Fertilization-dependent auxin response in ovules triggers fruit development through the modulation of gibberellin metabolism in Arabidopsis


  • Eavan Dorcey,

    Search for more papers by this author
    • Present address: Department of Plant Molecular Biology, University of Lausanne, Biophore Building, CH–1015 Lausanne, Switzerland.

  • Cristina Urbez,

    1. Instituto de Biología Molecular y Celular de Plantas (IBMCP), Consejo Superior de Investigaciones Científicas, Universidad Politécnica de Valencia (CSIC-UPV), Avenida de los naranjos s/n, 46022 Valencia, Spain
    Search for more papers by this author
  • Miguel A. Blázquez,

    1. Instituto de Biología Molecular y Celular de Plantas (IBMCP), Consejo Superior de Investigaciones Científicas, Universidad Politécnica de Valencia (CSIC-UPV), Avenida de los naranjos s/n, 46022 Valencia, Spain
    Search for more papers by this author
  • Juan Carbonell,

    1. Instituto de Biología Molecular y Celular de Plantas (IBMCP), Consejo Superior de Investigaciones Científicas, Universidad Politécnica de Valencia (CSIC-UPV), Avenida de los naranjos s/n, 46022 Valencia, Spain
    Search for more papers by this author
  • Miguel A. Perez-Amador

    Corresponding authorSearch for more papers by this author

For correspondence (fax +34 963877859; e-mail mpereza@ibmcp.upv.es).


Fruit development is usually triggered by ovule fertilization, and it requires coordination between seed development and the growth and differentiation of the ovary to host the seeds. Hormones are known to synchronize these two processes, but the role of each hormone, and the mechanism by which they interact, are still unknown. Here we show that auxin and gibberellins (GAs) act in a hierarchical scheme. The synthetic reporter construct DR5:GFP showed that fertilization triggered an increase in auxin response in the ovules, which could be mimicked by blocking polar auxin transport. As the application of GAs did not affect auxin response, the most likely sequence of events after fertilization involves auxin-mediated activation of GA synthesis. We have confirmed this, and have shown that GA biosynthesis upon fertilization is localized specifically in the fertilized ovules. Furthermore, auxin treatment caused changes in the expression of GA biosynthetic genes similar to those triggered by fertilization, and also restricted to the ovules. Finally, GA signaling was activated in ovules and valves, as shown by the rapid downregulation of the fusion protein RGA-GFP after pollination and auxin treatment. Taken together, this evidence suggests a model in which fertilization would trigger an auxin-mediated promotion of GA synthesis specifically in the ovule. The GAs synthesized in the ovules would be then transported to the valves to promote GA signaling and thus coordinate growth of the silique.


Successful plant reproduction depends entirely on fruit-set, an essential process that can be defined as the activation of a developmental program which will convert the pistil into a developing fruit. This transition comprises two different and coordinated processes, namely the fertilization of the ovule and the growth of the surrounding pistil and/or other structures to allocate the developing seeds (Gillaspy et al., 1993). In most species, coordination between these two events is achieved because the signal that promotes fruit growth originates only in developing seeds. However, the existence of fruits without seeds (parthenocarpic fruits) indicates that the fertilization of the ovules is not an absolute requirement, i.e. fruit development can be uncoupled from fertilization and seed development. Natural and induced parthenocarpy are basic tools for the analysis of fruit development, as they facilitate the unraveling of the molecular and genetic bases underlying fruit-set and early fruit development.

Hormones seem to have a prominent role in synchronizing fertilization and fruit growth. As early as the 1930s it was proposed that fruit development would be initiated as a consequence of hormones synthesized in developing seeds (Gustafson, 1936, 1939; Nitsch, 1950, 1952). Later work showed that early abortion of the fertilized ovules prevented fruit development in pea, but fruit growth was restored by application of several plant growth regulators (Eeuwens and Schwabe, 1975). The involvement of hormones in fruit-set is thus supported by the major observation that exogenous applications are sufficient to trigger fruit development, but also by the fact that parthenocarpic mutants are generally affected in hormone biosynthesis and/or signaling. Gibberellins (GAs), auxins and, in some cases, cytokinins have been shown to be particularly efficient in several species to trigger fruit growth. Treatment of unpollinated pea ovaries with auxin [2,4-dichlorophenoxyacetic acid (2,4-D) and 4-chloroindole-3-acetic acid (4-Cl-IAA)], GA (GA1 or GA3), or cytokinin [6-benzylaminopurine (BAP)] results in the development of parthenocarpic fruits (Garcia-Martinez and Carbonell, 1980; Ozga and Reinecke, 1999), although only GA3 treatment results in fruits that are nearly identical to fruits generated by pollination (Carbonell and Garcia-Martinez, 1985). Remarkably, GAs and auxins have a synergistic effect on fruit growth when applied simultaneously (Ozga and Reinecke, 1999). Unlike in pea, auxin is the hormone with a larger capacity to induce fruit-set in tomato, followed by GAs (Homan, 1964; Alabadi et al., 1996). Pistils of Arabidopsis also respond to exogenous GAs, auxins and cytokinins, although the treatments only induce partial fruit development (Vivian-Smith and Koltunow, 1999).

Endogenous bioactive GAs have recently been shown to play a fundamental role in fruit development in Arabidopsis: the GA biosynthetic enzymes GA 20-oxidases and GA 3-oxidases are required for silique elongation (Hu et al., 2008; Rieu et al., 2008b), and blocking GA inactivation, by knocking out the five inactivating enzymes GA 2-oxidases, leads to the formation of parthenocarpic fruits in the absence of fertilization (Rieu et al., 2008a). Moreover, tomato mutants pat, pat2 and pat3/pat4 show enhanced expression of GA biosynthetic genes and increased levels of GAs, which in turn induce parthenocarpic development (Fos et al., 2000, 2001; Olimpieri et al., 2007). In addition, GA concentration is much higher in the parthenocarpic citrus variety satsuma than in the non-parthenocarpic self-incompatible clementine, which again suggests that endogenous GA content is a limiting factor for parthenocarpic development (Talon et al., 1992).

