9 vector that contained the conventional multiple sgRNA expression cassettes or the polycistronic tRNA-sgRNA cassette (PTG) by targeting the sugar-related tonoplastic monosaccharide transporter (TMT) family members TMT1 and TMT2, and the overall editing efficiencies were higher than 10%. The simultaneous editing of TMT1 and TMT2 resulted in reduced sugar levels, which indicated the role of these two genes in sugar accumulation in grapes. Moreover, the activities of the VvU3, VvU6, and UBQ2 promoters in tobacco genome editing were demonstrated by editing the phytoene desaturase (PDS) gene in Nicotiana benthamiana leaves. Our study provides materials for the optimization of the CRISPR/Cas9 system. To our knowledge, our simultaneous editing of the grape TMT family genes TMT1 and TMT2 constitutes the first example of multiplex genome editing in grape. The multiplex editing systems described in this manuscript expand the toolbox of grape genome editing, which would facilitate basic research and molecular breeding in grapevine.


Introduction
The advent of the clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) system has been revolutionizing genome editing and genetic therapy 1,2 . In recent years, CRISPR/Cas9 has been more rapidly and widely applied in numerous plant species due to its simplicity, high efficiency, and versatility compared with previous genome editing technologies [3][4][5] . The most popular CRISPR/Cas9 system was initially derived from the adaptive immune system in Streptococcus pyogenes 6 , and an engineered system in which a single guide RNA (sgRNA) and the Cas9 nuclease are needed for genome editing 1 . The Cas9 nuclease can produce double-stranded DNA breaks, whereas the sgRNA specifically directs the Cas9 protein to its complementary DNA target site through RNA-DNA base pairing 1 . The expression of sgRNA and Cas9 has been shown to influence the editing efficiency of the CRISPR/ Cas9 system [7][8][9][10] .
In the CRISPR/Cas9 system, the expression of Cas9 is generally driven by an RNA polymerase II (Pol II) promoter. During the early applications of the CRISPR/Cas9 system in plants, the CaMV35S promoter was often used to express the Cas9 gene [10][11][12] . Recently, plant ubiquitin (UBQ) promoters have been isolated and employed in the CRISPR/ Cas9 system in place of the 35S promoter [13][14][15] . The Pol III promoters of small nuclear RNA (snRNA) genes, such as U3 and U6, are commonly used to drive sgRNA expression in plants and animals 16,17 . U6 snRNAs participate in the intron splicing of pre-mRNA in the nucleus 18 , whereas U3 snRNAs are involved in pre-rRNA processing 19,20 . Unlike other Pol III-transcribed genes that contain intragenic promoter elements, U6 snRNA genes only possess a conserved TATA box and the upstream sequence element (USE) in their promoters 18,21 . In contrast, the promoter structures of U3 snRNA genes vary among different species, particularly metazoans 20 . In higher plants, however, a canonical TATA-like box and well-conserved proximal USE are present in the U3 promoters 20,21 . Both of these promoters are capable of producing high levels of sgRNA, which typically have a length of~200 nucleotides 22 . In general, the Arabidopsis U6 (AtU6) and AtU3 promoters are extensively used in CRISPR/Cas9 vectors for genome editing in dicot plants, whereas the rice (Oryza sativa) U6 (OsU6) and OsU3 promoters are primarily applied in genome engineering in monocot plants 16,23 . Many CRISPR/ Cas9 toolboxes have been developed and optimized based on the Pol III promoters isolated from the two model plant species 16,21,23 . Recently, plant species-specific U6 promoters for CRISPR/Cas9-mediated genome editing were adopted in soybean 8 , cotton 10,24 , apple 25 , and wheat 26 , and these studies revealed that the use of plant species-specific Pol III promoters could contribute to increased sgRNA levels and thus, result in enhanced editing efficiencies [8][9][10] . However, the identification and application of plant species-specific Pol III promoters have not been fully explored, and U3 and U6 promoters in other plant species are expected to be characterized during the development and optimization of CRISPR/Cas9 systems for these species. As a fruit crop that is widely cultivated worldwide, grape (Vitis ssp.) is economically important. CRISPR/Cas9-mediated genome editing in grape (Vitis vinifera L.) was first reported in 2016 (ref. 27 ), and since then, the AtU6 or AtU3 promoters have been used to drive the expression of sgRNAs during grape genome editing [27][28][29][30][31] . Nonetheless, the grape Pol III promoter has not yet been identified.
