Distinct and overlapping roles of two gibberellin 3-oxidases in Arabidopsis development


*(fax +1 919 613 8177; e-mail tps@duke.edu).


Gibberellin (GA) 3-oxidase, a class of 2-oxoglutarate-dependent dioxygenases, catalyzes the conversion of precursor GAs to their bioactive forms, thereby playing a direct role in determining the levels of bioactive GAs in plants. Gibberellin 3-oxidase in Arabidopsis is encoded by a multigene family consisting of at least four members, designated AtGA3ox1 to AtGA3ox4. It has yet to be investigated how each AtGA3ox gene contributes to optimizing bioactive GA levels during growth and development. Using quantitative real-time PCR analysis, we have shown that each AtGA3ox gene exhibits a unique organ-specific expression pattern, suggesting distinct developmental roles played by individual AtGA3ox members. To investigate the sites of synthesis of bioactive GA in plants, we generated transgenic Arabidopsis that carried AtGA3ox1–GUS and AtGA3ox2–GUS fusions. Comparisons of the GUS staining patterns of these plants with that of AtCPS–GUS from previous studies revealed the possible physical separation of the early and late stages of the GA pathway in roots. Phenotypic characterization and quantitative analysis of the endogenous GA content of ga3ox1 and ga3ox2 single and ga3ox1/ga3ox2 double mutants revealed distinct as well as overlapping roles of AtGA3ox1 and AtGA3ox2 in Arabidopsis development. Our results show that AtGA3ox1 and AtGA3ox2 are responsible for the synthesis of bioactive GAs during vegetative growth, but that they are dispensable for reproductive development. The stage-specific severe GA-deficient phenotypes of the ga3ox1/ga3ox2 mutant suggest that AtGA3ox3 and AtGA3ox4 are tightly regulated by developmental cues; AtGA3ox3 and AtGA3ox4 are not upregulated to compensate for GA deficiency during vegetative growth of the double mutant.


Bioactive gibberellins (GAs) are important plant growth regulators which play essential roles in seed germination, stem elongation, leaf expansion and flower and seed development (Davies, 2004). The GA biosynthetic pathway has been well characterized and the majority of genes encoding enzymes in each step of the GA biosynthetic and catabolic pathways have been cloned (reviewed in Olszewski et al., 2002). These studies are allowing researchers to examine the expression patterns of GA biosynthetic genes using reporter genes and in situ hybridization techniques, circumventing prior difficulties of studying the regulation of GA concentration due to the low abundance of GAs in plants (Itoh et al., 1999; Kaneko et al., 2003; Silverstone et al., 1997; Yamaguchi et al., 2001). Such studies have begun to reveal the precise sites of GA metabolism in various plant species and the complex regulatory mechanisms in place to control the appropriate levels of bioactive GA for plant growth and development.

Our current understanding of the regulation of GA metabolism is limited; however, several studies suggest multiple layers of regulation in plants. The GA biosynthetic pathway (reviewed in Hedden and Phillips, 2000; Olszewski et al., 2002) is separated into distinct subcellular compartments. Gibberellin biosynthesis is initiated in the plastid, where geranylgeranyl diphosphate (GGDP) is converted to ent-kaurene in a two-step process catalyzed by the terpene cyclases, ent-copalyl diphosphate synthase (CPS) and ent-kaurene synthase (KS). ent-Kaurene is then oxidized to GA12 in the plastid envelope and endoplasmic reticulum (ER) by a series of steps catalyzed by cytochrome P450 monoxygenases. In the final stages of GA biosynthesis, GA12 is converted to bioactive GA in the cytosol by two classes of 2-oxoglutarate-dependent dioxygenases, including GA 20-oxidase (GA20ox) and GA 3-oxidase (GA3ox). In addition to subcellular compartmentalization of the pathway, the copy number of the genes encoding the GA biosynthetic enzymes at each step in the pathway adds an additional level of regulation. In several plant species, the enzymes involved in the early steps in the GA biosynthetic pathway are encoded by single genes (Sun and Kamiya, 1994; Yamaguchi et al., 1998a); however, the enzymes that catalyze the final steps in the pathway have been shown to be encoded by small multigene families (Itoh et al., 2001; Phillips et al., 1995; Rebers et al., 1999; Sakamoto et al., 2004; Thomas et al., 1999; Yamaguchi et al., 1998b). Although the enzymes that catalyze the later stages of GA biosynthesis show some degree of functional redundancy, they have distinct organ-specific and cell-type-specific expression patterns during growth and development. The differential regulation of gene family members catalyzing the final steps in GA biosynthesis is one mechanism that may be used by the plant to maintain the appropriate levels of GA required for plant development. Environmental stimuli such as light and temperature also specifically regulate the expression of genes encoding GA biosynthetic enzymes (reviewed in Yamaguchi and Kamiya, 2000).

Concentrations of GA in some plant tissues and organs during development appear to be regulated by activity in GA signaling that regulates GA biosynthesis by a feedback mechanism. Transcript levels of genes encoding enzymes that catalyze several steps in the later stages of the GA biosynthetic pathway are under feedback control (Chiang et al., 1995; Phillips et al., 1995). For example in Arabidopsis, the GA3ox1 gene, but not the GA3ox2 gene, is under negative feedback regulation (Chiang et al., 1995; Yamaguchi et al., 1998b). An additional layer of regulation may reside in the separation of the GA biosynthetic pathway into distinct cell types in tissues and organs that require GA for development. Previous studies suggest that the cell-type-specific expression of AtKO1 (ent-kaurene oxidase), AtGA3ox1 and AtGA3ox2 is different from that of the early GA biosynthetic gene AtCPS in germinating Arabidopsis seeds, supporting the idea that the synthesis of bioactive GA may require the intercellular transport of a pathway intermediate (Yamaguchi et al., 2001). Lastly, GA levels may be regulated in plants through long-distance and local transport mechanisms. Conventional grafting experiments suggest the active transport of bioactive GA and GA intermediates, and other studies have indicated the presence of GAs in xylem and phloem exudates (Katsumi et al., 1983; Proebsting et al., 1992). In addition, a recent study has demonstrated that the expression of GA-responsive genes is not restricted to the predicted sites of GA biosynthesis suggesting the involvement of local transport mechanisms (Ogawa et al., 2003).

