Changes in tomato ovary transcriptome demonstrate complex hormonal regulation of fruit set


Author for correspondence:
Wim H. Vriezen
Tel:+31 24 3652761
Fax:+31 24 3652490


  • • Plant hormones are considered to be important mediators of the fruit developmental signal after pollination. The role of phytohormones in tomato (Solanum lycopersicum) fruit set was investigated here.
  • • Transcriptome analysis of ovaries was performed using two complementary approaches: cDNA–amplified fragment length polymorphism (AFLP) and microarray analysis.
  • • The gene expression profiles obtained suggest that, in addition to auxin and gibberellin, ethylene and abscisic acid (ABA) are involved in regulating fruit set. Before fruit development, many genes involved in biotic and abiotic responses are active in the ovary. In addition, genes involved in ethylene and ABA biosynthesis were strongly expressed, suggesting relatively high ethylene and ABA concentrations before fruit set. Induction of fruit development, either by pollination or by gibberellin application, attenuated expression of all ethylene and ABA biosynthesis and response genes within 24 h.
  • • It is proposed that the function of ABA and ethylene in fruit set might be antagonistic to that of auxin and gibberellin in order to keep the ovary in a temporally protected and dormant state; either to protect the ovary tissue or to prevent fruit development before pollination and fertilization occur.


The shift from the static flower ovary to fast-growing young fruit is a phenomenon known as fruit set, and is an important step in the development of all sexually reproducing higher plants. In general, fruit set is induced after pollination and successful fertilization of the egg cells in the ovules (Gillaspy et al., 1993). The embryo or surrounding tissue possibly produces the first signals that stimulate growth of the placental tissue and the ovary wall (pericarp). Gustafson (1936, 1939) showed that auxin application can stimulate seedless (parthenocarpic) fruit growth in some species, and the role of auxin in fruit development has been firmly established by many other researchers (Nitsch et al., 1960; Sastry & Muir, 1963; Mapelli et al., 1978; Mapelli & Lombardi, 1982; Alabadíet al., 1996; Wang et al., 2005). An important function in fruit set has also been ascribed to gibberellins (GAs), whose application to tomato (Solanum lycopersicum) flowers also leads to parthenocarpic fruit growth (Asahira et al., 1968; Bünger-Kibler & Bangerth, 1982; Sjut & Bangerth, 1982; Alabadíet al., 1996; Fos et al., 2000). Auxin-induced fruits differ from GA-induced fruits in that they contain many more cells. GA stimulates growth by stimulating cell enlargement and to a lesser degree by stimulating cell division. Normal fruit growth probably requires the coordinated action of both hormones (Bünger-Kibler & Bangerth, 1982). In addition to auxin and GA, other plant hormones, such as cytokinins, abscisic acid (ABA) and ethylene, play a role in fruit development (Nitsch, 1970). The precise actions and interactions of these hormones in regulating normal fruit development are only partly understood. Thus, all the classical plant hormones probably influence fruit development, suggesting that many developmental and environmental factors that influence hormone concentrations are also linked to this process. Indeed, light conditions, humidity and temperature all have to be within a certain range to allow tomato fruits to develop (George et al., 1984). Detailed knowledge of the regulation of fruit set might help to optimize growth conditions or plant properties that might be beneficial for tomato cultivation.

At present, only a few components of the mechanism regulating fruit set have been identified. One class of important proteins includes the auxin/indole-3-acetic acid (Aux/IAA) transcriptional regulators that mediate many aspects of plant responses to auxin (Kim et al., 1997; Ulmasov et al., 1997). Tomato lines with inhibited IAA9 expression develop fruits without fertilization, suggesting a role for IAA9 in fruit set as a transcriptional repressor of auxin signaling (Wang et al., 2005). In Arabidopsis thaliana, another negative regulator of fruit set has been identified. AUXIN RESPONSE FACTOR 8 (ARF8), a transcription factor that interacts with Aux/IAA proteins, inhibits fruit development until fertilization has taken place (Goetz et al., 2006). In addition to auxin signaling, GA signaling in fruit set seems to be under the control of negative regulators. The parthenocarpic mutants pat-2 and pat-3/pat-4 contain higher GA concentrations, which induce parthenocarpic fruit development (Fos et al., 2000, 2001), suggesting that GA biosynthesis in wild-type plants is limited by the pat gene products. In addition, the silencing of a tomato floral organ specification gene, TM29, a SEPELLATA-like MADS (MCM1, AGAMOUS, DEFICIENCE and SRF) box gene, causes seedless fruits (Ampomah-Dwamena et al., 2002).

A few studies have recently been performed to study tomato fruit set and early tomato fruit development at the transcriptional level (Testa et al., 2002; Lemaire-Chamley et al., 2005). However, the first step of fruit development, fruit set, is largely unexplored. A transcriptome analysis of tomato (Solanum lycopersicum cv. Moneymaker) ovaries was performed to identify genes involved in fruit set. In light of the fact that plant hormone action and fruit set are tightly linked, transcriptomes were compared from pollinated ovaries and GA-treated ovaries. Gibberellin induces fruit set without seeds, which makes it possible to identify a set of genes that can induce fruit growth independently of pollination or fertilization. A cDNA–amplified fragment length polymorphism (AFLP) approach was used to analyze transcript abundances in the placenta with attached ovules and separately in the pericarp during the 3 d after fruit induction. In addition, the GeneChip Tomato Genome Array (Affymetrix, Santa Clara, CA, USA) was used to analyze the abundances of 9254 known tomato transcripts in the ovary. Both methods demonstrate that auxin signaling is induced by pollination and to a lesser degree during fruit set induced by GA, and many genes involved in ethylene and abscisic acid signaling appear to be modulated during fruit set. The possible role of the different hormones and the sequence of their actions are discussed.

Materials and Methods

Plant material

Tomato (Solanum lycopersicum L. cv. Moneymaker) plants were grown on soil in the glasshouse under a daily temperature regime of high day temperatures (20–25°C) and lower night temperatures (15–18°C). Plants were grown during the summer (June–September) and photoperiods were extended to 16 h with low-intensity light supplied by high-pressure sodium lamps (600 W; Philips, Eindhoven, the Netherlands). All flowers used were emasculated 3 d before anthesis (t = day – 3) when anthers were still green to prevent self-pollination. Ovaries from control flowers were treated at anthesis (t = day 0) with 2 µl of 10% (volume/volume (v/v)) ethanol, which corresponds to the ethanol concentration in the gibberellic acid (GA3) solution. Treatments (t = day 0) consisted of either pipetting 2 µl of 1 mm (0.6 µg) GA3 in 10% (v/v) ethanol on top of the ovary or pollination. Ovaries (10 per sample) were collected 24, 48 and 72 h (respectively t= days 1, 2 and 3) later, and using a dissecting binocular pericarp were separated from the placenta with ovules. Both tissues were frozen in liquid nitrogen and stored at –80°C until RNA extraction.

