Control of abscisic acid synthesis

Abstract

The abscisic acid (ABA) biosynthetic pathway involves the formation of a 9‐cis‐epoxycarotenoid precursor. Oxidative cleavage then results in the formation of xanthoxin, which is subsequently converted to ABA. A number of steps in the pathway may control ABA synthesis, but particular attention has been given to the enzyme involved in the oxidative cleavage reaction, i.e. 9‐cis‐epoxycarotenoid dioxygenase (NCED). Cloning of a gene encoding this enzyme in maize was first reported in 1997. Mapping and DNA sequencing studies indicated that a wilty tomato mutant was due to a deletion in the gene encoding an enzyme with a very similar amino acid sequence to this maize NCED. The potential use of this gene in altering ABA content will be discussed together with other genes encoding ABA biosynthetic enzymes.

Introduction

Abscisic acid (ABA) is a phytohormone which performs several specific functions in plant growth and development. There is now good evidence that ABA plays a role in seed development and primary dormancy (Frey et al., 1999). In addition, ABA levels in vegetative tissues can be elevated in response to various environmental stresses; most notably drought. For these reasons ABA is sometimes referred to as a stress hormone (Zeevaart, 1999).

Knowledge of the ABA biosynthetic pathway, combined with the cloning of genes directly involved in biosynthesis (Marin et al., 1996; Tan et al., 1997), have made the prospects of manipulating ABA levels in plants a realistic objective. Using antisense technology, it is now possible to investigate symptoms of ABA deficiency in species which are important from a physiological and/or economic viewpoint, but which do not have any conventional ABA‐deficient mutants. It now also appears to be possible to investigate the effects of ABA over‐production, using constructs designed to over‐express ABA biosynthetic genes.

In this review, important advances made in our understanding of the control of ABA biosynthesis will be discussed in relation to recent attempts to alter ABA concentrations in transgenic plants.

The ABA biosynthetic pathway

The general framework of the ABA biosynthetic pathway in plants is now well established. ABA synthesis represents a minor branch of the carotenoid pathway; as originally indicated by a number of pioneering experiments (Taylor and Burden, 1973). It has only recently become clear that the isopentenyl pyrophosphate (IPP) C₅ ‘building blocks’ from which ABA is formed, are themselves synthesized from glyceraldehyde phosphate and pyruvate inside plastids (Arigoni et al., 1997; Lichtenthaler et al., 1997). It is therefore important to note that the ultimate precursor of ABA is not mevalonic acid. This represents a major revision of the information given in many earlier reviews (Milborrow, 1974); as has been pointed out more recently by a number of authors (Milborrow and Lee, 1998).

The genes and enzymes involved in the assembly of the C₄₀ compound β‐carotene, from IPP precursor molecules, have been discussed in a recent comprehensive review (Cunningham and Gantt, 1998). The two rings of β‐carotene can be hydroxylated to form a xanthophyll known as zeaxanthin (Fig. 1). The ABA biosynthetic pathway can in some ways be regarded as starting from this xanthophyll (Taylor, 1991) since mutants blocked downstream of zeaxanthin show symptoms typical of ABA, rather than carotenoid, deficiency.

Important evidence relating to the validity of the pathway shown in Fig. 1 has been provided by a number of labelling studies (involving ¹⁸O₂, ²H₂O, ¹⁴CO₂, ¹⁴C‐xanthoxin, ¹³C‐xanthoxin, ²H‐ABA aldehyde etc), by stoichiometric experiments involving dark‐grown tissues and by the use of cell‐free biosynthesis systems. Endogenous ABA levels can be reduced using the carotenoid biosynthesis inhibitors, norflurazon and fluridone. However, the virtually white seedlings which can be obtained following norflurazon and fluridone treatments are deficient in a whole range of metabolic processes resulting from the photobleaching of chlorophyll. Against this background, the phenotypic effects of ABA deficiency seem trivial and ABA is not the limiting factor preventing continued growth. Comprehensive information on these important experiments has been provided recently by three excellent reviews (Cutler and Krochko, 1999; Liontenberg et al., 1999; Zeevaart, 1999) and therefore they are not described further in this review.

