Overexpression of a tomato miR171 target gene Sl GRAS 24 impacts multiple agronomical traits via regulating gibberellin and auxin homeostasis

1The GRAS family of plant proteins is responsible for regulating many aspects of growth, development and responses to the biotic and abiotic environment. GRAS family members are diverse proteins that typically have five conserved motifs in the C‐terminus (Bolle, ). GRAS proteins usually act as transcription factors, but not all GRAS protein functions have been described. We previously identified a tomato (Solanum lycopersicum) GRAS transcription factor gene SlGRAS24 as target of tomato miR171 (Huang et al., ) and aimed to discover more about its function in tomato development here.

2SlGRAS24 is phylogenetically clustered into the HAIRY MERISTEM (HAM) subfamily of GRAS genes and shows the highest sequence identity with Arabidopsis thaliana AtSCL6 (Huang et al., ). In 2002, a GRAS transcription factor named HAM was identified as a component of a novel non‐cell‐autonomous signalling pathway maintaining shoot indeterminacy in Petunia hybrida, ham mutants displayed arrest in lateral organ and stem production (Stuurman et al., ). In the same year, two Arabidopsis orthologs of Petunia HAM were proved to be endogenous targets of post‐transcriptional degradation by miR171, a member of a miRNA family conserved in different plant species (Llave et al., ). Actually, a total of three GRAS genes are regulated by miR171 in Arabidopsis, SCL6/SCL6‐IV, SCL22/SCL6‐III and SCL27/SCL6‐II (also known as the HAM or LOM (LOST MERISTEMS) genes because of their mutant phenotypes) (Reinhart et al., ). LOM1 and LOM2 genes promote incorporation of peripheral zone cells into leaf primordia and help to maintain a polar organization of the shoot meristem (Schulze et al., ). Further research showed that these miR171 target genes were not only required for shoot apical meristem maintenance, but for maintenance of root indeterminacy (Engstrom et al., ). More recently, it is found that HAM proteins act as conserved interacting cofactors with WUS/WOX proteins. They share common targets and their physical interaction is important in driving downstream transcriptional programmes and in promoting shoot stem cell proliferation (Zhou et al., ). Arabidopsis overexpressing miR171 and the triple scl6 mutants have similar pleiotropic phenotypes, where shoot branching, plant height, chlorophyll accumulation, primary root elongation, flower structure, and leaf shape and patterning were all altered (Wang et al., ). In barley and rice, overexpression of miR171 affects phase transitions and floral meristem determinacy (Curaba et al.,; Fan et al., ). miR171‐GRAS module controls flowering time (phase transition) and trichome distribution via inhibiting the activity of miR156‐targeted SPL proteins (Xue et al., ). This module is also critical for mediating GA‐DELLA signalling in the coordinate regulation of chlorophyll biosynthesis and leaf growth in light (Ma et al., ). Moreover, it has been extensively studied about the role of miR171 upon various stresses in different species, including Arabidopsis, barley, maize and Solanum tuberosum (Hwang et al.,; Kantar et al.,; Kong et al.,; Liu et al., ).

3HAM gene function may be conserved but the dramatic expansion in HAM homologs diversity in flowering plants strongly suggests the evolution of novel functions or functional subspecialization in angiosperms (Wu et al., ). Indeed, elevated rates of evolution in flowering plant HAM homologs indicate a refinement of HAM function in response to selective pressures (Engstrom et al., ). In tomato, six SlGRAS genes are clustered into the HAM subfamily, including SlGRAS24 and SlGRAS40, which are confirmed to be targeted for mRNA cleavage by miR171, and SlGRAS8, a suspected target gene whose translation is repressed by miR171 (Huang et al., ). SlGRAS24 contains a conserved MIR‐binding sequence which is perfectly matched with Sly‐miR171 (Huang et al., ). Despite the close evolutionary relationship between SlGRAS24 and SlGRAS40, their transcripts in different tissues/organs and in response to hormone and abiotic stress differ greatly (Huang et al., ), indicating that they might have different functions and participate in distinct physiological processes.

4In this study, the tomato GRAS transcription factor gene SlGRAS24, an ortholog of Arabidopsis AtSCL6 (Huang et al., ), was characterized. First, the expression pattern of the gene was studied in wild‐type (WT) and transgenic plants expressing β‐glucuronidase (GUS) under the SlGRAS24 promoter (proSlGRAS24‐GUS). Transgenic tomato plants overexpressing SlGRAS24 exhibited phenotypes similar to those observed in other species suggesting functional conservation among these homologs. Some new characteristics such as abnormal axillary bud emergence, reduced fruit set ratio, arrested fruit and seed development were also observed, indicating that SlGRAS24 has additional specific functions in tomato. Moreover, the promoter of SlGRAS24 was analysed and found to be associated with both gibberellin and auxin signalling pathways. Hormone‐related transcripts were thus quantified and the floral transcriptome was analysed in plants overexpressing SlGRAS24. Collectively, our study of SlGRAS24 protein unravels its role in vegetative growth and reproductive tissues and advances our understanding about hormone crosstalk in tomato.