RNA isolation and cDNA-AFLP

Total RNA was isolated from the frozen tissue samples using an RNA isolation kit (Qiagen RNeasy mini kit; Qiagen, Valencia, CA, USA). Five micrograms of total RNA in a maximum volume of 10 µl of demineralized RNase-free water was mixed with 40 µl of lysis buffer and 0.05 µl of biotinylated oligo-dT (both included in the mRNA Capture Kit; Roche Diagnostics, Basel, Switzerland). This mix was transferred to the streptavidin-coated PCR tubes (supplied with the mRNA Capture Kit), incubated at 37°C for 5 min in a thermo-cycler (Biometra T1, Biometra, Göttingen, Germany), allowed to cool to room temperature and stored on ice. During this step, the poly-A tails of the mRNA hybridize to the biotinylated oligo-dT which binds with its biotin molecule to the streptavidin-coated PCR-tube wall. The liquid was discarded and the bound mRNA washed twice with 100 µl of washing buffer (supplied with the mRNA Capture Kit). Reverse transcription was started by adding 50 µl of first-strand cDNA synthesis mixture (31.3 µl of water, 10 µl of 5 × first-strand buffer, 5 µl of 0.1 m dithiothreotol (DTT), 2.5 µl of 10 mm dNTPs and 1.2 µl of Superscript II reverse transcriptase (200 units µl−1); all components were from Invitrogen (Carlsbad, CA, USA) and the reaction mix was incubated for 2 h at 42°C. Thereafter, tubes were stored on ice and 10 µl of the reaction volume was discarded to obtain a total volume of 40 µl. A mixture containing 87 µl of demineralized water, 16 µl of 10 ×Escherichia coli ligase buffer, 6 µl of 0.1 m DTT, 3 µl of 10 mm dNTPs, 1.5 µl of E. coli ligase (all from Invitrogen), 4.25 µl (5 units) of E. coli DNA polymerase I, and 1.6 µl (5 units) of RNase H (both from Amersham Bioscience, Uppsala, Sweden) was added and incubated for 1 h at 12°C and subsequently for 1 h at 22°C in a thermocycler (Biometra T1). After incubation, the reaction mixture was discarded, while the double-stranded cDNA was still attached to the tube wall. The cDNA was washed three times with 200 µl of washing buffer. The first cDNA digestion was started by adding a mixture containing 38.8 µl of water, 10 µl of 5 × restriction-ligation buffer (5 × RL buffer: 50 mm Tris-HAc, pH 7.5, 50 mm MgAc2, 250 mm KAC and 25 mm DTT) and 1.2 µl of BstYI (10 units µl−1; New England Biolabs, Beverly, MA, USA). After incubation for 2 h at 60°C, tubes were washed three times with 100 µl of washing buffer. The second digestion was initiated by adding a mixture of 38.8 µl of water, 10 µl of RL buffer and 1.2 µl of MseI (10 units µl−1; New England Biolabs) followed by incubation for 2 h at 37°C. Liberated BstYI- MseI fragments were used for the next steps of the cDNA-AFLP procedure, which was essentially carried out as described by Breyne et al. (2002). Gene expression in the RNA samples was surveyed using 128 primer combinations for selective amplification. The cDNA-AFLP experiment was repeated with RNA from ovaries (placenta with ovules and ovary wall not separated) collected 3 d after treatments in an independent experiment.

Characterization of cDNA-AFLP fragments

The majority of bands showed no change in intensity between the samples (Supplementary Material Fig. S1). Changes in the intensity of individual bands did not affect others in the same lane, indicating that product accumulation was not affected by the concentration of individual substrates in the reaction. In some cases, a gene appeared to be induced in only one of the 18 samples; this finding generally could not be confirmed by the control experiment. These inconsistent bands were mainly observed in the region of the gel with very small DNA fragments (< 70 bp) and were therefore considered as aspecific amplification products. Bands corresponding to differentially expressed genes were cut out from the gel and eluted DNA was re-amplified under the same conditions as for the selective amplification. Fragments were subsequently ligated in a T-tailed EcoRV digested phagemid pBluescriptII SK(+) (Stratagene, La Jolla, CA, USA) and sequenced using the CEQ™ DTCS Quick Start Kit and CEQ2000 DNA Analysis System (both Beckman Coulter, Fullerton, CA, USA). Fragments that did not have the expected size, based on the height on the acryl amide gel from which they were isolated, were discarded. All cDNA-AFLP expression patterns displayed in the Results section were confirmed either by real-time quantitative PCR or using the microarray data.

Microarray analysis

The GeneChip Tomato Genome Array (Affymetrix) consists of over 10 000 S. lycopersicum probe sets for interrogation of over 9200 S. lycopersicum transcripts. Sequence information for this array was selected from public data sources including S. lycopersicum UniGene Build #20 (3 October 2004) and GenBank® mRNAs up to 5 November 2004. Detailed information can be found at Flowers were given the same treatments as in the cDNA-AFLP experiment. RNA was isolated from whole ovaries (10 per sample) 3 d after treatments. This procedure was repeated once to obtain two sets of RNA samples representing fully independent biological and technical repeats. Probe synthesis, chip hybridization and primary data analysis were performed by ServiceXS (Leiden, the Netherlands). RNA quality and quantity were analyzed using the Agilent 2100 Bioanalyser (Agilent Technologies, Santa Clara, CA, USA). Two micrograms of total RNA was used for the Affymetrix One-Cycle Target Labeling following the manufacturer's protocols (Affymetrix). The concentration and integrity of the cRNA were determined using the Agilent 2100 Bioanalyser (Agilent Technologies). For the hybridization, 10 µg of fragmented cRNA was used following the Affymetrix protocols. Data analysis was performed using rosetta luminator Software (Rosetta Inpharmatics, Seattle, WA, USA). The rosetta biosoftware error models calculate the total error for each measurement, and derive a P-value (a measure of confidence) for a given hypothesis. The P-value derived from the total error calculated by the error model is used to accept or reject the null hypothesis (H0) that a gene is not differentially expressed. The ‘fold change’ is the increase or decrease in the transcript abundance of a sequence in an experimental condition vs a baseline condition, measured on a linear scale (rosetta resolver; res4010-ug-07162004; Only genes whose expression was changed after treatment more than threefold were included in the analysis described in this paper. The P-values of all microarray data were used to calculate a q-value. The q-value of a particular feature can be described as the expected proportion of false positives among all features as or more extreme than the observed one (Storey & Tibshirani, 2003). The q-values were calculated using a computer program (qvalue; To obtain a false discovery rate (FDR) < 0.01, P-values for significant expression data after GA treatment should be < 0.003, and those for the expression data after pollination should be < 0.007.

Real-time quantitative PCR

Total DNA-free RNA was isolated from tomato ovaries (n > 5) using the Qiagen RNeasy mini kit together with the RNase-Free DNase Set (Qiagen). A PCR reaction with RNA- and DNA-specific primers (Supplementary Material Table S7) based on an intron in the S. lycopersicum actin gene Tom51 (GenBank accession number U60481) was performed to ensure the absence of contaminating genomic DNA. Total DNA-free RNA (1 µg) was used for cDNA synthesis (iScriptTM cDNA Synthesis Kit; Bio-Rad Laboratories, Hercules, CA, USA) in a total volume of 25 µl. PCR reactions were carried out in 25 µl containing 0.125 µl of cDNA synthesis reaction mixture, 400 nm of each primer and 12.5 µl of iQ SYBR Green Supermix (Bio-Rad Laboratories). PCRs were performed in a 96-well Bio-Rad iCycler (Bio-Rad Laboratories) using a temperature program starting with 3 min at 95°C followed by 40 cycles consisting of 15 s at 95°C and 45 s at 57°C, and finally the melting temperature of the amplified product was determined to verify the presence of a specific product. In addition, a fraction of the PCR mixture was analysed on a 1% agarose/ethidium bromide gel to check the size of the amplified DNA fragment. The primers that were used for the real-time quantitative PCR reactions were designed using a computer program (beacon designer 5.01; Premier Biosoft International, Palo Alto, CA, USA) to obtain primers that have close to identical melting temperatures and do not form secondary structures with each other in the given PCR conditions. Primer sequences are listed in Supplementary Material Table S7. In addition, to enhance primer efficiency, primer binding sites were chosen such that secondary structures of the template were avoided. All reactions were performed on two independently collected series of RNA samples. The gene expression patterns presented were comparable in the two series. The figures show one series with error bars based on one technical repeat.