Gibberellin 3-oxidase catalyzes the final step in the GA biosynthetic pathway leading to the production of bioactive GAs. The GA3ox genes have been cloned from several dicot species (Chiang et al., 1995; Itoh et al., 1999; Lester et al., 1997; Martin et al., 1997; Yamaguchi et al., 1998b) and more recently, from a monocot (Itoh et al., 2001). For most plant species, GA3ox is encoded by multigene families, although it is unclear how many GA3ox genes are present in each genome. In rice (Oryza sativa) there are presumably only two GA3ox genes, OsGA3ox1 and OsGA3ox2, and each gene has a distinct tissue-specific expression pattern (Itoh et al., 2001; Kaneko et al., 2003; Sakamoto et al., 2004). In Arabidopsis, four GA3ox genes have been identified (AtGA3ox1AtGA3ox4). AtGA3ox1 (formerly GA4; Chiang et al., 1995) and AtGA3ox2 (formerly GA4H; Yamaguchi et al., 1998b) have been shown biochemically to encode GA3ox (Williams et al., 1998; Yamaguchi et al., 1998b). AtGA3ox3 (At4g21690) and AtGA3ox4 (At1g80330) were identified through database searches following the completion of the Arabidopsis genome sequence (Hedden et al., 2002; Yamaguchi et al., 1998b). A null allele of AtGA3ox1, ga3ox1-2 (formerly ga4-2), exhibits a semi-dwarf phenotype (Chiang et al., 1995). This leaky phenotype is in contrast to the severe dwarf stature of the GA-deficient ga1-3 mutant (Koornneef and van der Veen, 1980). GA1 encodes the only Arabidopsis CPS, the enzyme that catalyzes the first committed step in GA biosynthesis (Sun and Kamiya, 1994). These results suggest that the less severe phenotype of the ga3ox1 mutant is due to functional redundancy, and illustrates the importance of determining the function of each AtGA3ox gene in plant development. To investigate the potential sites of synthesis of bioactive GA in plants, we examined the developmental expression profile and the level of expression of all four AtGA3ox genes in selected organs by quantitative real-time PCR. We also generated transgenic Arabidopsis that carried AtGA3ox1GUS and AtGA3ox2GUS fusions. Comparisons of the GUS staining patterns of these plants with that of AtCPSGUS (Silverstone et al., 1997) revealed the possible physical separation of the early and late stages of the GA pathway in organs other than germinating seeds. Using a reverse genetics approach, we determined the function of AtGA3ox1 and AtGA3ox2 in plant growth and development. We have isolated and characterized ga3ox1 and ga3ox2 single mutants, constructed a ga3ox1/ga3ox2 double mutant, characterized the mutant phenotypes and measured levels of endogenous GA, as compared with wild type (WT) and the GA-deficient dwarf mutant, ga1-3.


Developmental expression profile of the Arabidopsis GA3ox gene family

In Arabidopsis, GA3ox is encoded by a small gene family with at least four members (AtGA3ox1AtGA3ox4) (Hedden et al., 2002). It was previously shown by RNA blot analysis that AtGA3ox1 is expressed throughout development whereas AtGA3ox2 is primarily expressed in germinating embryos and very young seedlings (Yamaguchi et al., 1998b). To determine the organs in which bioactive GA is being synthesized in the plant and to help elucidate the function of each GA3ox gene in regulating plant development we analyzed the developmental expression profile and quantified the level of expression of all four GA3ox genes using gene-specific primers and real-time PCR.

Figure 1 depicts the absolute levels of expression of each Arabidopsis GA3ox gene in selected organs. A housekeeping gene UBIQUITIN11 (UBQ11) was used as the standard (Tyler et al., 2004). Relatively high levels of expression of GA3ox1 were detected in all organs examined. GA3ox1 is the predominant GA3ox gene expressed in the stem, consistent with the requirement of bioactive GA for stem elongation and the semi-dwarf phenotype conferred by the ga3ox1 mutant (Chiang et al., 1995). GA3ox2 was also expressed in all organs examined; however, in contrast to GA3ox1, GA3ox2 was expressed at very high levels during the early stages of development and then declined to extremely low levels during later stages of development. For example, GA3ox2 was detected at >20 000 copies/105 copies of UBQ11 in germinating seeds and <10 copies/105 copies UBQ11 in the stem of 35-day-old plants (Figure 1 and Supplementary Table S1). Among the four GA3ox genes, only GA3ox1 and GA3ox2 were expressed in roots of 5-day-old seedlings, suggesting a unique role for these genes in the production of bioactive GA for root development. The GA3ox3 and GA3ox4 genes displayed a much more restricted organ-specific pattern of expression during development. Expression of GA3ox3 was only detected in flower clusters and siliques. GA3ox4 was primarily expressed in siliques, with a much lower level of expression detected in flowers, germinating seeds and young seedlings. Our results clearly demonstrate that the Arabidopsis GA3ox genes are differentially regulated and show unique organ-specific expression patterns, indicating that they probably have distinct roles in plant development. On the basis of their expression patterns, we can predict that both GA3ox1 and GA3ox2 are involved in the production of bioactive GA for vegetative growth and development, whereas GA3ox1, GA3ox3 and GA3ox4 function in the production of bioactive GA for reproductive development.

Figure 1.

Transcript levels of GA3ox1–GA3ox4 throughout development.
The level of gene expression in each tissue sample, as determined by quantitative PCR, is shown as the number of cDNA copies per 105 copies of UBQ11 cDNA. The organs and developmental stages tested were, from left to right in the graph, 12- and 24-h imbibed seeds, 2-, 5- and 10-day-old seedlings, roots of 5-day-old seedlings, 14-day-old rosettes and the following organs from 35-day-old plants: cauline leaves, stems, flower clusters and siliques. The means of three experiments ±SE are shown.

Expression pattern of the AtGA3ox1 and AtGA3ox2 genes during plant development

Among the four Arabidopsis GA3ox genes, only GA3ox1 was expressed at relatively high levels in all stages of plant development examined (Figure 1). To determine the potential sites of synthesis of bioactive GA during plant development, we analyzed the tissue-specific and cell-type expression pattern of the GA3ox1 and GA3ox2 genes. Transgenic Arabidopsis plants that carried GA3ox1–GUS fusions were generated (Figure 2). Because a number of the GA biosynthetic genes require introns for proper expression (Sakamoto et al., 2001; Silverstone et al., 1997), two different GA3ox1GUS fusions were constructed (Figure 2). p3ox1-TC-GUS contained 3 kb of the 5′ upstream region from the translational start site of GA3ox1 fused transcriptionally to the GUS reporter gene. The second construct, p3ox1-TL-GUS, was a translational fusion that contained both the 3 kb 5′GA3ox1 upstream region and a portion of the GA3ox1 coding region (including exon1, intron 1 and part of exon 2) fused to the GUS gene. More than 10 homozygous transgenic lines were generated for each construct, and most of them showed a consistent expression pattern in the embryo axis of germinating seeds and also at later developmental stages. The developmental expression pattern of the GA3ox1 gene throughout Arabidopsis development was compared with the expression pattern of the GA3ox2GUS transgenic lines, p3ox2-TC-GUS and p3ox2-TL-GUS, which were generated previously but were not characterized beyond germination (Yamaguchi et al., 2001). Among the GA3ox2GUS lines, most of the TC and TL lines also had consistent expression patterns. Several pieces of evidence suggest that these GA3oxGUS lines reflect an accurate representation of the tissue- and cell-specific expression pattern of GA3ox1 and GA3ox2 during development. The GUS staining patterns of our GA3oxGUS lines observed in germinating seeds (Figure 3) were consistent with the cell-type-specific expression patterns determined previously by in situ RNA hybridization (Yamaguchi et al., 2001). Furthermore, GUS activity correlated with the developmental expression profile and level of expression of GA3ox1 and GA3ox2 as determined by our quantitative real-time PCR experiments. Lastly, the expression of GA3ox1 cDNA driven by the regulatory sequences in p3ox1-TC-GUS and p3ox1-TL-GUS completely restored the ga3ox1-3 phenotype to WT (data not shown). In addition, complementation of the ga3ox2 mutant phenotype by a GA3ox2 genomic DNA fragment (in the ga3ox1 mutant background, see below) suggests that the 6 kb 5′ non-coding region in p3ox2-TC-GUS and p3ox2-TL-GUS is sufficient as a 5′ regulatory sequence for proper GA3ox2 expression. In general, the p3ox-TL-GUS lines had weaker GUS activity than the p3ox-TC-GUS lines, although their expression patterns are consistent with each other. The weaker GUS activity of the translational fusions may be attributed to the extra amino acids interfering with the catalytic activity of the GUS enzyme. The only exception was that the p3ox2-TL-GUS lines, but not the p3ox2-TC-GUS lines, showed faint staining in anthers. The homozygous lines p3ox1-TC-GUS 9-5-1 and p3ox2-TC-GUS 3-2-5 were used for further analyses on the expression pattern of GA3ox1 and GA3ox2 throughout plant development. To show the anther staining of GA3ox2-GUS, p3ox2-TL-GUS 5-1-5 [in the Landsberg erecta (Ler) background] was also included (Figure 4h).

