VviPLATZ1 is a major factor that controls female flower morphology determination in grapevine

Plant genetic sex determinants that mediate the transition to dioecy are predicted to be diverse, as this type of mating system independently evolved multiple times in angiosperms. Wild Vitis species are dioecious with individuals producing morphologically distinct female or male flowers; whereas, modern domesticated Vitis vinifera cultivars form hermaphrodite flowers capable of self-pollination. Here, we identify the VviPLATZ1 transcription factor as a key candidate female flower morphology factor that localizes to the Vitis SEX-DETERMINING REGION. The expression pattern of this gene correlates with the formation reflex stamens, a prominent morphological phenotype of female flowers. After generating CRISPR/Cas9 gene-edited alleles in a hermaphrodite genotype, phenotype analysis shows that individual homozygous lines produce flowers with reflex stamens. Taken together, our results demonstrate that loss of VviPLATZ1 function is a major factor that controls female flower morphology in Vitis.

T raits associated with flower development and fertilization are key for increasing yield in agriculture and horticulture crops 1 . Plants evolved multiple reproductive systems to promote outcrossing, including self-incompatibility, dioecy and dichogamy 2,3 . In a dioecy mating system, individual plants produce either male or female flowers that are morphologically distinct [4][5][6] . During plant speciation, the dioecy reproductive system independently evolved multiple times indicating that the underlying mechanisms that control female and male flower development are diverse. While approximately 5-6% angiosperms are dioecious, a number of economically important crops utilize this mating strategy [5][6][7] .
In the past decade, one-and two-gene sex determination systems were characterized in a subset of dioecious plants 6 . A one-gene model of sex determination described in Diospyros and Populus is dependent upon the duplication of a single gene, which gives rise to small non-coding RNAs that alter flower sex identity [8][9][10] . Actinidia and Asparagus are two species that exhibit a two-gene model of sex determination involving DNA polymorphisms that results in: (1) ectopic expression of a malepromoting factor that suppresses pistil formation and (2) a recessive mutation in a single gene required for tapetum development and pollen fertility [11][12][13][14] .
Grapevine cultivars derived from Vitis vinifera are a major perennial crop of economic significance throughout the world for wine, table and dried grape production 15 . Modern domesticated grapevines produce hermaphrodite flowers with functional stamens and pistils for self-pollination, which is a fundamental trait for productivity acquired during domestication 16 . Interestingly, the wild progenitor of Vitis vinifera grapevine cultivars, as well as other Vitis species, are dioecious 17 . In wild Vitis species, females initiate flowers that produce short reflex stamens that bend away from the receptive pistil, which likely reduces self-fertilization 18 . The reflex stamen phenotype is utilized by grapevine breeders and biologists to identify female vines for cultivar development. While pollen derived from female flowers is viable, fertility is absent due to an impairment in germination 19 . In males, flowers produce erect functional stamens, while pistil formation is aborted early in development, which prevents self-fertilization 19,20 .
In Vitis species and hybrids, the SEX-DETERMINING REGION (SDR) is located on chromosome 2 [21][22][23][24][25] . Depending on the Vitis species/hybrid accession and cultivar, the size of the locus varies from 111 kb to 837 kb; however, gene identity and content are conserved [26][27][28][29] . One hypothesis predicts that grape sex determination is mediated by the two-gene model 27,29,30 . According to this hypothesis, female flower formation is mediated by a recessive mutation that controls stamen development. Therefore, the genotype of female individuals is f/f. In contrast, male flower development is determined by a dominant male-promoting factor (M) that suppresses pistil development. Lastly, domesticated hermaphrodite grapevines with an H/H or H/f genotype produce perfect flowers that contain a modified and less dominant Mfactor 27 .