5We hypothesized that the SlGRAS24 gene and its regulator Sly‐miR171 would have similar functions in tomato as their respective orthologs in Arabidopsis (Wang et al., ). We investigated this first by quantifying the expression of Sly‐miR171 and SlGRAS24 in the ‘Micro‐Tom’ cultivar (WT) by qRT‐PCR. Both genes were detectable in all tissues with the highest expression levels found in flowers (Fig. 1A). The levels of both transcripts dropped during fruit development and the lowest level of expression was at the ripening stage. Overall, Sly‐miR171 and SlGRAS24 mRNAs had similar transcription patterns, which were consistent with research on their Arabidopsis orthologs (Wang et al., ). However, floral organs were an exception because SlGRAS24 mRNA was most abundant in stamens where Sly‐miR171 mRNA expression was at its lowest level. The expression data here suggest that miR171‐SlGRAS24 regulatory networks are needed throughout vegetative and reproductive development in tomato.

6Spatial expression of SlGRAS24 was monitored by histochemical staining of transgenic tomato in which GUS reporter's expression was driven by the upstream promoter sequence of SlGRAS24 gene (Fig. 1B). In proGRAS24‐GUS homozygous seedlings, GUS reporter activity was strong in leaf primordia and root tips, but quite weak in cotyledons and hypocotyls. GUS activity was also expressed in flower primordia, in the young leaves’ margins of shoot apices, in internodes and in axillary buds. In reproductive tissues, SlGRAS24 was highly expressed in stamens and stigmas, and predominantly expressed in seeds of young fruits. These results suggest that SlGRAS24 expression is spatiotemporally regulated and SlGRAS24 may have specific functions in developmental processes.

7To determine the subcellular localization of SlGRAS24 protein, the vector 35S‐SlGRAS24‐GFP was transiently expressed in tobacco protoplasts. Confocal imaging of protein fluorescence showed that the green fluorescence signal of 35S‐SlGRAS24‐GFP was exclusively detected in the nucleus, whereas the cells transformed with the vector containing GFP alone displayed fluorescence throughout the cells (Fig. 1C). A yeast two‐hybrid experiment was used to examine the transcriptional activity of SlGRAS24. A GAL4 DNA‐binding domain SlGRAS24 fusion protein was expressed in yeast cells, which were then assayed for their ability to activate transcription from the GAL4 sequence. SlGRAS24 promoted yeast growth in the absence of histidine and adenine, and showed X‐α‐gal activity, whereas the control vector pGBKT7 did not (Fig. 1D). These results suggest that SlGRAS24 has transcriptional activity and is targeted to the nucleus in plant cells.

8To assess the physiological importance of the SlGRAS24‐encoded protein, the tomato ‘Micro‐Tom’ genotype was transformed with sense or antisense constructs of the SlGRAS24 gene to produce several independent overexpressing (SlGRAS24‐OE) or underexpressing (SlGRAS24‐AS) homozygous lines. qRT‐PCR was performed to evaluate the expression of SlGRAS24 in transgenic plants, and the results showed that the SlGRAS24 was successfully up‐regulated or down‐regulated in all transgenic lines tested (Table 1). Thereby, two most up‐ or down‐regulated lines of each genotype were then chosen for the following characterization. Interestingly, SlGAR24‐AS lines did not differ from WT tomato (Fig. 2A and B). This was not unexpected as no phenotype had been observed in Arabidopsis plants in which only one of the HAM genes was mutated (Wang et al., ). We hypothesized that there is functional redundancy among GRAS family members or multiple miR171 target genes in tomato. We tested this by generating transgenic plants overexpressing the precursor of tomato miR171 to silence miR171 target genes including SlGRAS24 (Table 1). SlGRAS24 expression was inhibited to a similar degree in SlGAR24‐AS and SlymiR171‐OE plants comparing to WT, but only SlymiR171‐OE demonstrated taller plants with earlier phase transition time (Fig. 2C and D). The SlGRAS24‐OE lines showed pleiotropic phenotypes, some of which were in line with the phenotypes of transgenic Arabidopsis plants overexpressing SlGRAS24 orthologs AtHAMs (Wang et al., ). Flower opening was significantly delayed in SlGRAS24‐OE, which was consistent with the delayed phase transition time from vegetative to reproductive development. The tissue sections of apical meristems showed no formed floral primordium in SlGRAS24‐OE lines, 25 days postgermination (dpg) (Fig. 2E). In line with that, transcript levels of SlFT and SlCO1, two key regulators controlling flowering time, undergone absolutely opposite expression tendency in samples harvested at three different stages (20 dpg, 30 dpg, and 40 dpg) (Fig. 2E). The leaves of transgenic plants were both shorter and narrower than WT leaves and leaf margins were not serrated (Fig. 2F). Microscopic analysis showed that leaves were thicker in SlGRAS24‐OE lines, which might be attributed to having much larger lower epidermal cells (Fig. 2F). WT plants had axillary buds at the internodes, but abnormal axillary bud emergence was observed on stems of SlGRAS24‐OE plants (Fig. 2G). Besides, SlGRAS24‐OE plants had more lateral branches and abnormal flower bud emergence (Fig. S1). Primary and lateral root growth of SlGRAS24‐OE lines was strongly suppressed compared to WT (Fig. 2H). More detailed information about the phenotypes of SlGRAS24‐OE transgenic tomato plants is shown in Table 2. Noticeably, the fruit set ratio of SlGRAS24‐OE tomato plants severely decreased, and fruit development was defective (Table 2).