Multiplex genome editing can be achieved by stacking multiple sgRNAs in a single CRISPR vector 13 . The simplest and most common approach involves the use of multiple U3 and/or U6 promoters for the expression of different sgRNAs 32 . Additionally, approaches based on self-cleaving ribozymes (RZ), tRNA, or Csy4 have also been used to produce multiple sgRNAs 33,34 . Multiplex genome editing, which enables the simultaneous targeting of several related or unrelated genes, has been applied in many plant species 35 . In this method, different alleles of the same gene, homoeoalleles in polyploid plants, or even a gene family could be simultaneously edited through multiplex gene editing 35,36 . The greatest potential of multiplex genome editing has been shown by editing different genes involved in the control of distinct traits in plants. For instance, the CRISPR/Cas9-mediated targeting of multiple genes that control the time of flowering and plant morphology can accelerate the domestication of wild tomato or directly develop customized tomato for urban agriculture 37,38 . In recent years, an increasing number of reported studies have implemented targeted genome editing in grape, and almost all the studies were performed with one or several sgRNAs targeting single genes [27][28][29][30][31]39 . The simultaneous editing of different genes of interest could improve different traits of grapes, which would be extremely significant for shortening the period of grape breeding. However, CRISPR/Cas9-mediated multiplex genome editing involving different genes has not been explored in grape.
In the present study, we first identified four VvU3 and VvU6 promoters and two UBQ promoters in grape and investigated their efficacy and efficiency in genome editing by targeting the grape phytoene desaturase (PDS) gene. The use of grape VvU3/U6 and UBQ2 promoters can clearly improve the editing efficiencies in grape cells and stable transgenic plants by increasing the expression of sgRNA and Cas9, respectively. We then optimized the CRISPR/Cas9 system using the VvU6 and UBQ2 promoters instead of the AtU6 and 35S promoters, respectively, and developed a multiplex genome editing system containing the traditional multiple sgRNA (multi-sgRNA) expression cassettes or the polycistronic tRNA-sgRNA cassette (PTG). The simultaneous editing of the grape tonoplastic monosaccharide transporter (TMT) genes TMT1 and TMT2 demonstrated the efficacy of the two systems. Additionally, the successful editing of the tobacco NbPDS gene suggests that the identified VvU3 and VvU6 promoters, as well as the UBQ2 promoter, might be applied in other dicot plants.

Identification of U3, U6, and UBQ promoters in grapevine
In previous studies on genome editing in grape, AtU6 and 35S promoters were commonly used to drive the expression of sgRNAs and Cas9, respectively 27,29,31,40 . In an effort to optimize the CRISPR/Cas9 system in grape, we conducted BLAST searches of the grape genome (V. vinifera L.) using the Arabidopsis AtU6-26 and AtU3b genes as the queries, and identified four VvU6 and VvU3 snRNA genes showing high sequence similarities to the queries ( Table 1). The two VvU6 (VvU6.1 and VvU6.2) and two VvU3 (VvU3.1 and VvU3.2) snRNA genes exhibited relatively conserved snRNA transcript sequences compared with the corresponding AtU6 and AtU3 genes. However, the promoter regions of these genes were divergent between the two species, with the exception of the USE and TATA-like box, which are required for transcription (Fig. 1a). The presence of the USE and TATA-like box in the promoters of the VvU3 and VvU6 genes suggests that these Pol III promoters might be involved in effective transcription. We therefore cloned the promoter regions upstream of the transcription initiation site of the four VvU3 and VvU6 genes as the VvU3 and VvU6 promoters for subsequent experiments (Table 1 and Supplementary Fig. S1). In addition, we screened two grape UBQ genes that are constitutively expressed in grapevine (Table 1 and Supplementary Fig.  S2). In this study, 748-bp and 845-bp DNA fragments upstream of the start codon "ATG" were amplified as putative UBQ promoters ( Supplementary Fig. S3).
Assay of the activities of the VvU3, VvU6, and UBQ promoters via transient expression in tobacco and grapevine leaves To investigate whether the amplified VvU3 and VvU6 promoters could drive the expression of sgRNA, we used these Pol III promoters instead of the AtU6 promoter in the pCACRISPR/Cas9 vector 27 containing a sgRNA targeting the grape PDS gene. The CRISPR vectors were then introduced into Nicotiana benthamiana leaves via Agrobacterium-mediated infiltration for transient expression. The expression of PDS sgRNA was inspected by quantitative real-time PCR (qRT-PCR). According to the results, the VvU3 and VvU6 promoters successfully promoted the expression of PDS sgRNA in tobacco leaves, and the sgRNA expression levels driven by the VvU3 and VvU6 promoters were comparable or higher than those coordinated by the AtU6 promoter (Fig. 2a).