Through sequence analysis of M-, H-and f-haplotypes derived from domesticated and wild Vitis species, the structure of the SDR was determined and candidate sex-determining genes were identified 28,29,31,32 . Genetic association analysis of M-and f-specific DNA polymorphisms and phylogenetic studies of sex-linked genes indicates that male and female flower development is specified by genetic determinants located in the 5′ and 3′ region of the SDR, respectively 29,32 . The male sex-determining region (MSDR) consists of two genes, Vitis vinifera YABBY3 (VviYABBY3) and VviSKU5, while 11 genes from TREHALOSE-6-PHOSPHATE PHOSPHATASE (TPP) to WRKY21 reside in the female sex-determining region (FSDR) 32 . The H-haplotype is structurally similar to the f-haplotype from VviYABBY3 to the aldolase gene but structurally related to the M-haplotype from TPP to the WRKY21 transcription factor 29,32 . Initial studies suggested that VviAPT3 was the candidate M-factor that controls male flower formation through the inactivation of cytokinin 26,28,31,33 , which is required for pistil development 34 . In support of these studies, transcript levels for VviAPT3 are significantly increased in male flowers compared to hermaphrodite and female flowers 28,29,31 and the mRNA for this gene localizes to cells in the male flower meristem that gives rise to the pistil 33 . Consistent with the increased expression of VviAPT3 in the male flower meristem, applications of cytokinin restore pistil development 35 . However, recent studies indicate that VviAPT3 is not located in the MSDR 32 . Therefore, VviYABBY3 emerged as the primary candidate M-factor 29,32 as this is one of two genes located in the MSDR and the expression of this gene associates with the formation of male flowers. Further, YABBY transcription factors are critical for the development of lateral organs, including carpels 36 . Genetic studies indicate that a candidate DNA polymorphism located in Vitis INAPERTU-RATE POLLEN1 (VviINP1) is the underlying female sterility determinant, as this gene is predicted to play a role in pollen development 28,29,31 . While candidate flower sex genes have been identified from the above studies, functional analysis demonstrating a role for these genes in pistil and pollen development have yet to be determined.
Here, we show through genetic and molecular studies that the VviPLATZ1 (plant AT-rich sequence-and zinc-binding protein1) transcription factor is a candidate factor involved in female flower development. To this end, VviPLATZ1 localizes to the FSDR and low expression of this gene correlates with the formation of reflex stamens. To validate the role of VviPLATZ1 in reflex stamen development, the CRISPR/Cas9 system is used to generate multiple alleles of this gene lacking the conserved PLATZ domain in a rapid cycling hermaphrodite genotype. Phenotype evaluation shows that homozygous individuals harboring gene-edited alleles of VviPLATZ1 produce flowers that initiate stamens with a reflex architecture. Taken together, results show that the formation of reflex stamens is mediated by the loss of VviPLATZ1, which is a factor involved in controlling female flower morphology in Vitis.

Results
Genetic and molecular analysis of VviPLATZ1. As shown in Fig. 1a, association genetics indicates that the FSDR is comprised of 11 genes from TPP to WRKY21 32 . To further narrow down the number of genes that reside in the FSDR, 21 SNPs spanning the PN40024 SDR were identified (Supplementary Table 1). Three genetic mapping resources were genotyped with this SDR SNP set: (1) Vitis vinifera cultivars with known flower sex genotypes (f/ f, H/f or H/H), (2) an F 1 00C001V0008 (f/f) x Ugni Blanc (H/f) mapping population 37 and (3) self-fertilized hermaphrodite microvine (H/f) progenies (S1 to S5). However, as the PN40024 SDR contains a mixture of H-and f-sequences, this reference genome was not suitable for fine genetic mapping of the reflex stamen phenotype, as gene content between the H-and f-haplotypes differs in the SDR 27,31 . Genomic studies indicate that gene content and order in the SDR of f-haplotypes is conserved 28,29,32 . Therefore, we posited that the previously well characterized Cabernet Sauvignon f-haplotype sequence 29 was a suitable reference for fine genetic mapping of the FSDR by identifying the likely relative positions of the genic SNP markers developed in our study (Supplementary Table 1). By using this genetic mapping strategy, the genotype information derived from the Vitis vinifera cultivars indicated that the 5′ boundary of the FSDR was located in TPP, at the VvMT_54 SNP ( Fig. 1b (Fig. 1b).