Ovary transcriptome analysis during fruit set

Fruit growth normally starts after pollination and successful fertilization. Pollen tube growth in vivo was analyzed to determine the time it takes for tomato pollen to germinate and to grow through the stigma and style to reach the ovary. Pistils were harvested at several time-points after pollination and observed by epifluorescence after being fixed and stained with decolorized aniline blue (Kho & Baer, 1968). The first pollen tube tips reached the ovary between 12 and 15 h after pollination (data not shown), suggesting that fertilization of the first ovules starts after 12 h and probably continues for a few hours thereafter. To investigate early ovary gene expression and changes caused by fertilization, ovary tissues were collected 24 h, 48 and 72 h after pollination or after treatment with GA3. To distinguish between genes induced by the pollination/fertilization event and genes involved in fruit growth, other ovaries were treated with GA3. The application of GA3 induced fruit set and the development of unpollinated ovaries in tomato, which is in agreement with previous results in different tomato cultivars (Bünger-Kibler & Bangerth, 1982; Sjut & Bangerth, 1982; Alabadíet al., 1996). The ovary wall and placenta together with attached ovules were separated and collected for visualization of differences in gene activity in these tissues.

The cDNA-AFLP fragments ranged in length from 50 to 500 bp; for each primer combination, 65–80 bands were observed, with ∼9000 fragments being analyzed. Supplementary Material Fig. S1 shows corresponding gel parts from the two experiments. Expression patterns appeared to be closely comparable, although in many cases differences in expression levels were attenuated in the experiment with complete ovaries, while in other cases they seemed to be more pronounced. Fragments were selected based on clear visual differences in intensity and on consistency of observed changes. High-quality DNA sequences were obtained for 283 of 319 isolated differentially expressed fragments. The sequences were compared with those in the GenBank database (, the Dana-Farber Cancer Institute (DFCI) Tomato Unique Gene Indices ( and the SOL Genomics Network (SGN) unigene database ( using the blast program (Altschul et al., 1997). Accessions were found for 202 sequences (Supplementary Material Table S1) and another 35 sequences were highly identical to sequences from Arabidopsis thaliana or Solanum tuberosum. All sequences with their corresponding database accession numbers and tentative annotations (if present) are listed in Supplementary Material Table S1.

The availability of the Affymetrix GeneChip Tomato Genome Array made it possible to validate the results obtained in the cDNA-AFLP experiment. The results of quantitative analysis of the ovary transcriptome after GA3 application and pollination are displayed using a Venn diagram (Fig. 1; Ruskey, 1997). The number of genes whose expression was significantly (P < 0.003) changed more than threefold in the ovaries was 689 after GA3 treatment and 874 (P < 0.007) after pollination. As most of the GenBank accessions corresponding to these genes were not annotated, they were all compared with the GenBank database, DFCI Tomato Unique Gene Indices and the SGN unigene database using the blast program (Altschul et al., 1997). The GA3-modulated genes together with the tentative annotations are listed in Supplementary Material Table S2, and the additional genes that were only modulated after pollination (Fig. 1; hatched area in the Venn diagram) are listed in Supplementary Material Table S3.

Figure 1.

Quantification of gibberellic acid (GA3)- and pollination-induced ovary transcriptome profiles. The numbers of genes on the Tomato Genome Array that changed intensity by more than threefold after treatments are shown. Upward and downward pointing arrows indicate decreased and increased gene activity, respectively.

The cDNA-AFLP patterns (induced or reduced mRNA abundances after treatment) of the genes displayed in the figures were all confirmed by the second cDNA-AFLP experiment. In addition, they were reconfirmed either by the microarray data or by real-time quantitative PCR in cases in which the gene was not present on the GeneChip.

Hormone-related genes

The main goal of this study was to identify genes involved in hormone signaling during fruit set. Table 1 lists a subset of genes that were modulated after fruit set and involved in GA and auxin signaling. Table 2 lists genes involved in ethylene and ABA signaling. Ethylene and ABA are both hormones known to be important at later stages of tomato fruit development (reviewed by Gillaspy et al., 1993) and have no known function during or before fruit set.

Table 1.  Genesa identified by microarray analysis involved in gibberellin and auxin biosynthesis, signaling and response
GenBank accession numberGA3-treated vs controlTentative annotationbDFCI accession numberPollinated vs control
Fold changeP-value (P < 0.003)Fold changeP-value (P < 0.007)
  • a

    Only genes whose activity, as a result of one of the treatments, significantly changed more than threefold compared with the untreated sample are listed. ‘Fold change’ values in bold are not significantly different from the control values.

  • b

    Tentative annotations are based on the Dana-Farber Cancer Institute (DFCI) Tomato (Lycopersicon esculentum) Gene Index (LeGI).

  • AUX, auxin; GA, gibberellin; GAST, GA-stimulated; GST, glutathione S-transferase; IAA, indole-3-acetic acid; IAR, IAA-alanine resistent; 3OH, 3β-hydroxylase; 20ox, 20-oxidase; parC, protoplast auxin responsive.

Gibberellin-related genes
AI771164–8.131.94E-07Similar to gibberellin 3-oxidase 1TC157125–2.865.84E-03
AB010991.1–2.320.43Le3OH-1 mRNA for 3β-hydroxylaseTC1670046.161.69E-03
AF049898.11.40.73Le20ox1 mRNA for GA 20-oxidaseTC16111035.51.23E-21
AW650375–1.80.45Some similarity to SLEEPY1 proteinTC164844–8.513.42E-04
BG12342212.231.21E-29Probable cell wall protein (GA-induced)TC164485–1.80.15
AW6496599.952.82E-04Similar to GAST1 protein precursor 24.282.60E-18
BG6268824.292.44E-10Gibberellin-regulated protein GAST1TC1632121.833.11E-03
Auxin-related genes
Transport and homeostasis
BE4502071.230.75AUX1-like amino acid permeaseTC1611055.636.58E-03
BI925563–4.17.02E-07Auxin efflux carrier protein familyTC158913–6.111.70E-11
AI4848581.850.66Similar to auxin efflux carrier family proteinTC16399411.772.72E-07
AW220773–2.883.37E-03Similar to IAA-amino acid hydrolase 3 (IAR3)TC156740–11.623.58E-12
AF022013.1–1.160.9AUX/IAA family (IAA2) 26.196.51E-23
BI205027–1.040.97AUX/IAA family (IAA5)TC1630484.293.30E-03
AF022020.1–1.440.38AUX/IAA family (IAA9)TC159789–3.655.60E-04
AW034122–1.210.87AUX/IAA family (IAA14)TC16340133.930
CN385509–3.451.35E-09AUX/IAA family (IAA18)TC163812–1.845.14E-04
BT013639.11.240.58Similar to auxin response factor (ARF9) 7.415.18E-12
AF332960.1–8.542.37E-04Auxin-regulated dual specificity cytosolic kinaseTC155316–3.820.02
BG627684–2.10.37Similar to GST (auxin-regulated protein parC)TC1678124.031.16E-05
BT013446.12.950.27Similar to auxin- and ethylene-responsive GH3-like proteinTC16794911.031.49E-06
AW224185–30.661.66E-10Auxin- and ethylene-responsive GH3-like proteinTC156174–5.655.70E-04
AF416289.1–10.277.02E-04Auxin-regulated proteinTC162697–4.161.09E-03
BF113559–4.012.31E-04Some similarity to auxin-induced proteinTC154355–2.657.18E-03
AW6237523.712.78E-04Auxin-induced (IAA-induced) proteinTC1693291.290.56
Table 2.  Genesa identified by microarray analysis involved in ethylene, abscisic acid (ABA) and cytokinin biosynthesis, signaling and response
GenBank accession numberGA3-treated vs controlTentative annotationbDFC accession numberPollinated vs control
Fold changeP-value (P < 0.003)Fold changeP-value (P < 0.007)
  • a

    Only genes whose activity, as a result of one of the treatments, significantly changed more than threefold compared with the untreated sample are listed. ‘Fold change’ values in bold are not significantly different from the control values.

