The gibberellin biosynthetic genes AtGA20ox1 and AtGA20ox2 act, partially redundantly, to promote growth and development throughout the Arabidopsis life cycle

Authors

  • Ivo Rieu,

    1. Plant Science Department, Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UK
    Search for more papers by this author
    • Present address: Laboratory of Molecular Genetics and Biotechnology of Plants, University of Freiburg, 79 104 Freiburg, Germany.

  • Omar Ruiz-Rivero,

    1. Plant Science Department, Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UK
    Search for more papers by this author
    • Present address: Instituto de Biología Molecular y Celular de Plantas (IBMCP), Universidad Politécnica de Valencia-CSIC, Avda de los Naranjos s/n, 46022 Valencia, Spain.

  • Nieves Fernandez-Garcia,

    1. Plant Science Department, Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UK
    Search for more papers by this author
    • §

      Present address: CEBAS-CSIC, PO Box 164, 30100 Espinardo, Murcia, Spain.

  • Jayne Griffiths,

    1. Plant Science Department, Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UK
    Search for more papers by this author
    • Present address: CSIRO Plant Industry, Canberra, ACT 2601, Australia.

  • Stephen J. Powers,

    1. Biomathematics and Bioinformatics Department, Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UK
    Search for more papers by this author
  • Fan Gong,

    1. Plant Science Department, Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UK
    Search for more papers by this author
  • Terezie Linhartova,

    1. Plant Science Department, Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UK
    Search for more papers by this author
    • **

      Present address: Palacky University & Institute of Experimental Botany, ASCR Slechtitelu 11, 783 71 Olomouc, Czech Republic.

  • Sven Eriksson,

    1. Umeå Plant Science Centre, Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, S-90183 Umeå, Sweden
    Search for more papers by this author
  • Ove Nilsson,

    1. Umeå Plant Science Centre, Department of Forest Genetics and Plant Physiology, Swedish University of Agricultural Sciences, S-90183 Umeå, Sweden
    Search for more papers by this author
  • Stephen G. Thomas,

    1. Plant Science Department, Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UK
    Search for more papers by this author
  • Andrew L. Phillips,

    1. Plant Science Department, Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UK
    Search for more papers by this author
  • Peter Hedden

    Corresponding author
    1. Plant Science Department, Rothamsted Research, Harpenden, Hertfordshire AL5 2JQ, UK
      (fax +44 1582 763010; e-mail peter.hedden@bbsrc.ac.uk).
    Search for more papers by this author

(fax +44 1582 763010; e-mail peter.hedden@bbsrc.ac.uk).

Summary

The activity of the gibberellin (GA) biosynthetic enzymes GA 20-oxidases (GA20ox) is of particular importance in determining GA concentration in many plant species. In Arabidopsis these enzymes are encoded by a family of five genes: AtGA20ox1AtGA20ox5. Transcript analysis indicated that they have different expression patterns and may thus participate differentially in GA-regulated developmental processes. We have used reverse genetics to determine the physiological roles of AtGA20ox1 and AtGA20ox2, the most highly expressed GA20ox genes during vegetative and early reproductive development. AtGA20ox1 and AtGA20ox2 act redundantly to promote hypocotyl and internode elongation, flowering time, elongation of anther filaments, the number of seeds that develop per silique and elongation of siliques, with AtGA20ox1 making the greater contribution to internode and filament elongation, and AtGA20ox2 making the greater contribution to flowering time and silique length. Pollination of the double mutant with wild-type pollen indicated that the GA promoting silique elongation is of maternal origin. The ga20ox2 phenotype revealed that GA promotes the number of stem internodes that elongate upon bolting, and does so independently of its effect on internode elongation. Comparison of the phenotype of the double mutant with that of the highly GA-deficient ga1-3 mutant indicates that other GA20ox genes contribute to all the developmental processes examined, and, in some cases such as root growth and leaf expansion, make major contributions, as these processes were unaffected in the double mutant. In addition, the effects of the mutations are mitigated by the homeostatic mechanism that acts on expression of other GA dioxygenase and GID1 receptor genes.

Introduction

Studies of mutants with lesions in the biosynthesis or signal transduction pathways of the gibberellin (GA) plant hormones have revealed the wide range of physiological processes in which these compounds participate (Ross et al., 1997). These processes include elongation or expansion of numerous tissues, including roots, hypocotyls, leaves, stems, stamens and pistils. Furthermore, GAs may induce certain developmental switches (Baurle and Dean, 2006), such as between the juvenile and adult phases (Telfer et al., 1997) or between vegetative and reproductive development (Eriksson et al., 2006). Gibberellin signalling responds to environmental cues, including changes in light conditions, temperature or stress, allowing GAs to translate these extrinsic signals into developmental changes.

Within the GA signalling pathway, metabolism of the hormone is particularly sensitive to developmental and environmental factors. The biosynthetic pathway is well characterized, particularly in Arabidopsis (Hedden and Phillips, 2000), for which it has been shown that GA4 is the primary bioactive GA for growth processes (Sponsel et al., 1997) and for flower induction (Eriksson et al., 2006). The production of this compound from the common diterpene precursor trans-geranylgeranyl diphosphate requires the sequential activities of six enzymes, comprising terpene synthases, cytochrome P450 monooxygenases and 2-oxoglutarate-dependent dioxygenases (Hedden and Phillips, 2000). This last group includes the GA 20-oxidases (GA20ox), which remove C-20 to form the C19 structure, and the GA 3-oxidases (GA3ox), which, by insertion of a 3β-hydroxyl group, catalyze the final step in the formation of the biologically active product. A third class of dioxygenases, the GA 2-oxidases (GA2ox), insert a 2β-hydroxyl into GA4 or a precursor forming an inactive product, or one that cannot be converted to an active form, and thereby serve to reduce the concentration of the bioactive hormones.

The dioxygenases are important sites of regulation in GA biosynthesis (Hedden and Phillips, 2000; Yamaguchi and Kamiya, 2000). They are encoded by gene families, the members of which have distinct expression patterns and may thus be involved in providing GA for different developmental processes. Expression of the GA dioxygenase genes is regulated by a number of intrinsic and environmental factors. Importantly, certain members of the GA20ox and GA3ox gene families are downregulated by GA signalling, whereas some GA2ox genes are upregulated by GA (Chiang et al., 1995; Phillips et al., 1995; Thomas et al., 1999). Thus there is buffering of GA concentration, although large changes in GA content can occur in response to developmental and environmental signals. For example, red light and exposure to cold induce GA3ox expression and suppress GA2ox expression in imbibed seeds of Arabidopsis, in which germination is stimulated by the resulting increase in GA production (Oh et al., 2006; Yamaguchi et al., 1998, 2004). Higher levels of GA20ox expression in light contributes to the induction of flowering by long days in plants such as Arabidopsis (Xu et al., 1997) and spinach (Wu et al., 1996).

Overexpression experiments show that GA20ox activity limits GA biosynthesis in Arabidopsis (Coles et al., 1999; Huang et al., 1998), indicating that these enzymes are likely to be particularly important for regulating the pathway. On the basis of sequence homology, the GA20ox gene family in Arabidopsis has five members (Hedden et al., 2002), three of which, GA20ox1, GA20ox2 and GA20ox3, have been shown in vitro to catalyze all steps in the conversion of the C20 intermediate GA12 to GA9, the immediate precursor of GA4 (Phillips et al., 1995), whereas GA20ox4 and GA20ox5 also possess GA20ox activity, catalyzing at least the first step in the reaction sequence (Pérez-Gómez, 2003). Northern blot analysis showed differential expression of GA20ox1, GA20ox2 and GA20ox3:, with GA20ox1 being the only GA20ox gene to be expressed in stems, GA20ox2 expressed in flowers and siliques, and GA20ox3 expressed only in siliques (Phillips et al., 1995). Consistent with this expression pattern, ga5, which contains a loss-of-function mutation in GA20ox1 (Xu et al., 1995), has reduced stem height, but otherwise appears to be developmentally normal (Koornneef and van der Veen, 1980). This relatively mild GA-deficient phenotype, compared for example with that of the highly GA-deficient ga1-3 mutant, indicates the involvement of one or more of the other GA20ox paralogues in all phases of development. On the basis of their differential expression it might be anticipated that the GA20ox gene family members will differ in their relative contribution to GA biosynthesis in different plant organs (Phillips et al., 1995). Combining detailed phenotypic analysis of wild-type plants and those with mutations in different gene family members with high-resolution data on the distribution of gene expression and GA content would be an effective way to determine the spatiotemporal relationship between GA synthesis and physiological function.