Figure 2.

Structure of the AtGA3ox1 promoter–GUS reporter gene fusion constructs.
Black boxes correspond to exons 1 and 2, and the gray box corresponds to an intron. AtGA3ox2 promoter–GUS reporter gene fusion constructs have been described previously (Yamaguchi et al., 2001).

Figure 3.

AtGA3ox1GUS and AtGA3ox2GUS expression during Arabidopsis vegetative development.
The following tissues and developmental stages of the p3ox1-TC-GUS 9-5-1 line and p3ox2-TC-3-2-5 line were assayed: (a, k) 16-h germinating embryos, (b, l) 2-day-old seedling, (c, m) 3-day-old seedling, (d, n) 5-day-old seedling, (e, o) root of 5-day-old seedling, (f, p) thin section through the shoot apical meristem of 5-day-old seedling, (g, q) 10-day-old seedling, (h, r) aerial rosette of 14-day-old plant, (i, s) higher magnification of main stem showing strong AtGA3ox1-GUS staining at the node, (j) tip of cauline leaf, (t) thin section through the root apical meristem of a 3-day-old seedling.
Cm, columella cells; Hyd, hydathode; Hyp, hypocotyl; Lr, lateral root; N, node; Qc, quiescent center; Rt, root tip; Sam, shoot apical meristem; Sm, stem; Vs, vasculature. Concentrations of potassium ferricyanide in the X-gluc solution: 1 mm in (a)–(j) and (q)–(s); 2 mm in (k)–(p) and (t).

Figure 4.

AtGA3ox1-GUS and AtGA3ox2-GUS expression during Arabidopsis reproductive development.
The following tissues and developmental stages were assayed: (a) flower cluster of 5-week-old plant, (b, c) higher magnification of developing flowers, (d–f) developing siliques and mature seeds, (h) primary inflorescence of 4-week-old plant, (g, i) different stages of embryogenesis.
AtGA3ox1GUS images shown are from p3ox1-TC-GUS line 9-5-1. All AtGA3ox2GUS images shown are tissues from p3ox2-TC-GUS line 3-2-5, except (h) which is from p3ox2-TL-GUS line 5-1-5. An, anther; St, stamen; Sp, sepal; Sr, silique receptacle. Concentrations of potassium ferricyanide in the X-gluc solution: 1 mm in (a)–(g) and (i); 0 mm in (h).

Although GA3ox1GUS and GA3ox2GUS exhibited the same expression patterns in germinating seeds (Figure 3a,k; Yamaguchi et al., 2001), they are differentially expressed in many tissues at later developmental stages (Figures 3 and 4). In 2- to 5-day-old GA3ox1GUS seedlings, GUS staining was observed predominantly in developing cotyledons, vegetative shoot apical meristem (SAM), and non-meristematic, non-elongation regions of the roots (Figure 3b–f). In contrast to GA3ox1GUS, GA3ox2GUS was expressed predominantly in the hypocotyl and roots of young seedlings (Figure 3l–p). In the SAM, GA3ox1GUS was expressed predominantly in the rib meristem with weaker staining in corpus cells and leaf primordia (Figure 3f). Thin sections confirmed that GA3ox2GUS was not expressed in the SAM (Figure 3p). In roots, GA3ox2GUS expression was similar to GA3ox1GUS expression in the non-meristematic and non-elongation zones of the root, but unlike GA3ox1GUS, GA3ox2GUS was also expressed in the elongation zone (Figure 3l,m), the quiescent center cells and the columella cells of the primary root tip (Figure 3t) and lateral roots (Figure 3o).

In rosette leaves, GA3ox1GUS expression remained primarily in the interveinal regions with higher concentrations of GUS staining in the vasculature near the tips of the leaves (Figure 3d,g,h). GA3ox2GUS expression occurred in the petioles and vasculature of rosette leaves (Figure 3q,r). Strong GUS expression was observed for GA3ox1GUS in the stem, with higher concentrations near the nodes (Figure 3i). GA3ox2GUS expression declined to very low levels in the stem (Figure 3s).

In flowers, GA3ox1GUS expression was predominantly associated with the development of the stamens and was detected at high levels at the base of flower buds extending up into the stamen filaments (Figure 4a–c). GA3ox1GUS expression was also detected in the vasculature at the tips of the sepals (Figure 4c). In the developing siliques, GA3ox1GUS staining was restricted to the receptacle (Figure 4d–f). Consistent with quantitative real-time PCR results, GA3ox2GUS expression declined to very low levels in reproductive organs. In flower clusters from 5-week-old GA3ox2-TC-GUS plants, weak expression was observed in the vasculature of sepals (not shown). Weak expression was observed in the anthers of the primary inflorescence of 4-week-old GA3ox2-TL-GUS lines (Figure 4h), but not GA3ox2-TC-GUS lines (not shown). During seed development, neither GA3ox1GUS nor GA3ox2GUS expression was detected in the endosperm. To examine GUS expression in developing embryos, we dissected embryos at various developmental stages following staining. Each GA3ox gene displayed a unique expression pattern in developing embryos. GA3ox1GUS was expressed at the end of the embryonic axis corresponding to the SAM from the heart stage to mature seed (Figure 4f,g). GA3ox2GUS expression was detected at the opposite end of the embryonic axis corresponding to the root apical meristem (Figure 4i).

Although GA3ox1GUS and GA3ox2GUS were expressed in the same cellular locations in the cortex and endodermis of the embryo axis of germinating seeds (Yamaguchi et al., 2001), and both function to produce bioactive GA required for the germination of seed (see below), each gene demonstrated a unique tissue-specific expression pattern and probably plays a unique role in later stages of plant development.

