The GAs (gibberellins) comprise a large group of diterpenoid carboxylic acids that are ubiquitous in higher plants, in which certain members function as endogenous growth regulators, promoting organ expansion and developmental changes. These compounds are also produced by some species of lower plants, fungi and bacteria, although, in contrast to higher plants, the function of GAs in these organisms has only recently been investigated and is still unclear. In higher plants, GAs are synthesized by the action of terpene cyclases, cytochrome P450 mono-oxygenases and 2-oxoglutarate-dependent dioxygenases localized, respectively, in plastids, the endomembrane system and the cytosol. The concentration of biologically active GAs at their sites of action is tightly regulated and is moderated by numerous developmental and environmental cues. Recent research has focused on regulatory mechanisms, acting primarily on expression of the genes that encode the dioxygenases involved in biosynthesis and deactivation. The present review discusses the current state of knowledge on GA metabolism with particular emphasis on regulation, including the complex mechanisms for the maintenance of GA homoeostasis.
The name GA (gibberellin) encompasses a large group of tetracyclic diterpenoid carboxylic acids with the ent-gibberellane (C20) or 20-nor-ent-gibberellane (C19) carbon skeletons, as exemplified by the simplest members of each class, GA12 and GA9 respectively (Figure 1). Although first identified as secondary metabolites of the fungus Gibberella fujikuroi, they are ubiquitous in higher plants, in which certain members of the group function as endogenous growth regulators. They are also present in certain species of endophytic bacteria, several fungal species and some lower plants, but their function in these organisms is unclear. Currently 136 gibberellin structures have been identified from all sources and are assigned the trivial names gibberellin A1–A136, abbreviated to GA1 etc., numbered in order of discovery . Their structures can be found at http://www.plant-hormones.info/gibberellins.htm. There are certainly further GAs still to be discovered, but due to their very low abundance, their chemical characterization is far from trivial, depending on comparison with chemically synthesized standards, so that identification of new GAs has become an increasingly rare event.
Structures of selected GAs
Relatively few GAs have intrinsic biological activity in higher plants, the most common active forms being GA1, GA3 and GA4 (Figure 1). The universal occurrence of GA1 and GA4 in plants suggests that these are the functionally active (hormonal) forms, at least for growth promotion; they co-occur with their biosynthetic precursors and metabolites, which are often present at much higher concentrations than the hormones themselves. Gibberellin A3, also known as gibberellic acid, is best known as the main GA product of G. fujikuroi (now reclassified as Fusarium fujikuroi), from which it is produced for commercial application, but it is also endogenous in some higher plant species. Developing seeds often contain very high concentrations of GAs, as well as a broad range of different structural forms. In fact, the vast majority of the different GA structures were identified from immature seeds, whereas vegetative and early reproductive tissues contain a relatively restricted and conserved array of structures, representing members of the biosynthetic pathways to GA1 and GA4 and their metabolites. Although there is clear evidence for a regulatory role for GAs in early seed development , the function of the very high concentrations of active and inactive GAs present in endosperm, cotyledons and/or testa of seeds at later developmental stages is unclear. Seeds approaching maturity often contain high levels of GA-inactivating activity, ensuring against concentrations of bioactive GAs in mature seed that could provoke premature germination and abnormal seedling growth .
Physiologically, a major function of GAs in higher plants can be generalized as stimulating organ growth through enhancement of cell elongation and, in some cases, cell division. In addition, GAs promote certain developmental switches, such as between seed dormancy and germination, juvenile and adult growth phases, and vegetative and reproductive development. In this last case, GAs may promote the vegetative or reproductive state, depending on species. Gibberellins have an important role in fertility, as, in addition to allowing stamen elongation, they are necessary for the development, release and germination of pollen and for pollen tube growth. In carrying out their functions GAs act in response to developmental and environmental cues, which may regulate GA biosynthesis, turnover (deactivation), perception or signal transduction, sometimes acting at several points in the pathway.
Impressive progress has been made within the last decade in establishing details of the GA biosynthesis and signal transduction pathways. The advances in our understanding of the GA signal transduction cascade (reviewed in [4–7]) have stemmed particularly from the discovery that GID (GIBBERELLIN-INSENSITIVE DWARF) 1 encodes a soluble GA receptor in rice . A pathway has emerged in which growth repression is exerted by DELLA proteins through their regulation of transcription (; also reviewed in ). In response to environmental and developmental stimuli, this repression is relieved through the production of biologically active GAs, which rapidly promote DELLA degradation. Gibberellins bind within a pocket of GID1, causing a conformational change that facilitates the formation of a GA–GID1–DELLA complex and the subsequent ubiquitination of DELLAs and their destruction by the 26S proteasome ([10–12]; also reviewed in ). In the absence of a functional degradation pathway, binding of GA–GID1 has also been demonstrated to have a repressive effect on DELLA function, although the physiological significance has yet to be established [13,14]. There is currently little evidence that DELLAs act as transcription factors that directly control the expression of GA response genes. However, recent studies demonstrated that DELLA repression of hypocotyl expansion is mediated through their binding and sequestration of the PIF (PHYTOCHROME-INTERACTING FACTOR) 3 and PIF4 bHLH (basic helix–loop–helix) transcription factors [15,16]. There is now strong evidence that DELLAs also regulate the activity of other classes of transcriptional regulators, including the JAZs (JASMONATE ZIM-domain proteins). In this case, DELLAs were shown to modulate jasmonate signalling by binding to JAZs and thereby blocking their inhibitory effect on downstream transcriptional events .
The present review focuses on GA biosynthesis and deactivation: after a discussion of the pathways, we will highlight recent advances in our understanding of how they are regulated.
THE GA BIOSYNTHETIC PATHWAY
This topic was reviewed previously by Yamaguchi . In higher plants, GAs are formed primarily from the methylerythritol phosphate pathway , by which the hydrocarbon intermediate ent-kaurene is produced from GGPP (trans-geranylgeranyl diphosphate) in proplastids . The biosynthetic pathway from GGPP to the bioactive GAs is presented in Figure 2. It has been generally assumed that ent-kaurene synthesis competes with that of other GGPP-derived metabolites for a common pool of GGPP (e.g. ). However, the finding that loss of geranyl diphosphate synthase in tomato and Arabidopsis thaliana (hereafter referred to as Arabidopsis) reduces GA content without affecting the levels of carotenoids and chlorophylls suggests that ent-kaurene is produced from a separate pool of GGPP, via a GGPS (GGPP synthase) that, in contrast with the GGPS isozymes involved in pigment synthesis, has a requirement for GPP as substrate . This would require that GAs and the GGPP-derived pigments are produced in different tissues and is consistent with ent-kaurene synthesis in leaves occurring in immature chloroplasts associated with the vasculature . Plants contain relatively many GGPS genes, the Arabidopsis genome containing 12 such genes, with most of the encoded enzymes localized in plastids , and it is feasible that one or more of these may be dedicated to GA biosynthesis. The fungus F. fujikuroi contains two GGPS genes, one of which is located in the GA biosynthesis gene cluster and is used exclusively for GA biosynthesis .
