Mis-expression of a PISTILLATA-like MADS box gene prevents fruit development in grapevine

Authors

  • Lucie Fernandez,

    1. Instituto de Ciencias de la Vid y del Vino (Consejo Superior de Investigaciones Cientificas, Universitad de la Rioja, Gobierno de La Rioja), Logroño, Spain
    2. Montpellier SupAgro, Unité Mixte de Recherche Amélioration et Génétique de l'Adaptation des Plantes, Montpellier Cedex, France
    Current affiliation:
    1. Institut National de la Recherche Agronomique, UMR Biologie du Fruit et Pathologie, Villenave d'Ornon Cedex, France
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Jamila Chaïb,

    1. Commonwealth Scientific and Industrial Research Organisation-Plant Industry, Glen Osmond, SA, Australia
    Search for more papers by this author
    • These authors contributed equally to this work.

  • José-Miguel Martinez-Zapater,

    1. Instituto de Ciencias de la Vid y del Vino (Consejo Superior de Investigaciones Cientificas, Universitad de la Rioja, Gobierno de La Rioja), Logroño, Spain
    Search for more papers by this author
  • Mark R. Thomas,

    1. Commonwealth Scientific and Industrial Research Organisation-Plant Industry, Glen Osmond, SA, Australia
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Laurent Torregrosa

    Corresponding author
    1. Montpellier SupAgro, Unité Mixte de Recherche Amélioration et Génétique de l'Adaptation des Plantes, Montpellier Cedex, France
    • Instituto de Ciencias de la Vid y del Vino (Consejo Superior de Investigaciones Cientificas, Universitad de la Rioja, Gobierno de La Rioja), Logroño, Spain
    Search for more papers by this author
    • These authors contributed equally to this work.


For correspondence (e-mail laurent.torregrosa@supagro.inra.fr).

Summary

The FLESHLESS BERRY (Flb) somatic variant identified in the grapevine cultivar Ugni Blanc develops grape berries without flesh, suggesting a role for the altered gene in differentiation of flesh cells. Here we describe identification of the molecular defect responsible for this phenotype. Using a combination of genetic and transcriptomic approaches, we detected the insertion of a miniature inverted-repeat transposable element in the promoter region of the PISTILLATA-like (VvPI) gene, the grapevine homologue of Arabidopsis PISTILLATA. The transposon insertion causes specific ectopic expression of the corresponding VvPI allele during early fruit development, causing expression of genes specific for petal and stamen development within the fruit. A causal relationship between the insertion and the phenotype was demonstrated by phenotypic and molecular analyses of somatic revertants showing that ectopic expression and mutant phenotype were always linked to the presence of the transposon insertion. The various phenotypic effects of the flb mutation on ovary morphology, fruit set and fruit development, depending on the cell lineage affected, are presented for each phenotype, offering new insights into floral and fleshly fruit development. The results highlight the importance of VvPI repression after fertilization to achieve normal fleshy fruit development, and the complex genetic, genomic and cellular interactions required for the flower to fruit transition in grapevine.

Introduction

Fruits represent a key evolutionary innovation for seed dissemination in angiosperms, showing wide phenotypic diversity resulting from a range of morphogenetic processes (Ferrandiz, 2011). Fruit development is critical to sustain plant reproduction by protecting the offspring during early developmental stages, and also plays a major role in dispersion of mature seeds. Some fruit features depend on specific pre-anthesis flower traits. For instance, tomato size variation is determined by differences in the number of ovary wall cells and carpels (Frary et al., 2000; Munos et al., 2011). Nevertheless, many changes associated with fruit development and size variation result from post-pollination events, such as cell division, endo-reduplication and cell expansion through vacuolar storage. These changes are critical in some fruits, such as fleshy fruits, where the final fruit volume may be 3 × 105 greater than the initial ovary volume (Coombe, 1976).

Knowledge regarding the genetics of flower and fruit development has improved substantially in recent years (Giovannoni, 2004; Reeves et al., 2012). Fruit development mutants have been identified in several species, and the identification of underlying genes and gene variants provides new insights into the mechanisms associated with fruit development. In tomato, several studies have demonstrated that forward genetics is a successfull approach to identify genes involved in plant domestication and improvement processes. For instance, the OVATE gene, encoding a regulatory protein, was shown to be involved in the regulation of fruit shape (Liu et al., 2002), and one fw2.2 allele controls fruit size through pericarp cell division patterning in the ovary (Cong et al., 2002). However, fleshy pericarp formation from the fertilized ovary remains poorly documented despite being the major difference between dry and fleshy fruits. In grapevine, the fleshless berry (flb) mutation prevents fleshy pericarp development when present in the L2 cell layer (Fernandez et al., 2006a,b). When also present in the L1 cell layer, the flb mutation impairs fruit set (Fernandez et al., 2006a). The shoot apical meristem is organized in histogenic cell layers, including the L1 and L2 layers found in most lateral organs and displaying an almost independent cell proliferation. The flower and ovary maintain the same cell layered structure, well described in Arabidopsis (Jenik and Irish, 2000), with L1 contributing to the outer and inner epidermis (the latter lining the cavity of locules) and L2 contributing to carpel walls, gametes and embryo sacs. To date, such mutant phenotypes have not been described in other fleshy fruit species, and identification of the flb mutation should advance our understanding of the molecular mechanisms of fleshy fruit morphogenesis.