Isolation and characterization of Arabidopsis mutants defective in GA3ox1 and GA3ox2

Our real-time PCR data suggest that only GA3ox1 and GA3ox2 play an important role in the production of bioactive GAs in vegetative organs in Arabidopsis (Figure 1). To help define the physiological functions of these two GA3ox genes in plant development, mutants defective in GA3ox1 and GA3ox2 were isolated (Figure 5). The previously identified ga3ox1-2 (ga4-2) mutant contains a T-DNA insertion in the single intron within the GA3ox1 (formerly GA4) coding region (Chiang et al., 1995). This allele is in the Ler genetic background and has a semi-dwarf phenotype, but is not defective in germination or flower development. For the purpose of generating the ga3ox1/ga3ox2 double mutant (see below), we isolated a new ga3ox1 mutant (ga3ox1-3) and the ga3ox2-1 mutant in the Columbia-0 (Col-0) background. The ga3ox1-3 mutant was identified from the SALK T-DNA insertion collection. This new allele displayed a phenotype consistent with that of the semi-dwarf Ler ga4-2 mutant (Figure 6a). A DNA sequence analysis showed that in ga3ox1-3, two tandem T-DNA insertions are inserted in the 5′ end of exon 2 of GA3ox1 (Figure 5). Using gene-specific primers corresponding to exon 2 or spanning the insertion site we were not able to detect any GA3ox1 transcript by RT-PCR, although RT-PCR using primers within exon 1 revealed the presence of a truncated transcript (data not shown) in this mutant. Previous studies indicated that two highly conserved domains required for GA3ox enzymatic activity are encoded by exon 2 (Hedden and Phillips, 2000; Xu et al., 1997), suggesting strongly that a functional enzymatic product is not made in the ga3ox1-3 mutant. Genomic DNA blot analysis indicated that this mutant contains a single T-DNA insertion locus (data not shown).

Figure 5.

AtGA3ox1 and AtGA3ox2 T-DNA insertion mutant lines.
The schematic diagram depicts the genomic structure of AtGA3ox1 and AtGA3ox2. Each gene comprises two exons and a single intron. The numbers refer to the nucleotide length of the genomic sequence of each gene and the insertion site of each T-DNA. n.t., nucleotide.

Figure 6.

Phenotypic characterization of the ga3ox1, ga3ox2 and ga3ox1/ga3ox2 mutants compared with WT and ga1-3.
(a) Aerial portions of 19- and 40-day-old plants (top and bottom, respectively), as labeled.
(b) Rosette diameter. Measurements are the means ± SE for 15 plants per genotype.
(c) Final height. Measurements are the means ± SE from 15 plants per genotype. The ga1-3 mutant did bolt slightly, but bolting was much delayed after flowering, and only a fraction of the plants bolted.
(d) The percentage of seeds that germinated was determined for seeds for each plant line. Flowering time was measured by counting the number of days from sowing until floral buds were clearly visible or by counting the number of rosette and cauline leaves produced by the primary inflorescence stem after bolting. Measurements of flowering time are the means ± SE from 13 to 15 plants per line. Root lengths of 7-day-old seedlings are from the average ±SE of 60–100 seedlings.
(e) Effect of exogenous GA4 and GA9 on the vegetative growth of ga1-3 and 3ox1/3ox2 mutant plants. After germination was induced by exogenous GA4, seedlings were grown for 14 days on GA-free agar media and then transferred to agar media containing GA4 or GA9 (or no GA as a control) as indicated. The agar media contained 1 μm paclobutrazol throughout the experiment to inhibit endogenous GA biosynthesis. Photographs were taken 3 days after the transfer to the GA-containing media.
(f) Effect of exogenous GA4 and GA9 on seed germination of ga1-3 and 3ox1/3ox2 mutant seeds. Seeds were imbibed in the light at 22°C for 7 days.

The ga3ox2-1 mutant was isolated by screening T-DNA mutant lines at the Kazusa DNA Research Institute (Chiba, Japan) using a PCR-based method. The T-DNA is inserted at the 5′ end of exon 2 (Figure 5) and RT-PCR experiments using gene-specific primers spanning the insertion site confirmed that this is a null allele (data not shown). The ga3ox2-1 mutant was backcrossed once to Col-0 to remove a possible second mutation. Genomic DNA blot analysis indicated that all F2 individuals homozygous for the ga3ox2-1 mutation (n = 22) contained a single T-DNA insertion locus (data not shown). A selected ga3ox2-1 line after the backcross was used for the following experiments. Figure 6 shows that the homozygous ga3ox2-1 mutant did not have a visible phenotype, probably due to functional redundancy with GA3ox1.

In the rest of this paper we will refer to the new T-DNA insertion alleles (in the Col-0 background) as ga3ox1 and ga3ox2. To reveal the role of GA3ox2 in development, a ga3ox1/ga3ox2 double mutant was generated by genetic crossing between ga3ox1 and ga3ox2, and double homozygous mutants were identified in the F2 generation using allele-specific PCR markers. The ga3ox1/ga3ox2 double mutant displayed a severe defect in seed germination and root growth, and a dwarf phenotype (Figure 6). The overall phenotype of the double mutant is consistent with expression patterns of these two genes.

Phenotypes of ga3ox1, ga3ox2 and ga3ox1/ga3ox2

We compared the phenotypes of ga3ox1, ga3ox2, ga3xo1/ga3ox2 with WT (Col-0) and the severely GA-deficient mutant ga1-3.

Vegetative growth and development.  Bioactive GA plays a major role in the control of leaf expansion, stem and root elongation (Davies, 2004; Fu and Harberd, 2003; Yaxley et al., 2001), as is clearly demonstrated by the severely dwarfed phenotype of the GA-deficient mutant ga1-3. Because the original ga1-3 mutant is in the Ler ecotype, for our mutant characterization experiments we used a ga1-3 line that had been crossed for six generations to Col-0. The phenotype of ga1-3 (Col-0) is similar to a new ga1 allele (in the Col-0 background), which we isolated recently from the Salk T-DNA insertion mutant collection (data not shown).

The ga3ox1 mutant has a semi-dwarf phenotype with a 50% reduction in final height and a slightly reduced rosette size as compared with WT (Figure 6a–c). Although the ga3ox2 single mutant does not have any phenotype, the ga3ox1/ga3ox2 double mutant is 30% smaller in leaf diameter and 37% shorter than ga3ox1. Based on the real-time PCR data, GA3ox1 is the predominant GA3ox expressed in the stem with close to 700 copies/105 copies UBQ11 (Figure 1 and Supplementary Table S1). The only other GA3ox expressed in the stem is GA3ox2, and this is expressed at <10 copies/105 copies of UBQ11 (Figure 1). Therefore, production of bioactive GA by GA3ox2 during early vegetative growth may be sufficient for transport to the stem to induce the partial stem elongation observed in the ga3ox1 mutant. It is also possible that GA3ox3 and/or GA3ox4 expressed in flowers and siliques may provide GA for stem elongation. Our results suggest that GA3ox2 is acting in concert with GA3ox1 for the production of bioactive GA required for leaf expansion and stem elongation. The fact that the ga3ox1/ga3ox2 mutant phenotype is not as severe as ga1-3 indicates that in the absence of these two GA 3-oxidases, the plant still contains bioactive GA at this stage of development (see below for GA analysis).

Real-time PCR analysis confirmed that GA3ox1 and GA3ox2 are the only two GA 3-oxidases expressed in WT Arabidopsis roots, suggesting that they may play a role in root development (Figure 1). We found that the roots of ga3ox1 were slightly shorter than WT, whereas the ga3ox2 single mutant has a WT root length. The ga3ox1/ga3ox2 double mutant roots displayed a severe root growth phenotype similar to the ga1-3 mutant (Figure 6d). Interestingly, the severity in root length for the double mutant is in direct contrast to the severity of the phenotype observed in the above-ground plant when compared to ga1-3.

