Down-regulation of a single auxin efflux transport protein in tomato induces precocious fruit development.

The PIN-FORMED (PIN) auxin efflux transport protein family has been well characterized in the model plant Arabidopsis thaliana, where these proteins are crucial for auxin regulation of various aspects of plant development. Recent evidence indicates that PIN proteins may play a role in fruit set and early fruit development in tomato (Solanum lycopersicum), but functional analyses of PIN-silenced plants failed to corroborate this hypothesis. Here it is demonstrated that silencing specifically the tomato SlPIN4 gene, which is predominantly expressed in tomato flower bud and young developing fruit, leads to parthenocarpic fruits due to precocious fruit development before fertilization. This phenotype was associated with only slight modifications of auxin homeostasis at early stages of flower bud development and with minor alterations of ARF and Aux/IAA gene expression. However, microarray transcriptome analysis and large-scale quantitative RT-PCR profiling of transcription factors in developing flower bud and fruit highlighted differentially expressed regulatory genes, which are potential targets for auxin control of fruit set and development in tomato. In conclusion, this work provides clear evidence that the tomato PIN protein SlPIN4 plays a major role in auxin regulation of tomato fruit set, possibly by preventing precocious fruit development in the absence of pollination, and further gives new insights into the target genes involved in fruit set.

Introduction

The formation of seedless or parthenocarpic fruit is a desired trait in several crop species producing hard seeds (e.g. banana and grape) as well as in species where the pollination/fertilization process is strongly dependent on environmental conditions (tomato). Seventy years ago, it was demonstrated that exogenous application of auxins, cytokinins, and gibberellins (GAs) to tomato flowers leads to the formation of parthenocarpic fruits. It was further suggested that this particular fruit development is due to the deregulation of the hormonal balance within the ovary, which substitutes for pollination/fertilization (Schwabe and Mills, 1981; Gorguet ). The manipulation of auxin synthesis/signalling within tomato ovaries later emphasized the crucial role played by auxin in fruit set and parthenocarpy (Ficcadenti ; Carmi ; Pandolfini ). More recently, the fundamental discoveries on auxin signalling in Arabidopsis thaliana (reviewed in Chapman and Estelle, 2009; Kieffer ) paved the way for deciphering the mechanisms involved in auxin regulation of fruit set (Wang ; Goetz ; de Jong ; Wang ; Ren ).

The current scheme of ovary growth regulation involves an Aux/IAA (auxin/indole-3-acetic acid) complex which represses downstream auxin-responsive genes up to fertilization (Swain and Koltunov, 2006). Upon pollination and fertilization, an increase in auxin levels in the fertilized ovules/ovary would activate Aux/IAA degradation via the ubiquitin–proteasome pathway (Dharmasiri ; Mockaitis and Estelle, 2008). Consequently, free auxin response factor (ARF) proteins would regulate auxin-responsive genes, thus triggering fruit set (Swain and Koltunov, 2006; de Jong et al., 2009a). According to this model, alterations at any step of the auxin signalling cascade, either by changing the auxin level in the ovule area, by affecting the Aux/IAA–ARF repressive complex, or by modifying the expression of target genes may dissociate fruit development from ovule fertilization and trigger parthenocarpic fruit development. Although this scheme remains to be fully validated, it is consistent with the development of parthenocarpic fruit when the auxin level in the ovules/ovary is increased by the overexpression of genes of auxin biosynthesis (Ficcadenti ; Pandolfini ; Carmi ) or by application of auxin or auxin efflux transport inhibitors (Beyer and Quebedeaux, 1974; Serrani ). A recent work using a DR5rev::mRFPer synthetic reporter described the distribution of auxin in ovary and fruit tissues and showed that the auxin signal was localized in the ovule before anthesis and close to the funiculus after anthesis (Pattison and Catala, 2012). Experimental data confimed the participation of several Aux/IAA and ARF proteins in fruit set in tomato, including SlAux/IAA9 (Wang ), SlARF7 (de Jong ), and an orthologue of AtARF8 (Goetz ). Besides recent data indicate that the Aucsia silencing-induced parthenocarpy, which is connected to auxin accumulation in reproductive organs, is not linked to the down-regulation of SlAux/IAA9 and SlARF8 (Molesini ). Up to now, few auxin-regulated target genes responsible for the initiation of fruit development have been identified. Profound changes in the expression of auxin, GA, and ethylene signalling-related genes have, been described in the ovaries developing into parthenocarpic fruits (Vriezen ; Wang ; de Jong ). In addition, three MADS-BOX genes have been identified as potential candidate genes in the control of fruit set in tomato (Ampomah-Dwamena ; Wang ).

In Arabidopsis, polar auxin transport plays a crucial role in auxin signalling (Zazimalova ). It is responsible for the local distribution of auxin in plant tissues, due to the asymmetric location of auxin influx and efflux transporters (Kleine-Vehn ; Petrásek and Friml, 2009). Auxin efflux transport is achieved by the PIN-FORMED (PIN) protein family made up of eight members in Arabidopsis (Paponov ). PIN proteins are involved in various developmental processes including embryogenesis, shoot and root morphogenesis, gravitropism, and phototropism (Vieten ; Petrásek and Friml, 2009). In tomato, the application of auxin efflux transport inhibitors leads to parthenocarpic fruit development (Serrani ) while fruit parthenocarpy observed in the Aucsia-silenced plants is associated with an alteration of polar auxin transport (Molesini ), therefore suggesting the implication of auxin transport in fruit set.

Despite the crucial role of auxin signalling during fruit development in tomato and the likely implication of auxin efflux transport in this process, scarce data were available until very recently concerning the implication of PIN in fruit set and development in tomato (Lemaire-Chamley ; Kharshiing ; Nishio ; Pattison and Catala, 2012). The aim of the present study was to unravel the role of the SlPIN4 protein, which is highly expressed in developing ovary and fruit in tomato. Towards this end, RNA interference (RNAi) transgenic lines specifically silenced for the SlPIN4 gene were generated. Partial to full parthenocarpic fruit phenotypes were observed, thereby suggesting the involvement of the SlPIN4 gene in tomato fruit set. Transcriptome analyses of wild-type (WT) and transgenic ovaries by microarray and targeted quantitative RT-PCR (qRT-PCR) transcription factor (TF) profiling allowed the identification of downstream genes involved in auxin-dependent fruit set regulation in tomato and provided new insights into the fruit set process.