The GA-biosynthetic pathway from trans-geranylgeranyl diphosphate to GA1, GA3 and GA4
Synthesis of ent-kaurene from GGPP occurs in two steps via ent-copalyl diphosphate, the reactions catalysed by separate enzymes, CPS (ent-copalyl diphosphate synthase), a class II (proton-initiated) cyclase and KS (ent-kaurene synthase), a class I (initiated by phosphate ionization) cyclase respectively, in higher plants . These reactions are also catalysed by separate CPS and KS enzymes in the GA-producing bacterium Bradyrhizobium japonicum , whereas in fungi both activities reside in a single polypeptide . In Arabidopsis, both CPS and KS are encoded by single genes, loss of function of which results in severe GA-deficient phenotypes, characterized by extreme dwarfism, male and female infertility and non-germinating seeds . Several species, particularly cereals, have been found to express two or more CPS- and KS-like genes [30–33], although in rice only one member of each gene family is thought to be involved in GA biosynthesis, the others producing phytoalexins [30,33].
ent-Kaurene is relatively volatile and has been found to exchange with the external environment, prompting the suggestion that it may function as a mediator of plant–plant communication . However, since regulation of GA biosynthesis occurs mainly at later stages of the pathway , exogenous ent-kaurene is unlikely to have a major influence on GA content and the resulting growth.
The conversion of ent-kaurene into the bioactive forms involves the action of membrane-associated P450s (cytochrome P450 mono-oxygenases) and soluble ODDs (2-oxoglutarate-dependent dioxygenases). The formation of GA12, which is considered the common precursor for all GAs in plants , requires six oxidative steps, catalysed by two mono-oxygenases, KO (ent-kaurene oxidase) and KAO (ent-kaurenoic acid oxidase). KO belongs to the CYP701A P450 clade and KAO to CYP88A . These enzymes are localized in the endoplasmic reticulum and, in the case of KO, also in the plastid envelope . The three-step KO-catalysed oxidation of ent-kaurene to ent-kaurenoic acid via ent-kaurenol and ent-kaurenal has been shown to involve repeated hydroxylations on C-19, with the initial step rate-limiting and the intermediates retained at the enzyme active site . KO is encoded by a single gene in Arabidopsis, whereas rice contains a cluster of five KO-like genes , one of which, OsKO2, was shown by heterologous expression in yeast to possess KO activity . Mutations in OsKO2 cause severe GA-deficiency and dwarfism, confirming its involvement in GA-biosynthesis and suggesting it may be the only gene of the cluster to have this role .
The oxidation of ent-kaurenoic acid to GA12 by KAO occurs in three steps via the intermediates ent-7α-hydroxykaurenoic acid and GA12-aldehyde, requiring successive oxidations at C-7β, C-6β and C-7 (evidence summarized in ). In fungi and developing seeds of some plant species in which GAs are produced at very high levels, the first two of these reactions produce by-products (kaurenolides and fujenoic acids respectively) in substantial quantities, but these do not appear to be formed in tissues in which GAs are produced in hormonal quantities. The fungus F. fujikuroi also employs cytochrome P450s for the KO and KAO reactions, but they are unrelated to the higher plant enzymes, belonging to the CYP503 and CYP68 clades respectively, and appear to have evolved independently . The F. fujikuroi KAO produces mainly GA14 (3β-hydroxyGA12) rather than GA12 due to 3β-hydroxylation of GA12-aldehyde, an activity not present in the higher plant enzymes .
Whereas rice contains a single KAO gene , Arabidopsis contains two fully redundant copies . Two KAO genes have also been reported in pea and sunflower, but these genes differ in expression domains, with one gene being expressed in most tissues, whereas expression of the second gene is restricted to developing seeds and roots [45,46]. The pea enzymes were found to produce small amounts of the non-GA by-products in vitro , perhaps in response to high substrate concentrations. Lesions in KAO have been shown to be the cause of severe dwarfism in several species, including rice (d35 mutant) , maize (dwarf3) , pea (na)  and sunflower (dwarf2) . Pumpkin contains an ODD that oxidizes GA12-aldehyde to GA12 , but orthologues of this enzyme have not yet been found outside of the Cucurbitaceae.
In higher plants, GA12 lies at a branch-point in the pathway, undergoing hydroxylation on C-13 and/or C-20. The nature of the enzymatic activity responsible for 13-hydroxylation, by which GA12 is converted into GA53, is unclear. Both P450 and ODD 13-hydroxylase activities have been detected (reviewed in ), and it has been recently reported that rice contains two P450s that convert GA12 into GA53 . Interestingly, the enzymes present in Arabidopsis (CYP714A) and Stevia rebaudiana (CYP716D) 13-hydroxylate ent-kaurenoic acid rather than GA12 [50,51].
GA12 and GA53 are oxidized on C-20 by GA20ox (GA 20-oxidase), an ODD, converting these substrates in parallel pathways into GA9 and GA20 respectively. This is a three- or four-step process, depending on the mechanism for the loss of C-20, which has not been determined . In this reaction series, the C-20 methyl group is oxidized to the alcohol and then the aldehyde, from which it is lost with the formation of a γ-lactone between C-19 and C-10. The alcohol and aldehyde intermediates accumulate in plants, the aldehydes often to relatively high levels , consistent with them being released and having to rebind to the enzyme active site for the successive oxidations. In contrast with the accumulation of these intermediates, the intermediates in the P450-catalysed reactions earlier in the pathway are normally present in only trace quantities . The C-20 alcohol readily lactonizes with the C-19 carboxylic acid group, particularly at low pH, preventing further oxidation by the GA20ox . Moreover, some tissues contain a 20-oxidase activity capable of oxidizing C-20 in its lactone form , although there is little information on the nature of the enzyme responsible for this activity. It is not known to what extent lactonization of C-20 alcohols occurs in planta, because the acidic conditions used during GA analysis promotes lactone formation, but the existence of the lactone-oxidizing activity suggests that these δ-lactones may occur naturally.
Little is known about the mechanism by which C-20 is lost, although it appears to involve the formation of a free radical on C-10 that reacts with the C-19 carboxylic acid to form the lactone . The finding that C-20 is lost directly as CO2  necessitates the existence of an intermediate between the C20 aldehyde and the C19 γ-lactone product. However, no candidate intermediate has been identified and it has been suggested that it remains bound at the enzyme active site . A side reaction from the oxidation of the aldehyde results in the formation of a carboxylic acid. This reaction can occur from a substantial to a negligible extent depending on the nature of the GA20ox, and is most common in developing seeds (see, e.g., ). The resulting tricarboxylic acid product is not converted into C19-GAs and constitutes a biologically inactive end-product. Oxidative loss of C-20 in the fungi F. fujikuroi and Sphaceloma manihoticola is catalysed by a P450 rather than an ODD [58,59]. Apart from GA14, the first substrate for the fungal enzyme, the proposed C20-GA intermediates (GA37 and GA36) in the formation of GA4 accumulate to a much smaller extent in fungal cultures than do the equivalent intermediates in plants, and, in contrast with the plant intermediates, are not oxidized to C19-GA products when added to fungal cultures . This may indicate that these fungal intermediates do not leave the active site, and their presence in fungal cultures may be due to hydrolysis of enzyme-bound intermediates . However, cultures of F. fujikuroi grown in the presence of 18O2 or H218O incorporated 18O only from 18O2 into the intermediates, indicating that they were not formed by hydrolysis .