9SlGRAS24‐OE plants showed reduced fruit set and fruits were smaller with fewer seeds (Table 2, Fig. 3A, B and C). As tomato is the most important model for fruit development, it is meaningful to investigate the molecular mechanism underlying these defects. We found that SlGRAS24‐OE plants occasionally produced flowers with dehiscent stamens, but the ovary and ovules in SlGRAS24‐OE flowers were as in WT (Fig. 3D). However, SlGRAS24‐OE flowers had smaller pollen sacs and collapsed anthers (Fig. 3D). TTC staining for pollen viability showed that fewer pollen viable grains in SlGRAS24‐OE flowers undergoing anthesis (first day of flower opening) than that in WT flowers (Fig. 3D). Considering that SlGRAS24 transcripts were most abundant in anthesis flowers (Fig. 1A and B), expression pattern was studied in more detailed by examining GUS staining in flowers from ‐2 days postanthesis (dpa) to 4 dpa (Fig. 3E) and by qPCR analysing in flowers and fruits at different stages (Fig. 3F). For GUS staining, in stamens and stigmas, the strongest staining was at 0 dpa, and it became much weaker at 2 dpa, till there was almost no expression at 4 dpa. By contrast, almost no staining was observed in −2 dpa ovaries, but the expression increased since 0 dpa ovaries but was limited to ovules (Fig. 3E). Consistently, qPCR results also showed most abundant SlGRAS24 transcript in flowers at the anthesis stage (Fig. 3F).

10Cross‐fertilization assay was carried out to examine the fertility of transgenic flowers (Table 3). Fruit set ratios were 90%, 22% and 15% in self‐pollinated WT flowers, OE L10 flowers and OE L15 flowers, respectively. When WT flowers were used as the female recipient, the fruit set ratio increased slightly to 33% and 25% with OE L10 and OE L15 pollen, respectively. There was 100% fruit set when WT pollen was used to pollinate OE pistils. The results also showed that WT × WT fruits and WT × OE fruit (pollen × pistil crosses) contained more seeds than OE × WT and OE × OE fruit. These experiments demonstrate that SlGRAS24 is necessary for normal stamen development and overexpression of SlGRAS24 has a negative impact on fertilization in tomato.

11As SlGRAS24‐OE plants have smaller fruits, the histology of WT and SlGRAS24‐OE ovaries was analysed from day 3 to day 15 (Fig. 4A and Fig. S2). The pericarp of SlGRAS24‐OE ovaries contains smaller cells and fewer cell layers than WT pericarp, indicating that overexpression of SlGRAS24 led to an inhibition of cell division and expansion in early fruit development. We analysed transcript levels of four genes involved in cell division (CYCLIN DEPENDENT KINASE (SlCDKB2.1), CYCLIN (SlCycB2.1) and SlCycD3.1, which encodes a G1 cyclin) and cell expansion (XYLOGLUCAN ENDOTRANSGLUCOSYLASE/HYDROLASE 1SlXTH1) (Fig. 4B). Relative to WT ovaries, the expressions of the three cell division genes were low in 3 dpa ovaries and increased at 7 dpa and 15 dpa. By contrast, expression of the SlXTH1 cell expansion gene was always lower in SlGRAS24‐OE ovaries than in WT ovaries.

12The 2.2‐kb SlGRAS24 promoter sequence was analysed in silico using the PLACE program (http://www.dna.affrc.go.jp/PLACE/signalup.html). Several cis‐acting elements were identified including the canonical auxin response element (AuxRE) at position ‐599, two GA‐responsive elements (GARE) at positions −523 and −826, and other elements related to GA and auxin (Fig. 5A). This strongly suggests that SlGRAS24 is regulated by both two hormones. SlGRAS24 transcript levels were compared in leaves treated or untreated by GA3 or IAA, respectively. SlGRAS24 expression increased significantly within 1 h in response to either GA3 or IAA (Fig. 5B). GA and auxin responsiveness of the promoters were tested using proGRAS24‐GUS transgenic seedlings incubated in solutions containing GA3 or IAA for 3 h. Compared with untreated seedlings (mock), GUS staining revealed that GA3 or IAA treatment led to ectopic expression of the GUS gene (Fig. 5C). qPCR analysis showed that both GUS and SlGRAS24 transcripts significantly increased in response to GA3 or IAA treatment in proGRAS24‐GUS seedlings (Fig. 5D).