Materials and methods

Plant material

Tomato plants used for expression profile analyses (Solanum lycopersicum Ailsa Craig) were grown in a growth chamber as previously described (Mounet ). Transgenic plants and the corresponding WT [Solanum lycopersicum var. cerasiformae ‘West Virginia 106’ (WVA 106)] were grown in a greenhouse (Alhagdow ). The RNAi-mediated silencing of the tomato SlPIN4 gene (AB508932/HQ127078) was performed by stable transformation (Alhagdow ) using a 475bp DNA fragment located in the 3'-untranslated region of the cDNA amplified by the specific primers attB1SlPIN4 (AAAAAGCAGGCTGCCAAGTGAAGAAAGGAGGA) and attB2SlPIN4 (AGAAAGCTGGGTGGAAGTAAACAACTAGCAAACC), introduced into the destination vector pK7GWIWG2(1) as an inverted repeat under the control of the 35S promoter (Karimi ). Primary transformants were further checked for ploidy level by flow cytometry. Ten independent diploid primary transformants (T0) with single-copy T-DNA insertion were selected and grown in a greenhouse. Obligate parthenocarpic lines were maintained as cuttings, and homozygous lines were generated for two facultative parthenocarpic primary transformants (L-2 and L-21). Seeds of homozygous lines L-2 and L-21 were selected by segregation test in the presence of kanamycin (300 µg l–1).

Phenotypic analyses of P35S:SlPIN4 RNAi lines

Flower development in WT and transgenic plants was investigated by measuring flower length every day from 1mm bud to anthesis stage. Flower emasculation was performed on 10mm closed flowers. For cytological analyses of flower buds, samples were harvested at 4mm, 8mm, and anthesis stages from T0 plants, and processed as previously described (Bereterbide ). Mean carpel area was measured by using ImagePro-Plus software (Media Cybernetics) on at least 10 carpel transverse sections. The effects of IAA on seedling germination was tested on WT, and L-2 and L-21 homozygous lines grown in vitro. Sterilized seeds were sown on MS (1.1g l–1) agar (5g l–1) medium containing 7.5g l–1 sucrose (w/v), pH 5.8±0.5 µM IAA (Sigma). Phenotypes were observed on fifty 12-day-old seedlings.

Analyses of auxin content

Auxin analyses were performed on three biological replicates of flower buds (4mm and 8mm) and ovaries at anthesis harvested on T0 (P35S:SlPIN4 RNAi L-14) and homozygous T2 plants (P35S:SlPIN4 RNAi L-2 and L-21). A 25mg aliquot of fresh samples was extracted, purified, and analysed by gas chromatography–selected reaction monitoring–mass spectrometry as described previously (Edlund ). Calculation of the isotopic dilution factors was based on the addition of 50 pg of [6-13C]IAA mg–1 tissue.

qRT-PCR analyses

Expression profiles of PIN genes in WT tomato fruit (entire and separated tissues) and vegetative organs were performed on Ailsa Craig tomato plants. Samples collected from 15 plants were pooled and divided into three subsamples (technical replicates). Expression profiles in flower buds and ovaries of WT and P35S:SlPIN4 RNAi transgenic lines were performed on three biological replicates collected from three WVA 106 tomato plants (T0 plants maintained as cuttings for P35S:SlPIN4 RNAi L-14 and homozygous T2 plants for P35S:SlPIN4 RNAi L-2 and L-21). RNA extraction was performed on each subsample and each biological replicate, treated with DNase, and reverse transcribed as previously described (Mounet ). The reverse transcription product was diluted 10-fold in water, and 2 µl were used in qRT-PCR (25 µl final volume) in the presence of 0.2 µM specific primers using the iQ-SYBR Green Supermix (Bio-Rad, Marne La Coquette, France) according to the manufacturer’s instructions. qRT-PCR was performed on a CFX-96 (Bio-Rad). Data acquisition and analysis was done using Bio-Rad CFX manager software (version 1.1) using ∆∆CT normalization. The specific primers used are indicated in Supplementary Table S1 available at JXB online.

Microarray analysis

The transcriptome of P35S:SlPIN4 RNAi lines (L-2, L-21, and L-14) was compared with that of WT plants. The microarray experimental design consisted of two biological replicates with one dye swap for a total of four slides per P35S:SlPIN4 RNAi line/WT comparison. A 1 µg aliquot of DNA-free total RNA extracted from 6mm flower buds was used to prepare the labelled probes as already described (Mounet ). Hybridization on TOM2 oligonucleotide microarrays and scanning parameters were as described in Prudent .

Statistical analysis of microarray data was performed using the Bioconductor LIMMA package v2.15.16 (Smyth, 2005). After background correction using the Normexp method (Ritchie ), the data were normalized using the print-tip lowess (within-array normalization) and A-quantile (between-arrays normalization) functions (default parameters). Flagged spots were given a weight of 0.1 using the weight function. For the statistical analysis, a reference design analysis was performed (Smyth, 2005) and a linear model with a coefficient for each of the three factors (L-2, L-21, and L-14) was fitted. The P-values resulting from the moderated t-test were corrected for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) adjustment. Genes were considered to be significantly differentially expressed if adjusted P-values were <5×10−2. Spots with low expression (A <9) were eliminated from further analysis, resulting in 1460 spots with a significant variation of expression. Genes that changed >1.5-fold in at least one P35S:SlPIN4 RNAi line when compared with the WT were considered as differentially expressed [Log2(P35S:SlPIN4 RNAi/WT] < –0.58 or > 0.58; Supplementary Table S3 at JBX online).

TF profiling

TF profiling was performed on three biological replicates of 4mm flower buds and 0 days post-anthesis (DPA) ovaries harvested from three plants of each genotype (WT, T0 plants maintained by cutting in the case of P35S:SlPIN4 RNAi L-14, and homozygous T2 plants in the case of P35S:SlPIN4 RNAi L-2). RNA extraction and DNase treatement were as previously described (Mounet ). Control of RNA, reverse transcription, qRT-PCR, and collection of the raw data were performed as described in Rohrmann . TFs with a threshold cycle (CT) value <35 in two of the three replicates were considered as expressed in the sample considered. TF normalized expression (∆CT=CTTF–CTControl) was calculated as previously described, using for each independent run the mean value of the two replicates of the ubiquitin gene as a CTControl (Rohrmann ). TFs displaying statistically altered expression in flower buds or in ovary were found via the use of analysis of variance (ANOVA) testing with FDR-corrected P-values (10%) implemented with the corresponding data (three genotypes×three biological replicates). The relative expression of the transformant lines was calculated as 2–∆∆CT=2–( mean∆CTRNAi–mean∆CTWT). TFs that changed >1.5-fold in at least one P35S:SlPIN4 RNAi line when compared with the WT were considered to be differentially expressed [Log2(P35S:SlPIN4 RNAi/WT] < –0.58 or > 0.58; Supplementary Table S4 at JBX online).