The final step in the formation of the biologically active hormones is the 3β-hydroxylation of GA9 and GA20 to GA4 and GA1 respectively, catalysed by the ODD GA3ox (GA 3-oxidase). In shoots of dicotyledonous species this reaction tends to be highly regiospecific, with a single product being produced [61–63], whereas in several monocotyledons, in addition to GA1, GA3 is formed from GA20 via the intermediate GA5 by the action of a single GA3ox [64–66]. Developing seeds of certain dicotyledonous species contain GA3oxs of relatively low regiospecificity, possessing 2β-hydroxylase [65,67,68] and/or 2,3-desaturase [65,69,70] activities in addition to their 3β-hydroxylase activity. C-2,3-unsaturated GAs may be further oxidized by GA3oxs to the 2,3-epoxide or to the 3β-hydroxy-1,2-unsaturated products. Thus GA5 may be converted into GA6 or GA3, this latter conversion involving abstraction of the 1β-H, rearrangement of the 2,3-double bond to the 1,2-position and then addition of OH at C-3β . Whereas both desaturation and formation of the 3β-hydroxy-1,2-unsaturated product are catalysed by a single enzyme in cereals, separate enzymes are required for these reactions in developing Marah macrocarpus seeds . In an example of extreme lack of regiospecificity, two GA3ox enzymes, from wheat  and M. macrocarpus seeds , were shown to possess some 13-hydroxylase activity. This requires that the substrate binds to the enzyme active site in opposite orientations to be oxidized either on C-3 or C-13, and indeed C19-GA substrates, such as GA9, possess some symmetry, with the 6-carboxylic and 19,10-lactone moieties potentially able to exchange positions in the binding site. Although the 13-hydroxylase activity of the M. macrocarpus enzyme may have some significance in seeds of this species, in which low amounts of 13-hydroxy-C19-GA, but no 13-hydroxy-C20-GAs, are present, consistent with this reaction occurring late in the pathway, early 13-hydroxylation by a regiospecific enzyme appears to be the most common route in higher plants. In the fungus F. fujikuroi, 13-hydroxylation of GA7 to GA3 by a P450 is the final step in the biosynthesis of GA3 in this species .
In contrast with the earlier enzymes, the ODDs are encoded by multiple genes in all plant species that have been studied: for example there are five GA20ox enzymes in Arabidopsis and four in rice, whereas these species contain four and two GA3ox genes respectively . Members of these gene families vary considerably in their levels and patterns of expression, and are differentially regulated (see below). This multiplicity of genes may be related to the fact that the GA biosynthetic pathway is regulated primarily by the activity of the ODDs.
It is essential for plants to be able to regulate precisely their GA content and to have the ability to change this parameter rapidly in response to changes in their environment. In common with other plant hormones, GAs are inactivated and this provides a means to regulate homoeostasis and to allow a rapid reduction of GA concentration when required. Several mechanisms for inactivating GAs have been identified (Figure 3), the most prevalent being 2β-hydroxylation. The enzymes responsible for this activity, GA2oxs (GA 2-oxidases), are ODDs, which can be divided into two major groups according to function: one group acts on C19-GAs, including the bioactive compounds and their immediate non-3β-hydroxylated precursors , whereas another group acts on C20-GAs . The C19-GA2oxs can be divided into two subgroups on the basis of amino acid sequence [75–77] and it has been proposed that they differ also in biochemical function . Several C19-GA2oxs, all of which belong to the same subgroup, have been shown to be bifunctional, acting as 2β-hydroxylases, but also further oxidizing the 2β-hydroxy group to a ketone. The isolated 2-keto products, known as GA-catabolites , are dicarboxylic acids with a double bond at C-1,10, C-5,10 or C-10,11 (shown for the Δ1,10 isomer in Figure 3) that are thought to arise from the 2-ketone 19,10-lactones by rearrangement, probably during analysis . The catabolites can accumulate to high concentrations in seeds of some species, for example, pea . The second subgroup of C19-GA2oxs do not appear to possess the capacity to produce the GA-catabolites . The GA2ox enzymes form particularly large gene families, although, on the basis of sequence phylogeny, the C20-GA2oxs might be considered a separate family from those acting on C19-GAs . In common with the other ODDs, the genes encoding GA2oxs are major sites of regulation (see below).
Reactions that result in GA deactivation
Other mechanisms for GA deactivation have been recognized in recent years. A cytochrome P450 mono-oxygenase (CYP714D1), which is encoded by the EUI (ELONGATED UPPERMOST INTERNODE) gene in rice, was shown in vitro to convert GAs into their 16α,17-epoxides . The enzyme accepted GA12, GA9 and GA4 as substrates (other 13-desoxyGAs were not tested), but was much less effective with 13-hydroxylated GAs. Overexpression of EUI in rice caused extreme dwarfism and a reduction in GA4 content in the uppermost internode, confirming that epoxidation resulted in GA deactivation. However, it was not possible to identify the epoxides in plant tissues, even when EUI was overexpressed, but the accumulation of GA-16α,17-dihydrodiols indicates that the epoxides are hydrated in planta, either enzymatically or by a non-enzymatic acid-catalysed reaction. GA dihydrodiols have been found to occur naturally in many plant species (discussed in ), suggesting that epoxidation may be a general deactivation mechanism. Arabidopsis contains two CYP714 enzymes, one of which, CYP714A2, 13-hydroxylates ent-kaurenoic acid to steviol . Overexpression of both CYP714 genes resulted in dwarfism, with CYP714A1 producing the most severe effect , although the function of its encoded enzyme has not been reported.
Incorporation of an eui mutant allele into MS (male-sterile) rice varieties is used to improve their heading performance . MS varieties used in hybrid rice breeding suffered from failure of the panicle to emerge fully from the leaf sheaths, a syndrome that was effectively overcome by the extended upper internodes resulting from the presence of the eui mutation. The effect of this mutation is consistent with the expression pattern of EUI, which, on the basis of reporter gene analysis, occurs most strongly in the flowering spikelets and in the nodes and cell division zone of the expanding internodes . Rice anthers produce extremely high levels of GA4 , and it is possible that EUI may regulate the influx of GA4 into the stem from reproductive tissues, whereas GA1 produced in vegetative tissues  would be unaffected.