Seed germination.  Bioactive GA is required for seed germination as demonstrated by the non-germinating phenotype of the ga1-3 mutant (Koornneef and van der Veen, 1980). Gibberellin has been proposed to play multiple roles during germination, including a role in promoting embryo growth, as well as facilitating the breakage of the seed coat for protrusion of the radicle by inducing enzymes involved in the degradation of endosperm (Groot and Karssen, 1987; Leubner-Metzger et al., 1996). Both GA3ox1 and GA3ox2 are expressed at relatively high levels in the cortex and endodermis of the embryo axis of germinating seeds, as shown by in situ hybridization and promoter–GUS studies (Yamaguchi et al., 2001; Figure 3). Germination percentages of the ga3ox1 and ga3ox2 mutants were similar to WT in either the dark or the light (Figure 6d), indicating that GA3ox1 and GA3ox2 are functionally redundant. In contrast to the single mutants, the ga3ox1/ga3ox2 double mutant seed failed to germinate in the dark and had only a 5% germination frequency in the light (Figure 6d). The slight leakiness in germination of the ga3ox1/ga3ox2 double mutant is similar to the ga1-3 mutant (Silverstone et al., 1997), and may be attributed to damage to the seed coat or low levels of endogenous bioactive GAs (Silverstone et al., 2001).

Reproductive development.  In Arabidopsis, bioactive GAs promote floral induction in long-day (LD) conditions. Furthermore, GAs are essential for the development of the stamen, production of viable pollen and petal development (Koornneef and van der Veen, 1980). We determined the days to flower and the number of leaves formed on the primary inflorescence stem as a measurement of flowering time in LD (Figure 6d). The ga3ox1 mutant had the same number of leaves as WT, but it flowered 3 days later than WT. The same trend was observed for the ga3ox1/ga3ox2 double mutant, with the same number of leaves as WT, but flowering was delayed for 7 days. The GA-deficient mutant ga1-3 flowered much later (both developmentally and chronologically) than WT and ga3ox1/ga3ox2. No observable phenotype was noted in the flowers and the fertility (average number of seeds/silique on primary inflorescence stems) of the ga3ox1, ga3ox2 and ga3ox1/ga3ox2 mutant plants.

Endogenous GA levels in the ga3ox mutants and their response to exogenous GAs

To assess the status of the GA biosynthetic pathway in the ga3ox mutants, we determined concentrations of precursor (GA12,15,9), bioactive (GA4) and deactivated (GA51,34) GAs in WT, ga3ox1 and ga3ox1/ga3ox2 rosette plants (Figure 7a). The bioactive level of GA4 was decreased in the ga3ox1 mutant and more severely in the ga3ox1/ga3ox2 double mutant. The amount of GA4 in the ga1-3 mutant at a similar vegetative stage was lower than that in the ga3ox1/ga3ox2 double mutant (data not shown), consistent with their degrees of dwarfism (Figure 6a). In contrast to the level of GA4, the level of GA9 was increased in ga3ox1 plants, and more drastically in ga3ox1/ga3ox2 plants, relative to that in WT plants (Figure 7a). These data are in agreement with the predicted reduction in GA3ox activity in the mutants and with the results from initial characterization of the original ga3ox1 mutant allele (ga4-1; Talon et al., 1990). A reduction in the level of GA12 and an increase in the level of GA15 by the ga3ox1 and ga3ox1/ga3ox2 mutations suggest that 20-oxidation of GA12 is activated via feedback regulation.

Figure 7.

Levels of endogenous GA in the ga3ox mutants.
(a) Concentrations of non-13-hydroxylated GAs in aerial portions of rosette plants. Plants were 14–19 days old at the harvest. The level of GA4 is also shown in the inset with a different y-axis scale for clarity. Data are means ± SE from duplicate determinations. *GA24 could not be quantified reliably because of co-migration of impurities. **GA4 in the 3ox1/3ox2 mutant could not be quantified due to low abundance in the initial duplicate measurements, but was quantified in a separate experiment using a larger amount of plant material. g f.w.; gram fresh weight.
(b) Levels of GA4 in seeds imbibed under continuous white light. Germination (breakage of the seed coat) was initially observed around 32 h for WT and the single mutants. The level of GA4 in the 3ox1/3ox2 double mutant was determined only at 36 h. Data are means from duplicate experiments. For clarity, error bars are shown only for WT and the 3ox1/3ox2 samples. Values for the 3ox1 and 3ox2 single mutants deviated similarly to those for WT.

A small amount of GA4 detected in rosette plants of the ga3ox1/ga3ox2 double mutant suggests that GA9 is still converted to GA4 in the absence of functional GA3ox1 and GA3ox2. We examined the effect of exogenous GA9 on the growth of 14-day-old ga3ox1/ga3ox2 mutant plants to deduce the degree of blockage of GA3ox activity (Figure 6e). For the purpose of examining the effect of exogenous GA9, plants were grown in the presence of paclobutrazol, which blocks endogenous GA biosynthesis by inhibiting ent-kaurene oxidase. Figure 6(e) shows that GA9 was not as effective as GA4 in rescuing the dwarf phenotype of the ga3ox1/ga3ox2 mutant in contrast to the case with the ga1-3 mutant, which is able to convert GA9 to bioactive GA4. Nevertheless, a slight recovery of the dwarf phenotype in response to exogenous GA9 suggests that there is a minor level of GA3ox activity in 14-day-old ga3ox1/ga3ox2 mutant plants (Figure 6e).

In germinating Arabidopsis seeds, GA4 content increases prior to germination (breakage of the seed coat) (Ogawa et al., 2003). As shown in Figure 7(b), the level of GA4 increased in germinating WT, ga3ox1 and ga3ox2 seeds. However, GA4 was significantly less abundant in ga3ox1/ga3ox2 seeds at 36 h than in the other three samples, and was comparable in abundance to that at 12 h in WT seeds. These results are in agreement with the reduced ability of the ga3ox1/ga3ox2 double mutant to germinate (Figure 6d). The GA4 present in imbibed ga3ox1/ga3ox2 seeds may be attributed to weak GA3ox4 expression at this stage or to relatively high expression of GA3ox3 and/or GA3ox4 in siliques, which might provide bioactive GA to seeds (Figure 1).

Exogenous application of GA4, but not GA9, was able to induce germination of ga3ox1/ga3ox2 mutant seeds in a dose-dependent manner (Figure 6f). These results are consistent with the premise that the conversion of GA9 to GA4 is blocked in the double mutant.


GA 3-oxidases exhibit distinct tissue- and cell-specific expression patterns

In this study, we have taken a molecular genetics approach to examine the site of expression and the role of the GA3ox gene family in regulating bioactive levels of GA and plant development in Arabidopsis. We clearly demonstrated that each gene has a unique organ-specific expression pattern, suggesting that they have distinct roles in plant growth and development. GA3ox1 showed relatively high levels of expression in all organs examined. This was in contrast to the other three GA3ox genes, which showed expression patterns restricted to specific organs. In general, the expression patterns of the GA3ox genes suggest that GA3ox1 and GA3ox2 are the major genes required for germination and vegetative growth, and that GA3ox1, GA3ox3 and GA3ox4 are important for the development of reproductive organs.