We have taken a reverse-genetics approach to dissect the contributions to development of the two GA20ox genes, GA20ox1 and GA20ox2, which were reported to be the most highly expressed during vegetative and early reproductive development (Phillips et al., 1995). We describe a detailed developmental analysis of loss-of-function mutants for these two genes and of a double mutant lacking both enzymes. As well as shedding light on the different physiological roles of GA20ox1 and GA20ox2, the study produced some novel findings on GA function: for example, GA regulates the number of stem internodes that elongate upon bolting, and this role of GA can be genetically separated from its role in promoting internode elongation. Furthermore, the endogenous GA content is limiting for the number of elongated vegetative and reproductive internodes, whereas it is saturating for internode length.

Results

Expression profiles of the GA20ox gene family members

Functional divergence between the five GA20ox gene family members may arise from differences in their expression patterns. To study their expression in different tissues, and at different developmental stages, we monitored transcript levels using real-time quantitative RT-PCR. Absolute transcript levels were determined, allowing for comparison between samples and between genes (Figure 1). In general GA20ox1, GA20ox2 and GA20ox3 were the more highly expressed genes. GA20ox3 was the most highly expressed gene in dry and imbibed seeds, whereas GA20ox2 was the most abundant in 3- and 7-day-old seedlings. In the vegetative shoot of 11- and 24-day-old plants, GA20ox1 was the most highly expressed family member, especially in the stem where its transcript was over 1000 times more abundant than that of any other family member. GA20ox1 and GA20ox2 constituted the majority of GA20ox transcripts in the inflorescence, whereas GA20ox2 and especially GA20ox3 are the dominant genes expressed in siliques. GA20ox4 transcripts were mainly detected in root and GA20ox5 transcripts in 3-day-old seedlings and in siliques, but both at relatively low levels. To confirm our results, we also analysed the expression of the GA20ox genes as reported by the AtGenExpress consortium using microarrays (Schmid et al., 2005; http://www.weigelworld.org/resources/microarray/AtGenExpress). These data confirmed that, compared with other tissues, the GA20ox1 mRNA level was relatively high in stem tissues, and indicated high levels of expression of this gene in stamens of stage-15 flowers, which could account for the slightly elevated abundance in our inflorescence samples. GA20ox2 was most highly expressed in flower tissues, especially carpels, in young siliques (containing developing seeds) and in isolated developing seeds, whereas GA20ox3 transcript was mainly detected in siliques and developing seeds, consistent with our results. Signals for GA20ox4 and GA20ox5 were very low, and were slightly elevated only in mature pollen and developing seeds, but not in roots, as we found for GA20ox4, and in seedlings, as we found for GA20ox5. Although the AtGenExpress data are not absolutely quantitative between genes, the observation that GA20ox1 and GA20ox2 signals are highest in most tissues, except for the silique and seed samples where the GA20ox3 signal was highest, and that GA20ox4 and GA20ox5 signals are generally very low, strengthens our absolute quantification results. In conclusion, the expression profiles show that GA20ox1, GA20ox2 and GA20ox3 are the most highly expressed of the known GA20ox genes in Arabidopsis, and that expression of none of the GA20ox genes is completely specific for particular tissues or developmental stages. Based on these results, we would expect some functional redundancy between the GA20ox genes in many developmental processes.

Figure 1.

 Developmental expression profile of the Arabidopsis GA20ox genes.
mRNA levels of the five GA20ox genes were determined using real-time quantitative RT-PCR on three biological replicates for each tissue. Standard curves were used to calculate the numbers of GA20ox cDNA molecules in each cDNA sample, which were then normalized against three reference genes (see Experimental procedures). Results are plotted as the ratio to the lowest detected level (i.e. GA20ox2 in stem) ± SE. Note that the y-axis is in logarithmic scale. xNot detected.

GA20ox1 and GA20ox2 knockout mutants in Col-0

To reveal the physiological functions of GA20ox1 and GA20ox2, we obtained insertion lines in which these genes were disrupted. Sequence analysis of the left-border flanking sequence tag showed that in SALK line 016701 a T-DNA was inserted in intron 1; this line was named ga20ox1-3 (Figure 2a; also see Hisamatsu et al., 2005). GABI-KAT line 734G06 has a transposon insertion in intron 2 of GA20ox2: this line was named ga20ox2-1. RT-PCR with primers flanking the disrupted introns showed that no spliced transcript was produced from these alleles (Figure 2b; Hisamatsu et al., 2005). Because the affected introns are located inside regions of the sequences that code for the catalytic domain of the 2-oxoglutarate-dependent dioxygenases, to which the GA 20-oxidases belong, the ga20ox1-3 (hereinafter referred to as ga20ox1) and ga20ox2-1 (referred to as ga20ox2) alleles are likely to encode non-functional proteins. To study possible redundant functions of GA20ox1 and GA20ox2, the ga20ox1 ga20ox2 double mutant was generated. Figure 2(c) shows that the mutations have different effects on overall plant growth: the ga20ox1 line has a semidwarf phenotype, whereas ga20ox2 is only slightly smaller than the wild type. The ga20ox1 ga20ox2 double mutant is more severely dwarfed than ga20ox1, but not as much as the GA-deficient mutant ga1-3(Col). A semidwarf phenotype has also been reported for ga20ox1-1 (or ga5), which is in the Ler background (Koornneef and van der Veen, 1980; Xu et al., 1995).

Figure 2.

 Molecular description and phenotype of ga20ox1 and ga20ox2 knock-out mutants.
(a) Gene model of GA20ox1 and GA20ox2 indicating the sites of mutations and insertions of mutant alleles. Numbers between brackets indicate the relative positions of the features in nucleotides.
(b) RT-PCR analysis to determine the level of knockdown of the GA20ox1 and GA20ox2 gene expression in various tissues of ga20ox1-3 and ga20ox2-1, respectively. PCR primers were designed to amplify fragments spanning the intron carrying the insertion. Amplification was monitored using SYBR-green in a real-time quantitative PCR system, and the cycle at which the fluorescence passed the threshold (Ct; normalized with reference genes) is indicated. –, not detected. Identical results were obtained using RT-PCR with a second set of primers and analysis of products on a gel (data not shown).
(c) Phenotype of ga20ox mutants compared with Col-0 and ga1-3(Col). Plants, grown in soil for 8 weeks, show the increasingly dwarfed statures of ga20ox2, ga20ox1, ga20ox1 ga20ox2 and ga1-3(Col).
(d) Rosettes of plants grown on soil for 14 days, in lateral view, showing decreased leaf hyponasty in ga20ox1 ga20ox2 and ga1-3(Col).

GA20ox1 and GA20ox2 and seed germination

GAs have an important role in seed germination (for a review, see e.g. Kucera et al., 2005; Peng and Harberd, 2002). Germination of seeds is accompanied by de novo GA biosynthesis, which is stimulated either by light or by cold applied during seed imbibition (stratification) (Derkx et al., 1994; Shinomura et al., 1994; Toyomasu et al., 1998; Yamaguchi et al., 1998, 2004). In Arabidopsis, exposure of imbibed seeds to cold results in increased expression of GA20ox1 and GA20ox2 (Yamauchi et al., 2004). To test whether GA20ox1 and GA20ox2 are necessary for light and/or cold stimulation of germination, imbibed seeds of wild-type and the ga20ox1 and ga20ox2 single and double mutants were incubated in the dark as a control, or were treated with cold and/or a light pulse and incubated in the dark. As expected, the germination rate of wild-type seeds was very low in the dark (Figure 3). Pre-treatment with cold did not enhance germination of the wild-type in our experiments, whereas the light pulse increased germination to 50% after 4 days. Cold and light had an additive effect on germination rate, and more than 90% of wild-type seeds receiving this combined treatment germinated. Germination of the ga20ox1 and ga20ox2 single mutants was indistinguishable from the wild type. Germination of ga20ox1 ga20ox2, however, was significantly more efficient than that of the wild type. Already 5% of the seeds of this genotype germinated in the dark, and cold treatment increased this to 30%. Treatment with a light pulse resulted in almost 100% germination, as did the combined treatment with cold and light. Analysis of germination curves in continuous light also showed that ga20ox1 ga20ox2 seeds germinated more readily than wild-type seeds, and this could be enhanced by a cold pre-treatment (Figure S1).