Two members of the SABATH group of methyl transferases, named GAMT (GA methyl transferase) 1 and GAMT2, present in Arabidopsis, were shown to be specific for GAs [86,87]. The enzymes methylate the 6-carboxy group of C19-GAs, a process that abolishes biological activity . Thus ectopic overexpression of these genes in Arabidopsis resulted in dwarfism. The GAMT genes are expressed in developing seeds, where they apparently function to regulate GA content, since gene knockout lines accumulated GAs . Their physiological roles may be restricted to seed development and the control of precocious germination.
Ester and ether GA conjugates have been identified in a range of plant species, with glucose the predominant conjugate partner . Although there has been interest in these compounds as potential storage and/or transport forms of GA, it was difficult to assign to them a definite physiological role, and attention to this topic has waned in recent years. Consequently, there is limited information on the enzymes involved in their formation or hydrolysis. Some GA conjugates were shown to have biological activity in bioassays, but this was thought to be due to hydrolysis to the aglycone in the assay system . Evidence for the reversibility of GA conjugate formation adds to the feasibility of their having a storage role , providing a mechanism for rapid release of bioactive GA that would, for example, stimulate seed germination. A recent report that ectopic expression of a β-glucosidase in transplastomic tobacco resulted in increased levels of several hormones including GA1 and GA4  suggests that plastids may act as a store of hormone conjugates. Conjugation might also provide a potentially reversible deactivation mechanism, although evidence for this is lacking. It is of interest that 2β-hydroxyGAs form glucosyl ethers linked through the 2β-hydroxyl group . In this case, conjugation may serve to solubilize and trap these inactive end-products in the vacuole.
REGULATION OF GA BIOSYNTHESIS
Sites of synthesis
In general, the highest levels of bioactive GAs are found in actively growing organs, for example developing leaves and expanding internodes . Such organs contain high expression levels of GA-biosynthetic genes, indicating active GA synthesis [23,94]. On the basis of the coincidence of expression of GA20ox and GA3ox genes with that of the GA signal transduction gene SLR1 (SLENDER1) in growing organs of rice, Kaneko et al.  concluded that GAs are synthesized at their site of action in such organs. There are several examples of different reactions of the pathway occurring in separate tissues, requiring the movement of intermediates between cells. On the basis of promoter–GUS (β-glucuronidase) reporter analysis and in situ hybridization, Yamaguchi et al.  found that, whereas CPS was expressed in the provasculature of the embryonic axis in germinating Arabidopsis seeds, KO and GA3ox genes were expressed in the cortex and endodermis. This would require the movement of an intermediate, potentially ent-kaurene, from the provasculature, a highly plausible scenario given the demonstrated release of this intermediate into the environment . The synthesis of ent-kaurene from GGPP can be considered the gateway into the GA biosynthetic pathway and as such would control the metabolic flux into the pathway . On the basis of incubations with cell-free systems from wheat tissues, Aach et al.  concluded that ent-kaurene is synthesized in proplastids within regions of cell division, but not in fully developed chloroplasts in elongating or mature cells. In the stem, they found the highest activity in the node, assumed to be associated with the intercalary meristem. This region was also reported to be the major site of TaGA20ox1 expression in the upper wheat stem, with TaGA3ox1 expression divided more evenly between the nodes and internodes [64,96]. Since cell elongation in the internode requires GA signalling , this distribution of activities would require movement of GAs or precursors from the meristem to the elongation zone, or alternatively GAs and intermediates may remain in the dividing cells as they move into the elongation zone. In contrast with this result, Kaneko et al.  reported promoter–GUS expression for the rice GA20ox2 paralogue in both the node and elongation zone, indicating marked differences between these monocot species.
In addition to growing tissues, CPS–GUS staining was found by Silverstone et al.  in the vasculature of mature leaves, in which the chloroplasts are immature and may thus be still capable of ent-kaurene synthesis. Indeed, Ross et al.  demonstrated that mature pea leaves and internodes were active for GA biosynthesis, but contained low levels of GA1 and GA20 due to high rates of 2β-hydroxylation. The potential association of GA biosynthesis with the vasculature raises the possibility that this tissue may act as a source of GAs for export via the phloem. There is in fact considerable evidence from experiments involving grafting and application of labelled GAs for long-distance movement of GAs in the vasculature (reviewed in ). Results from such experiments with pea  indicated that GA20 rather than GA1 or earlier intermediates is the mobile form in this species. It has been demonstrated in a number of species that leaves produce GAs as mobile signals for induction of flowering. In Arabidopsis, flowering in short days was associated with an increase in GA4 concentration at the shoot apex, assumed to have been exported from the leaves , whereas in Lolium species, the mobile signal, produced in the leaves in response to long days, was suggested to be GA5 and GA6, which would escape metabolism by GA2oxs . A discussion of the regulation of GA biosynthesis by photoperiod is given below.
Certain tissues, which are particularly active in GA biosynthesis, function as sources of GA to support the development of neighbouring tissues. In germinating cereal grains, the scutellum epithelium exports GAs to the non-GA-producing aleurone cells , which respond by synthesizing and secreting hydrolases to the endosperm as well as undergoing PCD (programmed cell death) . Within flowers, the highest concentrations of GAs occur in anthers, in which the tapetum has been identified as an important site of synthesis [84,105]. Interestingly, the tapetum also undergoes PCD in response to GA, to release components that are incorporated into the pollen wall . Some flower organs that are non-autonomous for GAs, particularly the petals, are dependent on the anthers, and potentially also the receptacle, to supply these hormones for normal development [105,107]. The suspensor, an organ present in the seed during early embryogenesis that supports growth of the young embryo , has been shown in several species to be an extremely rich source of GAs, which are thought to be important for early embryo growth (reviewed in ). Later in development, the immature seed of many, but not all, species contain high levels of GAs, which may be produced in the endosperm and/or the cotyledons. These organs can produce a vast range of different GA structures not found in other tissues , the function of which is unclear. A characteristic of these high-GA-producing tissues is the recruitment of several ODD paralogues. For example, in the rice scutellum and tapetum, OsGA20ox1 and OsGA3ox1 are expressed in addition to OsGA20ox2 and OsGA3ox2, the first two genes not being expressed elsewhere . Tissues of developing seeds typically express genes not active in other tissues, as for example found in pumpkin . These genes are presumably not subject to the regulatory constraints imposed by the normal mechanism for GA homoeostasis (see below).
Consistent with its proposed role as the ‘gatekeeper’ for GA biosynthesis, CPS expression is highly developmentally regulated as noted above . Although CPS activity may regulate the flux through the GA biosynthetic pathway, the activity of the ODDs more precisely determines the concentration of bioactive GA . The different ODD paralogous genes typically show distinct, but overlapping, expression patterns and as such contribute unequally to different developmental processes. For example, in Arabidopsis, loss of AtGA20ox1, but of no other individual GA20ox, results in a reduction in stem internode length, whereas additional loss of AtGA20ox2 causes further internode shortening, indicating that the latter also contributes to internode length . However, this contribution is exaggerated in the absence of AtGA20ox1, due to compensatory mechanisms (see below). A similar relationship exists between AtGA3ox1 and AtGA3ox2 in the regulation of stem extension .
