For in situ hybridization and RT-qPCR, inflorescences at phenological developmental stages B, D, G, and H (Baggiolini, 1952) were collected from all V. v. sylvestris parental individuals of a collection, composed by 22 female (F) and 11 male (M) individuals. The same developmental stages were sampled from 12 hermaphrodites V. v. vinifera cv. Touriga Nacional (Her) (Figure 1) in Dois Portos (Lisbon district, Portugal) developmental stages B–D could not be used due to the woody nature of their tissues. However, at later stages (D and subsequent) stages the plants have several developmental flower stages within the same inflorescence.

Individual closed and open flowers (without flower cap) were collected from the three flower types at the later stage H, and carpels were dissected from open flowers and used in RT-qPCR (Figure 1). Additionally, leaves from V. v. sylvestris and V. v. vinifera, were collected for DNA extraction. Samples were collected during April (2014 and 2015).

The DNA was extracted with the DNeasy^® Plant Mini Kit (Quiagen, USA) following the manufacturer’s instructions. Total RNA was extracted with the Spectrum^TMTotal RNA Kit (Sigma–Aldrich, Inc, USA) following the manufacturer’s instructions. The obtained DNA was stored at 4°C and the extracted RNA stored at -80°C between handling. The DNA and RNA concentration was determined using a Synergy HT Nanodrop system (Biotek, Germany), with the software Gen5^TM (Biotek, Germany).

cDNA from male, female and hermaphrodite inflorescences from the four flower developmental stages, as well as cDNA from closed flowers, open flowers and from carpels (Figure 1), was synthesized with the RETROscript^® kit (Ambion, Life Technologies, Spain), following the manufacturer’s protocol and amplified by PCR with specific primers (Table 1).

To determine the most efficient cDNA concentration to be used for RT-qPCR, a serial of decimal dilutions was tested. Amplification reactions were performed with two biological samples per phenological stage in triplicates containing 5 μL of master mix (SsoFast_EvaGreen Supermix, Bio- Rad, USA), 0.4 μM of specific primers and 0.21 μg of cDNA in a 20 μL reaction, according to Ramos et al. (2014). The following program was applied: initial polymerase activation, 95°C, 2 min; 40 cycles at 95°C for 15 s (denaturation); 57°C for 30 s (annealing); 76°C for 30 s (extension) with a fluorescence reading at the end of each cycle. The run was completed with a melting curve analysis to confirm the amplification specificity. To confirm amplicon size, RT-qPCR products were run on 2.2% (w/v) agarose gel. RT-qPCR runs were performed, for each gene, with decimal dilutions series of a precisely calculated number of plasmid copies to create a calibration ruler. Cqs (threshold cycles) obtained from RNA samples was matched against the calibration ruler to estimate the number of transcripts. The absolute number of transcripts was calculated using the following formula: Number of transcripts = C × NA/M; where number of transcripts corresponds to the number of molecules μL^-1 contained in the purified cDNA; C, concentration of the purified cDNA (g μL^-1); M, the molecular weight of the cDNA gene fragment; NA, Avogadro’s number = 6.023 × 1023 molecules mole^-1 (Whelan et al., 2003).

The sequence of ADENINE PHOSPHORIBOSYILTRANSFERASE (VviAPRT) gene was first obtained from Genoscope 8x genome version (GSVIVT00007310001)³. Since a blast with this gene sequence against Arabidopsis thaliana database (TAIR⁴) revealed great homology with AtAPRT3 (Fechter et al., 2012), from here on the gene will be referred to as VviAPRT3. Sequences from genes belonging to chromosome 2 in the previous identified 143 kb region (between 4,907,434 and 5,050,616 bp) (Fechter et al., 2012; Picq et al., 2014) and unknown chromosome were obtained from CRIBI Grape Database annotation 12x_v2.16⁵ (Supplementary Table 1). For clarity, the meaning of acronyms used for gene annotation is explained; for instance, VIT_202s0154g00200, referred here as VviFSEX, was identified on Vitis vinifera subs vinifera genome, on the sequencing 12x version v2.1 annotation (VIT_2), on chromosome 2 (02), scaffold s0154g and received the gene number 00200.