Figure 3.

 Germination characteristics of the wild type and ga20ox1 and ga20ox2 single and double-mutant seeds.
Seeds were germinated in the dark for 4 days, without or with a pre-treatment of 4 days of imbibition at 4°C and/or a 1-min pulse of white light. Values represent the average of six independent seed batches ± SE. Logit-transformed values were used for statistical analysis (see Experimental procedures) and are given in Table S1. *Significantly different from the wild type with the same treatment (P < 0.01).

GA20ox1 and GA20ox2 play a minor role in elongation of seedling organs

GAs are known to be major stimulators of elongation growth. To analyse the role of the GA20ox1 and GA20ox2 genes in this process, we determined the effect of the ga20ox1 and ga20ox2 mutations on the growth of different organs. Table 1 shows that hypocotyl growth was reduced slightly in ga20ox2, as compared with the wild type. The double mutant was significantly affected, but still much less than ga1-3(Col). To ensure that all developmental abnormalities in the ga20ox mutants were the result of mutations in the GA20ox genes, measurements were also taken from plants treated with GA3. Table S2 shows that the hypocotyl phenotype and other phenotypes described in Table 1 were completely, or almost completely, reversible by GA treatment. Root length and the final diameter of the rosette, although strongly reduced in ga1-3(Col), were not affected in the single and double ga20ox mutants (Table 1). However, Figure 2(d) shows that rosette leaves of the double mutant were severely epinastic.

Table 1.   Phenotypic characterization of the ga20ox mutants
GenotypeHypocotyl length (mm)Rosette radius (mm)Root length (mm)No. branches
wild-type2.33 ± 0.05 (0.83)a53.9 ± 0.9 (3.98)a56.0 ± 0.88.2 ± 0.2
ga20ox12.39 ± 0.03 (0.87)52.4 ± 0.8 (3.96)56.2 ± 0.510.1 ± 0.4*
ga20ox22.10 ± 0.04 (0.74)53.2 ± 0.9 (3.97)54.4 ± 0.67.9 ± 0.3
ga20ox1 ga20ox21.68 ± 0.04 (0.51)*53.4 ± 1.0 (3.97)59.4 ± 0.711.9 ± 0.4*
ga1-3(Col)0.70 ± 0.01 (−0.36)*22.8 ± 0.6 (2.84)*28.5 ± 0.5*n.d.
LSD(1%) (d.f.)0.162a (10) 0.069a (32) 4.34 (30) 1.17 (30)
 Vegetative internode length (cm)No. vegetative internodesInflorescence internode length (cm)No. inflorescence internodesFinal plant height (cm)
  1. See Experimental procedures for details on design. The measurements are the means ± SE.

  2. aLog-transformed values (shown in parentheses) were used for statistical analysis (see Experimental procedures), and the least significant difference (LSD)(1%) corresponds to these values.

  3. *Significantly different from the wild type (P < 0.01); d.f., degrees of freedom; n.d., not determined; n.a., not available.

wild-type4.2 ± 0.32.4 ± 0.10.82 ± 0.0251.9 ± 0.852.5 ± 0.6
ga20ox11.2 ± 0.1*2.7 ± 0.20.52 ± 0.01*44.0 ± 0.7*26.0 ± 0.4*
ga20ox24.3 ± 0.51.4 ± 0.1*1.02 ± 0.03*40.2 ± 1.0*46.4 ± 0.7*
ga20ox1 ga20ox20.9 ± 0.1*0.9 ± 0.2*0.35 ± 0.01*40.8 ± 0.9*15.1 ± 0.4*
ga1-3(Col)n.a.n.a.n.a.n.a.n.a.
LSD(1%) (d.f.)0.92 (29)0.64 (30)0.058 (30) 3.36 (30) 2.23 (30)

GA20ox1 and GA20ox2 promote floral transition

In Arabidopsis, GA stimulates the transition from a vegetative to a reproductive growth habit, with the effect being particularly important when plants are grown in short days, when there is an absolute requirement for GA (Wilson et al., 1992). The contributions of GA20ox1 and GA20ox2 to the production of the GA that affects flowering time were investigated by counting the number of leaves produced before the floral transition in the different genotypes (Figure 4). Under long-day conditions, flowering was slightly, but significantly, delayed in ga20ox2, and was more severely delayed in the ga20ox1 ga20ox2 double mutant. The ga1-3(Col) mutant flowered even later, the flowering time being double that of the wild type in terms of the number of leaves at flower appearance. GA treatment restored the flowering time of each genotype to that of the wild type. Similar results were obtained when taking the number of days to floral bud appearance, rather than leaf number, as a measure of flowering time (Figure S2). Under short-day conditions, both the ga20ox1 and ga20ox2 mutations delayed flowering. In contrast to ga1-3, the ga20ox1 ga20ox2 mutant still flowered, albeit much later than the wild type and the single mutants (Figure 4).

Figure 4.

 Flowering time of the ga20ox mutants.
Time to flowering was determined as the total number of leaves in the primary shoot. Values shown are the means ± SE. *Significantly different from the wild type within the same treatment (< 0.01); **significantly different between treatments (within a genotype) (P < 0.01).
(a) Plants grown under long-day conditions, untreated or treated with 100 μm GA3. Least significant difference, LSD(1%) = 1.1 (30 d.f.), 1.7 (40 d.f.) and 1.4 (40 d.f.) for within −GA, within +GA, and between −GA and +GA treatments, respectively.
(b) Plants grown under short-day (SD) conditions. n.a., not available. ga1-3 did not flower in SD during the course of the experiment (20 weeks). LSD(1%) = 9.6 (20 d.f.).

GA20ox1 and GA20ox2 have distinct roles in stem elongation

Upon flower induction, a number of vegetative stem internodes elongate (i.e. bolting). The elongated vegetative internodes of ga20ox1, and more so of ga20ox1 ga20ox2, were significantly shorter than those of the wild type, whereas those of ga20ox2 were of normal length (Table 1; Figure 2c). The same was true for the internodes in the main inflorescence (i.e. the part of the plant above the uppermost cauline leaf on the primary stem), with the exception that in this organ, the ga20ox2 mutation resulted in slightly longer internodes. Final plant height depends not only on the length, but also on the number of elongated internodes. The number of vegetative stem internodes that elongate upon bolting was significantly reduced in ga20ox2 and in the ga20ox1 ga20ox2 double mutant. In Arabidopsis, all internodes in the inflorescence elongate to some extent, with the number of elongated internodes equalling the number of flowers produced. This number was reduced in both single mutants and in the double mutant. Together, these differences in the number and length of vegetative and inflorescence internodes resulted in a slight reduction in final plant height of ga20ox2, and in the more severely dwarfed statures of ga20ox1 and ga20ox1 ga20ox2 (Table 1, Figure 2c). Treatment with GA increased stem height in all lines, including the wild type: the increased height of the wild type resulted from a greater number of elongated vegetative internodes, which were almost doubled in the GA-treated wild type relative to untreated plants (Table S2). Although vegetative internode length of ga20ox1 and ga20ox1 ga20ox2 was restored to that of Col-0 by GA treatment, there was no further increase in that of Col-0 and ga20ox2. GA treatment increased the number of inflorescence internodes in all lines, resulting in an increase in the height of the inflorescence in ga20ox1 and ga20ox1 ga20ox2 to that of the wild type, but there was no change in the height of the Col-0 and ga20ox2 inflorescence, as the increase in internode number was balanced by a decrease in internode length in both cases.

ga20ox1 and ga20ox1ga20ox2 have more stem branches

A characteristic phenotype of severely GA-deficient plants is their bushy appearance, which is caused by an increase in the number of axillary stem branches (Silverstone et al., 1997). To assess the roles of GA20ox1 and GA20ox2 in this process we counted the total number of branches originating from the basal part (i.e. at the level of the rosette) and the elongated part of the main stem. Table 1 shows that the ga20ox1 mutation resulted in a slight increase in the number of branches, whereas this phenotype was exaggerated by the addition of the ga20ox2 mutation. Again, GA treatment rescued these phenotypes, such that there were no significant differences from the wild type (Table S2).