Results

Characterization of the tomato PIN auxin efflux transport protein family

Phylogenetic analysis of tomato (HQ127074–HQ127083, Supplementary Table S2 at JXB online) and Arabidopsis PIN proteins highlighted likely tomato orthologues for AtPIN1, AtPIN2, AtPIN6, and AtPIN8 (Fig. 1A). In agreement with the very recent study of Pattison and Catala (2012), six tomato PIN sequences were clustered with the group of Arabidopsis plasma membrane PIN proteins, composed of AtPIN1, AtPIN2, AtPIN3, AtPIN4, and AtPIN7 (Petrásek and Friml, 2009). Expression analysis revealed that only five of these putative plasma membrane PIN genes were expressed in tomato fruit, with SlPIN4 being the most highly expressed (Fig. 1B). The SlPIN4 protein is close to AtPIN3, AtPIN7, and AtPIN4 (76, 75, and 73% amino acid sequence identity, respectively). The relative expression of SlPIN4 increased durng flower development up to the anthesis stage and decreased during fruit development (Fig. 1C), and suggests its potential role in fruit set. The SlPIN3 gene displayed a very similar pattern of expression during flower and fruit development but its expression was one-tenth that of SlPIN4 in fruit at 0 DPA (Fig. 1B, 1C). Other PIN genes were characterized by little variation (SlPIN1 and SlPIN7) or even a decline (SlPIN9) in expression during flower development. In agreement with published data (Nishio ; Pattison and Catala, 2012), the expression of all PIN genes decreased during fruit development, except for SlPIN7 whose expression increased at the onset of fruit ripening (Fig. 1C). The SlPIN4 gene was expressed in all plant organs, with a higher expression in sepals, ovary, young leaves, and petals (Fig. 2). The other PIN genes were also expressed in the various plant organs analysed, except SlPIN3, which was mainly expressed in petals. As previously shown (Lemaire-Chamley ; Pattison and Catala, 2012), SlPIN4 gene expression increased gradually from the outer part of the fruit (exocarp) to the central part of the fruit (columella, locular tissue, Fig. 2).

Fig. 1.

PIN homologues in tomato. (A) Phylogenetic tree of Arabidopsis PIN proteins and of tomato putative PIN proteins. The phylogenetic tree was constructed based on a complete protein sequence alignement of PINs by the Neighbor–Joining method with bootstrapping analysis (5000 replicates) using MEGA 4.0 software (http://www.megasoftware.net/) after sequence alignment using ClustalW. (B) Expression analysis of tomato plasma membrane PIN genes in 0 DPA fruit by qRT-PCR. The relative expression of each gene (arbitrary units) corresponds to gene expression normalized with the expression of actin, β-tubulin, and EiF4a. Bars represent the standard deviation (n=3). (C) Expression profile of PIN genes during tomato reproductive development. Relative expression of PIN genes was measured by qRT-PCR in 2–10mm flower bud, flower at anthesis (A), and in 0–45 DPA fruit. The relative expression of each gene (arbitrary units) corresponds to gene expression normalized with the expression of β-tubulin and EiF4a. Error bars represent the standard deviation (n=3).

Fig. 2.

Expression profile of PIN genes in tomato tissues. The relative expression of PIN genes was measured by qRT-PCR. The relative expression of each gene (arbitrary units) corresponds to gene expression normalized with the expression of actin, β-tubulin, and EiF4a. Bars represent the standard deviation (n=3). Sep, sepal; Pet, petal; Sta, stamen; 0, fruit at anthesis; YL, young leaf; ML, mature leaf. E, exocarp; M, mesocarp; LT, locular tissue; C, columella; S, seed.

Silencing of SlPIN4 leads to precocious ovary development and to parthenocarpic fruit

Since the SlPIN4 gene was the major PIN gene expressed in tomato ovary and young fruit, its role in fruit set was characterized. Recent attempts to alter fruit set or development by silencing SlPIN4 were unsuccessfull (Pattison and Catala, 2012). However, in contrast to the study presented herein, which targeted specifically SlPIN4, at least two genes (SlPIN3 and SlPIN4) and up to five SlPIN genes were significantly silenced in the lines studied by Pattison and Catala (2012). Considering the complexity of the regulation of fruit set through auxin transport in the ovary and from the apical shoot (Serrani ), it is likely that co-silencing multiple SlPIN genes in both ovary and stem may lead to complex and, possibly, opposing effects on fruit set. Therefore, in the study described herein, transgenic lines were generated in which SlPIN4 was specifically RNAi silenced by targeting its 3'-untranslated region under the control of the 35S promoter. The risk of silencing potential off-targets was checked beforehand (Xu ). Screening the tomato genome for sequences sharing at least 21 nucleotide identity with the SlPIN4-targeted region revealed the absence of such potential off-targets. Ten independent primary (T0) transgenic lines were generated that presented a range of 2- to 4.3-fold reduction in SlPIN4 expression (Supplementary Fig. S1 at JBX online). Among these plants, the severely affected line L-8 displayed wiry leaves and a low number of flowers, which generally aborted. The nine other T0 plants did not show any obvious phenotype at the vegetative level and were able to develop flowers and fruits. Among these T0 plants, four were sterile, their flowers giving rise to the development of seedless (parthenocarpic) fruits (Fig. 3, L-7, L-9, L-13, and L-14) similar to the previously described transgenic plants in which auxin signalling genes were down-regulated (Wang ; de Jong et al., 2009a). The other T0 plants (L-2, L-3, L-4, L-5, and L-21) showed facultative parthenocarpy, giving rise to parthenocarpic and seeded fruits (Fig. 3), with a normal or reduced number of seeds (data not shown). Emasculation of flowers before anthesis in obligate and facultative parthenocarpic P35S:SlPIN4 RNAi lines also led to the development of seedless fruits (Fig. 4), whereas in WVA 106 WT plants emasculation of flowers always led to flower abortion within a week after anthesis. These results indicate that in the transgenic lines fruit set was independent of ovule fertilization and was due to the precocious development of the ovary into a fruit. Failure to set seeds was not associated with male fertility defects since no significant difference in pollen germination rate was detected between WT and P35S:SlPIN4 RNAi lines (data not shown). In addition, cross-pollination experiments using obligate parthenocarpic P35S:SlPIN4 RNAi pollen on WT ovaries triggered seed formation, whereas cross-pollination using WT pollen on obligate parthenocarpic P35S:SlPIN4 RNAi ovaries led to seedless fruits (Fig. 4).

Fig. 3

. Down-regulation of SlPIN4 in P35S:SlPIN4 RNAi lines triggers the development of parthenocarpic fruits. Fruit and fruit transverse sections of WT and P35S:SlPIN4 RNAi lines at the red ripe stage. The bar represents 1cm.

Fig. 4.

Parthenocarpic fruit development in P35S:SlPIN4 RNAi transgenic lines is due to the uncoupling of fruit development from ovule fertilization. Emasculation was performed by removal of stamens on flowers from lines L-2, L-21, and L-13. Pollen from parthenocarpic lines L-13 and L-14 was used to fertilize carpel on WT plants. WT pollen was used to fertilize carpel from lines L13 and L-14. The bar represents 1cm.