  • b

    Tentative annotations are based on the Dana-Farber Cancer Institute (DFCI) Tomato (Lycopersicon esculentum) Gene Index (LeGI).

  • AP, APETALA; ERD, early response to dehydration; ER, ethylene responsive; GA3, gibberellic acid; Pti, Pto-interacting protein; RAP, related to AP; RAV, related to ABI3/VP1.

Ethylene-related genes
AY079426.1–2.566.10E-13Lycopersicon esculentum ethylene receptor-like protein (ETR6)TC157353–4.124.43E-13
Biosynthesis     4.76E-08
U72389.1–3.211.40E-06ACC synthase (LE-ASC1A)TC154563–3.544.76E-08
M34289.1–26.837.23E-12ACC synthase (ACS2)TC155080–13.23.96E-11
AB013100.1–7.371.07E-09ACC synthase (ACS6)TC162669–11.614.12E-25
CN384809–49.210ACC oxidase (LE-ACO1)TC161922–19.880
X58885.1–15.376.38E-26Ethylene-forming enzyme (LE-ACO3)TC161924–7.21.60E-11
AB013101.1–1.581.31E-03ACC oxidase (LE-ACO4)TC154814–3.325.80E-07
AJ715790.1–20.462.80E-05ACC oxidase (LE-ACO5)TC158405–37.083.44E-09
BF112635–4.865.64E-09ACC oxidaseTC164031–5.325.45E-09
AW223067–2.25.95E-05ACC oxidase homolog (protein E8)TC168797–4.089.35E-08
AI775872–24.56E-05ACC oxidase homolog (protein E8)TC165186–3.915.65E-16
BI210054–5.252.39E-06Similar to ACC oxidase homolog (protein E8)TC154470–2.26.42E-03
AF328786.1–3.33.23E-15Ethylene-insensitive 3-related (EIL3)TC164031–3.821.53E-15
AY192367.1–4.11.56E-16Ethylene response factor 1 (ERF1)TC168797–4.411.09E-33
AY077626.12.670.38Ethylene response factor 1 (ERF1)TC1651866.574.81E-03
AY192369.1–3.227.26E-12Ethylene response factor 3 (ERF3)TC154470–2.299.57E-09
AY192370.1–3.521.21E-05Ethylene response factor 4 (ERF4)TC156366–2.362.93E-03
CN385340–5.641.89E-03Homologous to ethylene response factor 5 – 7.393.06E-06
AJ784436–3.880Ethylene-responsive element binding factor 9 identical to ERFTC154775–5.530
AF502085.1–3.820.27Ethylene-responsive element binding protein (EREB)TC159935–8.721.29E-03
U89255.1–9.994.35E-12Ethylene-responsive element binding factor 2 (Pti4)TC155398–6.281.94E-09
U89256.1–75.533.80E-25DNA-binding protein Pti5 mRNATC156842–81.997.11E-30
U89257.1–4.571.52E-06Pathogenesis-related genes transcriptional activator Pti6TC156737–4.171.51E-12
CN3856411.40.46Dehydration-responsive element binding protein (DREB1-like)TC166990–6.973.11E-03
CN385434–2.931.33E-03Dehydration-responsive element binding protein (DREB1-like)TC165298–5.691.70E-08
CK715216–7.891.67E-03Dehydration-responsive element binding protein (DREB1-like)TC159681–1.150.78
AF506825.1–2.51.29E-10Dehydration-responsive element binding protein 3 (DREB2-like)TC154229–3.191.55E-29
AF096246.11.180.8Ethylene-responsive transcriptional coactivator (ER24)TC156704–20.362.13E-08
BG643777–3.821.78E-04Similar to AP2 domain-containing protein RAP2.10TC156763–5.878.07E-07
AI489178–17.028.61E-20Similar to DNA-binding protein RAV1 (RAV1)TC156066–30.758.57E-26
CN3845491.751.27E-04Homolog to ER33 proteinTC1548043.569.62E-39
U77719.1–3.554.29E-07Ethylene-responsive late embryogenesis-like protein (ER5)TC154458–3.991.60E-10
BM412148–3.771.56E-03Similar to universal stress protein (USP)TC162458–52.643.78E-38
J04099.1–51.244.27E-20Ethylene-responsive proteinase inhibitor I (ER1)TC164904–19.891.36E-11
AF096251.11.130.8Ethylene-responsive heat shock protein cognate 70 (ER21)TC164785–15.91.40E-19
ABA-related genes
Biosynthesis and catabolism    –3.714.77E-06
AF254793.1–1.630.08Neoxanthin synthaseTC157111–3.714.77E-06
Z97215.1–4.381.66E-089-cis-epoxycarotenoid dioxygenase LeNCED1TC154656–3.576.64E-08
BM412445–1.270.62Member of cytochrome P450 CYP707A1 familyTC161981–5.15.80E-03
AY530758.1–2.490.02ABA-responsive element binding protein 1TC165124–5.978.68E-06
CN385178–5.780Dormancy-associated protein DRM1TC162095–4.44.59E-23
BM408601–1.880.02Similar to dormancy-associated protein DRM3TC162430–6.721.56E-16
BT014196.1–3.219.16E-04Similar to ABA-responsive proteinTC161842–2.912.90E-04
AW217297–3.053.73E-03Some similarity to ABA-responsive protein-likeTC156619–6.241.06E-08
AI772677–8.273.13E-07Responsive to desiccation protein (RD2)TC162725–3.762.17E-05
BG630475–1.934.76E-06Homolog to 25-kDa protein dehydrinTC161916–4.480
CN385851–5.513.49E-04Identical to dehydrin ERD10TC161919–19.54.16E-09
AW650088–4.272.27E-15Dehydration-induced protein ERD15TC164751–6.486.35E-18
BI203945–1.050.89Similar to salt-tolerance protein 5TC155237–6.081.04E-04
L13654.16.94.22E-03Peroxidase precursor TPX1TC15422444.510
Cytokinin-related genes    –5.477.38E-12
AW037686–2.383.23E-04Similar to zeatin O-xylosyltransferase (zeatin O-β-D-xylosyltransferase)TC164518–5.477.38E-12
BG629373–11.813.20E-05Similar to zeatin O-glucosyltransferase (trans-zeatin O-β-D-glucosyltransferase)TC155568–6.633.89E-04

Of the modulated genes listed, only a few could be assigned directly to the GA3 treatment (Table 1). GA3 application decreased the expression of a GA biosynthesis gene, a putative GA 3β-hydroxylase. Pollination did not modulate this gene very strongly but it induced another GA 3β-hydroxylase, Le3OH-1, and a GA 20-oxidase, Le20ox-1. Le3OH-2, Le20ox-2 and Le20ox-3 are also present on the GeneChip, but these genes were not significantly changed in expression. Thus, GA biosynthesis seems to decrease after GA3 application, whereas it is induced by pollination. Genes encoding the GA signal components SPINDLY (LeSPY) and GA INSENSITIVE (LeGAI) (BG631207 and AY269087) were not significantly modulated during fruit set. The only gene putatively involved in GA signal transduction is a gene with sequence homology to the Arabidopsis SLEEPY1 (SLY1) and rice (Oryza sativa) GA-INSENSITIVE DWARF2 (GID2) genes. Although the nucleotide identity was only 49% over 40% of the sequence (e-value 7.5e-20), this gene seems to be the sole tomato candidate ortholog of SLY1 present in the databases. The tomato SLY1-like transcript level was much lower in the ovary after pollination, and more detailed expression analysis with real-time quantitative PCR (Fig. 2a) also demonstrated decreased transcript levels after GA3 treatment in the ovary wall and to a lesser degree in the placenta with ovules. Both pollination and GA3 treatment induced the few known GA response genes that were identified in this screen, but they did not do so in a comparable way (Table 1). This is also evident from the expression patterns of the genes known to be related to auxin, such as the auxin influx and efflux genes which were up-regulated by pollination but not by GA3 application (Table 1). Moreover, most of the identified AUX/IAA and AUXIN RESPONSE FACTOR (ARF) genes were specifically modulated after pollination. Two AUX/IAA family members, putative orthologs of Arabidopsis IAA2 and IAA14, were strongly induced by pollination alone (Table 1). Further characterization of the mRNA levels of these genes with real-time quantitative PCR demonstrated that they were limited to the placental and ovular tissues in the ovary (Fig. 2a). In addition, expression patterns obtained with cDNA-AFLP (Fig. 2b) showed that one ARF gene (cDNA fragment 69; BT013639) was up-regulated by pollination alone and that another newly identified ARF-like gene (cDNA fragment 79) was specifically silenced in the placenta with ovules. At the same time, a gene for a protein with similarity to an Arabidopsis auxin-responsive family protein (homologous to AIR12; cDNA-AFLP fragment 41) was induced in the same tissue.