13To further study the role of SlGRAS24 in GA and auxin pathways, the expression levels of a panel of 21 tomato genes were monitored in WT and SlGRAS24‐OE seedlings in response to GA3 or IAA treatment (Fig. 6A). The panel was made up of 5 GA biosynthetic enzymes (SlGA20ox1, SlGA20ox2, SlGA20ox4, SlGA3ox1 and SlGA3ox2), 3 GA deactivating enzymes (SlGA2ox1, SlGA2ox2 and SlGA2ox4), a key regulator of GA signalling pathway (SlDELLA), 4 auxin/indole‐3‐acetic acid (Aux/IAA) transcription factors (SlIAA2, SlIAA4, SlIAA4 and SlIAA9), 4 auxin response gene (ARF) transcription factors (SlARF5, SlARF6, SlARF7 and SlARF8) and 4 PIN‐FORMED (PIN) auxin efflux transport proteins (SlPIN1, SlPIN3, SlPIN5 and SlPIN6). Without hormone treatment, 17 genes were down‐regulated and 2 genes were up‐regulated in SlGRAS24‐OE seedlings compared to WT, which suggested that overexpression of SlGRAS24 disrupts GA and auxin homeostasis in transgenic plants. Furthermore, some of these genes responded differently to GA3 and/or IAA in SlGRAS24‐OE seedlings as compared to WT. For instance, upon GA3 treatment, SlDELLA expression decreased in WT, but significantly increased in SlGRAS24‐OE seedlings. SlGA2ox4 was induced by both GA3 and IAA treatment in WT, while in SlGRAS24‐OE it was not induced in response to GA3 and was inhibited in response to IAA. SlIAA2 was up‐regulated 2.45‐fold by IAA treatment in WT, but about 12‐fold in OE seedlings. SlARF8 was down‐regulated under both hormone treatments in WT, while in SlGRAS24‐OE it increased in response to GA3 treatment and did not respond to IAA treatment. Comparing WT and SlGRAS24‐OE seedlings, the different responsiveness of the GA‐related genes during IAA treatment, and conversely the auxin‐related genes during GA3 treatment, might indicate that SlGRAS24 acts as an integrator between GA and auxin pathways.

14It is thus possible that SlGRAS24 plays a role in regulating the expression of hormone‐related genes in tomato, particularly genes associated with GA or auxin biosynthesis, transport and signal transduction. To assess whether changes in hormone levels accompanied changes in gene expression, endogenous IAA and GA3 were quantified using HPLC‐MS/MS (Fig. 6B). SlGRAS24‐OE seedlings contained more IAA and GA3 than in WT seedlings. This was somewhat unexpected as most GA/auxin‐related genes were down‐regulated.

15Exogenous IAA and/or GA3 were applied to WT and SlGRAS24‐OE seedlings to investigate whether overexpression of SlGRAS24 altered other aspects of GA and auxin responsiveness (Fig. 7A). Without hormone treatment, the primary roots of SlGRAS24‐OE seedlings were distinctly shorter than those of WT. In the presence of 1.0 μm IAA treatment, primary root growth was inhibited in both WT and SlGRAS24‐OE seedlings (Fig. 7A and B), while SlGRAS24‐OE seedlings had longer primary roots and fewer lateral roots than WT seedlings (Fig. 7A, B, and C), indicating that auxin responsiveness was reduced when SlGRAS24 was overexpressed. Synthetic auxin NAA stimulated more and longer adventitious roots to form from excised WT cotyledons than from SlGRAS24‐OE cotyledons (Fig. 7E), another indication of a reduced auxin response in SlGRAS24‐OE seedlings. Similarly, 50 μm GA3 inhibited the primary root growth of both WT and SlGRAS24‐OE tomato seedlings, reducing the initial difference in length (Fig. 7A and B). Besides, outgrowth of the first true leaves from the shoot apex was severely suppressed in SlGRAS24‐OE seedlings compared with WT (Fig. 7A and D).

16Under normal condition, germination was inhibited in seeds from SlGRAS24‐OE compared to that from WT (Fig. 7F). More sensitive phenotypes were observed in SlGRAS24‐OE seeds when 10 μm GA3 or 5 μm paclobutrazol, a GA biosynthesis inhibitor, was applied since no observation of marked difference when they were applied to WT seeds (Fig. 7F), implying that SlGRAS24 is likely involved in seed germination through modulating GA signalling. However, the germination rates were not fully recovered to WT level when GA3 was applied (Fig. 7F). Similarly, the dwarf phenotype and delayed flowering time of SlGRAS24‐OE plants were only partially rescued by spraying with 20 μm GA3 (Fig. 7G, H and I), suggesting that some GA‐independent pathways are involved in SlGRAS24‐mediated regulation of plant growth and seed germination.

17To detect the changes in transcript levels that may be involved in the flower–fruit transition, a comparative transcriptome analysis was conducted using flowers at the onset of anthesis (0 dpa) of two SlGRAS24‐OE transgenic lines (L10 and L15) and WT controls. Under the criteria of false discovery rate <0.05 and log2 fold change ≥1, a total of 1671 and 1436 unigenes were differentially expressed in L10 and L15, respectively, compared with WT controls (File S1 and S2). Functional annotation of putative gene products indicated that overexpression of SlGRAS24 affected multiple processes including transcription, signal transduction, primary and secondary metabolite biosynthesis, phytohormone biosynthesis, photosynthesis and stress responses, to name a few. Based on the properties of SlGRAS24 and the phenotypes of SlGRAS24‐OE transgenic plants, we focused on genes involved in pollen development, hormonal biosynthesis/signalling and genes encoding transcription factors (Table 4). A total of 11 genes were selected for supplementary qRT‐PCR analysis, including 3 pollen‐related genes, 5 hormone‐related genes and 3 transcription factor genes which were stamen development regulators (Fig. 8). For all the genes tested, qPCR analysis results validated the transcriptomic data.