GA20ox1 and GA20ox2 promote fertility and silique elongation

A striking phenotype of the ga20ox1 ga20ox2 mutant is reduced fertility of the first flowers on the main stem. Whereas sterility among the first ten flowers that are produced on the main stem was rare in Col-0, and even more so in ga20ox2, it was more prevalent in ga20ox1 and especially in ga20ox1 ga20ox2 (Figure 5a). To examine the cause of the reduced fertility of the ga20ox1 ga20ox2 mutant, we visually analysed the first flowers of the different genotypes. Figure 5(b) shows that the first flowers on the main stem of ga20ox1 ga20ox2 had short filaments compared with the wild type, and undehisced anthers, suggesting that the reduced fertility in these flowers resulted from reduced male fertility. This was confirmed by the observation that hand-pollination of these early flowers with ga20ox1 ga20ox2 pollen (from older plants) resulted in seed set (Figure S3).

Figure 5.

 Silique phenotype of the ga20ox mutants.
Analysis of various silique characteristics. *Significantly different from the wild type within the same treatment (< 0.01); **significantly different between treatments (within a genotype) (< 0.01). Gibberellin (GA) treatment was carried out with 100 μm GA3.
(a) Infertility within the first 10 flower buds on the main stem. Log-transformed values were used for statistical analysis (see Experimental procedures) and are indicated in parentheses above the bars. Least significant differences (LSDs) correspond to these values: LSD(1%) = 0.36 (30 d.f.), 0.51 (30 d.f.) and 0.43 (40 d.f.) for within −GA, within +GA, and between −GA and +GA treatments, respectively.
(b) Photo of dissected early flower buds of wild type and ga20ox1 ga20ox2 plants.
(c) Average silique length. LSD(1%) = 0.46 (30 d.f.), 0.65 (30 d.f.) and 0.55 (40 d.f.) for within −GA, within +GA, and between −GA and +GA treatments, respectively.
(d) Average number of seeds per silique. LSD(1%) = 5.5 (30 d.f.), 7.7 (30 d.f.) and 6.6 (40 d.f.) for within −GA, within +GA, and between −GA and +GA treatments, respectively.
(e) Number of seeds and length of individual siliques. The shaded area marks siliques that were used to compare the length of ga20ox1 ga20ox2 siliques with those of wild-type and ga20ox2, as indicated in the text.
(f) Length of wild-type and ga20ox1 ga20ox2 siliques (in mm) after reciprocal hand-pollinations.

Siliques elongate rapidly following fertilization. Measurement of final silique length showed that siliques had elongated less in the ga20ox2 single mutant, whereas an effect of the ga20ox1 mutation was only apparent in combination with the ga20ox2 mutation (Figure 5c). In Arabidopsis, a strong correlation exists between seed number and silique length (Cox and Swain, 2006). To see whether the reduction in silique length was the result of a reduction in the number of seeds per silique, we counted the number of seeds in the wild-type and mutant plants. Figure 5(d) shows that the number of seeds per silique was not reduced in ga20ox2, suggesting a seed number-independent effect of GA20ox2 on silique elongation. The average number of seeds per silique was reduced in ga20ox1 ga20ox2, providing a possible cause for the further reduced silique length in this genotype. However, Figure 5(e), in which seed number versus silique length of individual siliques is plotted, shows that some ga20ox1 ga20ox2 siliques had similar numbers of seeds as some wild-type siliques, but were still significantly shorter [when considering siliques with 60–70 seeds: ga20ox1 ga20ox2, 9.9 ± 0.2 mm (SE); wild type, 17.9 ± 0.2 mm; least significant difference (1%) = 0.81 at 12d.f.), confirming the seed-number independent effect that we had already noted in the ga20ox2 mutant. These ga20ox1 ga20ox2 siliques with ‘normal’ numbers of seeds were also significantly shorter than the ga20ox2 siliques (ga20ox2, 15.2 ± 0.4 mm), revealing that GA20ox1 participates in the seed number-independent effect, at least in the absence of GA20ox2.

Because GA20ox2 is highly expressed in developing seeds, we hypothesized that GA20ox2 and possibly GA20ox1 function in the seed to control silique elongation. To test this, we hand-pollinated ga20ox1 ga20ox2 with its own pollen or with wild-type pollen. Pollination with wild-type pollen should result in ga20ox1 ga20ox2 siliques containing seeds with embryos and endosperm of wild-type phenotype (heterozygous for ga20ox1 and ga20ox2). Figure 5(f) shows that pollination with wild-type pollen did not rescue the silique length of ga20ox1 ga20ox2, as compared with pollination by its own pollen. Furthermore, wild-type pollen was not able to rescue the number of seeds per silique [ga20ox1 ga20ox2 × ga20ox1 ga20ox2, 51.6 ± 1.7 (SE); ga20ox1 ga20ox2 × wild type, 47.5 ± 3.0; wild type × wild type, 64.0 ± 1.7; also compare with Figure 5d).

Together these data show that GA20ox1 and GA20ox2 act redundantly to promote silique elongation: (i) indirectly, by positively affecting the number of seeds that develop in a silique (probably by promoting male fertility), and (ii) directly, by a mechanism that is independent of seed number and to which GA20ox2 makes the larger contribution.

Feedback regulation of GA20ox gene expression in ga20ox1 and ga20ox2 single and double mutants

Several GA biosynthetic genes have been shown to be downregulated by GA, constituting a negative feedback loop of GA on its own biosynthesis. This type of regulation could partially compensate for a loss of GA20ox1 or GA20ox2 expression, and thus ameliorate the effects of the ga20ox1 and ga20ox2 single and double mutations. In order to determine which members of the GA20ox gene family show this type of regulation, we used quantitative RT-PCR to monitor the effect of GA on the expression of each GA20ox gene in the severely GA-deficient ga1-3 mutant. As the DELLA proteins GAI and RGA are the main repressors of GA-promoted vegetative growth (Dill and Sun, 2001; King et al., 2001), we also monitored the effect of the gai-t6 and rga-24 loss-of-function mutations on GA20ox gene expression in the ga1-3 background. As shown in Figure 6(a), GA20ox1, GA20ox2 and GA20ox3 expression in liquid-culture grown seedlings was strongly downregulated by GA, confirming earlier results by Phillips et al. (1995). In the presence of the gai-t6 and rga-24 mutations, expression of these GA20ox genes was significantly reduced, and GA20ox2 expression in ga1-3 gai-t6 rga-24 was still slightly responsive to GA. Expression of GA20ox4 and GA20ox5 was not regulated by GA treatment or by DELLA protein. To examine whether the GA20ox genes are primary GA-responsive genes, we tested their GA responsiveness in a GA-deficient line (a GA2ox-overexpressing line; see Experimental procedures) in the presence of a protein synthesis inhibitor, cycloheximide (CHX). Figure 6(b) shows that the expression of GA20ox2 and GA20ox4 was significantly increased 3 h after CHX treatment, and that GA did not downregulate GA20ox1, GA20ox2 and GA20ox3 transcript levels in CHX-treated plants. This might suggest that GA20ox1, GA20ox2 and GA20ox3 are not primary responsive genes.

Figure 6.

 Feedback regulation and the effect of the ga20ox mutations on expression of gibberellin (GA) biosynthesis and receptor genes.
Relative mRNA levels were determined using real-time quantitative RT-PCR. The relative mRNA quantity in each cDNA sample was normalized against three reference genes (see Experimental procedures). Values represent the average of three RNA replicates ± SE.
(a) Expression of the five GA20ox genes in ga1-3 upon treatment with 2 μm GA4 and in ga1-3 rga-24 gai-t6. Results are plotted as the ratio to the level at 0 h. *Significantly different from both the 0-h and the 2-h control samples (within gene and genotype). **gal-3 rag-24 gai-t6 0h is significantly different from gal-3 0h (P < 0.01). See Table S3 for mean threshold cycle (Ct) values and least significant differences (LSDs).
(b) Expression of the five GA20ox genes in GA2oxOE upon treatment with 2 μm GA4 and/or 100 μm cycloheximide. Results are plotted as the ratio to the value at 0 h. *Significantly different from the 0-h and 1-h EtOH controls (P < 0.01). See Table S4 for mean Ct values and LSDs.
(c) Expression of GA20ox1, GA20ox2 and GA20ox3 and (d) GA3ox1, GA2ox1 and GID1b in the ga20ox mutants. Results are plotted as the ratio to the level in the same tissue of the wild type. *Significantly different from the same tissue in the wild type; n.d., not determined; xnot detected; #no comparison with wild type possible. See Table S5 for mean Ct values and LSDs.