Here, we report identification of the mutation responsible for the original Flb phenotype observed in both fruit and flowers. Using a combined genetic and transcriptomic approach, a transposable element insertion in the promoter of the VvPI gene (NCBI Gene ID 100232978) was found to cause its ectopic expression. VvPI mis-expression in L2-derived cells during the first stages of fruit development appears to prevent normal development of berry flesh, probably driving alternative cell differentiation pathways in this tissue. Both VvPI mis-expression and Flb phenotypes disappear after reversion of the transposon insertion. The phenotypes of the mutation when present in the L1, L2 or both cell layers are discussed in relation to cell layer-dependent regulation of ovary differentiation.

Results

Flb and revertant phenotypes are cell layer-specific

The Flb somatic variant of cultivar Ugni Blanc displaying a fleshless berry phenotype was previously characterized as a periclinal chimera caused by a dominant mutation affecting the L2 cell layer (Figure 1) (Fernandez et al., 2006a,b). The Flb somatic variant was shown to have impaired cell division and enlargement of the ovary wall after fertilization. However, Flb mutant plants derived from a cross between cultivars Chardonnay and the Flb somatic variant and presenting the flb mutation in all cell layers failed to develop fruits (Figure 1) (Fernandez et al., 2006a).

Figure 1.

Cell layer structure of the grapevine apical meristem. The Flb somatic variant corresponds to a periclinal chimera with the flb mutation present in the L2 cell layer, while Flb mutants in the F1 are not chimeric with the mutation present in all cell layers.

In order to further advance the genetic analysis of the flb mutation, we generated an additional F1 segregating microvine population by pollinating a picovine line (Chaib et al., 2010) using the Ugni Blanc Flb somatic variant. As previously observed in the previous F1 Chardonnay × Flb population, approximately half of the F1 plants displayed typical mutant flowers with mis-shapen pistils, in agreement with the dominance observed for this phenotype (Figure 2a–c). In both progeny, the Flb mutant plants failed to develop fruit (Figure 2a), while the Flb chimeric parent was able to produce fleshless fruit with normal seeds. The fruit developmental failure segregating in the F1 was not caused by pollination defects, as pollen germination behaviour within the pistil was found to be normal after controlled fertilization assays (Figure 2d–f). Moreover, pollination of wild-type emasculated flowers using pollen from Flb mutant phenotype plants led to normal fertilization, seed and fruit development, indicating that pollen viability and function were normal. On the other hand, pollinated flowers of Flb mutants remained viable until 21 ± 3 days after emasculation, while normal flowers that were not pollinated usually dropped within a few days. Light microscopy observation showed that ovules in these Flb mutant flowers were able to initiate development, evidenced by an increase in size, embryo sac expansion, and, in some cases, formation of a nucellus (Figure 2g–l). No difference in integument growth was observed compared to ovules at equivalent dates. However, ovules from Flb mutant ovaries showed signs of degeneration of the embryo sac and nucellus (Figure 2l). Because flower viability was limited to 3 weeks after pollination in Flb mutants, ovules were not observed at later stages. Altogether these observations suggest that degeneration of the internal ovule tissues may limit subsequent embryo differentiation, resulting in fruit developmental failure after fruit set.

Figure 2.

Flower and ovule development in the F1 microvine progeny. (a) Inflorescences 6 days after pollination with Grenache pollen. Left: normal fruit set on the female picovine parent; centre: no fruit set on female Flb mutant; right: no fruit set on hermaphrodite Flb mutant. Scale bar = 2 mm. (b, c) Pollen delivery at flowering on wild-type hermaphrodite (b) and Flb mutant hermaphrodite (c) genotypes. Scale bar = 2 mm. (d–f) Pollen germination and tube growth on hermaphrodite Flb mutant 24 h after pollination. (g–l) Ovule development in the wild-type F1 genotype at anthesis (g, j) and 21 days after pollination (h, k), and in Flb mutant ovaries at 21 days after pollination (i, l). Arrowheads in (l) indicate degeneration of the nucellus and the embryo sac while the funiculus remains normal. Scale bars = 1 mm (g–i), 100 μm (j, k) and 250 μm (l).

Both the Flb somatic variant and Flb mutant plants showed high genetic instability. Reversion towards production of wild-type berries was previously reported for the Flb somatic variant (Fernandez et al., 2006a). However, Flb mutant plants derived from Flb crosses displayed two reversion phenotypes: reversion towards the production of fleshless berries and towards the production of wild-type berries (Figure 3). Similar somatic reversion ratios were observed in the microvine F1 population, with 29 of 126 (23%) phenotyped Flb mutant plants exhibiting reversion events, and in the Chardonnay x Flb somatic variant, with 10 of 28 (36%) plants exhibiting reversion events. Reversions affected one to many grouped berries inside the same cluster or the whole cluster (Figure 3). Reversions towards wild-type berries were more frequent than reversions towards fleshless berries.

Figure 3.