18The miR171‐GRAS regulatory network participates in various physiological processes, including shoot meristem maintenance, axillary bud formation, flowering time, chlorophyll biosynthesis and trichome distribution. Tomato plants underexpressing SlGRAS24 did not differ from WT, while overexpression of Sly‐miR171 did cause plants to grow taller and flower earlier (Fig. 2A), providing another clue about the functional redundancy among GRAS family members or miR171 target genes. Indeed, it has been shown the existence of functional redundancy among GRAS proteins in Arabidopsis. GRAS protein SCL3 mutant line scl3‐1 does not show any phenotype when compared with WT, but it confers important role in promoting gibberellin signalling by antagonizing master growth repressor DELLA protein (Zhang et al., ). Loss function of only one of the three miR171 target genes in Arabidopsis does not have visible effects on plant growth, while transgenic plants overexpressing MIR171c (35S‐109mmproMIR171c) and scl6‐II scl6‐III scl6‐IV triple mutant plants exhibit a similar reduced shoot branching phenotype (Wang et al., ). On the other hand, overexpression of SlGRAS24 caused alteration of many important agronomical traits such as plant height, flowering time, leaf architecture, inflorescence architecture, lateral branch number, root length, fruit set and development (Table 2), making it a good target gene for generating new elite crop varieties with optimal flowering time and plant architectures, thus to meet the increasing demand for food, feed and biofuel production. Interestingly, we found that the plant height of SlGRAS24‐OE lines was severely suppressed at the early development stage, but the suppression gradually disappeared due to the growth of later branches. Overexpression of SlGRAS24 led to apical dominance inhibition and this inhibition accompanied the whole life cycle of the OE lines. At the mean time, because of apical dominance inhibition, the lateral branches showed vigorous growth, in number and length. As we know, it has long been a target of breeding selection to affect tiller and panicle/ear branching complexity because it significantly affects crop yield (Springer, ).

19The relationship between GA and GRAS proteins has been extensively studied (Sun et al., ), but some GRAS proteins have also been found to be involved in auxin signalling. For instance, an auxin‐induced GRAS protein AtSCL15 plays crucial role during seed maturation programme (Gao et al., ). The SCR/SHR complex comprised of two GRAS members has been proved to take part in root development by modulating both GA and auxin signalling (Heo et al.,; Rovere et al.,; Zhang et al., ). Interactions between GA and auxin are established by intricate crosstalk and self‐regulatory mechanisms involving the expression of auxin transport and GA metabolism genes. However, many important aspects of this relationship remain undiscovered. Plant hormones are implicated in the growth and development of shoot apical meristem (SAM) and root apical meristem (RAM) (Benková and Hejátko,; Perilli et al.,; Shani et al., ). Important roles have been attributed to GA/auxin signalling in SAM and RAM (Weiss and Ori, ). It has been documented that HAM genes are required for the maintenance of both SAM and RAM (Engstrom et al., ), suggesting that HAM genes might exert functions in SAM and RAM by modulating GA and auxin signalling. In Arabidopsis, although it is found that Atham1, 2, 3 exhibits root apex auxin maxima that are comparable to WT in spatial expression and intensity (Engstrom et al., ), it did not link AtHAM gene action directly to auxin signalling. Here, we demonstrate that tomato SlGRAS24 is a key transcription factor for coordinated regulation of GA and auxin signalling pathways, which may provide clues to the mechanisms underlying the significantly increased height of Atham1, 2, 3 mutant plants and their shorter primary roots (Engstrom et al.,; Wang et al., ). Atham1, 2, 3 mutant plants also have altered lateral branch formation (Schulze et al.,; Wang et al., ), a process which is also regulated by auxin and GA (Martínez‐Bello et al.,; Müller and Leyser,; Ni et al., ). Recently, it was found that overexpression of a sunflower (Helianthus annuus L.) GRAS‐like gene altered the GA content and axillary meristem outgrowth of transgenic Arabidopsis plants (Fambrini et al., ). Tomato plants overexpressing SlGRAS24 were dwarf with short primary roots, fewer lateral roots and more lateral branches (Fig. 2 and Table 2). Thus, it is plausible that HAM genes are regulators of endogenous GA/auxin balance in SAM, RAM and axillary meristems to control meristem maintenance and organ production.