The recessive mutation resulting in the formation of reflex stamens is predicted to reduce the expression and/or function of a candidate gene required for stamen development. Of the seven genes located in the FSDR, transcript abundance for VviPLATZ1 appeared to correlate with the development of reflex stamens 28,29,31 . Despite the results from these expression studies, a role for the loss of VviPLATZ1 function in reflex stamen development was not recognized. Therefore, to support the hypothesis that loss of VviPLATZ1 results in the formation of reflex stamens, transcript levels for this gene was examined in 04C023V0006 (H/H) and compared to 04C023V0003 (f/f) flowers at eight, six, and four week prior to anthesis (WPA), as previously defined 38 . Results showed that transcripts for VviPLATZ1 were readily detectable in 04C023V0006 flowers at eight and six WPA (Fig. 1c), which correspond to a developmental time in which stamens differentiate and filaments elongate, respectively 38 . At four WPA, when flower development is nearly completed, the mRNA levels for this gene were significantly decreased. In addition, VviPLATZ1 mRNA levels were readily detected in male flowers of 03C003V0060 (M/f) during stamen development at eight and six WPA ( Supplementary Fig. 1). In contrast to 04C023V0006 and 03C003V0060, transcript levels for Vvi-PLATZ1 were significantly reduced in 04C023V0003 during stamen development at eight and six WPA ( Fig. 1c and Supplementary Fig. 1). Transcript levels for VviINP1, Exostosin, KASIII, FMO1, FMO2, and FMO3, which are located in the FSDR, did not correlate with the formation of reflex stamens 28,29,31 .
DNA sequence comparisons between 04C023V0006 (H/H) and 04C023V0003 (f/f) identified female-specific DNA polymorphisms in VviPLATZ1. First, a thymine to adenine point mutation and adenine insertion were identified in the first codon, resulting in a frame-shift mutation that altered the position of the start codon (Fig. 1d). In addition, two SNPs were identified at −5 bp and −13 bp in the 5′ UTR. Through the alignment of hermaphrodite, male and female VviPLATZ1 alleles from sequenced haplotypes of domesticated and wild Vitis species 29,31 , results showed that female DNA polymorphisms at −13 bp, −5 bp, and +2 bp were conserved ( Supplementary Fig. 2). However, the adenine insertion at +3 bp in the 04C023V0003 VviPLATZ1 allele was found only in a subset of female-specific VviPLATZ1 alleles. Taken together, the conserved female-specific DNA polymorphisms at −13 bp, −5 bp, and +2 bp may act to reduce transcription and/or alter mRNA decay of the female VviPLATZ1 allele. In addition to these DNA polymorphisms, seven nonsynonymous SNPs were identified in the 04C023V0003 VviPLATZ1 allele, which may reduce protein function, including a glutamic acid to lysine substitution in the PLATZ domain (Fig. 1d). To investigate whether these seven nonsynonymous SNPs were conserved, the 04C023V0003 VviPLATZ1 amino acid sequence was aligned with hermaphrodite, female and male VviPLATZ1 proteins derived from sequenced domesticated and wild Vitis species 29,31 . Results showed that all seven nonsynonymous SNPs were conserved in the female VviPLATZ1 alleles ( Supplementary Fig. 3). It should also be noted that the reduction in the number of transposable elements upstream of the female VviPLATZ1 allele compared to hermaphrodite and male alleles 28,31 , as well as differences in RNA editing between male and female VviPLATZ1 transcripts 39 , may also be important factors attributed to low VviPLATZ1 transcript accumulation in female flowers. As the reflex stamen phenotype is a recessive inherited trait, the female allele is referred to as Vviplatz1.
Tissues-specific expression of VviPLATZ1 during flower development. To gain insight into the proposed function of VviPLATZ1 in stamen development, the expression pattern of this gene was analyzed in 04C023V0006 (H/H) and 04C023V0003 (f/f) flowers during filament elongation at six WPA. At this time point, transcript abundance of VviPLATZ1 was significantly higher in 04C023V0006 hermaphrodite flowers compared to 04C023V0003 female flowers (Fig. 1c). In hermaphrodite flowers, VviPLATZ1 was primarily expressed in elongating filaments, as well as developing anthers and microspores ( Fig. 2a-c). A similar pattern of expression was detected in flowers derived from the male microvine 03C003V0060 (M/f; Supplementary Fig. 4). In contrast, VviPLATZ1 expression was relatively low in female filaments and anthers, as well as microspores ( Fig. 2d-f). The expression of VviPLATZ1 in ovules of hermaphrodite and female flowers (Fig. 2a, d, respectively), as well as male flowers (Supplementary Fig. 4) suggests that this gene may play a role in ovule development. Based on the molecular and gene expression studies described above, we propose that reduced VviPLATZ1 function is a major factor that results in the development of reflex stamens.