Fruit phenotype in wild-type and Flb hermaphodite F1 mutants showing reversion events. (a) Wild-type fleshy fruit development. (b) Reversion towards fleshless fruit development affecting almost the whole inflorescence. The black arrowhead indicates the expected mutant blocked fruit set phenotype for F1 progeny. (c) Chimeric bunch including reversion towards fleshless fruit development and a single wild-type fruit (white arrowhead) on the same inflorescence. (d) A typical Flb mutant inflorescence that failed to produce fruit except for a single flower that had reverted and produced a fleshless fruit (white arrowhead). (e) Reversion towards groups of wild-type fruit on a Flb mutant inflorescence. Scale bars = 5 mm.

Reversion events in the somatic variant are restricted to the L2-derived cells that carry the mutation, and always give rise to regular wild-type berries. However, reversions in the Flb mutant plants affected one or both of the two cell layers. Reversion events affecting the L1 of the Flb mutants resulted in phenotypes equivalent to those of the Flb somatic variant (Figure 3b,d). On rare occasions, a whole cluster reverting to the fleshless berry phenotype included a small group of wild-type berries (Figure 3c). This result may suggest the occurrence of successive reversions, first in the L1 layer and then in the L2 cell layer. As independent reversion events affecting both the L1 and L2 cell layers or cell layer invasion are rare events (Jenik and Irish, 2000), they do not explain the frequency of reversion towards wild-type berry observed (Figure 3e). It is possible that reversions in the L2 cell layer result in wild-type berries, given that this is the other phenotype observed in mutants showing phenotype reversions.

The grapevine PISTILLATA homologue co-localizes with the fleshless mutation and is over-expressed in plants displaying the Flb phenotype

Previous genetic analyses located the flb mutation at one extreme of linkage group 18 (LG18) using the Chardonnay × Flb variant population (Fernandez et al., 2006a). To refine the location of the flb mutation, we used the larger segregating F1 microvine population. Given the possibility that meiotic reversion may generate phenotypically wild-type plants carrying the mutant locus haplotype, we only considered for mapping purposes those plants (126) that displayed the Flb phenotype. Genetic analysis confirmed the position of the flb mutation in LG18 between the VVMT93 and VVMT98 markers (Figure S1). These two markers are 410 kb distant in the physical map of LG18. This region contained 26 ORFs. VVMT98 is a SNP marker within the exonic region of the PISTILLATA-like gene, previously identified as over-expressed in the Flb somatic variant (Fernandez et al., 2007). The PISTILLATA-like MADS box gene in Vitis vinifera has the NCBI Gene ID 100232978 and gene symbol MADS9, and has been referred to in previous papers as VvMADS9 and VvPI (Sreekantan et al., 2006; Poupin et al., 2007). In this paper, we refer to it as VvPI.

In order to obtain more comprehensive gene expression results, we used the 23K GrapeGen Affymetrix GeneChip® to re-analyse transcriptional changes taking place in the first stages of Flb fruit development. Among the 85 genes differentially regulated in the Flb somatic variant (Table S1), VvPI gene was the only one that co-localized with the FLB locus (Figure S1). Moreover, 34 F1 plants in the microvine population displaying the Flb phenotype were analysed for VvPI expression, and showed ectopic expression of this gene in their leaves (Figure S2). Given the complete association observed between ectopic VvPI expression and the Flb phenotype, as well as the genetic co-localization of the flb mutation with this gene, we performed a detailed sequence analysis of VvPI in the somatic variant to identify molecular changes that may be responsible for the Flb phenotype.

The Flb phenotype is caused by a transposon insertion in the VvPI promoter

We analysed the VvPI gene sequence in Ugni Blanc and the Flb somatic variant. Because grapevine cultivars are highly heterozygous at most of their loci, we cloned and sequenced VvPI in both genotypes to identify two alleles that were named as VvPIa and VvPIb. A 5 kb genomic fragment was sequenced for each allele, including 2.4 before and 2.6 kb after the ATG sequence. Both genotypes contained the same two alleles, which displayed 109 nucleotide polymorphisms among them. Most of those polymorphisms (93) were located in the distal promoter regions, and 16 of them were found after the ATG (Figure 4a). The only two polymorphisms located in exon regions corresponded to synonymous substitutions. Genotyping of the VvPI allele linked to the Flb phenotype in the microvine population confirmed the association of the phenotype with the VvPIb allele. In addition to VvPIa and VVPIb, a third VvPI allele sequence was identified in the Flb somatic variant, as a VvPIb derivative sequence containing a 1.5 kb DNA insertion in a position 730 bp upstream of the ATG (Figure 4a). This DNA insertion corresponded to a transposon insertion that was specific to the VvPIb allele in the Flb somatic variant. It was not detected in the original Ugni Blanc plant, in other clones of this cultivar such as Ugni Blanc 384, or in the Chardonnay and the picovine 00C00V0008, which are the other parents of the studied progenies. This third VvPI allele specific to Flb plants was named VvPIbm (Figure 4a).

Figure 4.