A role for auxin in promoting fruit-set and development has been revealed by recent work in Arabidopsis and tomato, species in which key elements in auxin signaling with a function in fruit initiation and development have been identified (Swain and Koltunow, 2006; Pandolfini et al., 2007). In tomato, loss-of-function of the IAA9 gene, encoding a nuclear-localized Aux/IAA protein, or of the auxin response factor ARF7 result in plants with parthenocarpic fruits, suggesting that IAA9 and ARF7 inhibit growth in the absence of fertilization (Wang et al., 2005; de Jong et al., 2008). Furthermore, in Arabidopsis, a loss-of-function allele of ARF8 (fwf or arf8-4) also provokes parthenocarpic fruit development (Vivian-Smith et al., 2001;Goetz et al., 2006, 2007). Proteins of the Aux/IAA and ARF families interact to mediate auxin signaling (Dharmasiri and Estelle, 2004), which leads to the hypothesis that both Arabidopsis and tomato possess ARF8- and IAA9-related proteins that physically interact to regulate fruit-set. Thus an ARF/IAA complex would be directly involved in fruit initiation (Swain and Koltunow, 2006; Pandolfini et al., 2007). Further evidence comes from the parthenocarpy displayed by transgenic plants with increased biosynthesis of the auxin indole-3-acetic acid (IAA) (Rotino et al., 1997; Yin et al., 2006; Costantini et al., 2007). Moreover, the tryptophan aminotransferase TAA1, a key enzyme for auxin biosynthesis, is expressed in the apical parts of embryos, coinciding in location and time with the predicted sites of auxin production upon fertilization (Stepanova et al., 2008).

The observation that both auxins and GAs are involved in early events that lead from fertilization to fruit development raises the question of whether these hormones act in parallel, regulating different aspects of the process, or in a sequential manner, and what is the hierarchy and mechanism of their interaction. Previous work suggests that GAs may act downstream of auxins. For instance, auxins have been shown to regulate the expression of several GA biosynthesis genes (Ross et al., 2000; Wolbang and Ross, 2001; O’Neill and Ross, 2002; Frigerio et al., 2006), and auxin is needed for GA signaling during root growth and apical hook formation in the hypocotyl (Achard et al., 2003; Fu and Harberd, 2003). Furthermore, GAs are required for auxin-induced fruit set in tomato (Serrani et al., 2008). Nevertheless, the application of the two different hormones does not seem to be equally efficient in many systems, pointing to specific roles for each hormone. Therefore, to uncover the molecular mechanism by which GAs and auxins regulate fruit-set in response to fertilization, and to determine the extent of crosstalk between these two hormones, we decided to study the temporal and spatial regulation of the signaling events that occur during fruit-set in Arabidopsis. We show that fertilization triggers an increase in auxin response in ovules, which can be mimicked by blocking polar auxin transport. This induces subsequent activation of GA metabolism specifically in the young seeds. Furthermore, GA signaling occurs in the valve, suggesting that GAs synthesized in the seeds must be translocated to the pod to promote cell expansion and other differentiation processes. Finally, we also describe the parthenocarpic phenotype of a quadruple-DELLA loss-of-function mutant, which genetically confirms the role of GAs in fruit-set.


Hormonal regulation of fruit-set in Arabidopsis

To dissect the specific role of the different regulators of fruit-set in autopollinated and self-compatible species it is necessary to avoid fertilization. This is usually achieved by emasculating the flowers, but to carry out large-scale analyses a more time-efficient procedure is needed. Therefore, we chose the conditional male-sterile cer6-2 mutant of Arabidopsis for our studies (Preuss et al., 1993; Fiebig et al., 2000). At low relative humidity, mutant pollen grains do not hydrate and fertilization does not take place. In our low-humidity growth conditions, no seed set was observed in the cer6-2 mutant (data not shown). Mutant plants do not show any other apparent phenotype, remaining morphologically identical to the Ler parental plant.

More importantly, the response of cer6-2 pistils to hormonal treatments was equivalent to that reported for emasculated flowers in the Ler and Col-0 backgrounds (Vivian-Smith and Koltunow, 1999). Exogenous application of GA3 or naphthylacetic acid (NAA) to cer6-2 flowers under low humidity at anthesis promoted parthenocarpic fruit development, measured as the increase in fruit length 7 days after the treatment (Figure 1). Dose–response experiments showed these concentrations to be optimal for inducing fruit-set (data not shown). In addition, treatment with the synthetic auxin 2,4-D also promoted fruit elongation, although a slightly shorter fruit was obtained when compared with NAA or GA3 treatments (Figure 1). As previously reported (Vivian-Smith and Koltunow, 1999), the final fruit length was approximately 30–60% of that of a pollinated pistil, although longer fruits were obtained when NAA and 2,4-D were applied 1 or 2 days after anthesis (data not shown).

Figure 1.

 Fruit-set is induced by blocking polar auxin transport with NPA.
cer6-2 flowers at anthesis were treated with mock, gibberellin (GA3), 2,4-dichlorophenoxyacetic acid (2,4-D), N-1-naphthylphthalamic acid (NPA), naphthylacetic acid (NAA), or were hand-pollinated with Landsberg erecta (Ler) pollen (Pol.). Pistils and fruits were harvested 7 days after the treatment.
(a) Images of representative pistils and fruits.
(b) Quantification of pistil and fruit length. Mean and SD were calculated from at least 25 pistils/fruits per treatment. The length of pistils and fruits was normalized to the length of mock-treated pistils. The experiment was repeated three times with similar results.