Because expression of GA20ox1, GA20ox2 and GA20ox3 was negatively regulated by GA, we tested whether expression of these three genes was upregulated in different tissues of the ga20ox1 and ga20ox2 single and double mutants. As shown in Figure 6(c), GA20ox1 expression in ga20ox2 was not increased with respect to that in the wild type, whereas GA20ox2 expression was strongly upregulated in the stem internodes of ga20ox1 plants. GA20ox3 was slightly upregulated in ga20ox1, but much more in the ga20ox1 ga20ox2 double mutant. As expected, expression of GA20ox4 and GA20ox5 was very low, and no differences in expression could be detected in the mutants (Figure S4).

Feedback regulation of GA3ox, GA2ox and GID1 gene expression in ga20ox1 and ga20ox2 single and double mutants

Feedback regulation is also known to reduce expression of the biosynthetic GA3ox genes, and feed-forward regulation is known to increase expression of the GA-inactivating GA2ox genes (Cowling et al., 1998; Thomas et al., 1999). Figure 6(d) shows that GA3ox1 was upregulated in the leaf and inflorescence of ga20ox2, and in all ga20ox1 ga20ox2 tissues tested. GA2ox1 was slightly reduced in ga20ox2, and was more strongly reduced in ga20ox1 and ga20ox1 ga20ox2.

Recently, it was shown that the feedback mechanism also operates at the level of GA perception, with GA negatively regulating expression of the GID1 GA receptor genes in Arabidopsis (Griffiths et al., 2006). Whereas the GID1b transcript level in ga20ox2 was similar to that in wild type, it was significantly elevated specifically in the stem internodes of ga20ox1 and ga20ox1 ga20ox2, and also in leaf and inflorescence of ga20ox1 ga20ox2 (Figure 6d). GID1 overexpression in rice has been shown to produce a large increase in sensitivity to GA, which resulted in a clearly elongated phenotype of plants overexpressing GID1 genes (Nakajima et al., 2006; Ueguchi-Tanaka et al., 2005). Similarly, overexpression of GID1 in Arabidopsis seems to enhance GA reponses, whereas knock-out analysis has shown that the absence of GID1 genes results in a loss of GA responsiveness (Griffiths et al., 2006; Iuchi et al., 2007; Willige et al., 2007). In order to determine whether upregulation of GID1 expression could ameliorate the effects of the ga20ox mutations, we overexpressed GID1b from the 35S-CaMV promoter in the ga20ox1 ga20ox2 mutant. The ga20ox1 ga20ox2 35S::GID1b lines were clearly taller than the double mutant (Figure 7), indicating that the observed increase in GID1b transcript level in ga20ox1 ga20ox2 (14-fold in internodes) may partially suppress the GA-deficient phenotype.

Figure 7.

 Phenotype of ga20ox1 ga20ox2 35S::GID1b lines.
Photograph of 8-week-old plants, showing partial suppression of the dwarf phenotype of ga20ox1 ga20ox2 by the 35S::GID1b transgene.

The effect of the mutations on GA content

The effect of the reduction in GA20ox1 and/or GA20ox2 gene expression on endogenous GA content was examined in whole shoots when the first flower was opened. This ensured that all plants were analysed at the same developmental stage. Table 2 shows that the concentration of the bioactive species, GA4 and GA1, were reduced by 37% and 67%, respectively, in the double mutant compared with the wild type. Although there was a small reduction in GA4 concentration in ga20ox2, there was no reduction in the levels of GA4 and GA1 in ga20ox1. However, the concentrations of GA34 and GA8, the deactivated metabolites of GA4 and GA1, respectively, were substantially reduced in the ga20ox1 and double mutants. In the case of GA34, this reduction was 3.6- and 5.8-fold, respectively. It would appear that downregulation of GA2ox expression as part of the GA-homeostatic mechanism (Thomas et al., 1999) may be stronger for the ga20ox1-containing mutants than for ga20ox2, perhaps reflecting tissue specificity. To examine whether the decrease in GA content was more apparent in a tissue that showed a stronger phenotype, we compared GA4 levels specifically in bolting stems of the wild type and ga20ox1 ga20ox2 plants. The reduction in the mutant was comparable with that found for whole shoots, with ga20ox1 ga20ox2 having ∼50% of the level of GA4 found in the wild type (wild type, 7.0 ng g−1; ga20ox1 ga20ox2, 3.6 ng g−1).

Table 2.   Concentrations of gibberellins (GAs) in shoots of the wild-type and mutant lines
LineGA24GA4GA34GA19GA1GA8
  1. The values are the means of three biological replicates, except where indicated, in ng g−1 dry weight (SD). GA9 and GA20 were also analysed, but, although the internal standards were detected, the concentrations of these GAs were below the level of detection.

  2. aFrom two biological replicates.

Col-024.4 (0.4)a3.59 (0.18)9.37 (0.10)3.70 (0.13)0.37 (0.01)0.36 (0.01)
ga20ox130.3 (1.9)a3.30 (0.24)2.62 (0.22)2.35 (0.16)0.43 (0.01)a0.07 (0.10)a
ga20ox224.5 (0.7)a2.60 (0.02)5.56 (0.37)2.53 (0.14)0.39 (0.17)0.30 (0.01)
ga20ox1/ga20ox222.1 (1.7)2.26 (0.09)1.62 (0.06)2.26 (0.09)0.12 (0.04)0.09 (0.08)

Discussion

Gibberellin 20-oxidase activity is a major determinant of GA production and GA-dependent development in Arabidopsis (Coles et al., 1999; Huang et al., 1998). We have taken a reverse genetics approach to study the contribution of two of the five known GA20ox genes, GA20ox1 and GA20ox2, to plant development throughout the Arabidopsis life cycle. As has been shown for the GA3ox genes (Mitchum et al., 2006), the GA20ox genes differ in their expression patterns (see also Phillips et al., 1995), indicating that there is specificity in their physiological roles. GA20ox1 and GA20ox2 are the major GA20ox genes expressed during most phases of vegetative and early reproductive development, and we show that they act redundantly to regulate most phases of development. However, the redundancy is partial for some developmental processes, the relative contribution of the enzymes probably reflecting the levels of expression of their respective genes.

Comparison of the ga20ox1 ga20ox2 double mutant with the highly GA-deficient ga1-3 mutant indicates that other GA20ox genes make substantial contributions to GA biosynthesis in some tissues. For example, there was no significant effect on root growth or on final rosette size in the double mutant when grown in long days, although this mutant grew more slowly than the other lines in long days, and leaf expansion was reduced in short days (data not shown). Interestingly, double-mutant seeds gave higher rates of germination compared with the wild type and single mutants. This may be related to the development of the testa, which acts as a barrier to germination (Debeaujon et al., 2000; Penfield et al., 2006). Loss of function of the GA biosynthetic gene GA3ox4 and the combined loss of GID1a and GID1b GA-receptor gene function were shown to result in abnormalities in the seed surface structure (Iuchi et al., 2007; Kim et al., 2005), which may weaken the resistance of the seed coat to penetration by the radicle and cotyledons. However, in preliminary experiments, scanning electron microscopy of the seed coat, and analysis of the seed mucilage using ruthedium red staining, revealed no obvious changes in the ga20ox double mutant (data not shown). It is possible that a combination of an altered testa, which normally requires a GA signal for penetration during germination (Penfield et al., 2006), and high levels of GA20ox3 expression at this stage (Figure 1), could account for the failure to observe a delay in germination in the ga20ox mutants, as expected for GA-deficient seed. Further research using seeds from reciprocal crosses with the wild type and including mutants of GA20ox3 should shed light on this phenotype.