VvPI sequence and organization in the Flb somatic variant and wild-type Ugni Blanc (UB). (a) VvPI structure with the exons and UTR indicated in yellow and grey, respectively. The position along the VvPI gene is referenced to the ATG codon. Polymorphisms between VvPIa and VvPIb alleles are indicated by blue and red lines for SNPs and indels, respectively. The organization and detailed positions surrounding the flb mutation region are shown for the VvPIb, VvPIbm (mutated) and VvPIbr (revertant) alleles. TSD, target site duplication; TIR, terminal inverted repeat. (b) Amplification of the VvPIb allele in UB and Flb leaves and in revertant fruit tissue of Flb (FLB rev). The sample cell layer composition is indicated as L2 [mesocarp (flesh)] or L1/L2 [exocarp (skin) with minor flesh contamination]. Electrophoresis (1% agarose gel) of multiplex amplification of VvPI alleles with (726 bp) or without (1237 bp) the transposon. (c) Schematic representation of VvPIb allele types in wild-type, Flb somatic variant and revertant fruits according to presence or absence of the transposon in the L1 or L2 cell layers.

Complete sequencing of the inserted DNA revealed a 1506 bp sequence flanked by a 8 bp target site duplication (CCTACAAG) and 18 bp imperfect terminal inverted repeats [TIRs; CA(A/T)GGATTGAAA(T/A)ATCGG] that are characteristic of class II transposable elements (Figure 4a). Alignment of this sequence against the transposable element sequence database from Repbase (http://www.girinst.org/repbase/) showed 94% homology with the unclassified TE-7-1_Vv element, also known as Mila (Benjak et al., 2009). Subsequently, we named the element inserted in the VvPIbm promoter as Mila-flb. BLASTN sequence comparisons (http:/1blast.ncbi.nlm.nih.gov) with the PN40024 genome sequence (12×) showed high sequence similarity (>98.5%) with 22 sequences, all presenting identical TIRs and localized at various chromosomal positions (Table S2). All the sequences were similar in size and did not contain a putative transposase ORF. These elements were generally found close to genes and in distinct gene regions: promoter, UTR, 3′ end gene or introns (Table S2). However, no insertion was detected within coding regions.

Mila-flb appears to be a defective element without gene-coding capacity based on sequence analysis; therefore, it has characteristics of a non-autonomous active mobile element in the Ugni Blanc genome that is probably mobilized by the presence of an active element encoding a functional transposase.

VvPI expression is the cause of the Flb phenotype

Further expression analysis of VvPI in the Flb somatic variant revealed its ectopic expression in a range of organs, including fruit, shoot and petiole, as well as young and adult leaves, while VvPI expression in wild-type plants is generally restricted to flower organ development (Sreekantan et al., 2006). In order to show a causal relationship between insertion of the element and VvPI ectopic expression and between this expression and the mutant phenotype, and exclude the possibility that ectopic expression and the mutant phenotype are due to another mutation, we analysed fleshy berries that developed in Flb clusters of the somatic variant. If the Mila insertion is the cause of the Flb phenotype, revertant tissue should show loss of the transposon insertion as well as restored wild-type VvPI expression in the fruit. Moreover, it was important to confirm that VvPI over-expression specifically affected the VvPIb allele that is linked to the transposon insertion. Due to the difficulty in obtaining pure L1-derived berry tissue compared to pure L2-derived pulp tissue, we performed these studies in the Flb somatic variant.

To focus the analysis on the mutated allele, we designed allele-specific primers to amplify the sequence of the VvPIb allele. As shown in Figure 4(b,c), the presence of Mila-flb in the VvPIbm promoter was only detected in the Flb somatic variant, while the size of the amplified DNA fragment in revertant berries was identical to the Ugni Blanc amplicon. However, the sequence of the PCR products from revertant berries revealed the presence of a short tandem 8 bp duplication corresponding to duplication of the target site duplication, which is a typical sequence footprint left by these transposons after excision (Figure 4a). The allele with the 8 bp duplication footprint was called VvPIbr.

To understand the effect of Mila-flb insertion (VvPIbm) and excision (VvPIbr), we investigated their consequences on VvPI expression. If the transposable element is the cause of the observed ectopic expression of VvPI in young fruits of the somatic variant, this expression should be allele-specific and restricted to the allele carrying the insertion in cis in its promoter region. In order to quantify the expression of each VvPI allele, we developed allele specific quantitative RT-PCR assays and analysed the transcriptional activity of both VvPI alleles in inflorescences, as a positive control for gene expression, and in berries of Ugni Blanc, the Flb somatic variant and revertant berries (Figure 5). Both the VvPIa and VvPIb alleles showed a high level of expression in late stages of flower development, with both being slightly up-regulated in Flb compared to the wild-type Ugni Blanc. However, VvPI expression in fruit was only detected in the Flb somatic variant and corresponded exclusively to expression of the VvPIb allele. No expression of VvPI was detected in revertant berry tissues. These results demonstrated that revertant tissues of the somatic variant have lost the Mila-flb element in the VvPI gene and VvPI ectopic expression in fruit.

Figure 5.

VvPIb allele-specific expression. Total VvPI, VvPIba-specific, VvPIb-specific quantitative RT-PCR expression in flowers and berries of wild-type Ugni Blanc (UB) and the Flb somatic variant (Flb), and in the skin and flesh of flb revertant berries (Flb rev). Values are means and bars correspond to SD for technical triplicates.