N-1-naphthylphthalamic acid (NPA), an inhibitor of polar auxin transport (Keitt and Baker, 1966; Geldner et al., 2001), provokes alteration of auxin levels in plant tissues (Ljung et al., 2001; Desgagne-Penix et al., 2005), and has been used to test the interaction of auxin with other hormones (Fu and Harberd, 2003). To test whether disruption of polar auxin transport had any effect on fruit-set, we applied NPA to cer6-2 flowers at anthesis. A significant increase in fruit size was observed 7 days after the application, reaching a final length similar to that obtained by GA3 or NAA treatments (Figure 1). This indicates that a block in polar auxin transport is able to induce the development of parthenocarpic fruits, mimicking the effect of auxin application. Interestingly, neither the single treatments nor combined treatments with auxin (2,4-D or NAA) and GA yielded parthenocarpic siliques of a size comparable with that of fertilized fruits (data not shown). This suggests that factors other than hormones might be necessary for full development of the fruit. Alternatively, continuous hormone synthesis in the developing seeds throughout fruit development, as opposed to a punctual treatment, might be required for fruits to attain their final size.

These results confirm that GAs and auxins promote fruit development in Arabidopsis, but do not clarify whether the two hormones regulate different aspects of fruit-set or act upon the same processes. To start dissecting this question, we examined the effect of the different treatments on the tissue structure of the pistil. As shown in Figure 2, transverse cryosections of the mock-treated pistils displayed the usual layered structure (Ferrandiz et al., 1999; Roeder and Yanofsky, 2006), composed of an external epidermis or exocarp, followed by three layers of chlorenchyma cells or mesocarp, and two layers of endocarp [the inner epidermis or endocarp a (ena), and the endocarp b (enb)] which becomes the lignified valve layer in mature fruits. This overall structure was maintained in the fruits obtained by hormonal application, with the exception of the ena cell layer which was absent in the GA-induced fruits. The collapse of ena is known to occur in pollination-induced fruit development, towards the end of developmental stage 17 (Roeder and Yanofsky, 2006); therefore GA3 treatment seems to accelerate the destruction of the ena layer. We have ruled out the possibility that the premature collapse of the ena is due to a specific effect of GA3 on the cer6-2 mutation, given that the same collapse was observed in emasculated flowers of wild-type Ler plants treated with GA3 (Figure 2). These results suggest that at least one aspect of fruit development is under specific regulation by GAs, and not auxins.

Figure 2.

 Hormone-induced fruits show similar valve structure to those induced by pollination.
Transverse cryosections of pistils and fruits from cer6-2 (first and second rows) or Landsberg erecta (Ler) (third row) at 7 days post-anthesis. Pistils were treated with mock, mock, gibberellin (GA3), 2,4-dichlorophenoxyacetic acid (2,4-D), N-1-naphthylphthalamic acid (NPA), naphthylacetic acid (NAA) or were hand-pollinated with Ler pollen (Pol.). Flowers from Ler plants were emasculated 1 day before anthesis and treated on the day of anthesis. ex, exocarp (yellow); me, mesocarp (green); enb, endocarp b (red); ena, endocarp a (blue). Absence of ena in GA3-treated pistils is indicated by an asterisk. Bars are 50 μm.

Constitutive GA signaling in a multiple DELLA loss-of-function mutant triggers fruit-set

The five members of the DELLA gene family in Arabidopsis are repressors of GA signaling (Fleet and Sun, 2005). In the presence of GAs, DELLA proteins are rapidly degraded, which alleviates the repression they impose. To confirm that GA-regulated DELLA proteins are involved in the control of fruit-set, we examined the pistil phenotype of a multiple DELLA knockout mutant. As shown in Figure 3(a), while the pistil length at anthesis of the quadruple gai-t6 rga-t2 rgl1-1 rgl2-1 mutant (Achard et al., 2006) was similar to the size of pistils in wild-type plants, mutant pistils grew longer after anthesis, reaching a final length similar to parthenocarpic fruits obtained by GA3 treatment. In addition, constitutive GA signaling in the mutant plants also confers impaired fertilization (seed set) and silique elongation, because fruits obtained by manual pollination were smaller in the quadruple-DELLA mutant than in the wild type (Figure 3a) and contained fewer seeds (Figure 3b).

Figure 3.

 Quadruple–DELLA mutant shows parthenocarpic fruit development.
(a) Quantification of pistil and fruit length. Flowers of quadruple-DELLA mutant (gai-t6 rga-t2 rgl1-1 rgl2-1) and parental Landsberg erecta (Ler) plants were emasculated and either allowed to grow in the absence of pollination (unfertilized) or hand pollinated (fertilized) with Ler pollen. Pistil and fruit length were measured at anthesis, at 7 days after anthesis for unpollinated pistils or at maturity (stage 18) for pollinated pistils. The mean and SD were calculated from at least 50 pistils or fruits per treatment.
(b) Correlation between seed number and length of pistil or fruit of the quadruple-DELLA and Ler. Flowers were emasculated and hand pollinated with different amount of Ler pollen to vary seed set. More than 40 fruits and pistils were used for each genotype. The correlation coefficient (R2) and adjusted regression line is indicated. The experiment was repeated twice with similar results.

Consistent with the GA3 treatments, the ena layer in pistils of the quadruple-DELLA mutant, which is present at anthesis, was absent 7 days later, while it remained unaltered in wild-type pistils (Figure 4). This observation strongly supports the hypothesis that the degradation of the ena cell layer is mediated by GAs.

Figure 4.

 Degradation of endocarp a (ena) is mediated by GAs.
Transverse cryosections of fruits and pistils from quadruple-DELLA mutant and Landsberg erecta (Ler) plants at 7 days post-anthesis. Cell layers of the pericarp are indicated as in Figure 2. Absence of ena in the parthenocarpic fruit of quadruple-DELLA is indicated by an asterisk. Bars are 50 μm.