Similarly, the activities of the UBQ promoters were first assessed in tobacco leaves, and the β-glucuronidase (GUS) gene fused to the UBQ promoters was used as the reporter. As shown in Fig. 2b, both UBQ1 and UBQ2 promoted the expression of the GUS gene in tobacco leaves, and the activity of UBQ2 was higher than that of UBQ1, as indicated by the finding that the leaf region infiltrated with the UBQ2:GUS construct exhibited stronger GUS staining intensity compared with the region transformed with the UBQ1:GUS vector (Fig. 2b). However, no GUS staining was observed in the grapevine leaves after infiltration with the UBQ1: GUS construct, whereas strong GUS staining was observed with the UBQ2 promoter ( Fig. 2c and Supplementary Fig. S4). Based on these results, the UBQ2 promoter was selected for further study.  ) and TATA-like box are indicated by green boxes. The nucleotides "A" (adenine) and "G" (guanine) recognized by the U3 and U6 promoters for transcription initiation are labeled in red boxes. Different colors denote different levels of sequence identity. The nucleotides with 100% identity are highlighted in black, and those with ≥75% and ≥50% identity are highlighted in red and blue, respectively. The multiple sequence alignments were performed using DNAMAN software The use of the VvU3/U6 and UBQ2 promoters contributes to genome editing in grape It has been reported that using plant species-specific promoters could increase the expression of sgRNA/Cas9, resulting in enhanced editing efficiencies [8][9][10] . To determine whether the use of grape VvU3/U6 and UBQ2 promoters results in improved genome editing efficiency in grape, we constructed optimized CRISPR/Cas9 vectors based on the backbone of the pCAMBIA2300 vector using the grape UBQ2 promoter to drive the expression of Cas9 and the VvU3/U6 promoters to express PDS sgRNA (Fig. 3a). The optimized CRISPR vectors were designated pCA-UBQ2-Cas9-VvU3/U6-PDS and introduced into embryogenic cells derived from 41B grapevine rootstock via Agrobacteriummediated transformation, and kanamycin-resistant cells were obtained after antibiotic-dependent selection (Fig. 3b).
Exogenous T-DNA insertions were identified by PCR using Cas9-specific primers (Supplementary Table S1). The results revealed that the kanamycin-resistant cells contained the exogenous Cas9 gene (Fig. 3c), which suggested that the CRISPR vectors had been successfully introduced into the grape cells. We therefore checked the target site of the PDS gene in these transgenic cells. Given that the presence of unedited (or wild-type) sequences from untransformed cells could influence the identification of mutated sequences, we employed the restriction enzyme (RE)/PCR approach for the detection of mutagenesis in the PDS gene. The isolated genomic DNA (gDNA) from the grape cells was treated with the SspI restriction enzyme to digest the wild-type sequences, and the mutated DNA sequences, which are recalcitrant to enzyme digestion, could be enriched via subsequent PCR amplification (Fig. 3d). The desired bands were purified and sent for Sanger sequencing. As expected, indel (deletions or insertions) mutations at the target site were observed in the VvU3-PDS and VvU6-PDS transgenic cells (Fig. 3e). Most of the mutations were short nucleotide (<10 bp) insertions or deletions (Fig. 3e), which is consistent with our previous results obtained using the AtU6 promoter 40 . In addition, mutations consisting of ≥18-bp deletions were also observed in the VvU3.1-PDS and VvU6.1-PDS transgenic cells (Fig. 3e). These results demonstrated the efficacy of the VvU3, VvU6, and UBQ2 promoters in CRISPR/Cas9-mediated genome editing in grape.