Some indications of the molecular mechanisms underlying the developmental regulation of GA biosynthesis have been reported. The LEC (LEAFY COTYLEDON) transcription factors LEC2 and FUS3 (FUSCA3), which are required for correct tissue specification during embryogenesis and for normal seed development, suppress GA biosynthesis in the developing embryo by down-regulating AtGA3ox2 expression in the epidermis . FUS3 was shown to bind to RY elements in the AtGA3ox2 promoter. The requirement to restrict GA production during embryogenesis is reinforced by the findings of Wang et al. , who reported that the GA-deactivation gene AtGA2ox6 was up-regulated in the developing embryo by the MADS [MCM1 (minichromosome maintenance 1), AGAMOUS, DEFICIENS and SRF (serum-response factor)] domain transcription factor AGAMOUS-LIKE 15. Unlike most other organs of Arabidopsis, developing seeds contain higher amounts of GA1 than GA4, indicating that 13-hydroxylase activity is high specifically at this stage of development . Another MADS domain protein AG (AGAMOUS) was shown to directly promote expression of AtGA3ox1 in developing flowers . AG function is necessary for correct floral organ specification, but also for their development, and although organ specification can occur in the absence of GA signalling , GAs are necessary for floral organ development . AtGA3ox1 was also shown to be a direct target of the bHLH transcription factor INDETERMINATE in the valve margins and septum of the Arabidopsis silique, as part of the process of pod opening and seed dispersal .
In the vegetative shoot apex, GA biosynthesis occurs in leaf primordia and in the elongating sub-apical region . However, in order to maintain a pool of indeterminant cells in the shoot apical meristem, it is proposed that bioactive GAs must be excluded from this region through the action of homeodomain transcription factors of the Class-I KNOX (KNOTTED-like homeobox) class, which suppress the expression of GA20ox genes [120,121]. The tobacco homeobox NTH15 (Nicotiana tabacum homeobox 15) was shown to interact with a sequence within the first intron of the NtGA20ox1 gene . In potato, suppression of StGA20ox1 expression was mediated by both the KNOX protein POTH1 and a BEL-type homeodomain protein, StBEL5, which bound as a heterodimer to the promoter . The expression of GA2ox genes has been localized to the base of the meristem in several species where the enzyme may function to protect the meristem from the influx of GAs [123–126]. Expression of the rice gene OsGA2ox1 is reduced when plants are grown in florally inductive short days, so permitting an influx of GAs for promotion of the floral transition . In Arabidopsis, expression of AtGA2ox2 and AtGA2ox4 was up-regulated by induction of the KNOX protein SHOOTMERISTEMLESS and by cytokinin, which was proposed to mediate this effect through KNOX-induced expression of the cytokinin-biosynthetic gene ISOPENTENYLTRANSFERASE . However, direct induction of GA2ox expression by KNOX was demonstrated in maize, with the homeodomain protein KNOTTED1 binding to the first intron of ZmGA2ox1 .
Regulation of GA biosynthesis by other hormones
There has been considerable interest in recent years in interactions between hormone signalling pathways . Such interactions are consistent with the overlapping physiological roles found for the different hormones, although this is not so clearly reflected at the molecular level, with, in some cases, the hormones regulating different, albeit functionally related, genes . Interactions through regulation of hormone biosynthesis have been invoked in a number of cases, although direct evidence for this is rarely presented and its occurrence may be exaggerated . However, there is strong evidence for regulation of GA concentration by auxin signalling at the level of ODD gene expression, with up-regulation of GA20ox and/or GA3ox genes and, in some cases, down-regulation of GA2ox genes resulting in increased GA content [130–134]. The molecular mechanism for auxin regulation of GA-biosynthesis gene expression is not known, but appears to be direct and does not involve DELLA proteins [130,132], although the effect may be masked by DELLA-mediated homoeostasis in some cases .
Although Ross et al.  could find no evidence for regulation of GA metabolism by ABA (abscisic acid) in pea internodes, an interaction between these hormones has been shown in the context of seed germination . The CHO1 (CHOTTO1) transcription factor, which enhances dormancy in imbibed Arabidopsis seeds, inhibits GA biosynthesis through suppression of AtGA3ox2 expression . Since expression of CHO1 requires ABA signalling , this transcription factor couples the regulation of GA biosynthesis to ABA signalling. There has been no evidence that CHO1 binds directly to the AtGA3ox2 promoter, so the regulation may be indirect or involve interaction with other transcription factors, either scenario being consistent with this regulation being specific to imbibed seeds despite expression of CHO1 occurring in other organs .
Even prior to the identification of genes encoding GA metabolic enzymes or GA signalling components, it was apparent that homoeostatic mechanism(s) existed to maintain optimal GA levels for co-ordinated plant growth and development. The present review will focus on recent advances in understanding the molecular events involved in this mechanism, summarized in Figure 4. However, it is important to acknowledge that it is part of a wider homoeostatic mechanism, also including the regulation of transcription of core GA signalling components [13,14,112,117,138,139], which potentially co-ordinates GA-responsive growth through maintenance of optimal levels of the growth-repressing DELLA proteins.
Homoeostatic regulation of GA metabolism and signalling
The biochemical study of GA biosynthetic and response mutants in several plant species demonstrated that GA levels are controlled through a process of feedback regulation (reviewed in ). The subsequent cloning of genes encoding GA biosynthesis enzymes led to the demonstration that this regulation was mediated through the transcriptional control of specific ODDs. For example, in Arabidopsis this includes three GA20ox genes (GA20ox1, GA20ox2 and GA20ox3) and a single GA3ox (GA3ox1) whose expression levels are all highly elevated when bioactive GA levels are low, but are then rapidly reduced when these plants are treated with exogenous GA [112,113,141,142]. In contrast, other members of these two gene families, in addition to those encoding enzymes catalysing all of the earlier steps in the GA biosynthetic pathway, do not appear to be subject to feedback regulation [112,113,143]. An additional level of regulation that functions in GA homoeostasis involves the feedforward regulation of GA2ox genes [73,75,138]. Studies in Arabidopsis have demonstrated that multiple members of both the C19 and C20 classes of GA2oxs are transcriptionally up-regulated in response to increased GA signalling output [73,138,144]. On the basis of these observations it is likely that GA-mediated transcriptional changes in the ODD genes translate to alterations in the levels of the metabolic enzymes they encode, ultimately leading to the changes in flux of GAs through this pathway that are necessary for controlling the levels of bioactive GAs. The low abundance and labile nature of these metabolic enzymes has hindered their detection and the subsequent confirmation of this hypothesis. An exception to this was a recent study of the photoperiodic regulation of a GA20ox gene in spinach, in which the authors were able to detect the endogenous SoGA20ox1 protein by Western blotting . Although they found no evidence for the transcriptional feedback regulation of this gene in petioles and shoot tips of plants treated with GA biosynthesis inhibitors, these treatments resulted in a clear increase in SoGA20ox1 protein levels. These findings allude to further complexities in GA homoeostasis through potential post-transcriptional regulatory mechanisms. In light of the importance of targeted protein degradation in regulating GA signal transduction, it is tempting to speculate that similar mechanisms may exist to control the regulation of GA metabolism. A precedence for this is the feedback regulation of ACC synthase stability by ethylene signalling .