20Although cis‐acting elements including AuxRE and GARE were found in the promoter region of SlGRAS24 gene and SlGRAS24 transcripts were up‐regulated under exogenous GA3 and IAA treatment (Fig. 5), we observed that most GA/auxin‐related genes were significantly down‐regulated in SlGRAS24‐OE tomato seedlings. This was despite these seedlings containing more endogenous IAA than WT (Fig. 6B) and raises the possibility that there is a negative feedback loop between SlGRAS24 expression and auxin metabolism. The reduced auxin responsiveness of SlGRAS24‐OE seedlings under IAA or NAA treatment (Fig. 7) implies that SlGRAS24 might function in the downstream signalling response rather than in the upstream biosynthesis. There are more than ten kinds of active GAs in plant (GA9, GA4, GA34, GA7, GA51, GA19, GA20, GA1, GA8, GA5, GA3, GA29, etc.). Although the GA3 was elevated in OE seedlings (Fig. 6B), we do believe that the total GA content is reduced since most GA synthetic genes were down‐regulated (Fig. 6A) and the observation of dwarfism phenotype and lower seed germination rate. Noticeably, the expression of SlDELLA, which acts as repressor of GA signalling, was down‐regulated (Fig. 6). In contrast to the slender and earlier flowering time phenotype exhibited by pro mutant (a point mutation in the DELLA gene) tomato plants (Carrera et al., ), dwarfism and late flowering time phenotypes were observed in SlGRAS24‐OE plants, indicating the inhibition of DELLA‐dependent GA responses. Moreover, application of GA3 to SlGRAS24‐OE plants only partially rescued the dwarf phenotype and the germination rate to the WT level (Fig. 7), suggesting that these typically GA‐related phenotypes are not merely due to the alteration of the GA signalling.

21As transcription factors, GRAS proteins have been shown to participate in various biological processes in dozens of higher plant species. However, less is known about their roles during reproductive developmental stages. In lily (Lilium longiflorum), a GRAS protein named LiSCL was found to be involved in microsporogenesis (Morohashi et al., ). In Arabidopsis, Atham1, 2, 3 mutants occasionally produced flowers with three or five petals during the very early stages of flowering (Wang et al., ), which indicated the potential active roles of miR171‐targeted GRAS genes in flower organ formation. Here, we found that SlGRAS24 played pivotal roles in late stamen development. SlGRAS24 transcripts accumulated most in flowers during anthesis, predominantly in stamens (Fig. 1A and B). Overexpression of SlGRAS24 in tomato impaired pollen sac and pollen development and as a consequence the fruit set ratio was lower than in WT (Fig. 3 and Table 2). Pollen development in flowering plants is a highly programmed process requiring many genes. In the nonexhaustive list of genes differentially regulated between SlGRAS24‐OE and WT flowers (Table 4), 15 are related to pollen development, 13 are associated with hormones and 19 are transcription factors. Most phytohormones play crucial roles in the regulation of pollen development either directly (Song et al., ) or indirectly (Dobritzsch et al.,; Ji et al., ). Our results demonstrated altered expression of genes related to all hormones except GA (Table 4). Young tomato flower buds have a high content in metabolites of GA pathways, which decrease progressively during ovary development. Active GA levels are very low in anthesis‐stage flowers during pollination, after which the total GA content within the ovary increases again (Fos et al., ). As we only compared the transcriptome of flowers undergoing anthesis, it is understandable why GA‐related gene expression did not differ between WT and SlGRAS24‐OE lines. Many transcription factors are involved in regulating pollen development in a dynamic regulatory network. In our work, the transcription factors differentially regulated by SlGRAS24 are mainly from MYB, WRKY, MADS, zinc finger and bZIP gene families. In Arabidopsis, a number of MYB proteins have been documented as important regulators of pollen development (Cheng et al.,; Higginson et al.,; Mandaokar and Browse,; Mandaokar et al.,; Yang et al., ). Two pollen‐specific transcription factors, WRKY34 and its close homolog WRKY2, are required for male gametogenesis (Guan et al., ). A subset of pollen‐specific MIKC*MADS box proteins (AGL30/65/66/94/104) are expressed preferentially during pollen maturation and double mutant combinations reveal the important roles these genes play in pollen germination and pollen fitness (Verelst et al., ). Interestingly, we found decreased expression of Solyc04g056360.2 and Solyc11g045310.1, which are the homologous gene of Arabidopsis WRKY34 (AT4G26440) and AGL30 (AT2G03060, a member of AtMIKC*MADS complexes), respectively. It has been proved that WRKY34 is one of the direct target genes of AtMIKC*MADS complexes in pollen. During pollen maturation, WRKY34 is suppressed by several MIKC*MADS box transcription factors (Verelst et al., ). Thus, we speculate that SlGRAS24 might participate in the MIKC*MADS‐WRKY34 regulatory network in tomato pollen development.

22Upon flower fertilization, fruit and seed development occurs concomitantly, orchestrated by various phytohormones (McAtee et al., ). In SlGRAS24‐OE transgenic plants, fruits were smaller with fewer seeds (Fig. 3A). It is believed that pollen quantity and/or quality are closely associated with fruit and seed set (Aizen and Harder,; Burd, ). In normal conditions, the developing seed continually sends signals to the surrounding tissue to expand and there is usually a positive correlation between seed number and fruit size (Nitsch, ). Therefore, we assumed that overexpression of SlGRAS24 led to impaired pollen sac development and pollen viability, which resulted in less efficient pollination/fertilization and hence smaller fruits with fewer seeds. Both cell division and cell elongation programmes were significantly suppressed in the smaller transgenic fruits (Fig. 4), as would be predicted by the ‘seed control’ hypothesis that the seeds communicate through hormones to the surrounding tissue(s) to promote fruit growth firstly through cell division and then cell expansion (Ozga et al., ).