The expression of sgRNA and Cas9 in transgenic grape cells was measured by qRT-PCR, and the results showed that the transcript levels of the PDS sgRNA driven by the VvU3 and VvU6 promoters were higher (>3-fold) than those obtained using the AtU6 promoter (Fig. 4a). In addition to sgRNA expression, the transcript abundance of Cas9 was significantly increased when using the UBQ2 promoter (Fig. 4a). The editing efficiencies in the transgenic cells were assessed using the restriction fragment length polymorphism (RFLP) method. The target fragment of the PDS gene was amplified from the AtU6-, VvU3.1-, VvU3.2-, VvU6.1-, and VvU6.2-PDS transgenic cells by PCR, and the PCR products were digested with the SspI enzyme. As mentioned above, successful targeted editing destroyed the SspI restriction enzyme site, resulting in accumulation of the mutated sequences. The indel frequency can be calculated according to the intensity of undigested PCR bands 40,41 . The RFLP analysis showed that the use of the VvU3 and VvU6 promoters resulted in higher editing efficiencies, and the indel frequencies ranged from 14.65% to 22.10% ( Fig. 4b and Supplementary Table S2). In contrast, the editing efficiency obtained with the AtU6 promoter was 13.67% ( Fig. 4b and Supplementary Table S2).
Interestingly, the editing efficiencies with the VvU3 promoters were slightly lower than those obtained with the VvU6 promoters (Fig. 4b). One possible reason is that the eukaryotic U3 promoter requires adenine (A) for transcription initiation, whereas the U6 promoter recognizes guanine (G) to initiate transcription. The additional "A" or "G" at the 5ʹ ends of the mature sgRNA could affect the editing efficiencies 42,43 . Therefore, we adopted VvU6-PDS transgenic cells for plant regeneration, and albino The CRISPR vectors were infiltrated into N. benthamiana leaves for transient expression, and 3 days after infiltration, the sgRNA expression level relative to that of the Cas9 gene was determined by quantitative real-time PCR (qRT-PCR). Non-infiltrated plants were used as a negative control (ctrl). The data are presented as the mean values ± SDs from three biological replicates; * and ** indicate significant (P < 0.05) and highly significant (P < 0.01) differences, respectively; b and c GUS staining of tobacco and V. amurensis leaves, respectively. The UBQ1:GUS and UBQ2:GUS constructs were infiltrated into N. benthamiana leaves, and the infiltrated regions are shown in red dotted ellipses. Agrobacterium cells containing no expression vector served as the control (ctrl), and the control regions are marked with black dotted ellipses. GUS staining was conducted 3 days after infiltration plants were obtained as expected (Fig. 4c). According to the results, the mutation rates of VvU6.1-PDS (34.48%, 10/29) and VvU6.2-PDS (43.24%, 16/37) were higher than that of AtU6-PDS (23.53%, 4/17), and homologous or biallelic plants were obtained with the VvU6.1 and VvU6.2 promoters at rates of 13.79% and 16.22%, respectively ( Fig. 4d and Supplementary Fig. S5). Collectively, these results suggested that the use of speciesspecific VvU3/U6 and UBQ2 promoters resulted in high editing efficiency in grapes.
High-efficiency multiplex genome editing in grape using grape promoters Given that the use of VvU3/VvU6 and UBQ2 promoters resulted in efficient genome editing, we hypothesized that these promoters can be used to express different sgRNAs for multiplex genome editing in grape. Multiplex genome editing enables researchers to study the functions of several related genes or even perform metabolic engineering in plants 36,44 . However, previous studies on genome editing in grape have primarily focused on single genes [27][28][29][30][31]39 , and editing studies involving different genes of interest have not been reported. To investigate the efficacy of the CRISPR/Cas9-based multiplex editing system in grape, we edited the TMT gene family in 41B grape cells, which have been successfully used to explore sugar uptake and accumulation 45 . Grape has three TMT genes, namely TMT1 (VIT_18s0122g00850), TMT2 (VIT_03s0038g03940), and TMT3 (VIT_07s0031g02270), which show high similarity to their homologs in Arabidopsis 46 . However, the role of these three genes in sugar accumulation in grape has not been explored. By investigating the expression profiles of the three genes in different tissues or organs at different developmental stages, For each sample, a total of 10 clones were sequenced. The mutation type and corresponding number of clones are shown in red and black on the right, respectively. ND, not detected. WT, wild-type sequences. Mut, mutated sequences. The PAM sequence is indicated in green, and the SspI enzyme site is underlined we found that TMT1 and TMT2 exhibited high transcript levels in grape berries, particularly at veraison and during ripening, whereas the expression of TMT3 was poor or not detected in berries ( Supplementary Figs. S6 and S7). Similar results were observed in 41B cells ( Supplementary  Fig. S8). These findings suggested that TMT1 and TMT2 play a vital role in sugar accumulation in vacuoles. Hence, we selected TMT1 and TMT2 as our target genes. Specific sgRNAs targeting TMT1 and TMT2 were designed and ligated to the VvU6.1 and VvU6.2 promoters, respectively. The two individual sgRNA expression cassettes were connected by overlapping PCR, and the combined multi-sgRNA expression cassette was used instead of the sgRNA expression cassette in the pCA-UBQ2-Cas9-VvU3/U6-sgRNA vector to generate the multiplex editing vector pCA-UBQ2-Cas9-VvU6-TMTs (Fig. 5a). Furthermore, we synthesized the polycistronic tRNA-TMT sgRNA cassette ( Supplementary Fig. S9) and developed the PTG/ Cas9 system, in which the expression of Cas9, sgRNAs, and tRNA was driven by the single UBQ2 promoter (Fig. 5a).