Figure 2.

Expression patterns of genes identified through microarray and cDNA–amplified fragment length polymorphism (AFLP) analysis of the placenta with ovules and the ovary wall, 1, 2 and 3 d after treatments. Control, RNA from emasculated flowers; GA3, RNA from emasculated flowers after gibberellic acid treatment; Pollinat., RNA from emasculated flowers after pollination. (a) Relative transcript levels of a tomato (Solanum lycopersicum) SLEEPY1 (SLY1)-like gene and two auxin/indole-3-acetic acid (AUX/IAA) family members, IAA2 and IAA14, determined with real-time quantitative PCR. (b) cDNA-AFLP expression patterns of auxin-related genes. The transcript corresponding to fragment 69 is also present on the microarray as BT013639 (Table 1).

The expression of one auxin- and ethylene-responsive GH3-like gene (Guilfoyle et al., 1998; AW224185) decreased strongly after the treatments (Table 1), which might be a result of a decrease in ethylene signaling despite the activated auxin signal. cDNA-AFLP results (Fig. 3) and microarray data (Table 2) showed a decrease in the mRNAs of several ethylene biosynthesis genes. Three 1-aminocyclopropane-1-carboxylate (ACC) synthase genes and eight ACC oxidase (ACO) genes appeared to be down-regulated (Table 2) after fruit induction. This was confirmed by cDNA-AFLP fragments 232 and 241, corresponding to two of these ACO genes (CN384809 and BF112635, respectively; Fig. 3a). Probably as a consequence of the low ethylene biosynthesis rate, the activities of several ethylene-responsive element binding factor (ERF) genes also decreased (Table 2, Fig. 3a). Among these were Pto-interacting protein 4 (Pti4), Pti5, and Pti6, also members of the ERF gene family, the products of which regulate defense-related gene expression (Zhou et al., 1997; Chakravarthy et al., 2003). cDNA-AFLP fragment 90 (Fig. 3a) corresponds to Pti4, and data for this fragement showed that the activity of Pti4 rapidly decreased after fruit set induced either by GA3 treatment or by pollination. In addition to the Pti genes, many other defense-related genes decreased their activity after fruit set, such as genes encoding pathogen-related (PR) proteins, endochitinases, wound-induced proteins and leucine-rich repeat (LRR) proteins (Fig. 3a, Supplementary Material Table S4). In addition to the biotic stress genes, some abiotic stress genes also appeared to be highly expressed in the ovary before fruit set, and some of these were ABA-induced (Table 2). Data on several genes characterized in the microarray and cDNA-AFLP experiment suggest a decrease of the ABA signal with the initiation of fruit set. Table 2 shows that the ABA biosynthetic genes encoding neoxantin synthase (NSY; AF254793) and 9-cis-epoxycarotenoid dioxygenase (LeNCED1; Z97215) are expressed at lower levels after fruit set. In addition, a tomato gene similar to Arabidopsis CYP707A4 encoding an ABA 8′-hydroxylase was strongly induced after pollination specifically in the sample containing the placenta with ovules (cDNA-AFLP fragment 143; Fig. 3b). Decreased biosynthesis and increased degradation of ABA possibly lead to lower hormone concentrations, with a consequently reduced expression of ABA-responsive genes such as dehydration-induced (ERD10; CN385851) and dormancy-associated (DRM1 and DRM3; Table 2, Fig. 3b) genes.

Figure 3.

Expression patterns of genes identified through microarray and cDNA–amplified fragment length polymorphism (AFLP) analysis of tomato (Solanum lycopersicum) placenta with ovules and ovary wall, 1, 2 and 3 d after treatments. Control, RNA from emasculated flowers; GA3, RNA from emasculated flowers after gibberellic acid treatment; Pollinat., RNA from emasculated flowers after pollination. (a) cDNA-AFLP expression patterns of two ethylene biosynthesis genes (cDNA-AFLP fragments 232 and 241) and four response genes (cDNA-AFLP fragments 127, 90, 120 and 188). Transcripts corresponding to fragments 232, 241 and 90 are also present on the microarray as CN384809, BF112635 and U89255, respectively (Table 2). (b) cDNA-AFLP expression patterns of an ABA catabolic gene (cDNA-AFLP fragment 143), two signaling genes (cDNA-AFLP fragments 166 and 301), and one response gene (cDNA-AFLP fragment 199). The transcript corresponding to fragment 199 is also present on the microarray as CN385178 (Table 2). (c) Relative transcript levels of two 1-aminocyclopropane-1-carboxylate (ACC) oxidase (ACO) genes, dormancy-associated 1 (DRM1) and early response to dehydration 10 (ERD10), in ovaries during bud development compared with the levels in the RNA samples used for microarray analysis (shaded bars). Ovaries were collected from 3–4-mm buds (1), 4–5-mm buds (2), 5–6-mm buds (3), 7–8-mm buds with petals that were starting to open (4), and open flowers (5). ‘Control’, ‘GA3’ and ‘Pollinat.’ RNA was from ovaries 3 d after treatment of emasculated flowers. ERF, ethylene-responsive element binding factor; Pti, Pto-interacting protein.

To determine whether experimental treatments such as emasculation were the cause of the induced defense response, the expression of the ACO genes in the ovary during flower bud development was analyzed. Ovaries were collected from untreated flower buds ranging from buds 3 mm in size to fully grown buds at anthesis. Figure 3(c) displays the results of a real-time quantitative PCR analysis showing that ACO, ERD10 and DRM1 mRNA levels increased during normal flower development to a level comparable to that in ovaries from unpollinated and emasculated flowers, suggesting that the defense response was not induced by the treatments but was part of the normal developmental program.

To summarize, fruit set induced either by GA3 application or by pollination induced GA and auxin responses, and at the same time ethylene and ABA responses were attenuated.

Fruit growth-related genes

A change in the hormone balance is one of the first events inducing fruit development. The first visible changes in the ovary are cell enlargement and cell division. In general, tomato parthenocarpic fruits induced by GA application are smaller than seeded fruits and consist of a relatively small number of very large cells (Bünger-Kibler & Bangerth, 1982). Figure 4(a) shows that pollination induced several genes involved in the cell cycle in the placenta with ovules after 48 h and that GA3 application did not have such a strong effect. By contrast, in the pericarp the two treatments had comparable effects on the cell cycle genes, although they responded more quickly to GA3 application than to pollination. It is probable that, rather than pollination itself, fertilization, which occurs 12–15 h later, induces cell division. Table 3 lists all the cell cycle-associated genes that were identified on the microarray and whose expression levels changed significantly by more than threefold after treatment. All these genes were more strongly induced by pollination than by the GA3 treatment. Cell wall biosynthesis and extension are both processes that have to be activated to allow growth. Figure 4(b) and Table 3 show that several gene families encoding wall-modifying proteins were involved in early fruit growth. Some of these genes had a higher expression level after pollination than after GA3 application, such as the α-expansins 5, 9 and 12 and some genes encoding xyloglucan endotransglucosylase (XTH) proteins (Table 3). In contrast, GA3 treatment induced some pectate lyase genes more stronger than pollination. In general, most genes involved in wall extension were induced by GA3 as well as by pollination. Differences in expression are possibly related to differences in cell division rate. However, as the microarray showed the mRNA level at only one time-point after induction of fruit growth, it is also possible that differences in some cases were temporal and did not reflect average gene expression over a longer period.