23Tomato plants (Solanum lycopersicum cv. Micro‐Tom) were grown on soil in controlled glass house conditions with 14‐h light: 10‐h dark cycles, 25 °C day: 20 °C night temperature, 60% relative humidity and weekly irrigation with plant nutrient solution. For gene expression analysis, roots, stems and leaves were collected from 1‐month‐old plants, and flowers were harvested at the bud and anthesis stages and fruits at the immature, breaker and red stages. Receptacles, sepals, petals, stamens and ovaries were harvested from flowers at the anthesis stage. For each tissue/organ type, samples were collected from at least six healthy plants, mixed and then frozen in liquid nitrogen immediately. Sampling was done three independent times.

24Four DNA fragments, the SlGRAS24 promoter, the precursor of miR171, the full‐length SlGRAS24 coding sequence and a partial SlGRAS24 coding sequence were amplified from tomato genomic DNA or cDNA. Primer sequences used for amplification are listed in Table S1. The SlGRAS24 promoter sequence was fused with GUS in an expression vector. The miR171 precursor and SlGRAS24 full‐length coding sequence were cloned into the modified binary vector pLP100 in the sense orientation, while the partial SlGRAS24 coding sequence was cloned in the antisense orientation, all under the CaMV 35S promoter. Transgenic plants were generated by Agrobacterium tumefaciens‐mediated transformation according to Huang et al.. For each construct, more than 6 independent lines with consistent phenotypes were obtained. Homozygous lines from T2 or later generations were used for experiments.

25The SlGRAS24 open reading frame without the stop codon was amplified and cloned into the pGreen0029 vector. The recombinant plasmid containing the SlGRAS24‐GFP fusion gene and the control plasmid with GFP alone were transformed into tobacco (Nicotiana tabacum L.) protoplasts according to Ren et al.. For transactivation assays, the coding region of SlGRAS24 was amplified and ligated into the yeast expression vector pGBKT7 (Clontech, Japan) to produce pBD‐SlGRAS24. According to the manufacturer's instructions, pBD‐SlGRAS24, pGBKT7 (plasmid for negative control) and pGBKT7‐53+pGADT7‐T (plasmid combination for positive control) were transformed separately into the yeast strain AH109. Transformants were selected on SD/‐Trp or SD/‐Ade/‐His/‐Trp medium (Clontech, USA). The transactivation activity of each protein was evaluated by comparing growth on permissive and selective medium and the activity of X‐α‐Gal (5‐bromo‐4‐chloro‐3‐indoxyl‐α‐d‐galactopyranoside).

26GUS activity was assayed by submerging plant samples in 0.5 mg/mL X‐Gluc solution (0.1 m sodium phosphate buffer pH 7.0, 10 mm EDTA, 0.1% Triton X‐100, 0.5 mm potassium ferrocyanide, 0.5 mm potassium ferricyanide), infiltrating them under vacuum and incubating them at 37 °C. Samples were destained in 70% ethanol.

27Histological preparations were performed according to Gabe. Specific tissues/organs were embedded in FAA solution (50% (v/v) ethanol, 5% (v/v) acetic acid and 3.7% (v/v) paraformaldehyde). Samples were then placed under vacuum for 10 min, incubated at room temperature for 24 h, then dehydrated in an ethanol gradient and embedded in paraffin (Paraplast Plus, Sigma). Observations were carried out under a light microscope (OLYMPUS BX‐URA2, Japan).

28Two‐month‐old WT tomato plants were sprayed with 100 μm GA3 or 100 μm IAA. Leaves were harvested 0, 1, 3, 6, 12 and 24 h after spraying, frozen in liquid nitrogen and stored at −80 °C until RNA extraction. Meanwhile, 15‐d‐old proSlGRAS24‐GUS transgenic seedlings were soaked in liquid MS/2 medium containing 20 μm GA3 or 20 μm IAA for 3 h, and then the whole seedlings were used for GUS staining analysis or directly frozen in liquid nitrogen and stored at −80 °C. Similarly, 10‐d‐old WT and SlGRAS24‐OE transgenic seedlings were soaked in liquid MS/2 medium containing 20 μm GA3 or 20 μm IAA for 3 h, and whole seedlings were frozen in liquid nitrogen and stored at −80 °C. Seedlings soaked in liquid MS/2 medium without hormone were used as controls. All treatments were performed three independent times.

29Surface‐sterilized seeds of WT and each T2 transgenic tomato line were germinated in the dark. After the emergence of the radicle, seeds were transferred onto MS/2 medium containing 10−6 m IAA and/or 5 × 10−5 m GA3. The seedlings were grown in a controlled growth chamber for 2 weeks with a 16‐h light: 8‐h dark photoperiod and 25 °C day: 20 °C night temperature. For auxin dose–response experiments, cotyledon explants from 10‐d‐old WT and SlGRAS24‐OE seedlings were incubated on MS/2 medium containing the indicated auxin (α‐naphthalene acetic acid, NAA) concentrations in the same growth chamber conditions for 10 d.