The constructed vectors were introduced into 41B cells. After transformation and antibiotic-dependent selection, transgenic grape cells were recovered for mutation identification. Both the multi-sgRNA/Cas9 and PTG/ Cas9 systems could successfully induce the desired mutagenesis in the target genes. As shown in Fig. 5b, mutated sequences were detected in the amplicons of both the TMT1 and TMT2 target sequences. Single nucleotide insertions and nucleotide deletions were observed at the target sites (Fig. 5b), consistent with previous reports 27,31,40 . The T7EI assay revealed that the multi-sgRNA/Cas9 system resulted in editing efficiencies of 10.51% and 20.85% at the TMT1 and TMT2 target sites, respectively ( Fig. 5c and Supplementary Table S2).  Multiplex editing in grape using multi-sgRNA/Cas9 and PTG/Cas9 systems. a Schematic diagrams of the multi-sgRNA/Cas9 and PTG/ Cas9 vectors and the editing targets. b Sanger sequencing results of the editing at TMT1 and TMT2 target sites. The gDNA was extracted from transformed 41B cells for amplification of the target sequences. A total of 10 amplicons were sent for sequencing. A representative chromatogram and the mutated sequences identified for each target are shown. The target regions are underlined, and the mutation sites are indicated with red arrows in the chromatograms. The mutation types, as well as the corresponding number of sequences, are shown on the right. The inserted nucleotides are shown by green boxes. Mut, mutated sequences. c T7EI assay of the editing efficiencies at TMT1 and TMT2 target sites. The PCR amplicons containing the TMT1 or TMT2 target sequences were treated with the T7EI enzyme, and the digested products were separated on an EtBrstained agarose gel (1.5%). The wild-type sequences of TMT1 (WT-TMT1) or TMT2 (WT-TMT2) served as controls. The intensities of the bands were quantified using ImageJ. The indel efficiencies calculated from the band intensities 40 are shown as the mean values ± SDs from three technical replicates; +, T7EI digestion; −, no digestion Comparable mutation efficiencies were also generated using the PTG/Cas9 system at the TMT1 (10.38%) and TMT2 (17.78%) sites ( Fig. 5c and Supplementary Table  S2). Intriguingly, the TMT2 sgRNA had a higher editing efficiency than the TMT1 sgRNA in both the multi-sgRNA/Cas9 and PTG/Cas9 systems (Fig. 5c), which suggested that the sgRNA sequence had an effect on the editing efficiency. To evaluate the effect of TMT1 and TMT2 mutagenesis on sugar accumulation, we measured the sugar content in the edited 41B cells. Compared with the cells transformed with empty vector (EV), the contents of maltose, glucose, and fructose in the tmt1tmt2 cells were significantly reduced (Fig. 6), which suggested that the mutation of TMT1 and TMT2 affected the extent of sugar accumulation in grape cells.
The VvU3/U6 and UBQ2 promoters are suitable for genome editing in tobacco As mentioned above, the VvU3/VvU6 and UBQ2 promoters can drive the expression of sgRNA and GUS reporter genes in N. benthamiana leaves (Fig. 2a, b), which suggests that these promoters might be used for genome editing in tobacco. To confirm the efficacy of the VvU3/U6 and UBQ2 promoters in tobacco genome editing, the grape PDS sgRNA in the pCA-UBQ2-Cas9-VvU3/U6-PDS vectors (Fig. 2a) was replaced by the NbPDS sgRNA (Fig. 7a). The modified vectors were infiltrated into N. benthamiana leaves for transient expression. The tobacco leaves were sampled for mutation identification. Sanger sequencing assays showed that the use of the VvU3/U6 and UBQ2 promoters resulted in successful targeted mutagenesis at the target site in the NbPDS gene (Fig. 7b), which confirmed the efficacy of the grape Pol II and Pol III promoters in tobacco genome editing.