The characterization of ABA‐deficient mutants impaired at various points in the pathway, has provided very strong evidence that the biosynthetic steps shown in Fig. 1 are mainly correct; although some ambiguities persist and some evidence appears to indicate the existence of alternative/substitute routes (see Cutler and Krochko, 1999, for some discussion on this). Unfortunately, no ABA‐deficient mutant plants have yet been shown to be impaired in the various xanthophyll isomerization steps taking place at the C₄₀ level of the ABA pathway (Fig. 1). For this reason, considerable uncertainty remains concerning the precise route(s) followed prior to the formation of the first C₁₅ intermediate, xanthoxin. None of the following in vivo alternatives can be formally ruled out:‐

(A) Selective oxidative cleavage (defined by the proposal that only one of the two 9‐cis‐xanthophylls detected in plant tissues can be oxidatively cleaved to form the first C₁₅ ABA biosynthetic intermediate, xanthoxin). This could involve one or more of the following three routes:‐

Route A1:

selective cleavage of 9‐cis‐violaxanthin synthesized directly by a single isomerization step from all‐trans‐violaxanthin.

Route A2:

selective cleavage of 9′‐cis‐neoxanthin synthesized from all‐trans‐violaxanthin in two isomerization steps via 9‐cis‐violaxanthin.

Route A3:

selective cleavage of 9′‐cis‐neoxanthin synthesized from all‐trans‐violaxanthin in two isomerization steps via all‐trans‐neoxanthin.

(B) Non‐selective oxidative cleavage (defined by the proposal that either of the two 9‐cis‐xanthophylls detected in plant tissues can be oxidatively cleaved to form the first C₁₅ ABA biosynthetic intermediate, xanthoxin). This could involve one or more of the following two routes:‐

Route B1:

all‐trans‐violaxanthin is first converted to 9‐cis‐violaxanthin, most of which is further isomerized to form 9′‐cis‐neoxanthin. Either of these two 9‐cis‐xanthophylls could be used indiscriminately as a substrate for oxidative cleavage.

Route B2:

all‐trans‐violaxanthin is simultaneously subjected to two alternative isomerizations, leading to the formation of 9‐cis‐violaxanthin and all‐trans‐neoxanthin. Only the latter would be further isomerized to form 9′‐cis‐neoxanthin. Either of these two 9‐cis‐xanthophylls could be used indiscriminately as a substrate for oxidative cleavage.

At present there is no clear experimental evidence available to demonstrate which of these alternative routes occurs in planta. It is important for this uncertainty to be resolved. Several of the other steps in ABA biosynthesis have been characterized using ABA‐deficient mutants obtained in a number of different plant species.

The location of genetic lesions in the ABA biosynthetic pathway

ABA‐deficient mutants have been obtained in both monocotyledonous and dicotyledonous species. Various mutants have been listed in Table 1 together with a Chlamydomonas mutant affecting xanthophyll isomerization which blocks the formation of the allene (neoxanthin). To date no mutants have been identified which impair the isomerization of all‐trans‐xanthophylls to form the 9‐cis‐xanthophylls which are important ABA precursors. Characterization of mutants at this isomerization step would be likely to end some of the uncertainty which continues to surround this part of the pathway (see the previous section and also Fig. 1).

The plant genetic lesions included in Table 1 all result in a wilty phenotype, which is more severe in some cases than others. They undergo phenotypic reversion in response to exogenous ABA and can therefore be regarded as ABA‐deficient mutants according to earlier definitions (Taylor, 1991). Mutants affecting the C₄₀ level of the pathway are referred to in Table 1 as forming part of ABA/xanthophyll biosynthesis. Because xanthophylls form important components of the photosynthetic apparatus it is misleading to consider them simply as precursors of ABA. Nevertheless, it is reasonable to point out that the most obvious abnormalities of mutants which are impaired in the epoxidation of zeaxanthin (Duckham et al., 1991; Rock and Zeevaart, 1991;Marin et al., 1996) are associated with their ABA deficiency, rather than their reduced levels of epoxyxanthophylls.