Figure 4.

Expression patterns of genes identified through microarray and cDNA–amplified fragment length polymorphism (AFLP) analysis of tomato (Solanum lycopersicum) placenta with ovules and ovary wall, 1, 2 and 3 d after treatments. Control, RNA from emasculated flowers; GA3, RNA from emasculated flowers after gibberellic acid treatment; Pollinat., RNA from emasculated flowers after pollination. (a) cDNA-AFLP expression patterns of cell cycle-associated genes. Transcripts corresponding to fragments 169 and 239 are also present on the microarray as AJ297916 and BE462633, respectively (Table 3). (b) cDNA-AFLP expression patterns of cell wall metabolism. Transcripts corresponding to fragments 256 and 16 are also present on the microarray as BG629750 and D16456, respectively (Table 3). CDK, cyclin-dependent kinase; Cel1, cellulase; XTH, xyloglucan endotransglucosylase.

Table 3.  Genesa identified by microarray analysis involved in cell division and cell wall synthesis
GenBank accession numberGA3-treated vs controlTentative annotationbDFCI accession numberPollinated vs control
Fold changeP-value (P < 0.003)Fold changeP-value (P < 0.007)
  • a

    Only genes whose activity, as a result of one of the treatments, significantly changed more than threefold compared with the untreated sample are listed. ‘Fold change’ values in bold are not significantly different from the control values.

  • b

    Tentative annotations are based on the Dana-Farber Cancer Institute (DFCI) Tomato (Lycopersicon esculentum) Gene Index (LeGI).

  • CDK, cyclin-dependent kinase; MAD, mitotic arrest deficiency; TCP, TB1, CYC and PCFs.

Cell cycle
AJ243451.13.392.19E-12Cyclin A1 (CycA1 gene)TC1667546.690
AJ243452.15.81.50E-05Cyclin A2 (CycA2 gene)TC1695586.942.67E-06
AJ243454.12.790.17Cyclin B1 (CycB1 gene) 10.291.74E-04
AJ011108.111.998.15E-08B-type cyclinTC15841831.438.05E-21
BI9294539.21.01E-06Similar to CycB1-1TC15872349.093.22E-44
BG73482518.174.84E-29Similar to B-type cyclinTC15794239.30
BG1298871.981.54E-03Weakly similar to G2/mitotic-specific cyclin 1TC1665354.71.43E-22
BG1348707.829.75E-10Similar to cyclin 2 20.220
AJ515747.12.23.88E-03Proliferating cell nuclear antigen (PCNA gene)TC1633384.698.59E-15
BI9293611.886.69E-03Similar to mitotic checkpoint proteinTC1639173.61.83E-22
BI2039013.972.70E-09Similar to mitotic spindle checkpoint protein (MAD2)TC1560026.712.24E-28
BE46263312.331.56E-05Syntaxin of plants SYP111 (KNOLLE)TC16159753.511.61E-31
BT013238.16.131.83E-07Homologous to cell cycle switch protein CCS52aTC16483017.751.69E-24
BT012879.13.560.05TCP family transcription factor; CycloideaTC15636316.261.07E-13
BT013252.12.513.32E-10Histone H4TC1627303.660
BI2102743.714.61E-11Histone H4TC1541047.870
BG6311354.81.77E-16Histone H2BTC1552578.430
Cell wall
BG6297504.471.07E-08Similar to pectate lyaseTC1567611.650.05
U50985.13.82.67E-03Pectin methylesterase PME2.1TC1635121.320.56
BG13007410.984.54E-15Similar to pectate lyaseTC1551972.364.04E-03
BI9244873.921.30E-07Similar to pectate lyaseTC1558763.296.64E-06
BT013600.111.591.44E-15Similar to pectinesterase-like proteinTC16572218.563.29E-37
BI9242419.546.28E-09Similar to pectinesterase family proteinTC16507930.181.46E-19
CK7147743.626.47E-08Tomato (fragment) pectinesteraseTC1541115.651.53E-20
CK7159743.633.19E-05Similar to pectinacetylesterase familyTC1555326.251.60E-12
AW6490763.561.49E-16Similar to pectinesterase PPE8B precursorTC1575382.26.31E-09
AJ010943.1–2.194.88E-04Homolog to invertase/pectin methylesterase inhibitorTC163422–4.945.07E-16
AF096776.13.351.83E-05Expansin (LeEXP2)TC1633612.276.66E-04
AF059489.17.839.75E-28Expansin precursor (EXPA5)TC16295130.181.46E-19
AJ243340.12.823.37E-03Expansin 9 (EXP9)TC1566765.391.24E-08
AJ560647.11.630.02Expansin 12 (EXP12)TC1593203.821.02E-24
D16456.139.290Endo-xyloglucan transferase (LeXTH1)TC16261135.070
AY497475.1–4.633.61E-06Xyloglucan endotransglucosylase-hydrolase (XTH5)TC161980–6.26.40E-08
AY497479.13.973.52E-13Xyloglucan endotransglucosylase-hydrolase (XTH9)TC1626602.842.50E-07
CK7157564.12.76E-03Probable xyloglucan endotransglucosylase/hydrolaseTC16374812.487.81E-19
AW0929176.352.27E-04Probable xyloglucan endotransglucosylase/hydrolaseTC1640336.694.00E-06
AF176776.12.820.33Xyloglucan endotransglycosylase (LeXET2)TC1635405.031.33E-03
BT013529.1–1.730.3Similar to endo-1,4-β-glucanaseTC1545575.747.93E-11
AJ7855253.813.92E-15Endo-β-1,4-D-glucanase, putative 2.42.24E-08
BT013449.132.62E-07Similar to endo-1,4-β-glucanaseTC1545571.953.65E-04
X99148.15.954.46E-20Arabinogalactan (LeAGP1)TC1630024.843.71E-20
X55681.16.721.45E-09Extensin (class II) precursorTC16569818.072.64E-24
Z46675.1–2.131.47E-07Extensin (Lemmi11) –3.73.24E-04
L13654.16.94.22E-03Peroxidase (TPX1), cell wall biosynthesisTC15422444.510

In addition to the genes encoding proteins whose function can be placed somewhere in the processes leading to fruit set, there were many more genes that were strongly modulated that have a putative role in fruit set. These included genes involved in polyamine metabolism such as ornithine decarboxylase (cDNA-AFLP fragment 128; Fig. 5) and spermidine synthase (Supplementary Material Table S5). Polyamines have a stimulating effect on parthenocarpic fruit growth (Fos et al., 2003). Genes involved in phospholipid biosynthesis such as myo-inositol-1-phosphate synthase and phospholipase/esterase (cDNA-AFLP fragment 231; Fig. 5 and Supplementary Material Table S5, respectively) were also strongly regulated. Interestingly, almost all the genes involved in calcium signaling were expressed at a lower level after fruit growth induction than before the treatments. Finally, a large number of genes encoding proteins active in signal transduction were regulated during fruit set. These were receptor protein kinases, mitogen activated protein (MAP) kinases, phosphatases, proteins involved in 26S proteasome protein degradation and transcription factors belonging to different gene families (Supplementary Material Table S6). One interesting transcription factor identified has high homology to GROWTH-REGULATING FACTOR 2 (GRF2), a member of the transcription activator gene family, and was induced in the ovary after pollination (cDNA-AFLP fragment 266; Fig. 5).