Discussion
As an RNA-guided DNA endonuclease system, sgRNA activity and the expression of sgRNA/Cas9 greatly influence the efficiency of CRISPR/Cas9-mediated genome editing 47,48 . The application of plant species-specific promoters is an effective strategy for improving the efficiency of genome engineering in plants 8,10 . To date, the widely used eukaryotic U3 and U6 promoters in monocot and dicot plants have been isolated from rice and Arabidopsis, respectively 21,23 . In grapevine, the AtU6 and 35S promoters are commonly used to drive the expression of sgRNA and Cas9, respectively, in genome editing 27,31,40 . In this study, we characterized four Pol III promoters (VvU3.1, VvU3.2, VvU6.1, and VvU6.2) and one Pol II promoter (UBQ2) in grape and verified their efficiency in grape genome editing by targeting the PDS gene (Fig. 3). We established that both the UBQ2 and VvU3/U6 promoters exhibited strong activities in promoting the expression of Cas9 and sgRNA in grape (Fig. 4a), which resulted in high editing efficiencies (Fig. 4b, d). The results demonstrated that the use of species-specific promoters could increase the genome editing efficiency in grape.
In addition, the amplified VvU3 and VvU6 promoters can be used for the expression of several distinct sgRNA Fig. 6 Sugar content in the tmt1tmt2 cells. Edited cells obtained using the multi-sgRNA/Cas9 system were collected as examples for determination of the sugar content, and grape cells transformed with empty vector (EV) were used as a control. The data are shown as the mean values ± SDs from four biological replicates; ** indicates a highly significant (P < 0.01) difference Fig. 7 Targeted editing of the NbPDS gene in tobacco leaves using grape promoters. a Schematic diagram of the CRISPR vectors and target selection for tobacco genome editing. The primers NbPDS-F/R were used to amplify the target sequence of the NbPDS gene. The MlyI restriction enzyme site is underlined. b Sequencing results of targeted mutations in the NbPDS. The CRISPR vectors harboring the VvU3 and VvU6 promoters were infiltrated into N. benthamiana leaves for transient expression. The mutated sequences (Mut) were analyzed by RE/PCR and Sanger sequencing assays. The mutation types and corresponding number of mutated sequences are shown in red and black, respectively, on the right modules, which can direct the Cas9 protein to edit different targets simultaneously. This Pol III promoterdependent strategy is commonly adopted for multiplex editing in plants 13,15,16,21 . In the present study, multiple sgRNA expression cassettes were assembled using the VvU6.1 and VvU6.2 promoters (Fig. 5a), because the first nucleotide of TMT1 and TMT2 sgRNAs is "G". This multi-sgRNA/Cas9 system was demonstrated to be effective in grape, with editing efficiencies of 10.51% and 20.85% at the TMT1 and TMT2 targets, respectively (Fig. 5c). Individual sgRNA expression cassettes are relatively easy to construct, and several assembly methods have been developed 16,49 . However, the use of many promoters results in a large expression vector, and size limitation remains a challenge when expressing an increased number of sgRNAs 32 . Plant Pol III promoters generally require conserved elements of the USE and TATA-like box for effective transcription, and shortened U3 and U6 promoters containing only the USE, TATA box, and several artificial monocot-specific promoter elements have been developed for genome editing in rice 21 . Our identified VvU3 and VvU6 promoters contained the required USE and TATA-like box elements (Fig. 1), which suggested that the current promoters could also be optimized in the future. An alternative strategy for multiplex genome editing is expressing multiple sgRNAs from a single transcript 34 . In this case, endoribonuclease Csy4, tRNA, or RZs are generally employed to generate different sgRNAs 34 . However, Csy4 should be coexpressed with an artificial sgRNA array when using the Csy4-based sgRNA expression method 34 . Furthermore, in rice, the RZ system exhibited relatively poor editing activity compared with the tRNA system 14 . The tRNA-based PTG system is a very efficient approach for the simultaneous editing of several genes of interest with high efficiencies in plants 36,50 . We also investigated the efficacy of the PTG/ Cas9 system in grape, and the results showed that the PTG/Cas9 system was as effective as the multi-sgRNA/ Cas9 system in editing the TMT1 and TMT2 genes (Fig. 5). The editing efficiencies using the PTG/ Cas9 system were slightly lower than those obtained using the multi-sgRNA/Cas9 system (Fig. 5c), and the difference could be explained by different promoters or processing efficiencies of the primary transcript.