Finally, sequencing of ectopic VvPI mRNA from Flb somatic variant leaves showed only the VvPIb allele and did not reveal any new transcription initiation sites or post-transcriptional modifications compared to previously published sequences.

VvPI expression in fruit promotes alternative cell identity

Transcriptional changes associated with the Flb phenotype may provide information on the gene(s) altered in the somatic variant. Despite the large number of genes screened with the 23K GrapeGen Affymetrix GeneChip® (58% of the grapevine genome), only 85 non-redundant grapevine probe sets were found to be differentially expressed in 8-day-old fruit when comparing the Flb somatic variant and Ugni Blanc wild-type plants (expression ratio>3 and P value <0.05, Table S1). Nineteen of probe sets corresponded to genes that are down-regulated in Flb berries, including HB13, which was previously identified as repressed in the Flb somatic variant (Fernandez et al., 2007). The other 66 genes were over-expressed in Flb berries, and included VvPI.

We investigated the flower expression pattern of these genes using the same 23K GrapeGen Affymetrix GeneChip®. The 85 genes that are mis-regulated in fruits of the Flb somatic variant showed comparable expression in Ugni Blanc and Flb flowers (Figure 6). Interestingly, all genes showing up-regulation in the Flb fruit exhibited strong expression in flowers, contrasting with the results for genes down-regulated in the Flb fruit, which were generally more specific to fruits (Figure 6a). In addition, the analysis showed a correlation between the level of gene up-regulation in the Flb fruit and flower expression specificity during normal grapevine development (Figure 6b). Based on these results, we hypothesize that VvPI ectopic expression induced up-regulation of flower-specific genes in Flb fruits. In fact, comparison of the functional annotation of genes up-regulated in Flb fruits with functional information on the corresponding Arabidopsis orthologues supported the possibility these VvPI targets are specific of floral development (Table 1). Several of these genes have been described as specific regulators of flower developmental processes in Arabidopsis, such as GASA5, a regulator of flowering time, the IAA19 auxin-induced repressor and two genes encoding transcription factors MYB24 and MYB108, which have specific functions in stamen and pollen maturation in response to gibberellic acid and auxin (Zhang et al., 2009; Reeves et al., 2012). Remarkably, the genes showing the strongest over-expression in Flb fruits were involved in terpene and lignin biosynthesis (Table 1). These aromatic compounds are known to accumulate during floral development, and previous work has reported common regulators between the flowering process and lignin biosynthesis (Shi et al., 2011). In addition, Auxin and MYB factors have been reported to promote volatile sesquiterpenes production in Arabidopsis flowers through control of the AtTPS21 gene (Reeves et al., 2012), and the grapevine homologue was also up-regulated in Flb fruits.

Table 1. Genes up-regulated in Flb fruit that also have a specific pattern during floral development
Unique ID (Grimplet et al., 2012)Functional annotationArabidopsis orthologue
  1. Corresponding Affymetrix probe set IDs are listed in Table S1.

Floral regulators
Vv18s0001 g01760PISTILLATA (PI) floral homeotic proteinAT5G20240
Vv14s0066 g01090myb domain protein 24AT5G40350
Vv05s0077 g00500myb domain protein 108AT3G06490
Vv18s0001 g09460GASA5AT3G02885
Vv09s0002 g05150IAA19AT3G15540
Lignin biosynthesis
Vv15s0048 g02480Caffeate 3-O-methyltransferase 1AT5G54160
Vv00s0615 g00020Cinnamyl alcohol dehydrogenaseAT4G37990
Vv03s0180 g00250Cinnamyl alcohol dehydrogenaseAT3G19450
DT011795Ferulate 5-hydroxylaseAT4G36220
Terpene biosynthesis
Vv12s0134 g00030E-β-ocimene synthaseAT4G16740
Vv00s0361 g00060(E,E)-α-farnesene synthaseAT4G16740
Vv19s0014 g04900(−)-germacrene D synthaseAT5G23960
Vv13s0067 g00050Myrcene synthase, chloroplasticAT3G25810
Vv19s0014 g04930Germacrene D synthase AtTPS21AT5G23960
Figure 6.

Expression patterns of the genes that are significantly up- and down-regulated in fruit of the Flb somatic variant. (a) Flower expression showing a linear correlation between the Flb somatic variant (Flb) and wild-type Ugni Blanc (UB). (b) Flower/fruit expression ratio in UB, indicating specificity of expression during wild-type grapevine development. Differences in the flower/fruit expression pattern in Flb are expressed on the vertical scale as a ratio between genotypes.

To conclude, expression of the VvPI gene in fleshy fruit induced up-regulation of putative downstream genes, many of which are probably involved in flower developmental processes.