Gene-editing of VviPLATZ1. Efficient validation of candidate genes implicated in flower development via reverse genetics is constrained by the perennial nature of grapevine, which displays a long juvenile phase and annual reproductive cycle 37 . The microvine model system has been utilized to overcome the genetic limitations of grapevine. This model system is highly amendable to rapid reverse genetics assessment, as the microvine can be stably transformed and displays a continuous flowering phenotype with a reduced generation cycle 37 . Therefore, the role of VviPLATZ1 in stamen development was functionally validated in a homozygous hermaphrodite microvine genotype, 04C023V0006 (H/H), using the CRISPR/Cas9 system. Two guide RNAs designated FS1 and FS4 were developed to induce indels at the 3′ end of exon 2 upstream of the coding region for the PLATZ domain (Fig. 3a). Eight edited alleles were identified in T 0 plants, two of which contained a single bp insertion, while the remaining six alleles contained single or multiple bp deletions (Fig. 3b). All alleles altered the reading frame resulting in a truncated protein lacking the conserved PLATZ domain, which likely abolished VviPLATZ1 function (Fig. 3c). Interestingly, one of the T 0 plants Fig. 1 Structure of the FSDR and molecular analysis of VviPLATZ1. a Diagram of the SDR structure derived from the Cabernet Sauvignon f-haplotype 29 . Genes localized within the FSDR are highlighted in yellow 32 . Gray arrows depict genes located outside of the FSDR. Black arrows were used to display genes flanking the SDR. b Fine genetic mapping narrowed the FSDR to seven genes, highlighted in yellow. The black vertical arrows indicate the location of the VvMT54 and VvMT72 SNPs in TPP and FMO3, respectively, which mark the boundaries of the FSDR. c Transcript levels determined for VviPLATZ1 in 04C023V0006 (H/H) and 04C023V0003 (f/f) flowers at eight, six and four weeks prior to anthesis (WPA), as well as leaves (L). Note: stamen development and pistil initiation occurred at eight WPA 38 . Filament elongation and ovule development occurrs at six WPA. Stamen development is nearly complete by four WPA. Data are presented as mean values ± SD. Significant differences determined by two-tailed Student's t test indicated by asterisks **p = 0.0089, ***p = 0.00027 (n = 3); not significant, ns. d Diagram of VviPLATZ1 genomic structure, which contains five exons (boxes). Exons 3 and 4 displayed in green encode the PLATZ domain. In the f-haplotype, DNA polymorphisms altered the nucleotides at −13 bp and −5 bp in the 5′ UTR, as well as the first codon (underlined nucleotides). The arrowhead and gray highlight illustrate the adenine insertion. The start codon for the hermaphrodite (H/H) and female (f/f) VviPLATZ1 alleles is highlighted in yellow. Point mutations that alter amino acid content in exons 3, 4, and 5 are displayed by asterisk (*). The GenBank accession number for VviPLATZ1 derived from 04C023V0006 is MW548436. For the Vviplatz1 allele derived from 04C023V0003, the GenBank accession number is MW548435. Source data underlying Fig. 1c are provided as a Source Data file.
identified harbored both Vviplatz1-1 and Vviplatz1-2 alleles and was referred to as Vviplatz1-1/2. The other six T 0 individuals were heterozygous for only one of the remaining mutated alleles. All seven T 0 plants were self-pollinated and the flowers of homozygous T 1 Vviplatz1-(1/2-8) plants were analyzed and compared to hermaphrodite and female flowers.
Loss of VviPLATZ1 function produced flowers with reflex stamens. To determine if reduced VviPLATZ1 function plays a role in female flower formation, stamen morphology was examined in hermaphrodite (H/H; VviPLATZ1/VviPLATZ1), female (f/f; Vvi-platz1/Vviplatz1) and homozygous T 1 plants. Flowers produced in hermaphrodite individuals initiated upright stamens with an elongated phenotype (Fig. 4a, g). The average length of stamens in hermaphrodite individuals was 3.0 mm ( Supplementary Fig. 5). In contrast, female flowers initiated reflexed stamens 98% of the time (Fig. 4b, g). The remaining 2% of female stamens displayed an upright growth pattern but were significantly shorter with an average length of 1.2 mm (Supplementary Fig. 5). Flowers produced in Vviplatz1-3, -5, -7, and -8 displayed an overall morphology strikingly similar to female flowers (Fig. 4c-f, respectively). Flower morphology displayed in Vviplatz1-1/2, -4, and -6 was similar to the structure of female flowers (Supplementary Fig. 6). In all seven Vviplatz1-(1/2-8) plants analyzed, 95 to 99% of stamens displayed a reflex architecture, while the remaining stamens were stunted with an average length of 1.4 mm (Supplementary Fig. 5). Taken together, loss of VviPLATZ1 transforms hermaphrodites into females that initiate flowers with reflex stamens during flower development.