There was a substantial reduction in hypocotyl length in the double mutant, but not in the single mutants. Coles et al. (1999) reported a reduction in hypocotyl length in a GA20ox1 antisense line, but this might have resulted from the downregulation of multiple GA20ox genes. Furthermore, leaves of the double mutant were epinastic in comparison with those of the wild type and the single mutants (Figure 2d), confirming that GA20ox1 and GA20ox2 both contribute to petiole growth. Hisamatsu et al. (2005) found that the relative contribution of GA20ox1 and GA20ox2 to petiole growth was influenced by day length and light spectral quality. In short days ga20ox1-3, but not transgenic lines with reduced expression of GA20ox2, exhibited a short petiole phenotype, indicating that GA20ox1 expression became limiting under these conditions. In far red-rich light the contribution of GA20ox2 to petiole growth becomes important because the expression of GA20ox2 is strongly induced under these conditions. However, in plants grown in red-rich fluorescent light, as in our experiments, the contribution of GA20ox2 to petiole growth would be relatively low.

AtGA20ox1 and AtGA20ox2 contribute to floral transition

It has recently been demonstrated that GA4 is the GA responsible for promoting flower initiation in Arabidopsis, although it is not clear if it is synthesized at the shoot apex or if it is transported to the apex from the leaves or other tissues (Eriksson et al., 2006). We show that GA20ox1 and GA20ox2 influence flower initiation redundantly, their expression having an additive effect. In long days, knock-outs of either gene alone had little effect on flowering time, whereas in short days flowering was delayed in both single mutants, with GA20ox2 having the greater effect. Flowering in the double mutant was delayed even in long days, but not by as much as in ga1-3, indicating the involvement of at least one further GA20ox in this process. The GA20ox expression patterns illustrated in Figure 1 or the AtGenExpress microarray data (http://www.weigelworld.org/resources/microarray/AtGenExpress) offer few clues as to the site of synthesis of the GA4 that induces the floral transition. Eriksson et al. (2006) did not detect a large increase in GA20ox expression at the apex over the period preceding floral induction, and suggested that the GA might indeed be transported from the leaves. However, GA20ox2 expression has a greater effect on flowering time than that of GA20ox1, whereas the latter is expressed more highly in rosettes. Finer spatial resolution of gene expression may be needed to resolve this issue.

GA has two, genetically separable, effects on stem growth

The most obvious phenotype of the ga20ox1-3 mutant is reduced stem height (Figure 2a), as was noted also for the ga5 (ga20ox1-1) mutant in the Ler background (Koornneef and van der Veen, 1980). Overall height reduction in ga20ox1-3 is a consequence of shorter internodes as well as fewer internodes in the inflorescence. The number of elongated internodes in the vegetative stem was the same as in the wild type. The stem phenotype is consistent with the gene expression profile, which indicates that GA20ox1 is the predominant GA20ox gene expressed in the vegetative stem (Figure 1). Indeed, GA20ox2 expression was very low in this tissue. However, although mean internode length in the ga20ox2 mutant was the same as in the wild type, in contrast to ga20ox1, fewer vegetative internodes elongated, which resulted in a shorter vegetative stem. A combination of reduced numbers and lengths of the internodes in the double mutant resulted in a very short vegetative stem. Gibberellin treatment restored the length in the mutants to that of the wild type, which did not elongate further with GA application, indicating that GA content in the wild type was probably already saturating for elongation growth. In contrast, GA treatment doubled the number of elongated vegetative internodes in all lines. These observations demonstrate that GA influences vegetative stem growth by regulating both the number and length of the internodes. Moreover, the activity of different GA20ox isoforms is limiting for these two processes, presumably as a consequence of distinct temporal and spatial distributions. As both processes involve the same organs, it is reasonable to assume that the two responses to GA, propensity to elongate and elongation itself, are temporally separated or involve different groups of cells within the internode. Such a clear dichotomy in the physiological function of the two genes during stem elongation was unexpected, because GAs have some mobility in plants (see Ross et al., 2006). Indeed because the double mutant has fewer and shorter internodes than either single mutant, it is possible that there is partial rescue of the single mutant phenotypes by GA movement.

The number of flowers and thus internodes in the inflorescence was reduced relative to Col-0 in each of the mutant lines. Internode length within the inflorescence, although reduced relative to Col-0 in ga20ox1 and ga20ox1 ga20ox2, was greater than that of the wild type in ga20ox2, such that the final height of this mutant was almost 90% that of Col-0, despite fewer internodes. Furthermore, GA treatment resulted in a decrease in inflorescence internode length in ga20ox2 and Col-0, as well as an increase in the number of internodes within the inflorescence. The reason for this apparent inverse relationship between internode/flower number and internode length is unclear, but may indicate coupling between flower initiation/development and internode elongation, perhaps in response to limiting the availability of nutrients.

Male fertility

Development of all floral organs is arrested in highly GA-deficient plants (Cheng et al., 2004; Goto and Pharis, 1999) or in plants incapable of responding to GAs (Griffiths et al., 2006). However, Goto and Pharis (1999) have shown that male fertility is particularly dependent on GA signalling, as rescue of stamen filament elongation and pollen development required a higher dose of GA than rescue of development of other floral organs. We observed complete loss of fertility in the first ten flowers of the ga20ox1 ga20ox2 double mutant, which correlated with both shorter stamens and incomplete anther development (Figure 5a). Slightly reduced fertility in the first flowers was also observed in the wild type and more so in ga20ox1, but almost all flowers of ga20ox2 were fertile. High levels of GA application or signalling have also been shown to cause reduced filament elongation and male sterility (Colombo and Favret, 1996; Ikeda et al., 2001; Jacobsen and Olszewski, 1993; Sawhney and Shukla, 1994), and it is possible that GA levels are supra-optimal for stamen development in the young inflorescence of the wild type following the burst of GA4 production that coincides with flower induction (Eriksson et al., 2006). Alternatively, the high GA levels might cause premature or excessive elongation of the pistil. In either case, GA levels seem to be optimal in young inflorescences of ga20ox2 and suboptimal in ga20ox1 and ga20ox1 ga20ox2. This is supported by the observed reduced fertility in the wild type and ga20ox2 after GA treatment. Restoration of fertility in later flowers, even in the double mutant, may be the result of optimization of GA levels in the older inflorescence; in the case of the double mutant, the background GA levels may have increased because of expression of other GA20ox genes. The extremely high levels of GA20ox3 expression in developing seeds (Kim et al., 2005), for example, could allow sufficient GA to be produced once some seed set has occurred.

Seed set and silique elongation

Combined loss of expression of ga20ox1 and ga20ox2 resulted in reduced seed set and shorter siliques. Lower seed numbers could result from the poor male fertility of the double mutant, and could account to some extent for the reduced silique elongation (Cox and Swain, 2006). However, pollinating double mutant pistils with wild-type pollen did not restore seed numbers, indicating a maternal contribution to fertility. The GA20ox genes are also required for the production of GAs that regulate silique elongation directly, as siliques of the double mutant with seed numbers similar to the wild type also failed to elongate normally. GA20ox2 makes the greater contribution to silique elongation as ga20ox2 siliques are significantly shorter than those of the wild type and ga20ox1, but contain normal seed numbers. Also, this phenotype could not be rescued by pollination with wild-type pollen, indicating that the GA regulating silique elongation is of maternal origin, probably being synthesized in the silique tissue itself. It has been shown in pea that pod elongation depends on GA20ox expression in the pericarp, which is induced by auxin synthesized in the developing seeds (Ngo et al., 2002). This would be consistent with the high levels of GA20ox2 transcript in carpels and developing siliques (http://www.weigelworld.org/resources/microarray/AtGenExpress), and the observation that GA20ox2 is the major contributor to silique elongation in Arabidopsis. Alternatively, the GA regulating silique elongation could come from the seed coat, which has also been shown to express GA20ox2 (Kim et al., 2005).