To date, the CRISPR/Cas9 system has been a powerful tool for crop improvement. For instance, the simultaneous editing of three different genes helps increase the yield and cold tolerance of rice 15 , and the engineering of three homoeoalleles in bread wheat confers resistance to powdery mildew 51 . In grapevine, however, the CRISPR/ Cas9 system has not been fully exploited, and the application of this technology is restricted to single gene editing due to the limited CRISPR toolbox available. Multiple genes generally control the important traits involved in the quality of fruits of grape, such as sugar and organic acid accumulation 46,52 . TMTs are important sugar transporters that are reportedly associated with sugar accumulation in sugar beets and watermelon 53,54 . Of the three TMT genes in grape, TMT1 and TMT2 are thought to be involved in sugar accumulation in grape berries 46 . Knockout of the TMT1 and TMT2 genes in grape cells significantly reduced the sugar content (Fig. 6). The simultaneous editing of the grape TMT1 and TMT2 genes provides evidence that can be used to deduce the function of these two genes in sugar accumulation in grape and constitutes, to our knowledge, the first example of multiplex genome editing in grape.
In conclusion, our study provides effective and robust grape Pol II and Pol III promoters for the optimization of the CRISPR/Cas9 system, and these promoters increase the editing efficiency in grape. Moreover, the developed multiplex genome editing systems, including the multi-sgRNA/Cas9 and PTG/Cas9 systems, expand the toolbox of grape genome editing and can thus facilitate the application of CRISPR/Cas9 technology to the future study of functional genes and mutant phenotypes in grape.

Plant materials and culture conditions
Grapevine plants of V. vinifera cv. Pinot Noir were grown in the germplasm resources garden at the Institute of Botany, the Chinese Academy of Sciences, Beijing, under natural conditions. The leaves were sampled for gDNA isolation and the prepared gDNA was used for promoter cloning. 41B (Vitis vinifera cv. Chasselas × Vitis berlandieri, a rootstock) cells were cultured in liquid glycerol-maltose (GM) medium as previously reported 40
To develop the GUS reporter vectors, the UBQ1 and UBQ2 promoters were amplified from the pLB vector and cloned into the pBI121 vector instead of the 35S promoter via the BamHI and HindIII sites through homologous recombination (HR) using the ClonExpress II One Step Cloning Kit (Vazyme). To inspect the efficacy of the VvU3 and VvU6 promoters in driving the expression of sgRNA, the grape PDS sgRNA was ligated to the VvU3 and VvU6 promoters by PCR, and the VvU3/U6-PDS expression cassettes were cloned into the pCACRISPR/Cas9 vector as previously described 27 . To construct the pCA-UBQ2-Cas9-VvU3/U6-PDS vectors, the Cas9 gene was amplified from the pCACRISPR/Cas9 vector and cloned into the HindIII-digested pCAMBIA2300 vector by HR. The UBQ2 promoter was then amplified and cloned into the pCAMBIA2300-Cas9 vector just upstream of the Cas9 gene via the BamHI and SalI sites through HR. The VvU3-PDS and VvU6-PDS expression cassettes were subsequently amplified from the well-constructed pCA-CRISPR/Cas9 vectors described above and ligated into the SmaI-digested pCAMBIA2300-UBQ2-Cas9 vector via HR. Similarly, to construct the CRISPR vectors for tobacco genome editing, the NbPDS sgRNA was fused to the VvU3 and VvU6 promoters by PCR, and the VvU3/ U6-NbPDS expression cassettes were introduced into the pCAMBIA2300-UBQ2-Cas9 vector by HR.
To construct the multi-sgRNA/Cas9 vector for multiplex genome editing, the TMT1 sgRNA and TMT2 sgRNA were fused to the VvU6.1 and VvU6.2 promoters by PCR, respectively, and the two sgRNA expression cassettes were combined by overlapping PCR. The combined sgRNA expression cassette was cloned into the SmaI-digested pCAMBIA2300-UBQ2-Cas9 vector via HR. For construction of the PTG/Cas9 vector, a DNA fragment containing a 50-bp polyA sequence and the polycistronic tRNA-TMT sgRNA expression cassette ( Supplementary  Fig. S6) was synthesized (Tsingke) and ligated into the HindIII-digested pCAMBIA2300-UBQ2-Cas9 vector by HR. All the primers used for cloning and vector construction are provided in Supplementary Table S1.