Discussion
Dioecious mating systems evolved mechanisms to promote outcrossing through the arrested development of male and female reproductive organs and haploid sex cells in female and male individuals, respectively 6,7 . Floral morphology is a key factor that influences outcrossing, as exemplified by the spatial separation of anthers and stigmas in heterostyly flowers 40 . The formation of reflex stamens that spatially separates anthers from stigmas is a unique dioecious feature displayed by Vitis female flowers 18 , which likely contributes to the outcrossing potential displayed in this dioecious mating system.
The formation of male and female flowers is influenced by hormones, which not only affects ovule and pollen development but also the morphology of floral organs, including petals, stamens, and pistils 41,42 . Experimental studies indicate that asymmetric distribution of auxin regulates organ bending during root and hypocotyl development 43 . In developing stamens, auxin biosynthesis, signaling and transport is critical for filament elongation 44,45 . Therefore, the symmetric distribution of auxin during filament elongation is predicted to cause organ bending for the reflex stamen phenotype 44,46 . It would be interesting to determine whether auxin is involved in the formation of reflex stamens and whether the distribution of this hormone is regulated by VviPLATZ1.
PLATZ genes encode plant-specific transcription factors that regulate a diverse array of reproductive processes including endosperm development, grain length and filling 47,48 . Experimental studies indicate that the Zea mays floury3 (fl3) and Oryza sativa GL6 regulate grain development through interaction with the RNA polymerase III (RNAPIII) complex, which transcribes transfer RNAs (tRNAs) and 5 s ribosomal RNA (5 s rRNA) [47][48][49] . Of the 17 PLATZ proteins identified in maize, 14 interact with a component of the RNAPIII complex indicating that PLATZ proteins regulate developmental processes via tRNA and 5 s rRNA biogenesis 49 . The role of VviPLATZ1 in stamen development further demonstrates that this class of plant-specific transcription factors regulates a wide range of reproductive processes. Further, it is tempting to speculate that loss of VviPLATZ1 function reduces the pool of tRNAs and 5sRNAs required for proper translation during stamen development. Lastly, phylogenetic analysis showed that VviPLATZ1 is a member of a clade that includes GL6, as well as uncharacterized PLATZ proteins from Zea mays, Oryza sativa and Arabidopsis thaliana (Supplementary Fig. 7).
While female flower morphology is determined by loss of Vvi-PLATZ1 function, the f-haplotype-specific 8-bp deletion in the Vitis  28,29,31 . Indeed, inaperturate pollen is produced in female flowers of Vitis riparia and wild Vits vinifera individuals 19,50 While pollen fertility is not affected in Arabidopsis inp1 plants 51,52 , studies in maize indicate that ZmINP1 is crucial for germination 53 . As a result, gene-editing of VviINP1, in grapevine, is necessary to validate a role for the 8-bp deletion in pollen fertility. However, VviPLATZ1 may also play a role in pollen function given its location within the FSDR (Fig. 1a, b), expression in microspores (Fig. 2c, f), and possible role in regulating the transcription of tRNAs and 5 s RNA, as protein synthesis appears to be a primary driver of pollen germination 54 . In support of this hypothesis, mutations in sex determination factors often display pleiotropic effects on other floral traits associated with unisexual flower development 55,56 .
In conclusion, results from this manuscript show that Vvi-PLATZ1 is a key regulator of female flower formation in grapevine. This is supported by the fact that VviPLATZ1 is expressed in the filaments and anthers of hermaphrodite flowers prior to and at the time in which stamens elongate. Moreover, using the CRISPR/Cas9 gene-editing system in the rapid cycling hermaphrodite microvine, functional analysis demonstrated that loss of VviPLATZ1 is a key factor that controls reflex stamen development during female flower formation.