Genomic DNA and cDNA sequences of VviAPRT3 and VviFSEX genes from the three flower types (male, female and hermaphrodite) were cloned into pGEM^® T-easy vector system (Promega, Leiden, The Netherlands) and transformed into Escherichia coli DH5α competent cells. After plasmid isolation (PureLink^TM Quick Plasmid Miniprep Kit, Invitrogen, Carlsband, CA, USA) and sequencing, the sequences were aligned using Clustal Omega (Sievers et al., 2011; McWilliam et al., 2013; Li et al., 2015).

Plant tissue fixation, clearing, and in situ hybridization experiments were performed as previously described (Coen et al., 1990) with the following modifications: for the probe synthesis the cDNA of VviAPRT3 and VviFSEX (VIT_202s0154g00200) were cloned into the pGEM^® T-easy vector system and amplified by PCR with the M13 forward/reverse primers and specific forward/reverse primers (Table 1). The PCR product was purified using the MinElute PCR Purification Kit (QIAGEN, Valencia, CA, USA), according to the manufacturer’s instructions and used as template for the riboprobe synthesis, which was carried with SP6 and T7 RNA polymerase to obtain the sense and antisense strand. The paraffin embedded material was sectioned at 7 μm and the tissue slices mounted with distilled water. Images were captured with a fluorescence microscope (Wild Leitz, Laborlux S) with an AxioCam HRM (Zeiss).

For the statistical analysis, expression values of VviAPRT3 and VviFSEX were transformed into log₂ and tested through ANOVA using the program Graphpad Prism 5 (GraphPad Software, Inc.).

The samples with a p-value of the ANOVA lower than 0.05 were submitted to an additional Tukey test. The statistically significant differences were accepted when Tukey’s test p-value was lower than 0.05.

Previous work suggests the existence of two alleles on chromosome 2 of V. v. vinifera due to the high homology of this region with one of the unknown chromosome (Picq et al., 2014). Due to the putative existence of two alleles, we have analyzed all the genes present in the 143 kb region that are assigned to the scaffold_154 of chromosome 2 genome version 12x_v0 (VIT_202s0154gxxxxx) (Supplementary Table 1). The analysis showed eleven genes with high homology with genes present in scaffold_233 of the unknown chromosome (VIT_200s0233gxxxxx) (Supplementary Table 1). A pairwise alignment between homologous genes in both scaffolds revealed some sequence differences that were used to design PCR primers able to discriminate these genomic loci and, potentially, providing molecular markers for plant sex identity. Subsequently, these primer pairs were used to genotype 22 female, 11 male and 12 hermaphrodite plants. Based on the presence/absence of the amplified PCR products and their respective sizes, we noticed that only VIT_202s0154g00200 (VviFSEX) displays the differences predicted in silico. Two fragments with 449 and 413 bp were amplified from the DNA of male and hermaphrodite plants while a single fragment of 449 bp was amplified from female plants (Figure 2A). This genotyping approach allows the discrimination of female from male and hermaphrodite plants, and thus the VviFSEX gene sequence may be used as a genetic marker for female plants. The exact size of PCR products amplified from male, female and hermaphrodite was determined by sequencing analysis. This analysis showed that the 449 bp fragment present in all the flower types, despite being homologous to the 413 bp fragment of male and hermaphrodite individuals, contains an extra 36 bp region (Supplementary Figure 1). The sequence with the additional 36 bp may correspond to an alternative allele of this gene, being homozygous in female plants and heterozygous in male and hermaphrodite plants. Picq et al. (2014) had already suggested that a region of unknown chromosome homologous to the 143 kb region of chromosome 2 (Fechter et al., 2012) may correspond to different alleles of the same genes. In the present work we also identified several genes in the same scaffold of unknown chromosome with great homology to genes found on chromosome 2 (Supplementary Table 1). The comparison between the genomic sequences revealed some differences that suggest the existence of possible alleles.