Characterization of GA3oxGUS lines has identified the potential sites of synthesis of bioactive GAs produced by the GA3ox1 and GA3ox2 genes during the growth and development of Arabidopsis. Our expression analysis indicates that the expression of these two GA3ox genes is restricted to specific cell types in rapidly growing tissues and organs undergoing cell expansion, elongation and division, including germinating seeds, shoot apices, roots, stems and stamen filaments. Although GA3ox1 and GA3ox2 were expressed in similar cell types (cortex and endodermis) in embryos of germinating seeds (Yamaguchi et al., 2001), the differential tissue and cell-type expression patterns of these two genes diverged during later stages of growth, suggesting that each gene plays a unique role in the production of bioactive GA for plant development. Although expression of GA3ox in roots has been reported in other plant species, the cell-type-specific expression patterns of GA3ox genes in the roots has not been clearly defined (Itoh et al., 1999; Kaneko et al., 2003). We observed an interesting pattern of expression of GA3ox genes in Arabidopsis roots. Our results suggest multiple sites of GA biosynthesis in roots, and possible separation of the GA biosynthetic pathway requiring the transport of an intermediate. Both AtGA3ox1 and AtGA3ox2 were expressed in similar cell types along the length of the vasculature of non-meristematic, non-elongation regions of roots (Figure 3c,m). Therefore, GA biosynthesis may also be occurring to a certain extent in the mature regions of roots, although, this would require transport of GA intermediates because AtCPS is not expressed in these cells (Silverstone et al., 1997). These results are consistent with the previous observation that a GA-deficient tomato mutant, gib-1, produces shorter and wider cortical cells in the elongating and mature regions in the roots (Barlow, 1991). In addition, Paquette and Benfey (2005) found that GA is involved in regulating the initiation of middle cortex during formation of ground tissue in the Arabidopsis root. Unlike AtGA3ox1, AtGA3ox2 also had a very distinct expression pattern in the root tip that was restricted to the elongation zone (Figure 3l,m), quiescent center (QC) and columella cells (Figure 3t). We observed AtGA3ox1, but not AtGA3ox2, expression in the SAM of Arabidopsis (Figure 3f,p). Strong AtGA3ox1 expression is observed in the rib meristem, with weaker staining in the corpus cells and leaf primordia. In the SAM, expression of the GA 20-oxidase gene has been shown to be restricted to the rib meristem and leaf primordia and occluded from the corpus cells due to transcriptional repression by KNOX homeodomain proteins (Hay et al., 2002; Sakamoto et al., 2001). These data suggest that GA biosynthesis is probably excluded from the corpus cells of the SAM. In addition, repression of AtGA3ox1 transcripts was not detected in plants overexpressing KNOX proteins, suggesting that GA20ox genes may be specifically targeted for repression of GA biosynthesis (Hay et al., 2002). Taken together, these observations suggest that exclusion of GA from the SAM may be mediated at different steps in the GA biosynthetic pathway. It should be noted, however, that due to the strong GA3ox1–GUS staining in the rib meristem we cannot rule out the possibility that the corpus cell staining may be due to slight diffusion from the rib meristem. AtGA3ox1 also showed a very strong and striking expression pattern in the stamen filaments of Arabidopsis flowers (Figure 4a–c). This cell-type specificity of AtGA3ox1 expression is consistent with a recent finding by Gomez-Mena et al. (2005), who detected AtGA3ox1 mRNA specifically in the stamen filaments in flowers by using in situ RNA hybridization. Our expression pattern of AtGA3ox1 was consistent with the requirement of bioactive GA for elongation of stamen filaments (Koornneef and van der Veen, 1980) and the strong expression observed for GA biosynthetic genes (OsGA20ox2 and OsGA3ox) in the stamen primordia of rice flowers (Kaneko et al., 2003). However, it is not consistent with AtCPS expression in the anthers of developing flowers and suggests a possible role for AtGA3ox3 and AtGA3ox4 during this stage of development.

Cell-specific expression patterns of early and late GA biosynthetic genes during plant development

Previous studies have demonstrated the physical separation of the early and late steps in the GA biosynthetic pathway in germinating Arabidopsis seeds (Yamaguchi et al., 2001). In germinating embryos, the cellular location of expression of AtGA3ox1 and AtGA3ox2 was shown to be different from the expression site of the early GA biosynthesis gene AtCPS. In addition, the expression of the GA3 gene (AtKO) was co-localized with AtGA3ox1 and AtGA3ox2, suggesting that the transported intermediate had to be either CDP or ent-kaurene (Yamaguchi et al., 2001). This study allowed us to further explore whether separation of the GA pathway occurs in stages of development other than germinating seeds by making some comparisons between the expression of the AtGA3ox genes and the early pathway gene AtCPS characterized previously (Silverstone et al., 1997). Our observations with whole mount tissues suggest that the GA biosynthetic pathway may also be separated into different cell types in roots as described above. In most vegetative and reproductive tissues examined, however, expression of the AtCPS gene overlapped with at least one of two GA3ox genes examined. These tissues and organs include cotyledons, shoot apex, petioles, stem, leaves, stamen filaments, silique receptacle and developing embryos. In flowers, GA3ox1 showed very strong expression at the base of flower buds and in the stamen filaments, similar to AtCPS. A discrepancy was noted in developing anthers that are a site of expression for AtCPS, but not for either AtGA3ox1 or AtGA3ox2. Although we did observe extremely low levels of GA3ox2-TL-GUS activity in anthers, the physiological significance of this remains to be shown. As discussed earlier, AtGA3ox3 and AtGA3ox4 are specifically expressed in flowers and siliques as determined by real-time PCR. Preliminary observations of GA3ox3GUS and GA3ox4GUS lines show that these two GA3ox genes are expressed at high levels in anthers and/or developing seeds (MGM, N. Barnaby and TPS, unpublished results). Further experiments will be required to confirm the degree of overlap in expression between AtCPS and the AtGA3ox3 and AtGA3ox4 genes in these organs.

Physiological roles of AtGA3ox1 and AtGA3ox2 in plant growth and development

We used a genetic approach to reveal the role of AtGA3ox1 and AtGA3ox2 in plant development by isolating Arabidopsis knockout mutants in each of these genes. Because these two AtGA3ox genes are expressed at the highest levels during germination and vegetative growth (Figure 1), we predicted that knocking out both AtGA3ox1 and AtGA3ox2 would confer a severe defect in seed germination and a severely dwarfed phenotype. The double mutant certainly had a severe defect in germination and root growth similar to ga1-3, but unexpectedly exhibited weaker GA-deficient phenotypes at later developmental stages in comparison with the ga1-3 mutant (Figure 6), despite the absence of detectable accumulation of AtGA3ox3 and AtGA3ox4 transcripts in WT plants at similar developmental stages (Figure 1). We also confirmed that GA3ox3 and GA3ox4 transcripts were not detected by RT-PCR in 10-day-old seedlings or 14-day rosettes of the ga3ox1/ga3ox2 double mutant (data not shown), indicating that the GA3ox3 and GA3ox4 genes are not upregulated to compensate for the low levels of bioactive GA at these developmental stages. Nevertheless, 14-day-old ga3ox1/ga3ox2 mutant plants responded slightly to exogenous GA9, suggesting that a small amount of GA4 detectable in the ga3ox1/ga3ox2 rosette plants (Figure 7a) might be attributed to trace GA3ox activity at this stage. Alternatively, AtGA3ox4 might contribute to the occurrence of GA4 in the ga3ox1/ga3ox2 rosette plants (Figure 7a), if its expression in young seedlings (Figure 1) produced a sufficient quantity of GA4 for later vegetative development of the ga3ox1/ga3ox2 mutant. This hypothesis should be tested by generating a ga3ox1/ga3ox2/ga3ox4 triple mutant. Analysis of levels of GA3ox protein may also help interpret these observations, because it has not been demonstrated how exactly the levels of GA3ox protein correlate with their transcript abundance. Although we have not identified additional Arabidopsis GA3ox genes by searching the Arabidopsis sequence database, there may be additional novel Arabidopsis enzymes with GA3ox activity which cannot be identified based on sequence homology alone.