Discussion

The fleshless phenotype is caused by a transposable element insertion causing VvPI ectopic expression

We have identified a transposon insertion causing ectopic expression of the VvPI gene associated with the fleshless berry phenotype in grapevine. Analyses of revertant berries demonstrated that this insertion is the cause of VvPI mis-expression and the mutant phenotypes observed in both flowers and berries. VvPI mis-expression prevents regular development of the berry flesh when restricted to the L2 cell layer. In addition, when the mutation is also present in the L1 cell layer, fruit development cannot proceed normally being blocked at fruit set. This mutation is the second example in grapevine of a natural transposable element-mediated gene cis-activation as the origin of somatic variation. The first example described was the rrm (reiterated reproductive meristem) mutation in the cultivar Carignan that is due to insertion of the Hatvine1-rrm element (Fernandez et al., 2010). In both cases, the transposable elements do not modify the transcription start of mutated alleles. A possible explanation is that the Mila-flb and Hatvine1-rrm elements enhance the expression of mutated alleles, either by enhancing transcription initiation at their own transcription start site, or by disrupting a promoter region associated with a repressor function.

The element inserted in the VvPI promoter is a 1.5 kb class II transposable element identified as a Mila element (Benjak et al., 2009). Mila elements have been described as unclassified MITEs (miniature inverted repeat transposable elements) flanked by target site duplications but lacking TIRs and without sequence homology to known transposable elements. However, we were able to identify the existence of imperfect TIRs [CA(A/G)GGATTGAAA(T/A)ATCG] in the Mila reference element TE-7-1_VV (Benjak et al., 2009) that were similar to the TIR sequences flanking Mila-flb as well as Hatvine1-rrm (CAA/TGGATTGAAAT/AATCGGA/TAAA). Moreover, the TIR sub-terminal regions of Mila elements also share extensive sequence similarities with Hatvine1-rrm terminal sequences (more than 500 bp with >80% identity). Only the internal region of the Mila element sequence is unrelated to known transposon sequences. MITEs are known to be a particular type of defective class II elements. These elements lack gene-coding capacity and are amplified in genomes by an unknown replicative mechanism (Feschotte et al., 2002). Most MITE families share sequence similarities with the class II transposons from which they are thought to be derived by internal deletion and putative capture of extra genomic regions through transduplication (Feschotte et al., 2003). Thus, the Mila family may correspond to MITE elements derived from grapevine Hatvine1 class II transposons. Among the 108 copies of Mila elements identified in the PN40024 grapevine genome (Benjak et al., 2009), 22 share high homology with Mila-flb. Most of these insertions are located in positions closely linked to gene sequences in the PN40024 genome. However, none of them were found within coding sequences, which may suggest some specificity in the insertion mechanism or may be the result of strong selection imposed during generation of the PN40024 near-homozygous line. In highly heterozygous species such as grapevine in which vegetative propagation is the usual multiplication strategy, recessive mutations resulting in loss-of-function alleles are phenotypically undetected, while dominant mutations causing conspicuous phenotypic alterations in heterozygous and chimeric states are rapidly identified.

VvPI alteration of fruit development

VvPI, the grapevine PISTILLATA (PI) homologue, has been molecularly characterized in grapevine (Sreekantan et al., 2006; Poupin et al., 2007). PI encodes a MADS box transcription factor that is present in all studied angiosperms and is required to specify petal and stamen identity in specific flower primordia (Goto and Meyerowitz, 1994; Krizek and Meyerowitz, 1996). The VvPI expression pattern in developing grapevine flowers and lack of expression in carpels and fruits suggest a similar function during flower morphogenesis as described for Arabidopsis PISTILLATA (Sreekantan et al., 2006; Poupin et al., 2007).

Analyses of chimera and Flb mutant plants showing lack of suppression of VvPI expression after fertilization in the L2 layer, the L1 layer or both cell layers demonstrate the differential effects of VvPI expression in different cell layers. In the L2 layer, VvPI expression prevents flesh development by preventing differentiation of the highly vacuolated cells that are characteristic of this mesocarp tissue. In contrast, when the mutation is present in all cell layers, as in F1 plants with no transposon excision, ovary development is blocked at fruit set, with flowers remaining viable for a longer period post-anthesis (at least 3 weeks). Interestingly, longer flower longevity has been previously reported for Arabidopsis plants over-expressing a PI homologue from Phalaenopsis equestris (Tsai et al., 2005). What is not known is the flower and berry phenotype if the mutation exists only in the L1 layer, as no chimeric plants with this genetic make-up were identified. It is possible that such a plant displays a similar phenotype to that observed when the mutation is present in both cell layers, or may alternatively have no phenotypic effect. As L1-derived cells are present in the stigma and part of the transmitting tract, we investigated whether the presence of the mutation in the L1 cell layer altered the pollination process and prevented fertilization and consequently fruit development. However, no conspicuous alteration of pollen germination and pollen tube migration was observed along the styles in Flb mutant plants, and ovules showed normal developmental initiation. However, observations at later stages of ovule development in these mutant plants showed precocious degeneration of embryo sac. The lack of development of the nutritive tissues that normally support embryo development may be a cause of ovule abortion and subsequent fruit set failure and may be the result of VvPI mis-expression in L1-derived tissue or in both L1- and L2-derived tissue. This result, and the observation that fruit set takes place and fruits show restricted flesh growth when the mutation is present only in the L2 cell layer, demonstrate that VvPI mis-expression affects at least two fruit developmental processes in a cell layer-dependent manner.