It is important to note that the existence of an unknown chromosome in the reference genome (PN40024) indicates that the annotation is not concluded and it requires an additional effort to improve gene chromosome assignment (Jaillon et al., 2007).

According to Fechter et al. (2012), the VviAPRT3 gene was located in a 143 kb region and it might have a role in flower sex specification. These authors used the second intron of VviAPRT3 as a marker able to differentiate female plants of several different Vitis species, including V. v. vinifera and V. v. sylvestris subspecies (Fechter et al., 2012). However, it was not possible to amplify the VviAPRT fragment previously described Fechter et al. (2012) in the plant material used in the current work, since the primers or the gene sequences were not provided. The VviAPRT3 marker was located on the chromosome 2 of the reference genome between 4,986,517 bp and 5,025,265 bp (Fechter et al., 2012). In the 8x genome reference sequence on Genoscope database⁷, VviAPRT3 was located 200 kb downstream of the reported location (GSVIVT00007310001). In fact, the sex region of 143 kb in the 12x_v2.1 was located roughly 100 kb downstream in the 8x version of the reference genome, although most of the genes were kept together (Supplementary Table 2). However, in the 12x_v0 and v2.1 versions, VviAPRT3 was allocated to the unknown chromosome (VIT_200s1847g00010) (Supplementary Table 2) with a different genomic structure (Supplementary Figure 2). In Vitis, the expression levels of VviAPRT3 was first reported in the transcriptome assembly of different stages of flower development (Ramos et al., 2014). The transcriptomic data revealed that VviAPRT3 gene is differentially expressed in male, female and hermaphrodite plants. The putative involvement of this gene in grapevine sex specification and the disagreement between annotation and sequencing versions led us to further investigate its genomic structure. To determine which annotation version stands correct regarding VviAPRT3 structure, several primers were designed to amplify VviAPRT3 (Supplementary Figure 2). PCR amplification of cDNA with primers for the first and last exons (F1 + R2) originated a fragment of 443 bp in the three flower types (Supplementary Figure 2). cDNA sequencing revealed no differences between the fragments amplified from the different flower types (data not show). Amplification with primers for the third and fourth exons (F2 + R1) resulted in a fragment of 689 bp, while the amplification with primers for the fourth and sixth exons (F2 + R2) resulted in a fragment of about 3,000 bp (Supplementary Figure 2). By using these specific primer combinations (Supplementary Figure 2), and sequencing the corresponding amplified fragments, we concluded that VviAPRT3 genomic structure, with six exons, five introns and a 2x repeat in the second intron, was accurately represented in the 8x genome version. Therefore, the VviAPRT3 sequence subject of this work and VviAPRT from Fechter et al. (2012) are most likely the same gene (Supplementary Table 2). Fechter et al. (2012) reported that VviAPRT3 was able to discriminate female flowers from the other flower types based on a 2x repeat present in the second intron in male and hermaphrodite plants. Thus, we designed primers flanking the 2x repeat in the second intron (F3 + R3) of VviAPRT3 (Table 1; Supplementary Figure 2) that amplify the tandem repeat region in the three flower types. Instead of distinguishing female from other plants, we were able to distinguish male from female and hermaphrodite plants. These plants displayed a fragment of 541 bp (Figure 2B), while male plants showed an additional fragment with 611 bp (Figure 2B). By sequencing the 541 bp fragments from each flower type it was possible to demonstrate that their genomic sequences are similar (Supplementary Figure 3). The fragment of 611 bp found solely in the male plants results from an insertion of 70 bp that is absent in female and hermaphrodite plants (Supplementary Figure 3).