Recent advances in our understanding of GA signal transduction have provided important insights into the mechanisms controlling GA homoeostasis. There is conclusive evidence demonstrating that transcriptional feedback and feedforward regulation of GA metabolism is mediated by the GA signalling pathway and is dependent on the levels of DELLA proteins. In the case of feedback regulation, this has been clearly illustrated in mutants lacking core GA signalling components including GID1 [8,117,147,148] and GID2/SLY1 (SLEEPY1) [149,150], which are necessary for targeting DELLA degradation in response to GA. In these mutants, the expression levels of feedback-regulated GA biosynthetic genes are highly elevated and are not down-regulated by exogenous GA treatment. Consequently, these mutants accumulate highly elevated levels of bioactive GAs [8,117,147,150]. The absolute requirement for DELLAs in controlling the feedback response has also been demonstrated extensively in many higher plant species by characterizing both loss-of-function and gain-of-function (in which the abnormal DELLAs are resistant to GA-mediated degradation) mutants. Notably, it was the study of DELLA gain-of-function mutants in maize, wheat and Arabidopsis that initially provided the evidence for feedback regulation of GA biosynthesis on the basis of the observations that they contained elevated levels of bioactive GAs [151–153]. Subsequently, it has been demonstrated that the transcriptional feedback regulation is perturbed in both classes of DELLA mutants, with transcript levels found to be low in those lacking DELLAs, but highly elevated in those containing stabilized forms [144,154–157]. Most of these studies have focused on the role of DELLAs in the feedback regulation of GA20ox and GA3ox. However, a recent study of a DELLA loss-of-function mutant in pea (la cry) has demonstrated that the GA-mediated feedforward regulation of PsGA2ox1 and PsGA2ox2 is also controlled by these DELLAs .
DELLAs are nuclear-localized proteins that function as transcriptional regulators [9,156,158,159]. Because they act as growth repressors, there is occasionally the misconceived perception that they are transcriptional repressors. In fact, an early study of the rice DELLA SLR1 suggested that it acts as a transcriptional activator in yeast [159a]. This was later confirmed in a detailed transcriptomics analysis of DELLA target genes in Arabidopsis, showing that the majority of these are up-regulated by DELLA (and down-regulated by GA) . Interestingly, those genes showing the largest changes in expression in response to alterations in DELLA levels included the feedback-regulated genes GA20ox2 and GA3ox1. This work and an earlier study by Gubler et al.  demonstrated that DELLA protein degradation occurs within 5–10 min following GA treatment of GA-deficient Arabidopsis seedlings and barley aleurone layers [9,160]. In Arabidopsis, the subsequent reductions in the transcript levels of GA3ox1 and GA20ox2 were detectable after 15 min, suggesting that these are probably DELLA primary target genes . However, Zentella et al.  were unable to detect binding of an RGA (REPRESSOR OF ga1)–TAP (tandem affinity purification)-tag fusion protein to the promoters of GA3ox1 or GA20ox2, suggesting that other factors are necessary to regulate transcription.
There has been considerable interest in attempting to understand the mechanisms and identify the transcription factors that are responsible for controlling the feedback regulation of GA biosynthesis. Despite these efforts the picture is still rather unclear, although several transcription factors that appear to have a role in this process have been identified (recently reviewed in ). The most well characterized of these transcription factors is RSG (REPRESSION OF SHOOT GROWTH), a bZIP (basic leucine zipper)-containing transcription factor identified in tobacco . The characterization of a dominant-negative form of RSG initially demonstrated that it acts as a transcriptional activator of NtKO. It was subsequently demonstrated that RSG also binds to and activates the expression of the feedback-regulated NtGA20ox1 gene . Binding of RSG to the NtGA20ox1 promoter was found to be rapidly suppressed by GA, indicating that it might control the transcriptional feedback regulation of this gene. Interestingly, there was no evidence that RSG controlled the GA-mediated feedback regulation of another GA biosynthetic gene, NtGA3ox .
The identification of RSG protein interactors has led to some important findings about the regulatory mechanisms controlling RSG transcriptional activity. These comprehensive studies have demonstrated that 14-3-3 proteins bind to RSG, sequestering it in a cytoplasmic localization in response to GA signalling and preventing its activation of NtGA20ox1 and other target genes [164,165]. This process is controlled by CDPK1 (Ca2+-dependent protein kinase 1), which in response to GA signalling, and potentially an elevation in intracellular Ca2+ levels, phosphorylates RSG on Ser114, enhancing its association with 14-3-3 proteins in the cytoplasm [164,166]. Although these studies provided important insights into the control of feedback regulation of NtGA20ox1, there are still many remaining questions. Notably, the role of DELLAs in the regulation of RSG activity is unclear. It is conceivable that DELLA signalling regulates intracellular Ca2+ levels, and indeed an increase in Ca2+ levels is associated with GA signalling in the cereal aleurone . A similar function for RSG orthologues in the regulation of GA biosynthesis has not been reported for other plant species, but if this turns out to be the case, the availability of improved genetic resources in model species should provide important tools for dissecting the mechanisms involved.
In Arabidopsis, characterization of the cis-acting sequences responsible for controlling the feedback regulation of AtGA3ox1 has led to the identification of AGF1 (AT-hook protein of GA feedback 1) as a potential regulator of this response . The presence of an AT-hook DNA-binding motif together with the demonstration that an AGF1–GFP (green fluorescent protein) fusion protein is nuclear-localized in transgenic plants supports a role of AGF1 as a transcriptional regulator. However, although the overexpression of AGF1 potentiates the transcriptional feedback regulation of AtGA3ox1, a clear role in regulating this response has yet to be established.