Only three of the categories of mutant shown in Table 1, can be regarded as exclusively affecting ABA biosynthesis. There is strong evidence that the vp14 mutant of maize (Tan et al., 1997; Schwartz et al., 1997a) and the not mutant of tomato (Burbidge et al., 1997b, 1999) impair the oxidative cleavage reaction which forms the first C₁₅ intermediate of ABA biosynthesis, i.e. xanthoxin. It is likely that the wilty mutant of pea also affects this first committed step in the ABA pathway, although wilty pea has not yet been fully characterized. The first step at the C₁₅ level of ABA biosynthesis, the conversion of xanthoxin to ABA aldehyde (ABAld), is blocked only by the aba2 genetic lesion of Arabidopsis (Leon‐Kloosterziel et al., 1996; Schwartz et al., 1997b).

In contrast, a much larger and more diverse collection of mutants are known to be impaired in the oxidation of ABAld to form ABA. In Table 1, only one group of mutants affecting this last step has been assigned directly to the ABA biosynthetic pathway. The sit mutant of tomato accumulates the biologically inactive compound trans ABA alcohol instead of ABA in response to water stress (Linforth et al., 1987). This mutant converts exogenously supplied (±) ABAld to a mixture of cis and trans ABA alcohol (Taylor et al., 1988); it is possible that only the natural (+) enantiomer of ABAld forms trans ABA alcohol. Virtually identical results were obtained for the droopy mutant of Solanum phureja, which appears to occupy an equivalent position in the potato genome to the sit gene locus on the short arm of chromosome 1 of tomato (Duckham et al., 1989). Marin and Marion‐Poll showed that sit tomato mutants were only defective for aldehyde oxidase (AO⁻) and were not impaired in two other enzymes which also require a molybdenum cofactor (MoCo); nitrate reductase (NR⁺) and xanthine dehydrogenase (XDH⁺) (Marin and Marion‐Poll, 1997). From this it can be predicted that both sit and droopy are defective in the apoenzyme of ABAld oxidase (ABAO). Recently, strong evidence has been obtained that a mutant of Arabidopsis, previously known as v11 and now renamed aao3, is also impaired in the apoenzyme of ABAO (Seo et al., 2000).

It should be noted that two additional categories of mutant are defective in the conversion of ABAld to ABA, but are not involved in encoding ABA biosynthetic enzymes. One group of mutants affects the supply of the co‐factor either by affecting synthesis of the molybdopterin (MPT) or by in some way restricting the transfer of Mo to MPT to form MoCo. In Table 1 these are classified as MoCo synthesis mutants which are pleiotropically defective in the three main types of molybdoenzymes (AO⁻, NR⁻, XDH⁻). The second category has been referred to as affecting MoCo maturation, following the review by Mendel and Schwarz (Mendel and Schwarz, 1999). These authors regard the cofactor containing the dioxo form of Mo as the end product of MoCo biosynthesis. This is the form of the cofactor which is incorporated into the nitrate reductase apoprotein. After parallel incorporation of MoCo into aldehyde oxidase and xanthine dehydrogenase apoproteins, a sulphurylase adds inorganic sulphur to the Mo centre converting dioxo MoCo to the monooxo form. The tomato mutant flc (Marin and Marion‐Poll, 1997), aba3 of Arabidopsis (Schwartz et al., 1997b) and Nicotiana plumbaginifolia mutant alleles at the aba1 locus (Leydecker et al., 1995; Akaba et al., 1998) all affect genes encoding this sulphurylase enzyme. These mutants have normal nitrate reductase (NR⁺), but are simultaneously impaired in aldehyde oxidase activities (AO⁻) and xanthine dehydrogenase (XDH⁻).

Mutants affecting MoCo synthesis and maturation have been discussed recently (Mendel and Schwarz, 1999). As they are indirectly, rather than directly, associated with ABA biosynthesis these mutants will not be considered further in this review. Of the remaining ABA‐deficient mutants listed in Table 1, two categories have already been characterized at the molecular level in some detail and a third is currently being analysed to identify the precise DNA sequence changes involved. These categories will be discussed in turn.