Methods
Plant materials. The Pinot Meunier L1 mutant, referred to as the microvine, contains a GIBBERELLIC ACID INSENSITIVE1-like (VviGAI1) gain of function allele. While Vvigai1 vines display a dwarf phenotype 57 , this mutation does not alter flower morphology 37 . The 04C023V0006 (H/H) and 04C023V0003 (f/f) microvines, which were derived from a Grenache (H/f) x L1 microvine cross (H/f), were used for molecular and gene expression analyses. Gene-editing was performed in 04C023V0006. Gene expression studies was also performed in 03C003V0060 (M/f), which was derived from 00C001V0008 (selfed L1 Pinot Meunier microvine) x Richter 110 (V. berlandieri cv. Boutin B x V. rupestris cv. du Lot). Plant material used for fine genetic mapping was derived from F1 00C001V0008 (f/f) x Ugni Blanc (H/f) individuals and successive self-pollinated microvines (H/f) from first (S1) to fifth generation (S5) 37 . These mapping lines were maintained in glasshouses and growth rooms under long-days (16 h light/8 h dark) with 27°C and 22°C day and night temperatures, respectively. Vitis vinifera cultivars used in this study were derived from the cultivar collection at the CSIRO vineyard in Merbein, VIC, Australia.
DNA marker development and fine genetic mapping of the FSDR. Sequences for the SSR markers, UDV027, VVIB23, VMC3B10 and VMC6F1, flanking the previously identified SDR [22][23][24]58 , were used to identify the genomic region of this locus in the 8X PN40024 59 and Pinot Noir 60 genomes. The genomic sequences of annotated genes regularly spaced between UDV027 and VMC6F1 in PN40024 and Pinot Noir were scanned for candidate SNPs using the basic local alignment search tool (https://blast.ncbi.nlm.nih.gov). Thirty-three candidate SNPs were identified in exons and/or introns of selected genes in the region from VIT_202s0025g04920 to VIT_202s0154g00230 (Supplementary Table 1). The Agena Bioscience MassAR-RAY platform (Agena Bioscience, San Diego, California, USA) was used to validate the 33 SNPs by genotyping Vitis vinifera cultivars (n = 33) and self-pollinated microvines (n = 37) outlined in Supplementary Data 1. In addition, the F 1 00C001V0008 x Ugni Blanc mapping population (n = 101) was also genotyped with this SDR SNP set. The MassARRAY genotyping was performed at the Australian Genome Research Facility (AGRF; https://www.agrf.org.au/genotyping). In this procedure, DNA was isolated from immature leaves using the NucleoSpin ® 96 Plant II DNA extraction kit (Macherey-Nagel, Düren, Germany, #740663.4). PCR was used to amplify the 33 SNP-targeted regions in each of the DNA samples. After dephosphorylating excess nucleotides in the PCR reactions, a single base extension reaction was performed using an extension primer for each of the 33 SNPs. The extension reaction utilized a single mass-modified dideoxynucleotide complementary to each of the SNP alleles. Each SNP allele was identified by the distinct mass of the extension primer using MALDI-TOF mass spectroscopy analysis 61  To perform the fine genetic mapping, the positions of the 21 validated SNPs were identified in the SDR sequences of the Cabernet Sauvignon H-and fhaplotypes. This was achieved by aligning a 50 bp PN40024 genomic sequence containing each SNP to the Cabernet Sauvignon H-and f-haplotypes using the basic local alignment search tool (Supplementary Table 1). Sequence and gene organization of the Cabernet Sauvignon f-haplotype together with the genotype information shown in Supplementary Data 1 was used to delineate the boundaries of the FSDR.