Activation of auxin response in ovules upon fertilization

The observation that impairment of polar auxin transport causes parthenocarpic fruit development (Figure 1) suggests that fertilization could cause a local increase in the concentration of endogenous auxin sufficient to trigger fruit development. To test this hypothesis, we examined the pattern of expression of the synthetic auxin-regulated promoter DR5rev (Ulmasov et al., 1997) fused to GFP (ProDR5rev:GFP) (Benkova et al., 2003) during fruit-set, as an indirect indicator of auxin level and/or response. The low-fluorescence signal detected at anthesis (data not shown), and 24 h after mock treatment (Figure 5) was associated with the lignin in the vasculature of the funiculus, while only a very low GFP signal was detected in ovules. No GFP signal was detected in the valves. More interestingly, 24 h after pollination, a strong increase in the GFP signal was detected in the funiculus, chalaza, and micropyle of the ovule, but not in the valve, which indicates that an increase in auxin response in the young seeds is an early event in fruit development. DR5rev was also strongly upregulated 24 h after 2,4-D application, and this expression extended into the valve, which seems to be the result of ectopic presence of the synthetic auxin. The upregulation in ovules mimics the effect of pollination, suggesting that an increase in auxin response in the ovules is concomitant with their fertilization. This is further supported by the observation that NPA treatment also caused an upregulation of DR5rev expression in ovules but not in valves (Figure 5). In contrast, ovules of GA3-treated pistils did not show any GFP signal, indicating that GA-induced fruit development is independent of auxin in the ovule. Two possible scenarios emerge: either auxin and GAs act through independent pathways to promote fruit development, or GAs largely mediate the promotion of fruit growth induced by auxin.

Figure 5.

 Auxin response is detected in ovules upon fertilization or by treatments with auxin or N-1-naphthylphthalamic acid (NPA), but not with gibberellin (GA3).
Expression of GFP under the control of the DR5rev promoter in ovules and valves of pistils and fruits of cer6-2 plants. Flowers were emasculated and treated with mock, GA3, 2,4-dichlorophenoxyacetic acid (2,4-D), or NPA, or hand-pollinated with Landsberg erecta (Ler) pollen (Pollinated). Images were taken 24 h later and are either the composite of fluorescent GFP and bright field images (first row), or the fluorescent GFP image (second row). Chlorophyll fluorescence is observed as red signal, while GFP appears as green signal. f, funiculus, ch, chalaza; i, integuments; m, micropyle.

Gibberellin metabolism is activated by auxin after fertilization

The possibility that GAs mediate the action of auxins in Arabidopsis fruit-set is supported by previous reports showing that exogenous auxin increases the production of GAs in pea fruit pericarp (Van Huizen et al., 1995, 1997; Ozga and Reinecke, 1999; Ozga et al., 2003). In tomato, it has recently been shown that GAs are required for auxin-induced fruit-set (Serrani et al., 2008). Furthermore, auxin upregulates GA-metabolism genes in Arabidopsis seedlings (Frigerio et al., 2006). These observations led us to hypothesize that, upon fruit-set, auxin accumulation in ovules would alter GA metabolism and GA levels, which would in turn lead to fruit growth.

Using quantitative real-time PCR (qRT-PCR) we analyzed the expression of the genes involved in GA metabolism in pistils. Out of the 16 genes for GA metabolism described in Arabidopsis (five encoding GA 20-oxidases, four encoding GA 3-oxidases, and seven encoding GA 2-oxidases) (Curaba et al., 2004; Frigerio et al., 2006), we could detect expression for 15 of them in pistils at the anthesis stage (Figure S1). The highest expression level was that of AtGA20ox2, followed by AtGA2ox2, AtGA2ox4, and AtGA2ox6. Other genes were expressed at similar levels to that of AtGA20ox1 (AtGA3ox1, AtGA2ox1, and AtGA2ox8), while the rest showed lower or very low expression. Overall, these data are in agreement with previous data for the expression of AtGA20ox1 and AtGA20ox2 (Rieu et al., 2008b) and GA 3-oxidases (Mitchum et al., 2006; Hu et al., 2008).

To investigate the spatial and temporal expression patterns of the genes involved in GA metabolism and to understand the contribution of the different tissues within the pistil to the coordination of fruit growth, we have determined their expression level by qRT-PCR in manually dissected pistils and fruits. Pistils were either hand-pollinated with wild-type pollen, or not pollinated, harvested 1, 2, and 3 days after anthesis, and dissected to separate valves and ovules/seeds. Expression of most of the GA biosynthetic genes was upregulated upon fertilization within different time frames (Figure 6); but most importantly, induction occurred exclusively in the fertilized ovules, not in the valves. One day after pollination, AtGA20ox1 and AtGA3ox1 were transiently upregulated, while other genes (AtGA20ox3, AtGA20ox5, AtGA3ox3, and AtGA3ox4) were upregulated later. AtGA20ox2 showed a slight transient increase 2 days after pollination. In contrast, expression of GA inactivation genes was upregulated in both valves (AtGA2ox1 and AtGA2ox8) and in fertilized ovules (AtGA2ox1, AtGA2ox3, and AtGA2ox4) (Figure 7). AtGA2ox1 was upregulated 24 h after pollination, and remained elevated for the next 2 days in both valves and seeds, while AtGA2ox3 and AtGA2ox4 were upregulated only in the seeds 72 h after pollination, and AtGA2ox8 was upregulated only in the valves during the first day (Figure 7).

Figure 6.