Down-regulation of the SlPIN4 gene resulted in visible morphological modifications of flowers in obligate parthenocarpic P35S:SlPIN4 RNAi lines and, to a lesser extent, in the facultative parthenocarpic lines (Fig. 5A). Sepals were longer than in WT plants and had a yellowish colour. The fused stamen cones were pushed away by the developing carpels, and flowers presented a protruding style. In the most severe cases, multiple carpels were fused within one flower (data not shown). According to the lines, flowers with altered morphology represented 0% (L-5 and L-21) to 100% of the flowers (obligate parthenocarpic lines L-7, L-9, L-13, and L-14), the facultative parthenocarpic lines L-2, L-3, and L-4 presenting an intermediate proportion of altered flowers (60, 85, and 10% respectively). Although the duration of flower development was not modified in the transgenic lines (data not shown), carpel size at anthesis was 1.5–3 times that of the WT (Fig. 5B). Histological analysis of the three P35S:SlPIN4 RNAi lines selected for in-depth analysis (facultative parthenocarpic line L-2 and L-21; obligate parthenocarpic line L-14) showed an increased development of vascular tissues in the flower buds and revealed that carpel size was larger than those of the control WT plants as early as the 4mm stage in the P35S:SlPIN4 RNAi lines (Fig. 5C).

Fig. 5

. Modification of flowers in P35S:SlPIN4 RNAi plants. (A) Flower phenotype with visible alterations of sepals and stamens. The white bar represent 1cm. (B) Mean carpel area of flowers at anthesis in the WT and in transgenic lines. Vertical bars represent the standard deviation and an asterisk indicates a significant difference from the WT (t-test, n=10, P < 0.001). (C) Cytological analyses on 4mm and 8mm flower buds and 0 DPA ovaries.

As root development is affected in the young seedlings from the Arabidopsis Atpin3, Atpin4, and Atpin7 mutants (Blilou ; Petrásek and Friml, 2009), experiments were also carried out to examine whether root development was similarly modified in the P35S:SlPIN4 RNAi lines. Seeds from facultative parthenocarpic lines L-2 and L-21 where sown in vitro in the presence and absence of IAA. In the absence of IAA, P35S:SlPIN4 RNAi lines showed no significant differences in root growth when compared with the WT (Table 1). In the presence of IAA, primary root length was shortened in the WT and in P35S:SlPIN4 RNAi lines (data not shown). In addition, L-2 and L-21 P35S:SlPIN4 RNAi lines presented a significant reduction in the number of lateral roots when compared with the WT (Table 1). This phenotype is similar to the inhibition of lateral root initiation following naphthylphthalamic acid (NPA) treatments in tomato and in Arabidopsis (Muday and Haworth, 1994; Casimiro ) and suggests that auxin transport could be impaired in the roots of P35S:SlPIN4 RNAi lines.

Table 1.

Number of lateral roots developing on 12-day-old seedlings germinated in the presence/absence of 0.5 µM IAA

± represents the standard deviation (n=50).

* Significant variation of transgenic lines compared with the WT (t-test, P=0.001).

	WT	L-2	L-21
– IAA	8.62±1.71	8.32±1.96	8.22±1.97
+ IAA	7.56±1.63	6.00±1.92*	6.15±1.40*

Auxin metabolism is slightly affected in the P35S:SlPIN4 RNAi lines

Expression analysis of PIN genes during flower bud and early fruit development (Fig. 6) confirmed that SlPIN4 expression was strongly down-regulated in lines L-2, L-21, and L-14 at each stage of floral and fruit development studied. In addition, it revealed that the expression of other PIN genes was slightly modified in the transgenic lines during flower bud development. The SlPIN1 and SlPIN7 transcript levels were significantly higher at the 4mm and 8mm stages in the obligate parthenocarpic line L-14, while SlPIN3, the closest homologue of SlPIN4, displayed an opposite behaviour in the obligate parthenocarpic line L-14 and in the facultative parthenocarpic lines L-2 and L-21.

Fig. 6

. Expression profile of PIN genes in the WT and P35S:SlPIN4 RNAi lines. Gene expression was measured by qRT-PCR in flower buds (4mm and 8mm) and in ovary at anthesis and 4 DPA. The relative expression of each gene (arbitrary units) corresponds to gene expression normalized with the expression of β-tubulin and EiF4a. Vertical bars represent the standard deviation and an asterisk indicates a significant difference from the WT (t-test, n=3, P < 0.05).

According to the severe down-regulation of SlPIN4 in the three transgenic lines studied (Fig. 6), studies were conducted to determine whether auxin levels were affected in flower buds and ovaries from these plants. In 0 DPA ovaries, no significant differences in free IAA, oxindole-3-acetic acid (OxIAA), or auxin–aspartate conjugate (IAAasp) levels were observed between transgenic and WT plants (Fig. 7). In 8mm flower buds, only the IAAasp level was affected in the obligate parthenocarpic line L-14. The 4mm flower buds displayed more contrasted changes, depending on the auxin compound considered. The free IAA level was not significantly affected whatever the transgenic line considered, though an increase was clearly perceptible in the obligate parthenocarpic line L-14. The OxIAA level increased only in the obligate parthenocarpic line L-14. In contrast, the IAAasp level increased considerably in all three transgenic lines (Fig. 7C), clearly indicating a modification of auxin metabolism in the very young flower buds of the P35S:SlPIN4 RNAi lines.

Fig. 7

. Quantification of free auxin and related compounds in WT and in P35S:SlPIN4 RNAi flower buds (4mm and 8mm) and in ovary at anthesis (0 DPA). (A) Free IAA levels. (B) OxIAA levels. (C) IAA–aspartate conjugate levels. Vertical bars represent the standard deviation and an asterisk indicates significant differences (t-test, n=3, P < 0.05).

Several transcription factors are misregulated in the P35S:SlPIN4 RNAi lines

In order to identify auxin targets involved in the regulation of precocious ovary development in the SlPIN4-silenced lines, the transcriptome of flower buds (6mm stage) of WT and transgenic lines L-2, L-21, and L-14 was analysed using TOM2 microarrays. Although they do not represent the whole genome, TOM2 microarrays are enriched in fruit-expressed genes, in pre- and post-anthesis ovary-expressed genes, and in TFs (Alba ; Fei ). Among the 1460 genes significantly expressed in the flower buds of either transgenic lines or the WT, 167 genes displayed a variation of expression >1.5-fold in at least one P35S:SlPIN4 RNAi line when compared with the WT (Supplementary Table S3 at JBX online). Some genes previously highlighted in other parthenocarpic lines (Vriezen ; Martinelli ; Wang ; de Jong ) were misregulated in P35S:SlPIN4 RNAi lines, such as chlorophyll a/b-binding proteins, lipid transfer protein, and expansin (Supplementary Table S3). As previously shown (Molesini ; Wang ), polyamine metabolism was particulary impaired in P35S:SlPIN4 RNAi parthenocarpic lines (Table 2). Genes involved in abscisic acid or brassinosteroid metabolism were up-regulated in the three P35S:SlPIN4 RNAi lines (zeaxanthin epoxidase and sterol methyltransferase), whereas one gene coding for IAA-amido synthase GH3.1, involved in auxin conjugation, was down-regulated in these lines. Because the parthenocarpic fruit set resulting from inhibition of auxin transport is mediated by GA (Serrani ), the expression of three genes involved in GA biosynthesis (SlGA20ox-1, SlGA20ox-2, and SlGA3ox-1) and one GA-responsive gene (SlGAST) was further analysed by the more sensitive qRT-PCR. As shown in Supplementary Fig. S2 at JBX online, SlGA20ox-1, SlGA20ox-2, and SlGAST were significantly down-regulated in 8mm flower bud and/or in developing fruit, while SlGA3ox-1 was significantly up-regulated in 8mm flower bud in the most severely affected line L-14. For the GA biosynthetic genes, these results are consistent with those observed in the parthenocarpic transgenic lines silenced for SlARF7, an ARF acting as a negative regulator of fruit set (de Jong ). However, cross-talk between auxin and GA in the regulation of fruit set remains complex. Indeed, at the opposite, Wang et al. (2009) did not detect any variation of GA-related genes in the tomato parthenocarpic lines silenced for SlAux/IAA9.