Molecular characterization of zeaxanthin epoxidase (ZEP)

The first major breakthrough in understanding the molecular genetics of ABA biosynthesis involved the cloning of zeaxanthin epoxidase (Marin et al., 1996). A programme of non‐targeted transposon mutagenesis was carried out in Nicotiana plumbaginifolia using the maize Activator (Ac) transposon. A plant was identified which exhibited an unstable wilty mutant phenotype. In progeny of this plant, a specific Ac insertion was found to co‐segregate with the mutant phenotype. Inverse PCR was used to generate a DNA fragment flanking this Ac element. This fragment was used to screen a cDNA library from which a hybridizing full‐length, single copy cDNA clone was isolated (Marin et al., 1996).

A stable mutant (aba2‐s1), selected from the Ac‐tagged line, was shown to have arisen from by an imprecise transposon excision. The stable mutant plants accumulated very high levels of zeaxanthin and had abnormally low concentrations of the epoxyxanthophylls, i.e. antheraxanthin, violaxanthin and neoxanthin (Marin et al., 1996). A virtually identical pattern of xanthophyll abnormalities was reported previously for the Arabidopsis mutant aba1 (Duckham et al., 1991; Rock and Zeevaart, 1991; Table 1). The wild‐type phenotype was restored to N. plumbaginifolia aba2 following transformation with a construct based on the ABA2 cDNA under the control of the CaMV 35S promoter. In addition, this construct complemented the Arabidopsis aba1 genetic lesion. These are orthologous mutants, both impaired in the synthesis of the enzyme zeaxanthin epoxidase (ZEP). The complete open reading frames encoding ZEP have been reported for pepper (Bouvier et al., 1996) and tomato (Burbidge et al., 1997a) in addition to DNA sequence data on N. plumbaginifolia and Arabidopsis.

Zeaxanthin epoxidase gene expression

An internal 1.4kb HindIII fragment for the ABA2 cDNA was used as a probe to investigate steady‐state ABA2 mRNA levels in various tissues of N. plumbaginifolia. Changes in transcript level during development at different times of day and also in response to water stress, were determined in detail in this species (Audran et al., 1998). Parallel Northern analyses of tomato plants using a probe from the full‐length LeZEP1 cDNA have also been reported recently (Thompson et al., 2000a). Similar general conclusions can be drawn from both of these studies.

Diurnal variation

Steady‐state NpZEP mRNA was analysed in leaves from plants at the rosette stage during a single day consisting of 16/8 h light/dark (Audran et al., 1998). The NpZEP mRNA levels increased dramatically during the light period reaching a maximum after 3–5 h. Steady‐state NpZEP mRNA then fell continuously during the later part of the light period and reached a minimum in the dark phase. Audran et al. also observed that there was a small increase in NpZEP mRNA levels 30 min before the next light period was due to start (Audran et al., 1998).

Expression patterns in leaves of tomato plants were assayed at 6 h intervals throughout 3 d consisting of 12/12 h light/dark followed by 2 d of continuous darkness (Thompson et al., 2000a). The highest LeZEP mRNA steady‐state levels were observed 5 h into the light period (in agreement with Audran et al., 1998). After 11 h of light the LeZEP mRNA had fallen substantially from the peak of abundance; minimum values were observed after 5 h of darkness. As with N. plumbaginifolia, in tomato leaves the LeZEP1 mRNA level was observed to increase in anticipation of the light period (Audran et al., 1998; Thompson et al., 2000a). The LeZEP mRNA was found to oscillate with a phase very similar to light harvesting complex II (LHCII) mRNA, although there was a slightly earlier decline in LeZEP mRNA in the later phase of the light period. In the tomato experiment, the rises in LeZEP mRNA and LHCII mRNA continued during a 48 h period of darkness, suggesting that ZEP mRNA may also be under the control of a circadian oscillator. This similarity between ZEP and LHCII mRNAs is consistent with the proposal that phases of mRNA accumulation are co‐ordinated for individual complexes (Oelmuller et al., 1995). Expression of genes encoding the enzymes of the xanthophyll cycle may be co‐ordinated with those of photosystem II, as a result of their involvement in photoprotection.