 Expression of gibberellin (GA) biosynthesis genes, GA 20-oxidases and GA 3-oxidases, is upregulated in the ovule upon fruit-set.
Unfertilized (white bars) and fertilized (grey bars) cer6-2 pistils were harvested at 0, 24, 48, and 72 h after anthesis. Expression was analyzed in whole pistils (P), as well as in manually dissected valves (V) and ovules (O) at each time point with the exception of the 0 h control sample.
(a) Expression of the five GA 20-oxidase genes.
(b) Expression of the four GA 3-oxidase genes. Each sample was a pool of at least 50 pistil/fruits. Quantitative RT-PCR data were normalized to the expression of ACT8 (At1g49240) and then to the expression of control unpollinated whole pistils at anthesis. Expression of AtGA3ox3 at 0 h was not detected and a minimum Ct of 34 was assigned at this time point for quantification purposes. Data correspond to mean and SD of three technical replicates. Two independent experiments were carried out, with similar results.

Figure 7.

 Expression of gibberellin (GA) inactivation genes, GA 2-oxidases, is upregulated in the ovule and valve upon fruit-set.
Samples were those used in Figure 6, and the experimental procedures are described in the figure legend.

Given that the expression of GA metabolism genes during fruit-set is under strict spatial control, the question arises whether auxin in the fertilized ovules directs this regulation. If this is the case, auxin application, and treatments with NPA, should also upregulate GA-metabolism genes in pollinated pistils. Application of auxin to unpollinated pistils indeed produced an overall induction of GA biosynthesis genes (Figure 8a, and Figures S2–S4 in Supporting Information). In a similar way, NPA treatment, that had been shown to increase auxin response in ovules (Figure 5), also resulted in the immediate upregulation of GA-metabolism genes. These effects were particularly evident for AtGA20ox1, AtGA20ox2, and AtGA3ox1. On the other hand, GA treatments caused a strong inhibition of the expression of GA biosynthesis genes as a result of negative feedback. Moreover, expression analysis in dissected pistils indicated that several of these genes responded specifically in ovules versus valves (Figures 8b and S5–7): while AtGA20ox1 expression was induced by auxin treatments in both ovules and valves, AtGA20ox2, AtGA20ox3, and AtGA3ox1, were induced in the ovules, and AtGA20ox4 and AtGA20ox5 were not significantly affected. Finally, the GA 2-oxidase genes showed little or no induction by auxin.

Figure 8.

 Expression of gibberellin (GA) biosynthesis genes is upregulated upon fruit-set induced by auxin in an organ-specific manner.
(a) Time course of induction. Gene expression is shown for AtGA20ox1, AtGA20ox2, and AtGA3ox1 in fruits of cer6-2 at 2, 6, 24, and 48 h after treatment with mock, GA3, N-1-naphthylphthalamic acid (NPA), or 2,4-dichlorophenoxyacetic acid (2,4-D).
(b) Tissue specificity of the expression. Gene expression is shown for AtGA20ox1, AtGA20ox2, and AtGA3ox1 in cer6-2 whole pistils (P), valves (V), and ovules (O) at 6 h after treatment with mock, GA3, NPA, or 2,4-D. Each sample was a pool of at least 50 pistils/fruits.
Quantitative RT-PCR data in (a) were normalized to the expression of PP2A (At1g13320) and then to the expression of mock-treated pistils at the same time-point. Data in (b) were normalized to the expression of ACT8 and then to the expression of mock-treated whole pistils. Data correspond to mean and SD of three technical replicates. Two independent experiments were carried out, with similar results.

All these data suggest that, upon fertilization, auxin in the ovule causes a rapid increase in expression of GA biosynthesis genes that, in turn, results in an increased production of GAs specifically in the seed. Although auxin treatment mimics the effect of fertilization, there are some differences, such as the auxin-induced expression of AtGA20ox1 in the valve. This coincides with the effect of 4-Cl-IAA treatment on PsGA20ox1 expression in pea pericarp (Ngo et al., 2002), and may be attributed to an ectopic induction due to the direct application of auxin to the pistil wall. Thus, auxins and GAs seem to act in a sequential manner to induce fruit growth after fertilization. Nevertheless, the existence of a parallel pathway, whereby GA biosynthesis would be activated directly by fertilization without auxin mediation, could not be completely ruled out based on our data.

Stability of RGA protein in the ovules is controlled by auxin

Gibberellins are thought to exert their action in the valves, where they promote cell expansion (and eventually the collapse of the ena layer). But the fertilization-dependent increase in auxin response occurs in the ovules (Figure 5), where it provokes changes in GA metabolism. One relevant question is whether this change in GA biosynthesis is accompanied by changes in GA signaling and, more importantly, whether GA signaling occurs only in the ovules or also in the valves.

Elevated GA levels trigger GA signaling by promoting the degradation of DELLA proteins in a cell-autonomous manner, relieving the restriction imposed by these negative regulators upon GA responses (Fleet and Sun, 2005). In a reciprocal manner, the examination of the stability of DELLA proteins is a suitable approach to infer changes in GA concentration, although there are publications that suggest that signals other than GA might affect DELLA levels. Therefore, we examined the level of the fusion protein GFP-RGA, expressed under the control of the endogenous RGA promoter (ProRGA:GFP-RGA) (Silverstone et al., 2001), in pistils in response to various treatments that induce fruit-set. At anthesis, or 1 day after mock treatment, GFP-RGA was localized in ovules, with the highest signal at the base of the ovule and funiculus (Figure 9). It could also be detected in the valves, although at a lower intensity. As expected, exogenous GA3 provoked the complete disappearance of GFP signal in ovules and valves 24 h after the treatment. More importantly, pollination and NPA and 2,4-D treatments also caused degradation of RGA in seeds/ovules and in valves, albeit with different kinetics. NPA was the most effective, so a low signal was detected 24 h after the treatment, while in 2,4-D-treated and pollinated pistils the signal decreased after 48 h. In the case of pollination, a clear correlation was found between the presence/absence of GFP-RGA and non-fertilized/fertilized ovules (Figure 9). In all cases, GFP-RGA was also degraded in the valve, with a time course similar to the disappearance in the ovule.

Figure 9.