Mechanisms for GA homeostasis moderate the effects of the GA20ox mutations

The observation that GA levels were only moderately reduced in shoots and bolting stems of the ga20ox mutants suggests that even small decreases in GA affect the plant growth significantly, or that the decreases are restricted either in time or in place, such that they were not detected to their full extend. This emphasizes the need for GA analysis at a higher spatiotemporal resolution, which could provide important information regarding the relationship between sites of GA synthesis and GA accumulation and action. The mild reductions in GA levels also indicate that substantial GA20ox activity is still present in the mutants, in which the effects of losing both GA20ox1 and GA20ox2 activities are offset to different degrees by expression of other GA20ox genes. On the basis of its gene expression profile (Figure 1), GA20ox3 would be expected to make a major contribution to GA biosynthesis in many tissues, particularly in seeds and roots, in which it is relatively highly expressed (Figure 1; Kim et al., 2005). The other two GA20ox genes of which we are aware (GA20ox4 and GA20ox5) are expressed at much lower levels than GA20ox1, GA20ox2 and GA20ox3 in most tissues analysed, and they might be expected to make a minor contribution to GA biosynthesis. The effects of the ga20ox mutations could also be moderated by the DELLA-protein-mediated mechanisms for GA homeostasis (see review by Thomas and Hedden, 2006). As a consequence of feedback regulation, the reduction in GA concentration in the mutants would be expected to result in increased expression of GA20ox1, GA20ox2 and GA20ox3 (Phillips et al., 1995). Indeed, in internodes of the ga20ox1 mutant, expression of GA20ox2 was increased 40-fold relative to Col-0, although this was from a very low basal level (Figures 1 and 6) and was not sufficient to fully rescue the ga20ox1 mutant phenotype. In contrast to GA20ox2 there was little change in GA20ox3 expression in ga20ox1 internodes, but expression of this gene was increased between four- and fivefold in internodes, leaves and inflorescence of the double mutant, relative to Col-0. The difference in response of the various GA-regulated genes to the consequences of the mutations (Figure 6c,d) may reflect differences in the sensitivity of the genes to changes in GA concentration and differing spatial expression patterns, such that they are exposed to different levels of GA.

Our results indicate that GA20ox4 and GA20ox5 are not subject to feedback regulation (Figure 6a). This seems to be in contrast to an earlier observation that GA20ox4 expression was higher in the GA-deficient ga1-3 mutant, and was reduced to wild-type levels in the triple mutant ga1-3 rga-24 gai-t6, in which there is a loss of DELLA function (Frigerio et al., 2006). However, these results could result from developmental differences between the genotypes. We also found that feedback repression of GA20ox1, GA20ox2 and GA20ox3 was abolished in the presence of cycloheximide, but as treatment with this compound caused upregulation of all five GA20ox genes (see also Frigerio et al., 2006), it is not possible to conclude that feedback repression requires protein synthesis (Bouquin et al., 2001) as downregulation by GA might be masked by cycloheximide-induced upregulation.

Homeostasis of GA signalling may also be maintained by regulation of expression of GA3ox (Cowling et al., 1998), GA2ox (Thomas et al., 1999) and GID1 (Griffiths et al., 2006) genes. Indeed we found that expression of GA3ox1 was slightly increased and that expression of GA2ox1 decreased, particularly in the double mutant (Figure 6c). Furthermore, expression of the GA receptor gene GID1b, which functions redundantly with GID1a and GID1c in most developmental processes, was strongly upregulated in internodes of ga20ox1 and the double mutant, the height phenotype of which can be moderated by a further increase in receptor level through transgenic overexpression (Figure 7). Thus, GA signalling can be buffered at several different levels: biosynthesis, deactivation and perception. Although this does not allow the full rescue of the phenotype when there are severe perturbations, it is likely that these homeostatic mechanisms mask smaller changes to the signalling pathway.

Experimental procedures

Plant material and growth conditions

Arabidopsis thaliana (L.) Heynh. ecotype Columbia (Col-0) was used as the wild type in this study. SALK line 016701 and GABI-KAT line 734G06 are both in the Col-0 background (Alonso et al., 2003; Rosso et al., 2003). Homozygous insertion lines were selected using PCR-based genotyping, and flanking sequence tags from the left borders were sequenced to determine the exact locations of the insertions. SALK line 016701 was renamed ga20ox1-3; this line has also been used by Hisamatsu et al. (2005), who refer to it as ga5-3. The seed batch obtained for GABI-KAT line 734G06 segregated plants with very short siliques as a dominant trait, and a homozygous insertion line lacking this phenotype was named ga20ox2-1 and was used in this study. Absence of correct GA20ox1 and GA20ox2 transcripts in the ga20ox1-3 and ga20ox2-1 lines, respectively, was confirmed by RT-PCR on cDNA from leaves, stem internodes and inflorescence using SYBR Green in a real-time detection system, as described below for real-time quantitative RT-PCR. Primers are listed in Table S6.

ga1-3 and ga1-3 rga-24 gai-t6 are in the Ler background (Dill and Sun, 2001; Koornneef and van der Veen, 1980). The ga1-3(Col-0) line is a ga1-3 that has been backcrossed to Col-0 six times (Tyler et al., 2004).

Unless stated otherwise, plants were grown on Levington F2 soil under long-day conditions (LD; 16-h light at 300 μmol m−2 sec−1 from Osram HQI-BT 400W/D metal halide lamps, 23°C, 65% relative humidity/8-h dark, 18°C, 70% relative humidity), watered regularly and treated with a standard plant nutrient solution once a week. Plants treated with GA were sprayed with 100 μm GA3 (Duchefa, http://www.duchefa.com) three times a week.

Production of transgenic plant lines

Constructs used to generate the GA2ox (GA2oxOE) and GID1b (GID1bOE) overexpressing lines were prepared by amplifying the Phaseolus coccineus GA2ox1 and Arabidopsis GID1b coding regions by PCR. The PcGA2ox1 and GID1b coding regions were amplified using the primers listed in Table S6. The amplified products were subcloned into pCR2.1 (PcGA2ox1; Invitrogen, http://www.invitrogen.com) or pCR-BluntII-TOPO (GID1b; Invitrogen), and the sequence was confirmed by DNA sequence analysis. The PcGA2ox1 and GID1b coding regions were excised from pCR2.1/pCR-BluntII-TOPO by SacI digestion, and were inserted into the SacI site of the binary vector pLARS120 (Curtis et al., 2000). Constructs containing the PcGA2ox1 and GID1b coding regions in the sense orientation were identified by restriction digest analysis, and were subsequently designated as pLARS124 and pGID1bOE, respectively.

Wild-type Col-0 and ga20ox1 ga20ox2 were transformed with pLARS124 and pGID1bOE, respectively, via Agrobacterium tumefaciens (GV3101) using the floral dipping method (Clough and Bent, 1998). The pLARS124 transformants were identified by plating seeds from infiltrated plants on MS media plates containing 50 μg ml−1 kanamycin sulphate (Sigma, http://www.sigmaaldrich.com) and 25 μm GA3. The pGID1bOE transformants were identified by plating seeds from infiltrated plants on MS media plates containing 150 μg ml−1 kanamycin. Resistant plants were transferred to compost after 14 days. The pLARS124 T1 transformants were sprayed with 10 μm GA3 at intervals of 4 days. Transgenic lines containing the T-DNA inserted at a single locus were identified by scoring kanamycin-sensitive and -resistant plants in the T2 generation. Those lines exhibiting a 3 : 1 (resistant to sensitive) ratio were self-pollinated, and homozygous T3 lines were identified. T4 plants derived from homozygous lines were used for phenotypical, biochemical and molecular characterization.

Mutant phenotype characterization

For the germination assay 100–200 seeds were spread on three layers of wet 3MM Whatman filter paper in the dark, and plates were wrapped in two layers of aluminium foil to prevent exposure to the light. Seed germination was scored by counting all seeds with radical protrusion after 4 days of incubation at 22°C. For cold pre-treatment, seeds were incubated at 4°C for 4 days, and for light pre-treatment seeds received 1 min of white light from fluorescent tubes (Philips Master TL5 54W/840 HO, 200 μmol m−2 sec−1; Philips, http://www.philips.com) 6 h after the start of incubation at 22°C. For generation of the germination curves, plates were incubated under continuous white light (as above) after receiving their respective pre-treatments, and seed germination was determined repeatedly with 24-h intervals.

Plants used for hypocotyl (n = 32) and root length measurements (n = 40) were grown in round or square Petri dishes, respectively, on full-strength Murashige and Skoog salts containing Gamborg B5 vitamins, 1% sucrose, pH 5.8, 0.8% Gelrite under continuous light at 22°C. For root growth, plates were put in vertical orientation. Either 50 or 200 nm GA4 was included in the growth medium for GA treatment of hypocotyls and roots, respectively. After 7 days, plates were scanned and measurements were taken using NIH ImageJ software.

Rosette radius (n = 24 for −GA; = 6 for +GA) was measured just before the start of leaf senescence on plants that were grown on soil in LD. Other measurements were performed on plants (n = 16 for −GA; n = 8 for +GA) that had stopped flowering. A ruler was used for rosette, internode and inflorescence length measurements. Rosette radius is the average of the two longest rosette leaves. When scoring branch numbers, minor basal branches with less then 25% of the length of the main branch were ignored. Flowering time in LD was scored as the total number of leaves produced by the main shoot, and also as the number of days before buds could be detected with the naked eye.