Table 2.

Genes involved in hormone metabolism and transcription factors differentially expressed between P35S:SlPIN4RNAi lines and the WT in flower buds revealed by TOM2 microarray expression analysis

Spot ID	Unigene ID	Annotation	Arabidopsisgene	Fold change P35S:SlPIN4 RNAi/WT
				L-2	L-21	L-14
Hormone and polyamine synthesis
LE13G23	U573534	IAA-amido synthase GH3.1	AT2G14960	0.76	0.72	0.64
LE15N13	U578173	Sterol methyltransferase SMT2	AT1G20330	1.29	1.44	1.86
LE21D21	U590155	Squalene epoxidase XF1	AT1G58440	0.95	0.89	1.51
LE15J06	U569421	Zeaxanthin epoxidase ABA1	AT5G67030	1.39	1.52	1.33
LE5C02	U593751	Spermidine synthase SPMS, SPDS3	AT5G53120	1.46	1.20	1.58
LE3O19	U569175	Spermidine synthase ACL5	AT5G19530	0.63	0.59	0.44
LE16K13	U577418	Adenosylmethionine decarboxylase	AT3G25570	0.76	0.50	1.42
Transcription factors
LE28D08	U573372		AT3G16500	0.98	1.01	0.62
LE23I19	U594329	bHLP transcription factor	AT1G01260	0.79	0.55	0.97
LE14K15	U578320	bZIP transcription factor AtBZIP44	AT1G75390	0.74	0.81	0.61
LE18G14	U573993	Homeobox protein AtHB23	AT5G39760	1.15	0.97	0.57
LE16K18	U577422	Cycling DOF factor CDF3	AT3G47500	1.56	1.81	1.63
LE1F12	U569398	MADS-box PI (PISTILLATA)	AT5G20240	1.24	1.56	0.87
LE33O18	U573183	MADS-box TPI	AT5G20240	1.55	2.12	0.22
LE21H22	U576253	MYB transcription factor AtMYB21	AT3G27810	1.42	1.79	1.49
LE32N01	U602726	MYB transcription factor MYB24	AT5G40350	1.36	2.07	1.38
LE28D07	U569474	MYB transcription factor AtMYB44	AT5G67300	0.65	0.57	1.18
LE8A24	U566039	NAC transcription factor AtNAP	AT1G69490	1.06	1.50	1.04
LE11K21	U568609	NAC transcription factor AtNAC2	AT3G15510	0.86	1.03	0.65

In agreement with results from Wang , the TF category was one of the main categories presenting a differential expression in the P35S:SlPIN4 RNAi lines. The 12 TFs significantly misregulated in the transgenic lines (Table 2) belong to different classes, the major ones being the MYB, the MADS-box, and the NAC classes. Three genes (one DOF and two MYB) were up-regulated and one bZIP was down-regulated in all three transgenic lines. Three additional genes (Aux/IAA26 and homologues of AtHB23 and AtNAC2) were specifically down-regulated in the most severely affected obligate parthenocarpic line L-14. Other genes displayed dissimilar or even opposite patterns (TPI) of expression in the facultative and obligate parthenocarpic lines.

These results and the reported importance of the coordinated expression of TFs in the regulation of the transition from ovary to fruit development (Wang ) and in auxin signalling (reviewed in Chapman and Estelle, 2009; Kieffer ) prompted an extension of the investigations on the role of TFs. To this end, a qRT-PCR platform, more sensitive than microarrays and able to profile a large set (1090) of putative tomato TFs (Rohrmann ), was used. Flower bud (4mm) and ovary at anthesis (0 DPA) RNAs from one facultative parthenocarpic line (L-2) and one obligate parthenocarpic line (L-14) were compared with WT RNA. The 4mm flower bud was chosen because auxin metabolism is altered in the transgenic lines at this stage of flower bud development (Fig. 7). Previous reports also indicated that ovary at anthesis is the developmental stage at which the largest differences in gene expression can be observed between parthenocarpic and WT ovary/fruit (Pascual ; Wang ). Among the TFs significantly expressed in flower buds (1015) and in ovary (1005), 172 TFs were differentially expressed between P35S:SlPIN4 RNAi and the WT in flower buds and/or in ovary. Particular attention was drawn to the TFs with a ≥1.5-fold change in gene expression in P35S:SlPIN4 RNAi when compared with the WT (Supplementary Table S4 at JBX online).

In flower buds, 56 TFs were differentially expressed in P35S:SlPIN4 RNAi lines (Fig. 8A; Supplementary Table S4), 12 of them being up-regulated in both transgenic lines (Group 1 and 2, Table 3). Among these genes, two homologues of AtMYB21, a GATA zinc finger, and a homologue of AtERF110, presented the highest variation in expression (8- to 13-fold). Only two Aux/IAA TFs were up-regulated, corresponding to the homologues of AtAux/IAA16 and AtAux/IAA19. Interestingly, CRABS-CLAW, a C2C2-YABBY TF, was up-regulated in this analysis (Table 3). Sixteen genes were down-regulated in both transgenic lines, the main TF classes represented being the MADS-box (four TFs) and the MYB/MYB-related TFs (three TFs, Group 5, Table 3). A homologue of AtWRK16 was the most affected TF, with a 10-fold reduction of expression in the obligate parthenocarpic line (Table 3). Additional differentially regulated genes showed either a specific up-regulation in the facultative parthenocarpic line (Group 3, four TFs), or a specific down-regulation in the obligate parthenocarpic line (Group 4, 24 TFs). The main TF classes represented in the latter group are MADS-box TFs, including TPI and Jointless, and MYB/MYB-related TFs (Supplementary Table S4 at JBX online).