Plant transformation, regeneration, and identification
For the stable transformation of 41B cells, the CRISPR vectors were introduced into the Agrobacterium strain EHA105, and Agrobacterium-mediated transformation of 41B cells and the regeneration of transgenic plants were conducted as previously described 31 . After transformation, 41B cells were cultured in liquid GM supplemented with 200 mg/L Timentin and 5 mg/L kanamycin, and the grape cells were subcultured every 4 weeks until the development of kanamycin-resistant cells. The resistant cells were sampled for gDNA isolation, and Cas9-specific primers (Supplementary Table S1) were used to identify T-DNA insertions. For the regeneration of transgenic plants, 41B cells were transferred to GM without phytohormones. The induced embryos were further germinated on woody plant medium (Duchefa Biochemie) under light conditions 31 . The regenerated plants were sampled for gDNA extraction, and the prepared gDNA was used for the PCR amplification of target fragments. The plants were analyzed by direct sequencing of PCR products followed by sequencing assays of individual amplicon clones 23 . For each plant, a total of 25 clones were analyzed to verify the mutation types.
For transient expression experiments, the CRISPR vectors and GUS reporter vectors were introduced into Agrobacterium strain GV3101. The bacterial cells were cultured in liquid lysogeny broth medium containing 50 mg/L rifampicin and 50 mg/L kanamycin at 28°C overnight. The bacterial cultures were centrifuged at 5000 × g for 10 min, and the resultant supernatant was discarded, and the bacterial cells were resuspended in transfection buffer (10 mM MES, 10 mM MgCl 2 , and 200 μM acetosyringone, pH 5.6). The concentration of the bacterial suspension was adjusted to a final OD 600 of 0.4. The suspension was incubated at room temperature for another 2 h before transformation. The leaves of 5-weekold tobacco plants were infiltrated using a 1-mL needleless syringe. Three leaves from three independent plants were harvested as three biological replicates for qRT-PCR assay or GUS staining. For transient expression in grapevine leaves, the leaves of 1-month-old V. amurensis plants were immersed in the bacterial suspension, and vacuum was applied for 5-10 min until the leaves became translucent. After infiltration, the grapevine leaves were washed with sterile water and placed on sterile watersoaked filter paper in 90-mm Petri dishes. At least three independent V. amurensis plants were used for transformation with each vector. GUS staining of leaves was performed 3 days after infiltration.

Mutation detection
For the detection of mutagenesis in the target genes, DNA fragments containing the target sites were amplified by PCR with gene-specific primers from grape or tobacco. gDNA was first treated with the SspI enzyme to digest the wild-type sequence, and the target sequences of the grape PDS gene were then amplified. The PCR products were purified and cloned into the pLB vector for Sanger sequencing. For evaluation of the editing efficiencies, the RFLP method and T7EI assay were performed as previously described 40 . The PDS gene contained an SspI site within the target sequence, and digestion of the PCR amplicons with the restriction enzyme generated undigested bands. The indel frequency was measured based on the intensity of the undigested bands 40 . However, the target sequences of TMT1 and TMT2 lacked proper restriction enzyme sites; thus, the T7EI assay was adopted for assessment of the editing efficiencies. The editinginduced mismatched amplicons could be cleaved by the T7EI enzyme, and the indel frequency can be calculated from the intensities of the digested bands 40 . The PCR products treated with SspI (NEB) or T7EI (NEB) enzyme were separated on an EtBr-stained agarose gel (1.5%). The mutation efficiencies were determined using ImageJ.

qRT-PCR assay
Total RNA was extracted using the HiPure HP Plant RNA Mini Kit, and cDNA was synthesized from 1 μg of total RNA using the HiScript II Q RT SuperMix for qPCR Kit (Vazyme) following the manufacturer's protocols. qRT-PCR assays were performed to determine the expression levels of the sgRNA and Cas9 genes in grape cells, and grape Actin 1 (accession no. AY680701) and GAPDH (accession no. XM_002263109) were used as internal controls. In tobacco leaves, the expression of sgRNA relative to that of the Cas9 gene was measured. The relative expression levels were calculated using the 2 −ΔΔCT method 55 . The significance of gene expression was determined using Student's t-test.