Plants used to determine short-day flowering time (n = 15) were grown at 23°C with 9 h of light at 130 μmol m−2 sec−1 from fluorescent lamps and 15 h of darkness. These plants were treated with a nutrient solution once a month. Leaf initiation was followed once a week by marking new leaves with a felt pen.

For analysis of early flower infertility (n = 24 for −GA; n = 12 for +GA), all short siliques were taken and absence of seeds was confirmed.

For silique characterization (n = 48 for −GA; n = 24 for +GA), the length of two fully elongated siliques chosen from silique number 10–30 on the main stem was measured, and the number of seeds per silique was counted.

Real-time quantitative RT-PCR

Tissue samples for the developmental series have been described previously (Griffiths et al., 2006). For the GA-response experiment, ga1-3 and ga1-3 rga-24 gai-t6 were grown in liquid culture. Seeds were surface-sterilized and imbibed at 4°C in 25 μm GA4 for 4 days. After thorough washing with water, ∼50 seeds were incubated in 250-ml flasks containing 100 ml of medium [half-strength Murashige and Skoog salts containing Gamborg B5 vitamins, 1% sucrose and 0.05% 2- (N-morpholine)-ethanesulphonic acid (MES); pH 5.8]. The cultures were kept on a rotary shaker at 160 rpm under continuous white light (Phillips Master TL5 54W/840 HO fluoresencent tubes) at 22°C. GA4 was resuspended in H2O (pH 7.0) to a concentration of 200 μm, and was added to 7-day-old cultures to a final concentration of 2 μm. For the cycloheximide experiment, the GA2oxOE line was grown in liquid culture as described above for ga1-3 and ga1-3 rga-24 gai-t6. Seven-day-old cultures were treated with 100 μm cycloheximide for 1 h before GA treatment. GA4 was added in 8 μl ethanol to a final concentration of 2 μm, control samples had 8 μl ethanol added. To compare transcript levels between the wild type and the ga20ox mutants, plants were grown on soil and leaves, and stem internodes and inflorescence were harvested when the first flower of the primary shoot had opened. Standard curves for the GA20ox genes were obtained from dilution series of known quantities of GA20ox DNA fragments. DNA fragments encompassing at least the full coding sequence of each transcript were first amplified using PCR, and were cloned into pGEM-T Easy (Promega, http://www.promega.com). The PCR primers used are listed in Table S6. The fragment was then re-amplified from these clones using the same primers, purified (Qiagen, http://www.qiagen.com), quantified (Nanodrop, http://www.nanodrop.com) and used directly as template for qPCR.

Total RNA extraction, DNase treatment, reverse transcription and real-time quantitative RT-PCR using SYBR Green have been described previously (Griffiths et al., 2006). In short, total RNA was extracted with on-column DNase treatment (RNeasy; Qiagen), treated again with DNase in solution (Turbo DNA-free kit; Ambion, http://www.ambion.com) and used for reverse transcription (SuperScript III Platinum Two-Step qRT-PCR Kit with SYBR Green; Invitrogen). The cDNA equivalent of 20 ng of total RNA was used in a 25-μL PCR reaction on an ABI 7500 Real Time PCR System (Applied Biosystems, http://www.appliedbiosystems.com) with Platinum SYBR Green qPCR SuperMix-UDG reagents (Invitrogen). In all experiments, three biological replicates of each sample type were tested, and reactions were caried out with two technical (PCR) replicates. Absence of genomic DNA and primer dimers was confirmed by analysis of RT-minus and water control samples, and by examination of dissociation curves.

To normalize the qPCR data, three or four reference genes (from Czechowski et al., 2005) were used in each experiment. For each experiment, stability of the reference genes across samples was tested using geNorm software (Vandesompele et al., 2002), and the most stable genes were included to calculate the normalization factors. In all cases, the sets of reference genes were highly stable (variability below the recommended cut-off value; Vandesompele et al., 2002). Reference genes for developmental series, At1g13320, At2g28390 and At4g34270 (V2/3 = 0.118); for GA-response series, At1g13320, At4g33380, At4g26410 and At5g25760 (V3/4 = 0.036); for the cycloheximide series, At2g28390, At4g05320 and At4g33380 (V2/3 = 0.096); for ga20ox mutant series, At1g13320, At4g26410 and At4g33380 (V2/3 = 0.089). For calculation of the absolute ratios in the developmental series, the number of GA20ox cDNA template molecules per sample was normalized against the corresponding geNorm normalization factor, and results from the three biological replicates were averaged. Primers for the following genes have been described before: At1g13320, At2g28390, At4g05320, At4g26410, At4g33380 and At4g34270 (Czechowski et al., 2005), and At5g25760, GA3ox1 and GID1b (Griffiths et al., 2006). All other qPCR primers were designed using Primer Express v.2.0 (Applied Biosystems) or GENOPLANTE SPADS software (http://urgi.versailles.inra.fr/tools/spads/), and are listed in Table S6. Primer pair efficiencies were estimated by analysis of the amplification curves with LinReg software (Ramakers et al., 2003), and the average efficiency of all reactions on a plate (which was always > 1.95) was used in calculations (see Cook et al., 2004).

Quantitative analysis of GAs

Whole shoots were harvested from plants grown in LD in trays (24 seedlings per tray) when the first open flower was visible. The trays, which were arranged randomly, each contained a single genotype and represented one replicate sample. Three replicate samples (0.5 g dry weight) were analysed by GC-MS, essentially as described in Griffiths et al. (2006) except that the selected ion monitoring was carried out using a MAT95XP mass spectrometer, operated at a resolution of 1000, coupled to a Trace GC (ThermoElectron, http://www.thermo.com). The MS source was at 220°C and the electron energy was 70 eV. GC-MS quantification of GA24 was performed on an 6890 Series Gas Chromatograph (Agilent Technologies, http://www.agilent.com), coupled to a Micromass GCT (Waters, Waters). GC conditions were as reported previously, except that a TR-5 column (30 m × 0.25 mm internal diameter × 0.25 μm film; ThermoElectron) was used. The GC interface was 250°C and EI+ (electron ionization) mass spectra were acquired from m/z 41–650 from 0–37.50 min, with an acquisition rate of one spectrum per second.

Statistical analyses

A completely randomized design was used for plants from which germination, hypocotyl and root length measurements were taken. The design for the other measurements in the characterization experiment was a randomized block with 16 blocks (for SD flowering measurements) or a split-plot in four blocks (for all other measurements), with pairs of trays making up each block, and with one tray treated with GA and the other kept as the control (Gomez and Gomez, 1984). For statistical analysis of qPCR data, reference-gene-corrected threshold cycle (Ct) values were used. These were obtained by converting the Ct values to relative quantities, which were corrected with the calculated normalization factors (see above) and were back-transformed to Ct values. Where experiments were balanced for comparison, analysis of variance (anova) was used; otherwise the restricted maximum likelihood (REML) method was applied. In order to correct for heterogeneity of variance, a logit transformation was applied to the germination data, and a natural log transformation was applied to the hypocotyl length and rosette radius data. Note that because a common transformation was applied to all observations, the comparative values between genotypes are not altered, and comparisons between them remain valid. Least significant differences at 1% were used to assess significance between particular pairs of genotypes or treatment combinations, on the transformed scale where appropriate. The GenStat statistical system (version 8.2; Lawes Agricultural Trust, http://www.rothamsted.ac.uk/corporate/LawesAgriculturalTrust.html) was used for all analyses.

Acknowledgements

The authors wish to thank Nathan Hawkins for help with the GC-MS analysis, Ian Pearman, Anthony Griffiths and other greenhouse staff for excellent plant material, Graham Shephard and the Visual Communications Unit for all photographs, and Tai-ping Sun for the gal-3 rga-14 gai-t6 and gal-3 (Col-0) lines. This work was supported by grants P16508 and P19317 from the Biotechnology and Biological Sciences Research Council of the United Kingdom. We are also grateful to the Spanish Ministry of Education and Science for a fellowship to NFG (Ref: EX2004-0398) and to the Czech Ministry of Education for a grant (MSM6198959216) to TL.

Ancillary