RNA extraction and RT-qPCR. Total RNA was extracted from 100 mg of flowers derived from at least three inflorescences from a single 04C023V0006 (H/H), 04C023V0003 (f/f) and 03C003V0060 (M/f) plant at eight, six and four WPA ( Supplementary Fig. 8) 38 , using the Spectrum Plant Total RNA Kit (Sigma-Aldrich Pty. Ltd., Sydney, NSW, Australia, #STRN250). After RNA extraction, samples were purified using the RNA Clean & Concentrator™ (Zymo Research, Irvine, California, USA, #R1013). Total RNA was also extracted and purified from 100 mg of leaf tissue derived from 4 immature leaves, approximately 25 cm 2 in size, from 04C023V0006, 04C023V0003, and 03C003V0060. Three biological replicates were used for quantifying VviPLATZ1 transcript abundance in developing flowers at eight, six and four WPA, as well as leaves, for 04C023V0006 and 04C023V0003. For 03C003V0060, only two biological replicates were used for determining Vvi-PLATZ1 expression in developing flowers at eight, six, and four WPA, as well as leaves (Note: each biological replicate was derived from a single plant). First strand cDNA was synthesized from 1 µg of total RNA using the SuperScript TM IV Reverse Transcriptase (Thermo Fisher Scientific, Waltham, Massachusetts, USA, #18090010) at 53°C for 1 hr. After denaturation at 80°C for 10 min, the cDNA samples were diluted 1:10 in H 2 O. RT-qPCR was performed with the LightCycler ® 480 SYBR Green I Master mix (Roche Diagnostics, Mannheim, Germany, #04887352001) using 2.5 µl of diluted cDNA sample with gene-specific primers for VviPLATZ1 (VvPla_RT_F: CCCCTGTTTCTCTCCGAACT) and VvPla_RT_R: GCGCTTTTTCTTCACGAACT), VviUbi 62 and VviActin2 63 with a final concentration of 0.75 µM. Amplification was performed using the Roche LightCycler 480 system with the following parameters: 5 min at 95°C, 20 s at 59°C and 20 s 72°C (45 cycles), 5 min at 72°C, followed by a melt cycle of 15 s at 95°C, 45 s at 50°C, continuous heating to 95°C at 11°C s-1 64 . To calculate DNA copy number, the amplified PCR product for VviPLATZ1, VviUbi and VviActin2 was cloned into Qiagen pDRIVE vector using the Qiagen PCR Cloning Kit (Qiagen, Chadstone, Victoria, Australia, #231222). After sequencing the clones, the DNA concentration (ng/µl) for each construct was determined using the Quanti-iT™ PicoGreen ® dsDNA Quantification Kit (Promega, Madison, Wisconsin, USA, #E2670) and the GloMax ® Discover Microplate Reader (Promega Madison, Wisconsin, USA, #GM3000). DNA copy number was calculated in each standard dilution 65 . After amplification, a standard curve derived from the absolute DNA copy number values from the dilution series for each gene was used to calculate transcript copy number for VviPLATZ1, VviUbi and VviActin2 in each of the samples using the Roche LightCycler 480 software. Transcript copy number for VviPLATZ1 was normalized against reference genes, VviUbi and VviActin2 for each sample.
The sections were mounted with glycerol and imaged using an optical microscope (Zeiss Axioskop2, Oberkochen, Germany). Three biological replicas were derived from 04C023V0006, 04C023V0003 and 03C003V0060 plants. Each biological replicate was derived from a single inflorescence.
CRISPR/Cas9 vector construction. The VviPLATZ1 DNA sequence was scanned for 20 bp guide sequences followed by the NGG protospacer adjacent motif (PAM) using Benchling (https://benchling.com), which also calculated on and off target scores 67,68 . The two guide RNA sequences, FS1 and FS4, that targeted exon 2 were selected, as gene-edited mutations would result in the production of truncated proteins lacking the conserved PLATZ domain. These guide RNAs were also selected for their high cleavage efficiency using the Guide-it sgRNA In Vitro Transcription and Screening kit (Takara Bio USA, Inc).
The CRISPR-associated protein-9 nuclease gene (Cas9)-single guide RNA (sgRNA) cassette was synthesized at GenScript Biotech (Piscataway, NJ, USA) then subcloned into the pCLB1301 binary vector (Supplementary Fig. 9) 37 . The Streptococcus pyrogenes Cas9 sequence containing nuclear localization sequences and the potato IV2 intron 69 was codon optimized for Vitis. Cas9 transcription was controlled by the Arabidopsis UBIQUITIN1 promoter and terminator 70 , which drives expression in vegetative and reproductive tissues 71 . The UBIQUITIN6 promoter was used to drive the expression of FS1 and FS4 sgRNAs 72 . The two binary vectors pVCAS9FS1 and pVCAS9FS4 contained the FS1 or FS4 guide RNA, respectively. Lastly, these binary vectors contained the hygromycin resistance gene (HygR), and the endoplasmic reticulum localized green fluorescent protein (GFP-5ER) reporter for selecting transgenic plants 73 .