 Gibberellin (GA) signaling occurs in both ovules and valves upon fruit-set.
Pistils of cer6-2 plants harboring the fusion protein GFP-RGA under the control of the endogenous promoter RGA (Silverstone et al., 2001) were either treated with mock solution, GA3, N-1-naphthylphthalamic acid (NPA), 2,4-dichlorophenoxyacetic acid (2,4-D), or hand-pollinated with Landsberg erecta (Ler) pollen. Images were taken 24 and 48 h later, and are the composite of fluorescent GFP and bright field images. Chlorophyll fluorescence is observed as a red signal, while GFP appears as green signal. u, unfertilized ovule; f, fertilized ovule.

Our data suggest that RGA degradation in fertilized ovules during fruit-set is most probably due to elevated GA levels, achieved by auxin-induced upregulation of GA biosynthesis genes after fertilization. Interestingly, RGA degradation is ovule/seed autonomous (Figure 9), suggesting that no GA diffusion occurs from fertilized to unfertilized ovules. In addition, degradation of RGA in the valves indicates the presence of GAs in that tissue. Since our expression analysis (Figure 6) does not support upregulation of GA synthesis in the valves, a likely possibility is that they are translocated from the fertilized ovules, where they are synthesized according to our gene expression data.


Our analysis of the molecular events associated with the action of auxin and GA during fruit-set indicates that: (i) auxin response in young seeds is an early event during fruit-set; (ii) auxin promotes the activation of GA metabolism specifically in fertilized ovules; (iii) GA signaling is detected in both seeds and valves; and (iv) constitutive GA signaling is sufficient to promote parthenocarpic fruit development.

Auxin response in the fertilized ovule is an early event during fruit-set

Through the analysis of ProDR5rev:GFP transgenic plants, we have shown that an auxin response is activated in ovules soon after fertilization, probably as a consequence of elevated auxin levels. Several explanations could account for this observation: a local increase in auxin synthesis, a blockage of auxin transport, interference with auxin response upon fertilization, or even the downloading of auxins from the pollen tube into the ovule (Sweet and Lewis, 1969). However, the fact that NPA treatment induces parthenocarpic fruit development by mimicking the same auxin response in the ovule seen upon fertilization rules out the latter two possibilities. Parthenocarpic development induced by a block in polar auxin transport was previously reported in cucumber (Robinson et al., 1971; Beyer and Quebedeaux, 1974), and it was proposed that the effect might be due to a blockade in the outward flow of auxin from the ovary, resulting in auxin accumulation sufficient to trigger fruit-set in absence of fertilization.

Whether the increase in auxin in the fertilized ovule is due to enhanced biosynthesis or interference in transport cannot be determined based on our results. Nevertheless, we have not observed any modification in the abundance and localization of several PIN auxin efflux carriers (Blakeslee et al., 2005) during pistil development and fertilization (data not shown). Thus, it seems likely that auxin synthesis is activated in the ovule early after fertilization and that this may initiate fruit development. Several pieces of evidence support this view. First, auxins are naturally present in developing ovules. We have detected a very low GFP signal driven by the ProDR5 in mock-treated ovules, which indicates that a basal level of auxin is present in unfertilized ovules. In addition, conjugated auxins have been detected in the ovules of unfertilized pistils (Aloni et al., 2006) and in developing ovules and embryos (Benkova et al., 2003; Friml et al., 2003). Second, exogenous auxin can induce parthenocarpy in a variety of species, including Arabidopsis. Finally, overexpressing IAA-biosynthesis genes in ovules or placenta of transgenic plants result in parthenocarpy in several species (for example, Rotino et al., 1997; Yin et al., 2006). In summary, either application of auxin or overexpression of auxin biosynthetic genes in ovules would promote the auxin response observed upon fertilization. It is then plausible to hypothesize that elevated levels of auxin in the ovule would initiate a cascade of events that would finally promote fruit growth. These events would include auxin signaling, through the function of Aux/IAA and ARF proteins (Swain and Koltunow, 2006; Pandolfini et al., 2007), and crosstalk with other hormones (Weiss and Ori, 2007).

Auxin action is mediated by gibberellins in the ovary

Although the efficiency of the different hormones in the induction of fruit development varies between species, the activation of GA metabolism after fertilization is a common theme. For instance, early fruit development in tomato plants is characterized by elevated mRNA levels of copalyl diphosphate synthase, a gene involved in the early steps of GA biosynthesis, LeGA20ox1 and LeGA20ox3, accompanied by decreased level of LeGA3ox2 (Rebers et al., 1999). Indeed, the expression of GA 20-oxidase genes seems to be a limiting step for the increase in GA concentration during fruit-set (Serrani et al., 2007). A similar trend has been observed in pea, where a GA 20-oxidase gene is strongly upregulated during early fruit development, both in the developing seeds and in the pod (Garcia-Martinez et al., 1997).

How does fertilization activate GA biosynthesis? Our results suggest that the initial event that causes this increase in GA concentration may be the activation of auxin signaling in the ovules. Again, this connection is not species-specific, and the particular steps in GA metabolism affected by auxin are equivalent between species. For instance, in pea the auxin 4-Cl-IAA induces expression of GA biosynthesis genes (PsGA20ox1 and PsGA3ox1) and accumulation of GAs (Van Huizen et al., 1995, 1997;Ozga and Reinecke, 1999; Ozga et al., 2003), which is in agreement with our observations that auxin upregulates AtGA20ox1, AtGA20ox2, AtGA20ox3, and AtGA3ox1 in early fruit development in Arabidopsis. Furthermore, auxin induces fruit-set and growth in tomato, at least partially, by enhancing GA biosynthesis and decreasing GA inactivation activity, leading to a higher bioactive GA content (Serrani et al., 2008). However, we cannot rule out the existence of a parallel pathway where GA biosynthesis would be activated directly after fertilization, in an auxin-independent manner.