A Western blot analysis was carried out in N. plumbaginifolia (Audran et al., 1998). Despite the diurnal rhythm for NpZEP mRNA no oscillations in steady‐state protein levels were detected. Although superficially surprising, this could have resulted from putative rises in newly synthesized ZEP being obscured by high background levels of stable ZEP protein. In parallel experiments with tomato LHCII protein, oscillations were observed by in vivo labelling (Riesselmann and Piechulla, 1992), but no changes in steady‐state protein levels had been detected by earlier workers (Piechulla and Gruissem, 1987).

Drought‐induced expression

Because of the strong diurnal rhythm, steady‐state ZEP mRNA levels in leaves can appear to either rise or fall during a period of several hours following the application of a water‐stress stimulus. It was found that NpZEP mRNA level decreased 2–3‐fold in N. plumbaginifolia leaves after 4 h of dehydration and 1.5–5‐fold after 8 h of dehydration (Audran et al., 1998). A similar fall was detected in steady‐state NpZEP mRNA in the leaves of non‐dehydrated N. plumbaginifolia plants over the same time‐course, as a result of the normal diurnal rhythm. This demonstrates the importance of designing experiments with suitable controls to allow for these time of day effects.

In both N. plumbaginifolia (Marin et al., 1996; Audran et al., 1998) and tomato (Burbidge et al., 1997a; Thompson et al., 2000a) leaves, no rise in steady‐state ZEP mRNA levels have been reported to occur in response to a variety of dehydration treatments. Because the products of ZEP, the epoxyxanthophylls (e.g. all‐trans‐violaxanthin) are present at high levels in leaves (Parry and Horgan, 1992), it has been concluded that ‘the availability of xanthophyll precursors might not be limiting for ABA synthesis, even under water stress. An increased gene expression would therefore not be needed or would be restricted to certain tissues and might not be detected’ (Audran et al., 1998).

The epoxyxanthophyll levels in roots are much lower than in the leaves (Parry and Horgan, 1992) and therefore could be limiting for sustained ABA synthesis. On this basis it might reasonably have been predicted that in roots it would be possible to increase steady‐state ZEP mRNA levels by water stress treatments. In the systems investigated so far, this appears to be true.

It was found that NpZEP mRNA levels in roots were increased by between 2.5–5‐fold after 4 h of dehydration, rising to between 3–7‐fold after an 8 h period of water stress (Audran et al., 1998). The elevated levels of steady‐state NpZEP mRNA were still maintained after 24 h of dehydration. Low and constant levels of steady‐state NpZEP mRNA were observed in unstressed roots over a 24 h period (Audran et al., 1998). A more gradually imposed drought treatment in tomato resulted in a 4‐fold increase in steady‐state LeZEP mRNA levels in the roots (Thompson et al., 2000a) confirming the earlier results obtained for N. plumbaginifolia. The full extent of the drought‐induced increase in LeZEP mRNA was masked by the presence of an additional hybridization signal when Northern blots were washed at 60 °C rather than at 68 °C. There appears to be an additional ‘ZEP‐like’ mRNA which is present in tomato roots and is not responsive to drought stress (Thompson et al., 2000a).

Developmental changes in expression

The abundance of NpZEP mRNA was found to change substantially during seed development (Audran et al., 1998). Steady‐state NpZEP mRNA levels increased to a maximum in seeds 7–10 d after pollination (DAP) and then fell substantially as the seed matured prior to capsule opening at 22 DAP. This followed the expected time‐course for changes in ABA content during seed development. In N. plumbaginifolia flowers the levels of ZEP mRNA were constant during development, but higher steady‐state NpZEP mRNA levels were observed during the later stages of flower maturation (Audran et al., 1998).

Artificially‐induced modifications in expression

The report of the cloning of NpZEP (Marin et al., 1996) offered the first prospect of manipulating ABA levels by genetic engineering. Antisense technology could be used to down‐regulate a specific step in ABA biosynthesis and to create the first ABA‐deficient transformants to add to the collection of ABA‐deficient mutants shown in Table 1.