Fig. 8

. Groups of transcription factors differentially expressed in a global expression analysis by qRT-PCR using a TF profiling platform. (A) Clustering of the 56 TFs differnentially expressed in 4mm flower buds in P35S:SlPIN4 RNAi lines compared with the WT. (B) Clustering of the 79 TFs differentially expressed in 0 DPA ovaries in P35S:SlPIN4 RNAi lines compared with the WT. TFs were manually assigned to an expression group according to the values of log2(L-2/WT) and log2(L-14/WT). The relative expression of each gene (arbitrary units) corresponds to log2 (normalized gene expression WT/normalized gene expression transgenic line). The bold line corresponds to the mean of the values in the group.

Table 3

TFs up-regulated or down-regulated in the flower buds of the facultative and the obligate parthenocarpic P35S:SlPIN4RNAi lines revealed by qRT-PCR

a Clustering group according to Fig. 8A.

ID	Unigene ID	Annotation	Arabidopsishomologue	Groupa	Fold-change P35S:SlPIN4 RNAi/WT
					L-2	L-14
J0905	U573092	GATA type zinc finger TF	AT1G01780	1	9.68	2.26
J0021	U584903	AP2/B3-like TF	AT3G06160	2	1.36	1.72
J0074	U577088	Ethylene response TF, ERF110	AT5G50080	2	2.89	8.43
J0125	U579607	IAA19, MSG2	AT3G15540	2	1.26	2.06
J0141	U580151	IAA16	AT3G04730	2	1.18	1.57
J0166	U575251	ILR3, bHLH105	AT5G54680	2	1.10	1.53
J0240	U576188	AtBZIP53	AT3G62420	2	1.54	1.65
J0306	U572646	CRABS-CLAW, CRC	AT1G69180	2	1.26	1.99
J0349	U576689	AtZFP2	AT5G57520	2	1.45	5.28
J0385	U585955	Zinc finger (CCCH-type)TF	AT3G51950	2	1.07	1.52
J0712	U576253	AtMYB21	AT3G27810	2	6.30	13.84
J0746	U602726	AtMYB21	AT3G27810	2	6.65	11.62
J0261	U593980	TGA10, bZIP65	AT5G06839	5	0.81	0.42
J0355	U575936	C2H2-type zinc finger TF	AT2G28200	5	0.39	0.51
J0407	U586223	Zinc finger (CCCH type) TF	AT5G06770	5	0.78	0.62
J0453	U564651	NF-YB5	AT2G47810	5	0.51	0.15
J0498	U571553	DNA-binding storekeeper protein	AT4G00390	5	0.79	0.59
J0632	U597825	MADS-box TF	AT1G17310	5	0.40	0.26
J0637	U572195	AGAMOUS-like 2AGL20, SOC1	AT2G45660	5	0.59	0.50
J0660	U585967	AGAMOUS-like 66, AGL66	AT1G77980	5	0.55	0.74
J0681	U583379	MADS-BOX AP1, AGL7	AT1G69120	5	0.60	0.19
J0742	U563618	AtMYB62	AT1G68320	5	0.45	0.17
J0750	U600259	AtMYB80	AT5G56110	5	0.20	0.20
J0761	U604577	Myb-like TF	AT5G61620	5	0.41	0.15
J0803	U603443	NAC TF, NAC2, anac078	AT5G04410	5	0.25	0.15
J0858	U600327	Calmodulin binding TF	AT4G16150	5	0.51	0.20
J1031	U600423	AtWRKY16	AT5G45050	5	0.36	0.10
J1087	U576656	Mini zinc finger, MIF3	AT1G18835	5	0.51	0.22

In ovary at anthesis, 79 TFs were differentially expressed in the P35S:SlPIN4 RNAi lines (Fig. 8B; Supplementary Table S4). In these samples, the majority of the genes were differently regulated in the facultative and in the obligate parthenocarpic lines. This may reflect both the differences in ovary development at anthesis between the various transgenic lines and the absence of fertilization in the obligate parthenocarpic line. In this regard, 16 TFs were up-regulated only in L-2 (Group 2; Supplementary Table S4), including four MADS-box and four C3H zinc finger TFs. Twenty TFs corresponding mostly to MYB/MYB-related and MADS-box TFs (Group 3, Supplementary Table S4) were up-regulated in L-2 and down-regulated in L-14. However, six TFs were up-regulated in both transgenic lines (Group 1, Table 4). Among these genes, one bZIP (homologue of AtTGA10) presented the highest variation in expression (19-fold in line L-2). In addition, two ERFs (ethylene-responsive factors) were present in this group. Conversely, five TFs were down-regulated in both transgenic lines, among which were two heat shock factors (homologues of AtHSFA6B and AtHSF4) and the MADS-box TAGL11 (Group 4, Table 4). Finally, a large number of TFs were down-regulated only in the obligate parthenocarpic line (Group 5, 32 TFs). The main TF classes represented in these groups correspond to MYB/MYB-related, zinc finger, and MADS-box TFs including TAG1 and TDR3 (Supplementary Table S4 at JBX online). CRABS-CLAW was also included in this group of TFs repressed only in the obligate parthenocarpic line.

Table 4.

TFs up-regulated or down-regulated in the ovary of the facultative and the obligate parthenocarpic P35S:SlPIN4RNAi lines revealed by qRT-PCR

a Clustering group according to Fig. 8B.

ID	Unigene ID	Annotation	Arabidopsis homologue	Groupa	Fold-change P35S:SlPIN4 RNAi/WT
					L-2	L-14
J0087	U577090	ERF	AT5G25190	1	1.72	4.09
J0107	U565667	ERF, SHN3	AT5G25390	1	1.12	2.06
J0182	U571566	bHLH TF	AT5G57150	1	0.92	1.96
J0208	U581766	bHLH TF	AT1G72210	1	1.13	2.28
J0261	U593980	TGA10, bZIP65	AT5G06839	1	19.64	11.57
J0301	U563209	GATA type zinc finger TF, GATA21	AT5G56860	1	2.09	1.21
J0050	U591075	RAP2.7, TOE1	AT2G28550	4	0.86	0.52
J0603	U590552	AtHSFA6B	AT3G22830	4	0.72	0.51
J0607	U566892	AtHSF4	AT4G36990	4	0.92	0.49
J0672	U585390	MADS-box TF, STK, AGL11	AT4G09960	4	0.92	0.47
J1067	U580500	AtWRKY75	AT5G13080	4	0.57	0.23

Discussion

Parthenocarpic fruit development in SlPIN4-silenced lines

Among the 10 homologues of Arabidopsis PIN proteins found in tomato (Fig. 1A; Supplementary Table S2 at JBX online; Pattison and Catala, 2012), SlPIN4 displays remarkable features regarding flower and fruit development in tomato. SlPIN4 is much more highly expressed than its close homologue SlPIN3 or any other PIN gene in tomato ovary and fruit (Fig. 1B). Its transcript abundance peaks at anthesis, when the ovary develops into a fruit following fertilization (Fig. 1C). The SlPIN4 gene therefore appears to be a likely candidate for controlling auxin efflux in the ovary and early developing fruit of tomato.