30For seed germination assays, surface‐sterilized seeds from T2 transgenic lines and WT were germinated on MS/2 medium containing 10 μm GA3 or 5 μm paclobutrazol (PAC), a GA biosynthesis inhibitor, then placed in the same growth chamber conditions described above. Seeds germinated on MS/2 medium were used as controls. Germination rates based on radicle tip emergence were scored daily for 7 d after sowing. GA3 was also applied by spraying the aerial part of 10‐d‐old plants grown in the glasshouse with 20 μm GA3 solution containing 0.1% Tween‐80 every other day for 3 weeks. Control plants were sprayed with the equivalent solvent solution.

31All experiments were repeated three times with about 30 seeds or 10 plants for each genotype.

32Pure IAA and GA3 were purchased from Sigma Chemical Co. (St Louis, MO). Isotopically labelled internal standards [2H5]IAA and [2H2]GA3 were purchased from ICON Isotopes (Summit, NJ). Plant hormones were extracted and quantified as previously described (Pan et al., ) with some modifications. Approximately 0.2 g of 10‐d‐old WT or SlGRAS24‐OE seedlings was frozen in liquid nitrogen and ground into powder. Two millilitres of isopropanol–HCl buffer solution (2:0.002, v/v) was added to the powder and shaken for 30 min at 4 °C. Dichloromethane (4 mL) was added, shaken for 30 min and then centrifuged at 13 000 g for 5 min. After centrifugation, the lower organic phase was transferred to a 10‐mL tube and evaporated in a constant stream of nitrogen. Each sample was kept in the dark and resolubilized in 150 μL methanol containing 0.1% formic acid, and then the solution was filtered using a 0.45 μm microfilter for HPLC‐MS/MS analysis. Samples were injected into a reversed‐phase column (C18 ZORBAX 300SB 3 μm, 4.6 × 150 mm, Agilent, CA) using a binary solvent system composed of methanol (solvent A) and water with 0.1% formic acid (solvent B) as the mobile phase. The column thermostat was set at 30 °C. Separations were performed using a solvent gradient which started at 20% methanol for 2 min, increased linearly to 80% over 14 min and maintained for 5 min, and returned to 20% methanol for 0.1 min, when it was allowed to equilibrate for 5 min. A hybrid triple quadrupole/linear ion trap mass spectrometer (SCIEX 6500 QTrap, Applied Biosystems, Foster City, CA) was used with nebulizer gas pressure set at 75 psi, drying gas pressure at 65 psi, curtain gas pressure at 15 psi, source voltage at 4.5 kV and source temperature at 500 °C.

33Pollen activity was evaluated by soaking pollen grains in 0.1% 2, 3, 5‐triphenyl‐2 h‐tetrazolium chloride (TTC) solution. Viable pollen stains red because the NADH/NADPH produced reduces TTC to 1,3,5‐triphenylformazan (TPF), which is red. Stained pollen grains were observed under the microscope.

34Total RNA was extracted from anthesis‐stage flowers of SlGRAS24‐OE plants (lines 10 and 15) and WT controls using RNeasy® Plant Mini Kit (Qiagen) following the manufacturer's protocol for RNA‐Seq. RNA quantity and quality were assayed in the Agilent 2100 Bioanalyzer (Agilent Technologies). Two independent RNA samples from transgenic or WT plants were sent to Illumina Cluster Station and Illumina HiSeq™ 2000 System (BGI Inc.) for RNA library construction and deep sequencing. RSeQC‐2.3.2 program (http://code.google.com/p/rseqc/) was used to assess the quality of RNA‐Seq data. Clean tags were obtained after quality filtering of sequences and mapped to the annotated genome sequence of S. lycopersicum in the Tomato Sol Genomic Network database (http://solgenomics.net/), and transcript abundance was also normalized by the fragments per kilobase of exon per million mapped reads (FRKM) method using Cuffdiff software (http://cufflinks.cbcb.umd.edu/) to identify differentially expressed genes (DEGs). The expression level of a gene from RNA‐Seq was normalized by the tags per million method. The criteria for defining differentially expressed genes were a false discovery rate (FDR) <0.05 with a P value <0.05. The raw transcriptome reads reported here have been deposited in the NCBI Short Read Archive under accession no. SRA473616.

35One microgram of total RNA (RNeasy® Plant Mini Kit, Qiagen) was used to synthesize first‐strand cDNA (RevertAid™ First Strand cDNA Synthesis Kit, Fermentas). Transcript levels were determined by absolute qPCR according to the methodology described in Huang et al. using specific primers. Primer sequences used for qRT‐PCR are listed in Table S2. Amounts of mRNA in samples were quantified using three biological replicates.

36All data in this report were obtained from at least three independent experiments with three technical replicates each. For data analysed with Student's t‐test, the differences between treatments were considered as significant when P < 0.05.