Figure 5.

Expression patterns of genes identified through microarray and cDNA–amplified fragment length polymorphism (AFLP) analysis of tomato (Solanum lycopersicum) placenta with ovules and ovary wall, 1, 2 and 3 d after treatments. Control, RNA from emasculated flowers; GA3, RNA from emasculated flowers after gibberellic acid treatment; Pollinat., RNA from emasculated flowers after pollination. cDNA-AFLP expression patterns of genes involved in polyamine biosynthesis, calcium signaling and phospholipid synthesis and a gene encoding a growth-related transcription factor are shown. The transcript corresponding to fragment 231 is also present on the microarray as BG627650 (Table S5).


In plant development, fruit set is dependent on successful pollination and fertilization. To fertilize the egg cell, pollen has to germinate and the pollen tubes have to grow through the stylar tissue to the ovule and into the embryo sac. The zygote possibly triggers the development of the ovary into a fruit. Fruit development (fruit set) is therefore dependent on stimulatory growth signals generated by pollination or by fertilization. Application of GAs to unpollinated tomato flowers causes an increased auxin concentration in the ovary (Sastry & Muir, 1963). Pollen need and produce GAs for pollen tube growth (Singh et al., 2002), and parthenocarpy (the formation of seedless fruits) can be induced in tomato by applying pollen extracts to the sides of the ovary (Gustafson, 1937). Gibberellins produced by the pollen may thus play a role in increasing auxin production in the ovary, which in turn may act as a signal for fruit set and activate cell division (Gillaspy et al., 1993). Our data demonstrate, however, that GA3 induced genes involved in the cell cycle within 24 h after application (cDNA-AFLP fragments 43, 64 and 169, encoding a cyclin-dependent kinase (CDK) regulatory subunit, a cyclin and a B-type CDK, respectively; Fig. 4), which was c. 1 d faster than the induction of these genes by pollination. The reason for this is probably that it takes c. 1214 h before the pollen tips reach the ovary after pollination. The sample obtained 1 d after pollination actually corresponds to < 12 h after fertilization. This also suggests that induction of fruit growth does not take place before the pollen reaches the ovary, and possibly not before fertilization has taken place. Several parthenocarpic lines have higher endogenous concentrations of auxins and GAs in the ovaries than normal (seed-producing) lines (Gustafson, 1939; Nitsch et al., 1960; Mapelli et al., 1978; Mapelli & Lombardi, 1982; Fos et al., 2000). Therefore, it appears that the sequential or cooperative action of GAs and auxin is part of a signal transduction chain that leads to fruit set and subsequent activation of cell division. The mechanism, in which GAs and auxin signals interact to regulate fruit set, remains to be established.

Many genes involved in hormone biosynthesis and signal transduction were identified by comparative transcriptome analysis of tomato ovaries. Comparison of genes induced by GA application or pollination can shed more light on the relationship between GA and auxin during fruit set. In addition, ABA and ethylene, which are hormones that normally inhibit growth, appeared to regulate a significant proportion of the genes that were found to be strongly modulated after fruit growth induction.

Gibberellin and auxin

A single treatment with GA3 induced fruit growth but not seed development, and seedless fruit development followed a comparable time schedule to that of the seeded fruits that developed after pollination. The responses to the two treatments, however, were not the same. Pollination induced GA biosynthesis genes, particularly GA 20-oxidase (Le20ox-1) and GA 3β-hydroxylase (Le3OH-1), suggesting an increase in GA biosynthesis rate after pollination, as has been described previously (Rebers et al., 1999). In general, 3β-hydroxylation of GA20 was suggested to be the rate-limiting step in GA1 biosynthesis (Bohner et al., 1988; Koshioka et al., 1994). GA3 application had the opposite effect on GA biosynthesis genes; they were not significantly induced. The observation that pollination led to the induction of GA biosynthesis, whereas GA3 application did not, can be explained by the negative feedback of GA in many plants on its own biosynthesis (Hedden & Kamiya, 1997).

A tomato gene with some similarity to the Arabidopsis SLY1 gene (SLEEPY1) was down-regulated. SLY1 is an F-box protein that interacts directly with REPRESSOR OF ga1–3 (RGA) and GAI (Dill et al., 2004), which are highly conserved negative regulators of GA signaling in Arabidopsis (Peng et al., 1997; Silverstone et al., 1998). It is possible that tomato SLY-like gene expression, and thus GA signaling, are subject to GA feedback. GA3 induced some genes not induced by pollination, suggesting that, although both treatments lead to fruit development, they do not induce the genetic program in precisely the same way. Pollination exerted a profound effect on GA biosynthesis, possibly via auxin signaling (Table 1). Conversely, the expression of auxin signaling genes seemed not to be greatly influenced by GA treatment (Fig. 2, Table 1). Two ARFs, transcription factors that bind to the auxin-response elements found in the promoters of many auxin-regulated genes (Kim et al., 1997; Guilfoyle et al., 1998), were modulated during fruit set after pollination but not after GA3 treatment. ARFs can interact with a class of transcription regulators, Aux/IAA, through conserved domains, forming homo- or heterodimers, in various combinations (Kim et al., 1997; Ulmasov et al., 1997). IAA9 (an AUX/IAA gene) was shown previously to be involved in the development of tomato fruits, and in IAA9-inhibited lines, fruit development is triggered before fertilization, giving rise to parthenocarpy (Wang et al., 2005). AUX/IAA genes such as the tomato IAA2 gene, which is known to be auxin induced (Nebenführ et al., 2000; Wang et al., 2005), and the tomato IAA5 and IAA14 genes, which we found to be induced during fruit set, might also play fundamental roles in tomato fruit set. IAA2 and IAA14 genes appeared only to be induced after pollination and specifically in the placenta with ovule samples (Fig. 2a). The expression patterns of these genes indicate that pollination induces auxin signaling. Auxin signaling as a consequence of the pollination event might induce GA biosynthesis but GA, in turn, does not modulate the auxin signal.


Interestingly, ethylene biosynthesis and signaling in the ovary were strongly attenuated 3 d after GA3 treatment or pollination compared with the unpollinated control. Llop-Tous et al. (2000) reported that, in the absence of pollination, tomato flowers started to senescence approximately 72 h after anthesis. This might explain why all the ethylene biosynthesis and signaling genes that were identified on the tomato GeneChip were highly expressed in the unpollinated control sample consisting of ovaries 3 d after anthesis. cDNA-AFLP expression data (Fig. 3a) showed, however, that two ACO genes, the tomato ortholog of AtERF2 and Pti4, were already active 24 h after anthesis in the unpollinated control ovaries. Therefore, we assume that the high ethylene signaling level was probably not attributable to senescence. In addition, we did not find senescence-associated genes with increased expression levels in the transcriptome analyses. Moreover, steady-state ACO mRNA levels in the ovary increased well before flower anthesis (Fig. 3c). It is known that pollination temporally induces ethylene biosynthesis in tomato pistils by inducing the ACC synthase genes LEACS1 and LEACS3 and that ethylene biosynthesis starts to decline again within 12 h (Llop-Tous et al., 2000). This explains why we found a decrease in ethylene signal 24–72 h after pollination compared with the signal in unpollinated ovaries (Table 2).