Members of the YABBY family of C2C2 zinc finger transcription factors have a well characterized role in the maintenance of polar outgrowth of lateral organs . Although extensive studies of the YABBY genes in Arabidopsis have not yet indicated involvement in the regulation of GA metabolism, there is evidence in rice implicating OsYAB1 (OsYABBY1) in the feedback control of OsGA3ox2 . OsYAB1 was shown to directly repress the expression of this gene, potentially by binding to a GA-responsive cis-regulatory element within its promoter. The expression of OsYAB1 transcripts was found to increase rapidly in rice seedlings after GA treatment, with paclobutrazol treatment having the opposite effect. These observations suggest that the GA signalling-mediated changes in OsYAB1 expression are responsible for regulating OsGA3ox2 transcript levels as opposed to a direct effect of OsYAB1 on DELLA (SLR1)-mediated transcription. Interestingly, the expression of another GA-regulated gene, OsGA20ox2, did not appear to be controlled by OsYAB1, supporting the existence of additional components responsible for its feedback regulation.
Important recent studies have identified SCL3 (SCARECROW-LIKE 3) as a novel GA signalling component that also has a function in regulating feedback control of GA biosynthesis in Arabidopsis [171,172]. Although SCL3 was identified as an early GA-response gene that is positively regulated by DELLAs through direct binding to its promoter , genetic analysis demonstrates that it acts as a positive regulator of GA-responsive root and shoot growth [171,172]. Consistent with this role, transcript levels of the GA feedback-regulated genes GA20ox1, GA20ox2, GA20ox3 and GA3ox1 are all elevated in the scl3 mutant . Similarly, the truncated SCL3 transcript produced in this T-DNA (transferred DNA) insertion mutant (upstream of the T-DNA insertion) is also found to be elevated compared with wild-type plants. These observations, and the important finding that DELLAs physically interact with SCL3, have led to a model in which SCL3 is proposed to attenuate the transcriptional activity of DELLAs through direct interaction. Like DELLAs, SCL3 is a member of the GRAS family of transcriptional regulators, but its mode of action is currently unclear. It is interesting that SCL3 regulation of GA-dependent root growth is mediated through its confined localization within the endodermis . This is consistent with studies demonstrating that GA signalling within the endodermis drives the cell elongation and cell division necessary for promoting GA-dependent root growth [173,174]. However, it is not entirely consistent with a role in controlling the feedback regulation of GA biosynthesis because some of these genes, including GA20ox1, GA20ox2 and GA3ox1, are predominantly expressed in other cell types within the root [175,175a]. Further work is necessary to understand the mechanisms involved.
How an SCL3/DELLA complex acts to regulate the expression of direct target genes is still unclear. ChIP (chromatin immunoprecipitation) studies suggest that this complex binds directly to regulate the expression of the SCL3 promoter [9,171], although this does not appear to be the case for the GA feedback-regulated genes GA20ox1, GA20ox2, GA20ox3 and GA3ox1. It is conceivable that regulation of these genes occurs through an alternative mechanism, for example, through the sequestration of a transcriptional repressor. The characterization of DELLA-interacting proteins is rapidly improving our understanding of how they control growth regulation, and it is likely that the identification of others will further improve this knowledge and potentially answer the question of how they control the expression of genes controlling GA homoeostasis.
In common with other phytohormones, GAs act as mediators of environmental signals, allowing plants to respond, often rapidly, to changes in light conditions, temperature, water and nutrient status, and other abiotic and biotic stresses. In many cases it has been shown that changes in the environment result in altered GA content and the molecular mechanisms underlying these metabolic changes are beginning to be resolved, primarily in relation to light and abiotic stress.
Depending on context, GA metabolism is sensitive to changes in light quantity, quality or duration, which may result in increased or decreased GA content. In early seedling development, GA signalling suppresses photomorphogenesis , with exposure of dark-grown seedlings to light, i.e. de-etiolation, resulting in a rapid reduction in GA content . Both phytochrome and cryptochrome light receptors are involved in this process, which results in up-regulation of GA2ox expression and, in the case of blue light, down-regulation of GA20ox and GA3ox genes (reviewed in ). Experiments with pea indicate that the rapid decrease in GA1 content in etiolated seedlings following exposure to light, due mainly to up-regulation of PsGA2ox2 expression, required involvement of the bZIP transcription factor LONG1, which is related to Arabidopsis HY5 (ELONGATED HYPOCOTYL 5), and, like HY5, appears to be regulated at the level of protein stability by the COP1 (CONSTITUTIVELY PHOTOMORPHOGENIC 1) orthologue LIP1 (LIGHT-INDEPENDENT PHOTOMORPHOGENESIS 1) .
Gibberellins promote flowering in LD (long day) species, in which exposure to LDs enhances GA production, primarily through up-regulation of GA20ox genes (reviewed in ). In Arabidopsis, up-regulation of AtGA20ox2 expression in the leaf petiole following exposure to far-red-rich LDs, in a phytochrome B-mediated process, is associated with enhanced petiole elongation  and floral induction . Although light-induced flowering in Arabidopsis on LDs involves mainly the CONSTANS pathway, which is independent of GAs, in the absence of this pathway on SDs (short days), floral induction is absolutely dependent on GAs . There is evidence from Arabidopsis to suggest that, under SDs, GA4 may function as a mobile signal from leaves to the shoot apex for floral induction , whereas in the grass Lolium temulentum, it is proposed that the mobile signals are GA5 and GA6, produced in leaves in response to as little as one LD . On account of their unsaturation on C-2, these last GAs escape deactivation by GA2oxs present below the shoot apex; following floral induction, GA2ox activity subsides and the C-2-saturated GAs, GA1 and GA4, are produced for promotion of floral development and elongation of the stem . Compared with GA5, applied GA1 and GA4 are only weakly florigenic in L. temulentum, presumably due to their rapid deactivation at the shoot apex . In rice, a SD flowering plant, following exposure to SDs, GA2ox expression below the meristem is reduced, suggested to allow influx of GA1 from leaves to promote floral induction . In an analogous, but converse, process, SD induction of tuberization in potatoes, a process that is inhibited by GAs, is associated with enhanced expression of StGA2ox1 adjacent to the stolon meristem prior to stolon swelling and tuber formation . Transfer of potatoes to SDs also results in reduced expression of StGA3ox2 in stolons, but increased expression of this gene in shoots . It is suggested that in SDs less GA is transported from the shoots to the stolons due to enhanced conversion of the mobile GA20 into the less mobile GA1, which accumulates and promotes shoot growth.
The red-light-induced germination of photoblastic seeds involves accumulation of GA through enhanced synthesis and suppressed deactivation as well as a reduction in ABA content . In Arabidopsis, this process is mediated by the bHLH transcription factor PIL5 (PIF3-LIKE 5), an inhibitor of germination that is degraded by the 26S proteasome after being targeted by phosphorylation through interaction with phytochromes A and B . PIL5 suppresses GA accumulation (and ABA degradation) by inhibiting expression of the GA biosynthetic genes AtGA3ox1 and AtGA3ox2, while promoting expression of the GA-deactivating AtGA2ox2, its action being relieved by the light-stimulated degradation (Figure 5). Two intermediaries in this pathway have been identified: the action of SOM (SOMNUS), a CCCH-type zinc finger protein, has the same effect on the GA metabolic genes as PIL5, which directly induces its expression , whereas DAG1 (DOF AFFECTING GERMINATION 1) inhibits expression of AtGA3ox1, but has no effect on expression of AtGA3ox2 and AtGA2ox2 . DAG1, expression of which is induced indirectly by PIL5, was shown to bind to the AtGA3ox1 promoter. AtGA3ox1 is also the target of cold induction of germination (stratification) , expression of the gene being suppressed in non-stratified seeds in the dark through the action of the light-stable bHLH protein SPT (SPATULA), although the mechanism is unclear .