In Arabidopsis, PI loss-of-function or gain-of-function mutations result in flower organ homeotic conversions (Krizek and Meyerowitz, 1996; Lamb and Irish, 2003). The slight up-regulation of VvPI during flower development in the Flb somatic variant is not sufficient to generate flower organ homeotic conversions, but it has effects on pistil shape in addition to those described in ovule and fruit development. The unusual transcriptomic pattern observed in young Flb fruits suggests that mutated cells cannot properly differentiate, as they are expressing genes that are normally involved in flower organ development. In fact, some of those genes are orthologues of Arabidopsis genes that have been reported to have specific biological functions during flower development, such as MYB24, MYB108, GASA5 and IAA19. It is tempting to propose that mis-expression of the VvPI transcription factor causes mis-expression of these various floral target genes whose function is incompatible with proper flesh morphogenesis. This possibility is supported by the observation that most of the genes mis-regulated in the Flb fruit are up-regulated, as for VvPI. PI regulates the expression of a cascade of genes, but so far only a small number of PI direct target genes have been identified, including APETALA1 (Sundstrom et al., 2006), BANQUO (Mara et al., 2010), GATA transcription factors (Mara and Irish, 2008) and NAP (Sablowski and Meyerowitz, 1998), which was also up-regulated in Flb fruit (Fernandez et al., 2007). Genome-wide identification of target genes in Arabidopsis suggested that PI regulates different sets of genes over a developmental time line (Wuest et al., 2012). Many of the transcripts identified here as up-regulated in Flb fruit probably correspond to downstream VvPI target genes. However, comparison of the gene set identified in the grapevine fruit with the AP3/PI targets reported in the Arabidopsis flower (Wuest et al., 2012) revealed a very poor level of similarity. Nonetheless, considering that the organs and developmental stages of the studies are different, our observations are consistent with those of Wuest et al. (2012). Similarly, Wuest et al. (2012) identified fewer than 2% of genes in common with those identified by Mara and Irish (2008). It remains unclear which targets are directly involved in the mutant phenotype. Consequently, the resulting Flb mesocarp tissue may have features of both flower and fruit organs.

Although, PI has been mainly found to be involved in the specification of petal and stamen organs in many angiosperm species (Lamb and Irish, 2003), several reports have proposed additional PI functions based on gene expression patterns or phenotypes associated with specific mutations or allele variants. For instance, the pea PI gene was proposed to regulate leaf development (Berbel et al., 2005) and the Taihangia rupestris (Rosaceae) PI orthologue was proposed to regulate the plant architecture (Lu et al., 2010). The MdPI loss-of-function mutant in apple showed a parthenocarpic (seedless) phenotype, probably as a consequence of flower organ alterations (Yao et al., 2001). A function of negative control of ovary and ovule development was proposed for the orchid PI homologue (Tsai et al., 2005). The phenotype of the Flb somatic variant and derived Flb mutants described here provides evidence of a direct negative effect of PI orthologue expression on fruit development. Previous grapevine studies showed that VvPI is drastically repressed after flower fertilization (Poupin et al., 2007; Dauelsberg et al., 2011). Moreover, in non-pollinated grapevine flowers, there was a correlation between the maintenance of VvPI expression in carpels and impairment of mesocarp growth, at least in the Red Globe cultivar (Dauelsberg et al., 2011). Because non-pollinated Red Globe berries do not have seeds, the impairment of mesocarp growth may be due either to the absence of seed tissues or the expression of PI. However, in the Flb somatic variant, in which seeds are normally formed, seed signals are present but are unable to promote normal mesocarp development, supporting a specific involvement of PI expression in flesh impairment.

The present analysis of the grapevine Flb somatic variant and Flb mutants conclusively demonstrates that VvPI expression in fruits negatively affects flesh development and fruit set, and that this effect is cell layer-specific.

Experimental procedures

Plant materials

The fleshless berry somatic variant (Flb) has been described previously (Fernandez et al., 2006b). A progeny with 478 individuals showing dwarf stature and precocious and continuous flowering was derived from the cross 00C001V0008 (genotypes: f/f and Vvgai1/Vvgai1) x Flb somatic variant (genotypes: H/f and VvGAI1/VvGAI1), and established in 3 litre pots in a greenhouse as described by Chaib et al. (2010). As flower sex segregates in the microvine population, all female flowers were manually pollinated to score fruit phenotype.

Green berries were collected 8 days after anthesis, corresponding to the developmental stage of fruit set, for Ugni Blanc and the Flb somatic variant, and 20 days after anthesis for fleshy revertant berries of the Flb somatic variant to phenotype the presence of flesh. Berries were used for subsequent DNA and RNA extractions, after removing seeds for the 20 days after anthesis revertant berries. As pericarp tissue includes the L1 and L2 cell layers, skin (L1 + L2) and pulp (L2) were separated for revertant analysis to obtain L2 pure fruit tissue.

Microscopic observation

Wild-type and Flb female flowers of F1 microvine plants were manually fertilized using pollen of either Grenache or Flb mutants. After 24 h, flowers were collected and stained using aniline blue as described by Singh et al. (2002).