Transformation and identification of gene-edited VviPLATZ1 plants. To edit VviPLATZ1, pVCAS9FS1, and pVCAS9FS4 were transformed into the Agrobacterium tumefaciens strain EHA105 and selected using the kanamycin resistance gene, KanR. Next, transgenic 04C023V0006 (H/H) plants were produced by incubating the EHA105 strains with somatic embryogenic callus derived from anthers 74 . Transgenic somatic embryos were selected using the HygR and GFP-5ER 75 . Shoot growth was induced in transgenic somatic embryos with 5 µM 6-Benzylaminopurine (Sigma-Aldrich, Sydney, New South Wales, Australia, #B327). After shoots developed, root growth was induced with 0.5 µM 1-naphthaleneacetic acid (Sigma-Aldrich, Sydney, New South Wales, Australia, #N0640). Finally, rooted transgenic plantlets maintained in tissue culture, were propagated, potted into soil and transferred to the glasshouse and/or growth chamber for further analysis.
To identify gene-edited plants, DNA isolation was performed on immature leaves from GFP-positive plantlets. The primers Pla_F (ATAAGGCTCAACCCCCACTT) and Pla_R (ACACCCCAATAAAACGCAAA) were used to PCR amplify the target regions for the guide RNAs in the second exon of VviPLATZ1. Amplicon Sanger sequencing followed by chromatogram analysis was used to identify indels in the GFP-positive plants. Using the next-generation sequencing service at AGRF, geneedited VviPLATZ1 were analyzed further using the MiSeq sequencing system (Illumina, San Diego, CA, USA). Subsequently, MiSeq sequencing data were processed through the CRISPResso pipeline to identify indels 76 . Mutation efficiency was calculated by counting the number of sequence reads with an indel at the guide RNA target region divided by the total number of sequence reads. The value was corrected by subtracting the data obtained for the untransformed 04C023V0006 (H/ H) control. After this analysis was completed, seven first-generation plants (T 0 ) containing gene-edited alleles were identified and self-pollination was performed to evaluate the flower sex phenotype in T 1 plants. After DNA isolation from T 1 root and cotyledon tissues, the guide RNA target region was amplified by PCR using Pla_F and Pla_R primers. Sanger sequencing performed at AGRF was used to genotype the embryos to identify T 1 plants homozygous for the mutated alleles (Vviplatz1-1/2-8).
Flower sex determination. Phenotyping for flower sex was performed by morphological scoring using the OIV descriptor, No 151, of recently opened flowers (https://www.oiv.int/public/medias/2274/code-2e-edition-finale.pdf). The architecture of each stamen was scored as (1) elongated and upright, (2) reflexed or (3) stunted and upright. A minimum of 30 flowers from at least 2 inflorescences were scored from each genotype. Images of open flowers and stamens were taken with a Spot 15.2 64 Mp camera mounted on a Zeiss Stemi 2000-C dissecting microscope (Zeiss Vision Australia Pty Ltd, Tonsley, SA, Australia) using the Spot V5.1 software (SPOT Imaging, Sterling Heights, MI, USA). Stamen length was measured in flower images from 04C023V0006 (H/H), 04C023V0003 (f/f) and T 1 homozygous plants using the ImageJ.app. (https://imagej.net/Fiji/Downloads). Analysis of variance and Tukey's honest significant difference analysis was performed using standard statistical packages in R.
Statistics and reproducibility. Two-tailed Student's t test was used to determine statistical differences with p values provided for all comparisons. This information is provided in figure legends with raw data provided in the Source Data file. Measurements were taken from multiple biological samples. For in situ hybridization, three independent inflorescences were fixed, processed, sectioned, and hybridized in separate experiments with similar results produced and a representative image displayed in Fig. 2.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All relevant data presented in this manuscript are available within the article and its Supplementary Information files. The GenBank sequence identifier for VviPLATZ1 derived from 04C023V0003 is MW548436. For Vviplatz1 derived from 04C023V0003, the GenBank sequence identifier is MW548435. Source data are provided with this paper.