Regulation of seed germination in Arabidopsis by red light and cold
Inhibition of growth in Arabidopsis by salt and cold stress through induction of GA deactivation
Among the complex responses of plants to stress, growth reduction is a common strategy that allows resources to be focused on withstanding the stress . Many hormone signalling pathways have been implicated in stress responses, including the accumulation of DELLA proteins, which restrain growth and promote stress resistance , although it is unclear to what extent these responses are causally related. An emerging theme is stress-induced up-regulation of GA2ox gene expression as a means of reducing GA content and allowing DELLA accumulation. Response to abiotic stress is mediated by the CBF (C-repeat-binding factor)/DREB [DRE (dehydration-response element)-binding] transcription factors, overexpression of which confers a degree of stress tolerance, but also causes severe growth restriction . In Arabidopsis, high salt-induced expression of six GA2ox genes, one of which, AtGA2ox7, was up-regulated specifically by the salt-responsive CBF/DREB1 protein DDF1 (DWARF AND DELAYED FLOWERING 1) . DDF1 was shown to bind to a DRE-like motif on the promoter of AtGA2ox7. Short-term exposure of Arabidopsis plants to cold resulted in increased expression of three GA2ox genes, two of which, AtGA2ox3 and AtGA2ox6, were also up-regulated by overexpression of the cold-responsive CBF1/DREB1b gene . Cold and CBF1 overexpression also increased expression of several GA biosynthetic genes, but this effect was abolished in a DELLA-deficient mutant, indicating that it occurred indirectly due to feedback regulation. Although both salt and cold stress stimulated expression of many GA2ox genes, the tested CBF/DREB transcription factors mediated induction of only subsets of these, indicating a high level of specificity and presumably the involvement of other transcription factors. As well as the accumulation of DELLA proteins resulting from reduced GA concentrations, it was shown that both salt and cold treatments up-regulated expression of RGL3 (RGA-LIKE PROTEIN 3), which encodes one of the five Arabidopsis DELLA proteins, the induction being mediated by the respective CBF/DREB factors [196,197]. Thus abiotic stress allows DELLAs to accumulate through protein stabilization and enhanced gene expression.
Seed dormancy in Arabidopsis in response to low temperature was shown recently to be associated with increased expression of AtGA2ox6, a process which is mediated by the dormancy factor DOG1 (DELAY OF GERMINATION 1) . Interestingly, both DOG1 and AtGA2ox6, as well as other GA2ox genes, were regulated by CBF1 in developing seeds, but this transcription factor was not cold-induced in this context, although it appeared to have a role in seed dormancy.
The positive growth response of plants to increasing ambient temperature was shown to be mediated by several hormone pathways, including GA signalling, with AtGA3ox1, AtGA20ox1 and AtGA2ox1 identified as targets in Arabidopsis seedlings . Plant growth is reduced under abnormal thermoperiodic conditions, i.e. lower temperatures during the light period than during the dark, a condition known as negative DIF (difference between day temperature and night temperature) that is exploited for control of plant stature . Stavang et al.  showed that stems of pea plants grown in negative DIF contained lower concentrations of C19-GAs and increased expression of PsGA2ox2 compared with plants grown in positive DIF and suggested that this gene was a major target for this response.
As GAs are major effectors of plant growth and development, their biosynthesis is subject to strict regulation throughout most of the plant growth cycle and in response to environmental stimuli. Recent research on GA metabolism has been directed towards understanding regulation, which occurs primarily on the steps catalysed by ODDs. Many molecular components that mediate GA metabolic regulation have been identified, although our understanding of these processes is far from complete. In particular, the mechanisms underlying GA homoeostasis are proving surprisingly complex, perhaps at least in part as the result of the promiscuous interactions of the DELLA proteins, which mediate this process. Understanding the nature of DELLA-mediated signalling and identifying the many partners of these proteins is currently a rapidly advancing area of research. There is emerging evidence that DELLAs may act as a point of convergence for different hormone signalling pathways, thereby providing a channel by which these pathways may influence GA metabolism. However, it is clear that auxin signalling and, in certain circumstances also that of ABA, modify GA biosynthesis directly without the intervention of DELLAs [132,136]. Details of these mechanisms at the molecular level should become clearer in the next few years.
There is also still much to learn about the sites of GA synthesis, particularly at the cellular level. The increasing evidence that GAs may act remotely from their source is stimulating renewed interest in GA movement, which may occur over long distances within the vasculature, particularly the phloem, or by diffusion between neighbouring cells. Improved physicochemical and molecular methods are required to probe GA biosynthesis and action at greater tissue resolution than is currently possible.
Although GA signalling influences most aspects of plant development, the specificity of expression domains and consequently physiological function of the ODD paralogues in GA metabolism has enabled the identification of GA-deficient mutants with desirable agricultural traits. For example, the sd1 mutant of rice, lacking OsGA20ox2, and le pea, with a mutant allele of PsGA3ox1, have reduced stem height, but normal fertility. Reverse genetic approaches such as TILLING (Targeting Induced Local Lesions in Genomes) provide an opportunity to introduce such beneficial mutations into many more agriculturally important species, although a fuller understanding of the physiological roles of the target genes is required. The rapidly expanding number of plant genome sequences available is simplifying the task of gene identification, allowing suitable targets to be selected for this purpose. Furthermore, the analysis of natural genetic variation in these genes between varieties or ecotypes will provide important information on their influence on plant phenotype. These and other developments are likely to enhance the already substantial impact that GA-related research has had on global agriculture.
AT-hook protein of GA feedback 1
basic leucine zipper
ent-copalyl diphosphate synthase
DOF AFFECTING GERMINATION 1
DWARF AND DELAYED FLOWERING 1
difference between day temperature and night temperature
DELAY OF GERMINATION 1
ELONGATED UPPERMOST INTERNODE
GA methyl transferase
trans-geranylgeranyl diphosphate synthase
ELONGATED HYPOCOTYL 5
JASMONATE ZIM-domain protein
ent-kaurenoic acid oxidase
MCM1 (minichromosome maintenance1), AGAMOUS, DEFICIENS and SRF (serum-response factor)
cytochrome P450 mono-oxygenase
programmed cell death
REPRESSOR OF ga1 (DELLA protein)
RGA-LIKE PROTEIN 3
REPRESSION OF SHOOT GROWTH
Rothamsted Research receives strategic support from the Biotechnology and Biological Sciences Research Council, U.K.