To characterize ovule morphology, developing ovules were extracted from growing ovaries and cleared in solution (20 g chloral hydrate, 20 ml water, 10 ml glycerol) before observation using differential interference contrast optics.

flb mapping

Molecular markers available from grapevine genetic maps and flanking the FLB locus were tested for polymorphism in the microvine population (Figure S1). In addition, SNPs in the target region were identified through alignment of the near-homozygous reference genome sequence (Jaillon et al., 2007) against the corresponding heterozygous cultivar Pinot Noir (Velasco et al., 2007). A total of 103 sequences featuring targeted SNPs were used for genotyping. A MassARRAY® iPLEX Gold assay with MALDI-TOF MS detection was used for SNP analysis by the Australian Genome Research Facility (www.agrf.org.au/). A genetic map of LG18 was constructed using MapMaker (Lander et al., 1987).

Transcriptomic analyses

RNA was extracted using a Spectrum plant total RNA kit (Sigma, www.sigmaaldrich.com) as recommended by manufacturer, including DNase treatment (Qiagen, www.qiagen.com).

Microarrays

Transcriptional profiling was performed on flowers and 8 days after anthesis berries from the Flb somatic variant and the Ugni Blanc cultivar, using three biological replicates per experiment. Synthesis of labelled probes, hybridization and scanning of the GeneChip® Vitis vinifera (grape) genome array version 1.0 (23K GrapeGen Affymetrix GeneChip®; Affymetrix, www.affymetrix.com) (Lijavetzky et al., 2012) were performed at the Genomics Service of the Centro Nacional de Biotecnologia - Consejo Superior de Investigaciones cientificas. Microarray data were processed as previously described (Fernandez et al., 2010). A multiple test correction was applied to the P value of the t-statistics to adjust the false discovery rate (Benjamini and Hochberg, 1995). Genes with an adjusted P value <0.05 and and expression change fold ratio ≥3 between Flb and Ugni Blanc were selected for further analysis.

Quantitative RT-PCR

Quantitative RT-PCR were performed as previously described (Fernandez et al., 2007). Amplification of total VvPI (DQ059750), ubiquitin (CF406001) and VvEF1α (BQ799343) genes was performed using the primers previously described (Sreekantan et al., 2006; Fernandez et al., 2007). Allele-specific expression for VvPI was analysed using specific primers (Table S3) for allelic discrimination assays, designed as recommended by Gupta et al. (2005). Relative quantification was performed using the ΔΔCT method for target genes, using the control VvEF1α gene as a standard.

cDNA RACE

The complete VvPI cDNA flb sequence was isolated from leave RNA using a GeneRacer™ kit (Invitrogen, www.invitrogen.com), Platinum® Taq high-fidelity DNA polymerase (Invitrogen) and VvPI primers (Table S3). PCR products were purified and cloned for further sequencing.

VvPI sequencing

Leaf genomic DNA was extracted using a DNeasy plant mini kit (Qiagen). Primers were designed to amplify 5 kb of the VvPI gene (GSVIVT01008806001, chr18_02291387_02296396, Genoscope 12X, www.genoscope.cns.fr) in the Flb somatic variant and Ugni Blanc using the Long-Range PCR enzyme mix (Qiagen) for cloning. A total of 9 and 15 VvPI inserts from the Flb and Ugni Blanc, respectively, were sequenced using the primers described in Table S3. The 5′ and 3′ ends of the transposon insertion were first sequenced from a plasmid containing the cloned VvPIbm allele from the Flb somatic variant. Identification of closely related sequences was performed using the Genoscope Blat server (www.genoscope.cns.fr/blat-server/cgi-bin/vitis/webBlat), to find a homologue reference sequence. Later, primers were defined based on the reference sequence and used to obtain the complete sequence of the transposon (Table S3).

Transposon presence genotyping

PCR amplifications were performed on leaf genomic DNA of Flb and Ugni Blanc and fruit tissue DNA from revertant berries, using a Multiplex PCR kit (Qiagen) and three primers (Table S3) to specifically amplify VvPIb alleles containing the transposon (726 bp amplicon) or without the transposon (1237 bp amplicon). Each purified PCR product was further sequenced.

Accession numbers

The Mila-flb transposon, VvPIa and VvPIb sequences have been deposited in the GenBank database with accession numbers JQ993317, JQ993315 and JQ993316, respectively. Affymetrix expression profiling data are available from PLEXdb (www.plexdb.org), accession number VV35.

Acknowledgements

The authors would like to thank Isidoro Laguna and Patricia Corena for assisting with cloning, Don McKenzie and Virginia Rodriguez for helping with RNA/DNA extractions, Maria Mrinak and Gilbert Lopez for taking care of the plants, the Centro Nacional de Biotecnologia Genomics Service staff for Affymetrix GeneChip hybridization experiments, and Christophe Rothan for helpful discussions. This project was funded by the Grape and Wine Research and Development Corporation, Commonwealth Scientific and Industrial Research Organisation - Plant Industry Department, the Institut National de la Recherche Agronomique Research Division for Plant Breeding and Genetics, and the Spanish Ministerio de Ciencia e Innovacion (projects BIO2008-03892 and BIO2011-26229).

Ancillary