The ERF family is part of the AP2/ERF superfamily of plant-specific transcription factors, which also contains the APETALA 2 (AP2) and related to ABI3/VP1 (RAV) families (Ohme-Takagi & Shinshi, 1995; Riechmann et al., 2000). The ERF family consists of two major subfamilies, the ERF and the C-REPEAT-BINDING FACTOR/dehydration-responsive element binding (CBF/DREB) subfamilies (Sakuma et al., 2002). The Pti4, Pti5 and Pti6 genes (ERF subfamily) are active in biotic defense (Zhou et al., 1997; Gu et al., 2002) and are strongly down-regulated after GA3 treatment and pollination. These genes are possibly responsible, together with other ethylene-induced transcription factors, for the high expression levels of several classes of genes encoding defense-related proteins such as pathogenesis-related proteins, glycosyl hydrolases, chitinases and others listed in Table 2 (and Supplementary Material Table S4). The decrease in their expression after induction of fruit growth suggests either that the ovary tissue of unpollinated pistils is under stress or that defense mechanisms are in a default active state in the reproductive organs of unpollinated flowers. This might help to protect the nongrowing flower tissue against pathogens while it is waiting for pollination to occur.

We also found several transcription factor genes belonging to the CBF/DREB subfamily. CBF/DREB1 genes are quickly and transiently induced by cold stress, and their products activate the expression of target stress-inducible genes and genes that are involved in acquired stress tolerance. (Zhu et al., 1995; Knight et al., 2004).


Many stress-inducible genes are regulated by the endogenous ABA that accumulates during drought and high-salinity stress (Shinozaki & Yamaguchi-Shinozaki, 2000). Genes involved in ABA biosynthesis such as neoxantin synthase (NSY; Bouvier et al., 2000) and 9-cis-epoxycarotenoid dioxygenase (LeNCED1; Thompson et al., 2004) are expressed at relatively high levels in unpollinated tomato ovaries, and expression decreases after fruit induction. Mutation in LeNCED1 causes a decrease in ABA concentrations and a wilty phenotype in tomato (Stubbe, 1958; Burbidge et al., 1999). In addition, a cytochrome P450 gene (cDNA-AFLP fragment 143), homologous to Arabidopsis ABA 8′-hydroxylase CYP707A (Kushiro et al., 2004; Okamoto et al., 2006), is strongly induced by pollination in the placenta with ovules, suggesting an enhanced breakdown of ABA in that tissue. ABA signaling is under the control of negative regulators encoded by the protein serine/threonine phosphatases 2C (PP2C) gene family in Arabidopsis (Leung et al., 1997; Merlot et al., 2001). cDNA-AFLP results showed that two tomato genes (corresponding to cDNA-AFLP fragments 166 and 301) with homology to members of this PP2C gene family, ABSCISIC ACID INSENSITIVE 1 (ABI1) and HOMOLOGY TO ABI1/ABI2 (HAB1), have much lower mRNA levels after fruit set. ABI1/ABI2 expression in Arabidopsis is induced by ABA (Leung et al., 1997). Moreover, in tomato several ABA response genes, such as a gene encoding a dehydrin (ERD10-like; Fig. 3c, Table 2), are expressed at a lower level after fruit set. Arabidopsis ERD10 is induced by ABA and is not influenced by GA or auxin (Kiyosue et al., 1994), suggesting that the decreased activity of the tomato ABI1/ABI2 and ERD10 orthologs is a consequence of lower ABA concentrations in the ovary after induction of fruit growth.

ABA and dormancy

ABA concentrations in ovaries from unpollinated garden pea (Pisum sativum) are relatively high and are thought to prevent parthenocarpic fruit growth (Rodrigo & Garcia-Martinez, 1998). From this point of view, ABA might keep the ovary tissue in a state of temporal dormancy comparable to the dormant state of axillary buds. In pea, dormant buds are as metabolically active as growing buds, and genes that might have a function in the maintenance of dormancy, called dormancy-associated (DRM) genes, have been isolated (Stafstrom et al., 1998). Homologous tomato DRM genes are expressed at high levels in ovaries, and mRNA levels decrease after fruit set, suggesting that unpollinated ovaries are in a temporally dormant state (Fig. 3b,c, Table 2). The proliferating phase of pea axillary buds is also characterized by the accumulation of several specific cell cycle gene transcripts (Devitt & Stafstrom, 1995; Shimizu & Mori, 1998). These genes, histone H4, cycB1;2 and cycD3;1 (cyclins), and the proliferating cell nuclear antigen (PCNA) gene, are indeed also induced in tomato ovaries after fruit set (Fig. 4a, Table 3).

Fruit growth

Several cell cycle genes were only induced by pollination, such as a homolog of Arabidopsis KNOLLE (cDNA-AFLP fragment 239; BE462633) and CYCLIN D3;3 (AJ002590; Table 3), which has highest homology to Arabidopsis CYCLIN D3;1 (CYCD3;1). In Arabidopsis, CYCD3;1 dominantly stimulates the G1/S transition, and controls cell cycle progression in response to cytokinin and sucrose availability (Riou-Khamlichi et al., 1999; Menges et al., 2006). Bünger-Kibler & Bangerth (1982) showed that cell number in tomato fruits initiated by pollination or auxin is much higher than in fruits initiated by GA. Cytokinins are probably responsible for stimulation of cell division in developing fruits (Mapelli, 1981; Bohner & Bangerth, 1988; Martineau et al., 1995). It is possible that fertilization and concomitant auxin signaling are responsible for increased cytokinin concentrations, as both pollination and auxin application simulate cell division (Bünger-Kibler & Bangerth, 1982). GA3 application does not strongly induce cell division. However, GA-induced fruit growth is mainly caused by cell enlargement (Bünger-Kibler & Bangerth, 1982), suggesting that neither auxin nor cytokinin signaling is stimulated by GA. Moreover, Baldet et al. (2006) showed that cell proliferation and final fruit size can be correlated to the activity of CYCD3;1, CDKB2;1 and CYCB2;1 expression during the early stages of fruit development. Our data show that the mRNA levels of the last two genes are higher in the ovary after pollination than after GA3 treatment, which correlates well with the smaller fruit size of GA-initiated parthenocarpic fruits.


Transcriptome analysis of tomato ovaries led to the identification of many genes with significantly different transcript levels before and after fruit induction. Based on the expression of these genes, we propose a model of the hormone interactions during fruit set (Fig. 6). In unpollinated ovaries ABA probably restrains cell division and together with ethylene it induces genes involved in protection against biotic and abiotic stress factors. Pollination leads to a decrease in ABA and ethylene signals as a result of an increase in auxin and GA signals. Genes involved in auxin signaling were specifically induced by pollination but not by GA3 application, suggesting that pollination leads to an increased auxin response and subsequently auxin induces GA biosynthesis (Fig. 6). In future work, to demonstrate the validity of the proposed model, we will modulate the expression of the identified auxin signaling genes and analyze the effect on GA concentrations and fruit set. In addition, we are currently analyzing the effects of changes in ABA and ethylene concentrations on flower receptivity and fruit set.

Figure 6.

Model for hormonal signaling in the tomato (Solanum lycopersicum) ovary after pollination/fertilization. The central shaded square shows auxin and gibberellin acting together to inhibit abscisic acid (ABA) and ethylene biosynthesis and signaling in order to stimulate fruit set, either directly or in cooperation with other factors such as polyamines and cytokinins. Dashed arrows show suggested interactions; auxin might regulate gibberellin biosynthesis and signaling, and ABA and ethylene might act together in a mechanism to prevent fruit set without fertilization. ACO, 1-aminocyclopropane-1-carboxylate (ACC) oxidase; ACS, ACC synthase; ARF, auxin response factor; CDK, cyclin-dependent kinase; CYP707A, ABA 8′-hydroxylase; DRM, dormancy-associated; EIL, EIN3-like; ERD, early response to dehydration; ERF, ethylene response factor; ETR, ethylene resistant; IAA, indole-3-acetic acid; IAR, IAA-alanine resistant; NCED, 9-cis-epoxycarotenoid dioxygenase; NPH, NON-PHOTOTROPHIC HYPOCOTYL; NS, neoxantin synthase; PP2C, protein serine/threonine phosphatases 2C; XTH, xyloglucan endotransglucosylase.


This work was supported by a European Union Marie Curie Reintegration Grant (MERG-CT-2004-506367).