A large amount of GA9, the major substrate for GA3ox, accumulated in the ga3ox1/ga3ox2 mutant (Figure 7a). Besides the block at the GA 3-oxidation step, a decrease in the level of GA12 in the ga3ox1/ga3ox2 mutant suggests that the abnormal accumulation of GA9 is accelerated by upregulation of GA 20-oxidase activity through feedback regulation (Phillips et al., 1995). It should be noted that in the presence of excess GA9 GA3ox would produce a larger amount of GA4 than that is produced by the same amount of enzyme using a lower level of the substrate. Therefore, it is possible that GA3ox3 and GA3ox4 (and any enzyme with GA3ox activity) synthesize GA4 more efficiently in vegetative tissues in the ga3ox1/ga3ox2 double mutant than in WT when they are expressed at the same level. Thus, we can speculate that even a very minor GA3ox activity would account for the low level of GA4 synthesis in ga3ox1/ga3ox2 rosette plants (Figure 7a). An analogous situation might in fact be seen for the level of GA51 (Figure 7a); GA 2-oxidase activity (conversions of GA9/4 to GA51/34) might be downregulated due to a reduction in the level of bioactive GA in the ga3ox1/ga3ox2 double mutant (Thomas et al., 1999), but GA51 was much more abundant in the double mutant than in WT, possibly because of the elevated supply of the substrate, GA9.

The fact that the ga3ox1 and ga3ox2 single mutants retain WT germination characteristics, but the ga3ox1/ga3ox2 double mutant displays a severe defect in seed germination, indicates that the AtGA3ox1 and AtGA3ox2 genes function redundantly in imbibed seeds at 22°C. However, recent work has shown the GA3ox1 gene, but not the GA3ox2 gene, is induced by cold temperature (stratification) in dark-imbibed seeds, and that ga3ox1 mutant seeds are defective in both cold-stimulated GA4 accumulation and germination unlike WT seeds (Yamauchi et al., 2004). These findings emphasize the importance of investigating the relative role of individual GA3ox genes further in defined environmental conditions and the usefulness of loss-of-function mutants for elucidating how each family member contributes to determining levels of bioactive GA.

A comprehensive picture of the function of the GA3ox genes and the sites of GA biosynthesis during Arabidopsis development is beginning to unfold. Isolation of ga3ox3 and ga3ox4 mutants and construction of triple and quadruple ga3ox mutant combinations are in progress. These studies should help us elucidate the potentially unique functions of each AtGA3ox in plant growth and development. A thorough characterization of the cell-type expression pattern of AtGA3ox3 and AtGA3ox4 in flowers and developing seeds will also be necessary to complete the picture of GA biosynthesis in Arabidopsis, and will help to elucidate whether separation of GA pathways occurs in these stages of development and whether transport of GA intermediate(s) or bioactive GAs plays an important role.

Experimental procedures

Plant material and growth conditions

The Col-0 genetic background was used as WT in this study unless specified. The ga1-3 line used in this study had been crossed for six generations to Col-0. The ga3ox1-3 line was obtained from the Salk Institute Genomic Analysis Laboratory (see below). The ga3ox2-1 line was isolated by PCR screening of T-DNA insertion lines generated in the Col-0 background at the Kazusa DNA Research Institute (see below). The GA-deficient mutants ga1-3 and ga3ox1/ga3ox2 were supplied with exogenous GA by incubation with 50 μm GA4 for 7 days during stratification at 4°C. Seeds treated with GA were washed thoroughly with water before planting. To grow ga3ox1/ga3ox2 mutant plants for GA analysis we used 100 μm GA3, instead of 50 μm GA4, to assist germination. We confirmed by gas chromatography (GC)-mass spectrometry analysis that this batch of GA3 did not contain other GAs that we measured in rosette plants later (Figure 7a). Wild-type seeds were incubated in water at 4°C for 3 days prior to planting. Plants were grown on MetroMix 200 (Scotts-Sierra Horticultural Products, Marysville, OH, USA) at 22°C under LD conditions (16 h light/8 h darkness). Seeds were incubated on wet filter paper as described previously (Yamaguchi et al., 1998a,b) for germination tests and GA measurements in imbibed seeds.

Identification of T-DNA insertion mutant lines

The ga3ox1-3 mutant was isolated by searching the SALK Institute Genomic Analysis Laboratory (SIGnAL) T-DNA express database for ga3ox1 T-DNA insertion mutants in the Col-0 background (Alonso et al., 2003). T3 seed corresponding to Salk_004521 was requested from the Arabidopsis Biological Resource Center and two tandem T-DNA insertions were found to be present in the 5′ end of exon 2 of the AtGA3ox1 gene by PCR using a T-DNA left border primer (JMLB1) and GA3ox1-specific primers (5′GA4 BamHI and 3′GA4 SalI). Homozygous mutant plants were identified by PCR using GA3ox1-specific primers (5′GA4 BamHI and 3′GA4 SalI) flanking the T-DNA insertion site. A ga3ox2-1 mutant was isolated by screening a collection of pooled T-DNA insertion lines at the Kazusa DNA Research Institute using a PCR-based method. A T-DNA-specific primer 5′-GGAGAGCCTTCACCGGTTAGGG-3′ and gene-specific primers, 3ox2/14F (5′-GCCTTTTAGCATGAGTTCAAC-3′) and 3ox2/13R (5′-AGATCATTATATCGGATGGTG-3′), were used for the screening. These primers were also used for routine genotyping by PCR. Homozygous ga3ox1-3 and ga3ox2-1 plants were crossed and individual F1 seeds were grown and self-fertilized to obtain the F2 generation. Identification of the ga3ox1/ga3ox2 double mutant was conducted using allele-specific PCR primers. Primer sequences were as follows: JMLB1: 5′-GGCAATCAGCTGTTGCCCGTCTCACTGGTG-3′; 5′GA4 BamHI: 5′-CGCGGATCCGTGTTTAGAGGCCATCCCATTCAC-3′; 3′GA4 SalI: 5′-ACGCGTCGACCAAACAAATCATATTGCTGAAATC-3′.