AtPIN1 and AtPIN3 are the most highly expressed PIN genes in Arabidopsis ovaries (Paponov and http://jsp.weigelworld.org/expviz/expviz.jsp). However, inactivation of either AtPIN1 or AtPIN3 in Arabidopsis and of SlPIN4 in tomato leads to different phenotypic alterations in reproductive organs. Floral traits are strongly affected in the Arabidopsis pin1 and pin3 homozygous mutants, which are characterized by a drastic reduction in flower numbers and by modifications of the number and structure of floral components (Goto ; Bennett ). In addition, the pin1 homozygous mutant is sterile because of the homeotic transformations of floral organs, for example the presence of petaloid sepals or stamenoid petals (Goto ). Silencing of SlPIN4 in tomato only resulted in modifications of the shape of the floral components and, in a few cases, in increased carpel numbers (Fig. 4). The most striking feature in P35S:SlPIN4 RNAi lines was the development of parthenocarpic fruits, a phenotypic alteration which was not observed the Arabidopsis pin mutants already described. A possible explanation is that large alterations of flower development (Goto ) and lack of emasculation experiments in the Arabidopsis pin1 mutant hindered further observations of silique parthenocarpy. Another likely explanation is that in tomato the SlPIN4 gene has evolved to fulfil specialized functions related to the control of fruit set by auxin. The strong expression of SlPIN4 in the ovaries (Fig. 1), which has recently been shown to be restricted to the ovules (Nishio ), as well as the lack of compensation of SlPIN4 silencing by other PIN genes expressed in flower buds (Fig. 6), support this hypothesis.

Indeed, specific silencing of SlPIN4 had profound effects on fruit set in tomato, and led to the development of parthenocarpic fruits in the transgenic lines displaying phenotypic alterations. In these lines, two types of parthenocarpy were observed. The most severely affected lines such as L-14, called obligate parthenocarpic lines, were unable to produce seeds. The moderately affected lines such as L-2 and L-21, called facultative parthenocarpic lines (Fig. 3), could produce seeds. However, seed production was irregular: it was not observed in all fruits, and seed number was highly variable. It is important to mention that the ‘WVA 106’ cultivar has long been used for studying tomato flower and fruit development (Joubès ; Gilbert ; Mathieu-Rivet ) and that in normal growth conditions the development of parthenocarpic fruits has never been observed in this cultivar. In the obligate parthenocarpic lines, the onset of fruit development is very precocious and takes place as early as the 4mm flower bud stage, thus leading to the development of seedless fruits (Figs 3, 5C). In the facultative parthenocarpic lines, the onset of fruit development also takes place before anthesis, but later than in the obligate parthenocarpic lines (Fig. 5C). In contrast to obligate parthenocarpic lines, seeds can be produced after manual pollination, indicating that precocious fruit development does not hamper fertilization. The irregular production of seeds in these lines probably stems from the morphological alterations of the flowers, in which protruding styles were often observed. In that case, self-pollination was no longer possible due to the exsertion of the stigmas from the anther cone (Fig. 5A), unlike WT flowers in which the cone of anthers encloses the ovary, style, and stigmas (Chen et al., 2007). Cross-pollination is unlikely in the insect-proof greenhouse and culture conditions used in these experiments and, indeed, has never been observed. The combination of precocious fruit development and lack of fertilization therefore provides a likely explanation for the occurrence of unseeded or few-seeded fruits in the facultative parthenocarpic lines.

The findings of this study do not support the research of Pattison and Catala (2012) who observed strong phenotypic alterations at the vegetative level when silencing SlPIN4. One likely explanation is that the simultaneous silencing of multiple PIN genes, including not only SlPIN4 but also the closely related SlPIN3, and, in one line, also the SlPIN1, SlPIN7, and SlPIN9 genes, led to the pleiotropic plant phenotypes observed at the vegetative level in Pattison and Catala (2012). Such strong alterations were absent in the lines used here, which were specifically silenced for SlPIN4 (Fig. 6). More surprising is the lack of ovary/fruit phenotype in the study of Pattison and Catala (2012). In the present study, all the lines displaying comparable down-regulation of SlPIN4 exhibited fruit parthenocarpy (Figs 3, 6). This phenotype is strongly reminiscent of the effect of the polar auxin transport inhibitor NPA, a well known inducer of fruit parthenocarpy in tomato (Serrani ; Pattison and Catala, 2012). Again, the combination of multiple PIN gene silencing in Pattison and Catala (2012) may bias the results observed, for example by redirecting the auxin fluxes in the plant, ovary, and fruit, therefore counter-balancing the effect of SlPIN4 silencing. An alternative explanation is that the SlPIN4 silencing effect is genotype dependent. The Pattison and Catala (2012) study used ‘Ailsa Craig’, an S. lycopersicum variety commonly used in laboratory studies. The present study used the cherry tomato cultivar ‘WVA 106’, a tomato variety used for years in early fruit development studies (Joubès ; Gilbert ; Mathieu-Rivet ). As recently shown (Ranc ), cherry tomatoes can be considered as hybrids between S. lycopersicum and Solanum pimpinellifolium, the ‘WVA 106’ cultivar being closer to S. pimpinellifolium. Whether these different genetic origins may explain why the ‘Ailsa Craig’ cultivar is less sensitive than the ‘WVA 106’ cultivar to modifications of ovule to carpel auxin fluxes in the ovary before anthesis remains an open question.

Regulation of precocious fruit development in SlPIN4-silenced lines

How modulation of auxin transport and responses in tomato ovary via SlPIN4 silencing affects fruit set remains an open question. In Arabidopsis, the genetic manipulation of auxin efflux carriers of the AtPIN1, AtPIN3, AtPIN4, and AtPIN7 subfamily revealed how modifications of auxin distribution in plant tissues may affect a wide range of developmental processes such as response to gravity (Blilou ; Petrásek and Friml, 2009; Zadnikova et al., 2010). Such a detailed analysis of the auxin fluxes between ovules, placenta, and carpel walls of the tomato ovary was beyond the scope of this study. Recent studies, however, suggested that the patterns of auxin distribution in tomato ovary and fruit were dependent on polar auxin transport (Nishio ; Pattison and Catala, 2012). Using the SlPIN4-silenced lines, it was shown that the induction of precocious fruit development in tomato can be achieved by modifying polar auxin transport, without the large increases in ovary auxin content induced by ectopic expression of genes encoding enzymes of auxin biosynthesis (Pandolfini ; Molesini ). Indeed, the phenotypes observed in SlPIN4-silenced lines are fully consistent with the effects of polar auxin transport inhibitors in tomato (Serrani ). It was further shown that redirection of the ovary developmental programme probably occurs very prematurely since alterations in auxin metabolism were detected at very early stages of flower bud development (Fig. 7). These results and the temporal and spatial distribution of SlPIN4 in tomato ovary (Fig. 1 and Nishio ) suggest that SlPIN4 interacts with the ovary to fruit transition by controlling local distribution of auxin in ovules and nearby tissues.