Interestingly, GA application triggers fruit development without any alteration in auxin response, based on the analysis of ProDR5rev:GFP expression. These results point to a hierarchy in the action of these hormones, according to which fruit development would be initiated after fertilization by the sequential action of auxins and GAs. If this is the case, the co-occurrence of GA biosynthesis and signaling would be an absolute requirement for fruit development, induced either by fertilization or by auxin treatments. Several pieces of evidence support this hypothesis. First, double loss-of-function of AtGA20ox1 and AtGA20ox2, or ectopic expression of a GA 2-oxidase gene, causes seed abortion and shorter siliques in Arabidopsis (Singh et al., 2002; Rieu et al., 2008b). Second, the use of inhibitors of GA biosynthesis hinders auxin-induced fruit-set in tomato, an effect that is reversed by GAs, suggesting that the effect of auxin is fully mediated by GA (Serrani et al., 2008). Third, expression in tomato of a gain-of-function allele of an Arabidopsis DELLA protein, gai-1D, reduces fruit development (Marti et al., 2007). Fourth, silencing of the single tomato DELLA gene (SlDELLA) results in the formation of parthenocarpic fruits in the absence of pollination (Marti et al., 2007), indicating that, at least in tomato, fruit initiation is mediated by DELLA proteins. The only observation that seems to contradict this hypothesis is the ability of auxin to induce fruit-set in the Arabidopsis gai-1D mutant (Vivian-Smith and Koltunow, 1999), although redundancy of DELLA proteins in Arabidopsis might account for this effect, i.e. auxin-induced fruit-set would be mediated not by GAI but by RGA or RGL proteins. However, it is also possible that auxins and GAs have specific independent roles in the promotion of fruit development, beyond the early and sequential action described above. Support for this scenario comes from the observation that only GAs promote certain differentiation processes early during fruit development, such as the collapse of the ena layer.

Coordinated hormone signaling in ovules and valves

Upon ovule fertilization, the initiation of embryo development triggers elongation of the pistil to host the developing seeds. Thus, a growth-induction signal has to be generated to promote growth of the valves. Most probably, this signaling molecule would be hormonal in nature, but where this molecule is generated and how it is transported is unknown. Our data indicate that: (i) fertilization-induced changes in auxin response are detected in young seeds, but not in valves; (ii) GA metabolism is activated upon fertilization or during auxin-induced fruit development, specifically in seeds, at least within the first 72 h upon pollination; and (iii) changes in GA signaling are detected in both seeds and valves. In this scenario, it is likely that auxin present in fertilized ovules promotes GA synthesis specifically in the young seeds, which suggests that there might be a GA-specific regulatory function in seed development.

Furthermore, we have shown that the different genes involved in GA biosynthesis act in a sequential manner: some genes are upregulated early upon fruit-set while others are activated later on, the latter probably being associated with seed development (Figure 10). The GAs synthesized in the seed would then be transported to valves to promote GA signaling and finally to promote coordinated growth. In other words, GAs would be necessary for both seed and valve development. In this regard, a close correlation between seed development and pod elongation has been reported in Arabidopsis, which implies that a seed-derived growth-promoting signal would be transferred through the funiculus to the adjacent valve structures and allow growth (Cox and Swain, 2006). It has been shown that AtGA20ox2 makes the greater contribution to fruit elongation, as its loss-of-function provokes shorter siliques, but with similar seed numbers, than in wild-type plants (Rieu et al., 2008b). A similar effect was observed for the loss-of-function of AtGA3ox1 (Hu et al., 2008). Since this phenotype could not be fully rescued by pollination with wild-type pollen, it was postulated that the GAs regulating silique elongation would be of maternal origin, i.e. that GAs would probably be synthesized (via AtGA20ox2 and AtGA3ox1 activity) in the silique tissue itself or in the seed endosperm (Hu et al., 2008; Rieu et al., 2008b). Based on our observation that the expression of these genes is localized in seeds but not in silique walls, it is possible that the maternal effect described above could be attributed to GAs being synthesized in the endosperm, but not in the valve tissue. The loss-of-function of AtGA3ox4, in combination with ga3ox1, results in shorter fruits that can be rescued by wild-type pollen, suggesting an embryo- or seed-effect (Kim et al., 2005; Hu et al., 2008). Nevertheless, the observation that a quintuple GA 2-oxidase loss-of-function mutant, in which biosynthesis of bioactive GA is unaltered but inactivation is blocked, develops parthenocarpic siliques (Rieu et al., 2008a) implies that even in the absence of fertilization the pistil must have access to GAs that would then accumulate and drive parthenocarpic fruit development. Our results suggest that these active GAs would have to be transported to the silique from other tissues, but specific experiments would be needed to address this question.

Figure 10.

 Schematic representation of the temporal and spatial expression of gibberellin (GA) metabolism genes upon fruit-set.
Arrows indicate the temporal expression frame of the corresponding gene, between anthesis and 72 h after fertilization, in the ovule (upper panel) and valve (lower panel).

Finally, our data show that while GA synthesis occurs only in the young seed (Figure 10), GA signaling is detected both in developing seeds and in valves, indicating that GAs could be the hypothesized growth-regulator that coordinates seed and fruit development. The role of GAs in the valves would be to promote fruit growth and cellular differentiation, including cell expansion and degradation of the inner endocarp cell layer. The collapse of ena occurs naturally towards the end of stage 17 of pollinated fruit development (Ferrandiz et al., 1999; Roeder and Yanofsky, 2006), as well as in auxin-induced fruits (our data), but seems to be accelerated in GA-induced fruits. Similar data were reported in emasculated spy-4 pistils induced to grow with GA3, but not in wild-type pistils (Vivian-Smith and Koltunow, 1999). These data suggest that degradation of ena prior to pod shattering is controlled by GAs, as is the case for other developmental processes that take place during fruit development, such as cell expansion in mesocarp cells (Vivian-Smith and Koltunow, 1999).