Analysis of endogenous GAs

Quantitative analysis of GAs was carried out by GC-selected ion monitoring (SIM) using 2H-labeled GAs as internal standards, as described previously (Gawronska et al., 1995). Briefly, a pre-purified ethyl acetate-soluble fraction containing GAs was subjected to HPLC purification using a reverse phase column (Capcell Pak C18 SG120; Shiseido Fine Chemicals, Tokyo, Japan). When necessary, fractions containing GA after reverse phase-HPLC were further purified through another round of HPLC using an ion exchange column (Senshu Pak N[CH3]2, 1151-N; Senshu Scientific, Tokyo, Japan) to ensure the removal of impurities for reliable quantifications. The purified fractions were subjected to GC-SIM analysis using a mass spectrometer (Automass Sun; JEOL, Tokyo, Japan) equipped with a GC (6890N; Agilent Technologies, Palo Alto, CA, USA) and a capillary column (DB-1; Agilent Technologies) after derivatization. Authentic GA samples and 2H-labeled internal standards were purchased from Professor Lewis Mander (Australian National University, Canberra). Approximately 20 g (fresh weight) of plant tissue was used to measure GA levels in rosette plants. To determine GA4 levels in imbibed seeds, 0.5 g of dry seeds was used as the starting material. Measurements of GA were performed twice using independent plant materials. At the rosette stage of the 3ox1/3ox2 double mutant GA4 was quantified in a separate experiment using 50 g (fresh weight) of the plant material.

Mutant characterization

For germination assays, seeds were washed briefly with 0.02% Triton X-100 followed by several washes with sterile water before sowing in Petri plates on moist filter paper. The seeds were incubated at 22°C under continuous white light and the percentage of seeds that germinated was measured after 7 days. Measurements of the final root length were taken from plants grown vertically at 22°C under LD conditions for 7 days on 1× Murashige and Skoog (MS) agar (containing 2% sucrose, 1.5% agar) in square Petri plates. The longest rosette leaf was determined as a measure of rosette radius. The flowering time was marked at the first visible sign of the flower bud to the naked eye. Images were captured with a Nikon 990 digital camera.

Construction of AtGA3ox1–GUSgene fusions

Two AtGA3ox1–GUS gene fusions were constructed. p3ox1-TC-GUS was a transcriptional fusion of approximately 3 kb of the AtGA3ox1 promoter region upstream of the translational start site fused to the GUS reporter gene. p3ox1-TL-GUS was a translational fusion containing approximately 3 kb of the AtGA3ox1 promoter region upstream of the translational start site including the first exon, intron and 186 nucleotides up to a native NcoI site of the second exon fused to the GUS reporter gene. Genomic DNA fragments were amplified by PCR from Arabidopsis BAC Clone T16N11 using primers containing BamHI and SalI restriction sites (for p3ox1-TC-GUS: 5′3ox13kbpr SalI and 3′3ox15′UTR BamHI-2 were used; for p3ox1-TL-GUS: 5′3ox13kbpr SalI and 3′3ox1ex2 BamHI were used) and directionally cloned into the binary vector pBI101.1 upstream of the GUS reporter gene. Plasmid constructs were sequenced to confirm that no mutations were introduced by PCR. Construction of AtGA3ox2GUS gene fusion constructs were described previously (Yamaguchi et al., 2001). Primer sequences: 5′3ox13kbpr SalI: 5′-ACGCGTCGACCACCAGAGTGTGTGCTACATGC-3′; 3′3ox15′UTR BamHI-2: 5′-CGCGGATCCAACACAGCAGGCAGCTTGCTC-3′; 3′3ox1ex2 BamHI: 5′-CGCGGATCCAGCTGCTAGACCCATGGCTCG-3′.

Plant transformation

Plasmid constructs were introduced into the Agrobacterium tumefaciens strain GV3101/pMP90, and then into WT or ga3ox mutant plants by the Agrobacterium-mediated floral dipping method (Clough and Bent, 1998). Transformants were selected on MS agar plates containing 50 μg ml−1 kanamycin. Plants containing a single insertion were determined by identifying plants that showed 3:1 kanamycin resistant versus kanamycin-sensitive segregation patterns in the T2 generation. Unless specified, at least 10 independent homozygous lines were identified in the T3 generation for each construct.

Quantitative real-time PCR

Absolute mRNA levels for AtGA3ox1AtGA3ox4 were assessed using real-time PCR in the following WT tissues: seeds imbibed in water for 12 or 24 h under continuous light, 2-, 5- and 10-day-old seedlings, roots of 5-day-old seedlings, entire aerial portions (rosettes) of 14-day-old plants, and stem tissue, flower clusters, siliques and cauline leaves from 35-day-old plants (Tyler et al., 2004). Stem segments were defined as the internodes from the base of the rosette to the base of the primary inflorescence.

Plant tissues were harvested, and DNA-free RNA was isolated as described (Tyler et al., 2004). The First Strand cDNA Synthesis Kit for RT-PCR (Roche Diagnostics, Mannheim, Germany) was used to make cDNA from 1 μg of RNA. Each cDNA reaction was purified on Qiagen Mini Quickspin Columns (Qiagen, Valencia, CA, USA). The LightCycler FastStart DNA Master SYBR Green I Kit (Roche, Mannheim, Germany) was used to prepare half-reactions (10 μl) containing the following: 1 μl SYBR Green I reaction mix, 3 mm MgCl2, 0.5 μm forward and reverse primers and 1 μl cDNA. A Roche LightCycler real-time PCR machine was used in all experiments according to the manufacturer's instructions. An annealing temperature of 58°C was employed for all reactions. Gene-specific real-time PCR primers were designed using the Roche LightCycler Primer Design Software. The following primer sets were used: for AtGA3ox1 (At1g15550), 5′-CCATTCACCTCCCACACTCT-3′ and 5′-GCCAGTGATGGTGAAACCTT-3′; for AtGA3ox2 (At1g80340), 5′-TGGTCCGAAGGTTTCAC-3′ and 5′-GGGTCGAGTCTGTATGG-3′; for AtGA3ox3 (At4g21690), 5′-TCCTACCCGGTTTGCC-3′ and 5′-ACGGTGCATTGTACTTC-3′; for AtGA3ox4 (At1g80330), 5′-GCCGATGACTCCTACC-3′ and 5′-ACACTTGTAGCCCTCC-3′; for UBQ11 (At4g05050), 5′-GCAGATTTTCGTTAAAACC-3′ and 5′-CCAAAGTTCTGCCGTCC-3′. The mean value of three replicates was normalized using the UBQ11 gene as the control after previously confirming consistent expression levels in all tissues examined (Tyler et al., 2004). Standard curves were generated using linearized plasmid DNA for each gene of interest and the copy number was calculated and normalized to UBQ11 levels (see Supplementary Table S1). A second set of experiments was conducted on an independent set of tissue as a control.

Histochemical GUS assays

β-Glucuronidase (GUS) assays were conducted as described previously (Jefferson, 1987; Yamaguchi et al., 2001). Whole mount photographs were taken under a SZX12 dissecting microscope (Olympus, Tokyo, Japan) and captured with a Nikon 990 digital camera.

For sectioning, X-gluc stained tissues were fixed and embedded in paraffin as described (Yamaguchi et al., 2001). Photographs of thin sections were taken using bright field microscopy.


The authors wish to thank Jinyoung Yang and Natalie Weaver for technical assistance. This research was supported by US Department of Agriculture National Research Initiative Competitive Grants 01-35304-10892 and 03-35304-13284.