In turn, this modification of auxin local distribution in the ovaries of P35S:SlPIN4 RNAi lines may affect GA biosynthesis and response. Indeed, GA has been shown to be involved in tomato fruit set (reviewed in Schwabe and Mills, 1981; Gorguet ; deJong et al., 2009a). Serrani recently proposed that the fruit parthenocarpy induced by inhibition of auxin transport is mediated by GAs (Serrani ). In this study, it was shown that the expression profiles of GA biosynthetic and response genes were significantly altered in the ovary buds from transgenic lines, as early as the 4mm stage, and later on during early fruit development (Supplementary Fig. S2 at JBX online). These results are consistent with those obtained in the SlARF7-silenced lines (deJong et al., 2010), but not with those from Serrani et al. (2008, 2010) who observed the up-regulation of GA biosynthetic genes following 2,4-D or NPA treatments. These apparent discrepancies reflect the complexity of fruit set regulation and highlight the need for additional studies, using various approaches (e.g. tissue- or cell-specific promoters with various target genes) to further our understanding of the hormonal regulation of fruit set.

Whatever the mode of action of SlPIN4, modification of auxin fluxes in the ovary from SlPIN4-silenced lines must trigger the precocious deregulation of genes associated with the parthenocarpic development of the ovary into a fruit. Various transcriptomic approaches have been undertaken to unravel changes in gene expression during fruit set (Pascual ; Serrani ; Vriezen ) or to characterize parthenocarpic fruits (Martinelli ; Molesini ; Pascual ; Wang ; de Jong ). The differentially expressed genes mainly detected were involved in hormonal signalling or in fruit growth processes such as the cell cycle and cell wall modification. Here, attention was focused on TFs, which are likely targets of auxin signalling with possible roles in fruit set.

In flower buds, a number of differentially expressed genes implicated in flower development such as AtMYB21 and AtMYB24 (Cheng et al., 2009), TGA10 (Murmu ), and Aux/IAA19 (Tashiro ) were identified (Table 2; Supplementary Table S3 at JBX online). These results are consistent with the role of auxin in flower development (Aloni ) and with the visible flower morphological alterations observed in the P35S:SlPIN4 RNAi lines (Fig. 5A). However, several of these genes, such as the homologues of AtMYB21 and TGA10, were misregulated in both flower buds and ovaries (Supplementary Table S3), suggesting that their role in tomato is not restricted to the three first flower whorls but is extended to ovary/fruit development.

In the ovary at anthesis, TF profiling allowed the identification of several categories of TFs with possible roles in the ovary to fruit transition (Table 3; Supplementary Table S4 at JBX online). Though auxin signalling responsible for the onset of fruit development is transmitted through a specific Aux/IAA–ARF complex (Goetz ; Swain and Koltunow, 2006; Pandolfini ; Wang , de Jong ), only minor changes in SlAux/IAA expression were detected in flower buds from P35S:SlPIN4 RNAi lines (Tables 2, 3). Molesini ) reported results similar to those in the present study in the SlAucsia-silenced plants displaying auxin transport deficiency. These authors therefore proposed that auxin transport-related parthenocarpy could be independent of SlAux/IAA9 and SlARF8, two genes implicated in auxin-dependent parthenocarpic fruit development (Goetz ; Wang ). A recent study also suggested that induction of fruit set in unpollinated tomato ovary treated with NPA, an auxin efflux transport inhibitor, may be mediated through changes in GA metabolism (Serrani ). Actually, most of the differentially expressed TFs identified through P35S:SlPIN4 RNAi profiling are likely to be implicated in the regulation of developmental processes. These include the MADS-box genes, which have a wide implication in the regulation of flower and fruit development (Dornelas ; Klee and Giovannoni, 2011). Indeed, a large number of MADS-box TFs were down-regulated in the ovaries of the facultative and/or obligate parthenocarpic P35S:SlPIN4 RNAi lines, including TAGL11, Fruitfull 1/TDR4, TDR3, TAG1, PI, homologues of AGL22, AGL42, and AGL68 (Table 4; Supplementary Table S4 at JBX online). These results are in agreement with previous studies showing the misregulation of Fruitfull 1/TDR4 (Vriezen ), TAG1, and TAGL6 (Wang ) in tomato parthenocarpic fruits. Previous functional analysis of TAG1 did not reveal any function for this gene in fruit set, probably because of the dramatic phenotypic changes linked to its homeotic function (Pnueli ). However, the up-regulation of Fruitfull 1/TDR4, TAG1, and TAGL11 in tomato carpel at anthesis (Busi ) is in agreement with their possible implication in the regulation of fruit set. Interestingly, MADS-box TFs are active as multimers, and numerous interactions have been identified between the MADS-box TFs identified in the present study and TM29 (Busi ; Leseberg ), a MADS-box implicated in the regulation of fruit set (Ampomah-Dwamena ). It has been recently proposed that the MADS-box network controls the identity of floral organs and the growth of floral components and fruit by targeting genes associated with cell proliferation and growth (Dornelas ). In Arabidopsis, one of the targets of the MADS-box AGAMOUS is CRABS-CLAW, a YABBY TF implicated in the control of carpel development (Alvarez and Smyth, 1999; Bowman and Smyth, 1999; Gómez-Mena ; Lee ). Similarly, the co-regulation of CRABS-CLAW and TAG1 in P35S:SlPIN4 RNAi lines (Supplementary Table S4 at JBX online) may suggest the implication of both genes in tomato fruit set.

In conclusion, this work provides clear evidence that the tomato PIN protein SlPIN4 plays a major role in the auxin control of tomato fruit set, possibly by preventing precocious fruit development in the absence of pollination. It further sheds new light on the regulation of fruit set and onset of parthenocarpic fruit development in tomato by highlighting the TFs involved in the regulation of fruit set and ovary/fruit development, which are likely targets of auxin signalling.

Supplementary data

Supplementary data are available at JXB online.

Figure S1. Expression profile of the SlPIN4 gene in leaves of T0 P35S:SlPIN4 RNAi lines.

Figure S2. Expression profile of genes involved in gibberellin signalling.

Table S1. List of the specific primers used for qRT-PCR analysis.

Table S2. Accession numbers of tomato PIN genes and mRNA sequences.

Table S3. List of the 1460 genes with significant expression in P35S:SlPIN4 RNAi lines in TOM2 microarray analysis flower buds at the 6mm stage.

Table S4. List of the TFs with a differential expression in P35S:SlPIN4 RNAi lines compared with the WT in qRT-PCR profiling analysis in flower buds at the 4mm stage and in ovaries at 0 DPA.