Experimental procedures

Plant material and hormone treatments

Arabidopsis thaliana cer6-2 seeds, in the Landsberg erecta (Ler) accession (Preuss et al., 1993; Fiebig et al., 2000), were obtained from the Arabidopsis Biological Resource Center (ABRC). Seeds from the lines ProDR5rev:GFP in the Columbia-0 accession (Col-0) (Benkova et al., 2003), the ProRGA:GFP-RGA in Ler (Silverstone et al., 2001), and the quadruple-DELLA mutant (gai-t6 rga-t2 rgl1-1 rgl2-1) in Ler (Achard et al., 2006) were obtained from Jiri Friml (Tübingen University, Germany), Tai-ping Sun (Duke University, NC, USA), and Nicholas P. Harberd (John Innes Centre, Norwich, UK), respectively.

Plants were grown in chambers at 22°C under a 16-h light/8-h dark photoperiod and at 50% relative humidity to avoid fertilization of cer6-2. For each plant, only flowers from the primary bolt, between 0 and 2 days post-anthesis (dpa) were used. For plants not harboring the mutant cer6-2 allele (i.e. the parental Ler, the transgenic line harboring the ProDR5rev:GFP construct, and the quadruple-DELLA mutant), self-pollination was avoided by emasculation 1 day before anthesis. Fruit-set was induced by hand pollination with Ler pollen or spray application of 330 μm GA3 (Fluka, http://www.sigmaaldrich.com/), 10 μm 2,4-D (Sigma Aldrich, http://www.sigmaaldrich.com/), 300 μm NAA (Sigma), or 50 μm NPA (Duchefa Biochemie, http://www.duchefa.com/). These concentrations were tested to be optimal to induce fruit-set. Tween-80 at 0.05% was used as the wetting agent, and the pH was adjusted between 6.5 and 7.0. Samples were harvested and processed at the indicated time-points. For dissection experiments, valves and ovules were collected by hand dissection under a stereoscope microscope with the help of a razorblade and acupuncture needles, placed in a vial with RNA extraction buffer, and RNA was immediately extracted (see below).

To test induction of fruit-set, pistil or fruit length was measured at 7 dpa or at full maturity. Pistils and fruits were harvested and scanned, and images were analyzed using Image J software (Abramoff et al., 2004).

Scanning electron microscopy (SEM)

Samples were harvested, mounted on the specimen holder of a CT-1000C cryo-transfer system (Oxford Instruments, http://www.oxford-instruments.com/), interfaced with a JEOL JSM-5410 scanning electron microscope, and frozen in liquid N2. Samples were fractured, and sublimated by controlled heating at −85°C. Finally, samples were observed at incident electron energy of 10 keV with 10 × to 100 × magnification.

Confocal microscopy

A Leica TCS SL confocal microscope (Leica Microsystems, http://www.leica-microsystems.com/) was used. Green fluorescent protein was excited at 488 nm and emission was detected between 500 and 520 nm. Endogenous chlorophyll was excited with the same wavelength but detected between 660 and 690 nm. The identity or specificity of each signal was confirmed with a λ-scan.

Quantitative RT-PCR analysis of gene expression

Total RNA was extracted using the RNeasy Plant Mini Kit (Qiagen, http://www.qiagen.com/). Genomic DNA was eliminated by treatment with 50 units of DNaseI (RNase-free DNase set; Qiagen) for 15 min at room temperature. Two micrograms of total RNA were used to synthesize first-strand cDNA, using the SuperScript® First-Strand Synthesis System for RT-PCR (Invitrogen Life Technologies, http://www.invitrogen.com/). The cDNA synthesis reactions were finally diluted in a volume of 80 μl.

Quantitative RT-PCR was carried out using the SYBR® GREEN PCR Master Mix (Applied Biosystems, http://www3.appliedbiosystems.com/) in an ABI PRISM 7000 Sequence Detection System (Applied Biosystems). The final reaction volume was 20 μl, with 1 μl of cDNA, 10 μl of SYBR® GREEN PCR Master Mix, and 9 μl of primer mixture, containing 0.66 μm of each primer. The PCR program consisted of an initial incubation of 2 min at 50°C followed by a denaturation at 95°C for 10 min, and 40 cycles of amplification of 15 sec at 95°C and 1 min at 60°C. In a single experiment, each sample was assayed in triplicate. Expression levels were calculated relative to the constitutively expressed genes ACT8 or phosphatase 2A (PP2A) subunit PDF2 (At1g13320) (Czechowski et al., 2005), which were tested to be constitutive in the different tissues used in each experiment (PP2A for whole pistils and fruits, and ACT8 for dissected ovules and valves). Normalization was carried out using the ΔΔCt method (Applied Biosystems), where ΔCt was calculated for each sample as the difference between Ct (gene of interest) and Ct (constitutive gene), and the final relative expression level was determined as inverse of log2 of Ct (sample) –Ct (reference sample). Normalization was as indicated in the figure legends and in the text. Primer sequences for amplification have been previously described (Curaba et al., 2004; Czechowski et al., 2005; Frigerio et al., 2006) and are shown in Table S1. All experiments were repeated twice, with similar results.


We thank Jiri Friml (Tübingen University, Germany), Tai-ping Sun (Duke University, NC, USA), and Nicholas P. Harberd for the gift of ProDR5rev:GFP, ProRGA:GFP-RGA, and quadruple–DELLA mutant seeds, respectively; J. L. Garcia-Martinez and D. Alabadi for critical reading of the manuscript; and Ms M. A. Argomániz for excellent technical assistance. ED was supported by an FPI Fellowship from the Spanish Ministry of Education and Science (MEC). MAPA received a post-doctoral contract from the ‘Ramón y Cajal’ program from MEC. MAB was supported by the EMBO Young Investigator Program. This work was funded by grants from the Spanish MEC BIO2002-04083-C03-02 and BIO2005-07156-C02-01.