Antisense constructs were produced based on a 1.36 kb HindIII internal fragment and a 3′ XbaI fragment of 0.81 kb obtained from the NpZEP cDNA (Frey et al., 1999). Northern analysis of leaves from homozygous transgenic N. plumbaginifolia plants revealed a substantially reduced steady‐state NpZEP mRNA level in an antisense line involving the 1.36 kb fragment and a much less obvious down‐regulation with the 0.81 kb construct. Despite the antisense suppression of mRNA in leaves, it was reported that none of the 19 antisense plants studied exhibited a wilty phenotype and that water loss from detached leaves was similar to that of wild‐type plants. Mutant aba2‐s1 plants, in contrast, were extremely wilty and were described as unable to grow under greenhouse conditions unless they were grafted onto wild‐type tobacco stocks (Frey et al., 1999). The lack of a wilty phenotype in the NpZEP antisense transformants of N. plumbaginifolia was therefore unexpected. It is conceivable that the regions of the NpZEP gene used to build the antisense constructs were for some reason sub‐optimal in providing a leaf phenotype. Nevertheless, it was observed that two out of five transgenic lines obtained with the 0.81 kb antisense construct and six out of 14 carrying the 1.36 kb antisense sequence, exhibited reduced seed dormancy (Frey et al., 1999). Data were reported for two of the lines which showed the greatest effect on seed dormancy. Half of the seed in these antisense lines had germinated after 6–7 d; this contrasted with wild‐type seed which took 13–14 d to attain this level of germination. This lack of seed dormancy was correlated with reduced steady‐state NpZEP mRNA during seed development and lower ABA levels in mature seeds.

Homozygous tomato transgenics carrying an LeZEP1 antisense construct have been produced (Parker et al., 1999). These tomato transformants did show a wilty ‘mutant’ phenotype, but were less severely affected than notabilis mutants. However, in combination with notabilis the antisense transgene produced an extremely wilty phenotype. A similar effect was described previously for notabilis/sitiens double mutants, which had a much more severe phenotype than either mutant alone (Taylor and Tarr, 1984). Because tomato seeds have almost no dormancy, in contrast to those of N. plumbaginifolia, it is difficult to detect an effect of the LeZEP1 antisense construct on germination rate. However, the mature flowers of tomato which are normally bright yellow have a much paler colour in the LeZEP1 antisense transformants. This was associated with the accumulation of zeaxanthin in the corolla (Fig. 2) at the expense of the epoxyxanthophylls, e.g. all‐trans‐violaxanthin and 9‐cis‐neoxanthin (Parker et al., 2000). The increased expression of NpZEP mRNA as the flowers mature in N. plumbaginifolia (Audran et al., 1998) may be part of a developmental programme to produce yellow flowers in the Solanaceae, which is no longer fully functional in this white‐flowered species.

Constructs were also designed in which the complete ORF of the NpZEP cDNA was placed under the control of a constitutive 35S promoter, in the sense rather than antisense orientation (Frey et al., 1999). Transgenic N. plumbaginifolia plants over‐expressing NpZEP mRNA were obtained. In homozygous over‐expressing lines a maximum delay in germination of about 3 d was reported in this species (Frey et al., 1999). This effect was correlated with increased steady‐state NpZEP mRNA levels during seed development and increased ABA levels in mature seed (Frey et al., 1999). This is consistent with the idea that epoxyxanthophyll levels in non‐photosynthetic tissues, such as developing seeds and roots, could be limiting for sustained ABA synthesis. Expression of ZEP genes may be involved in the regulation of ABA biosynthesis in these non‐photosynthetic tissues.

Molecular characterization of 9‐cis‐epoxycarotenoid dioxygenase (NCED)

The second major breakthrough in understanding the molecular genetics of ABA biosynthesis was the cloning of 9‐cis‐epoxycarotenoid dioxygenase (Tan et al., 1997). This involved characterization of the viviparous 14 (vp14) gene locus in maize (Zea mays). As shown in Table 1, mutants impaired at this gene locus are unable to convert the final C₄₀ precursor(s) (9‐cis‐violaxanthin and/or 9′‐cis‐neoxanthin) into the first C₁₅ intermediate of ABA biosynthesis (xanthoxin) in an oxidative cleavage reaction (Fig. 1). The argument has often been made (most recently and persuasively by Qin and Zeevaart, 1999) that ‘the limiting step of ABA biosynthesis in plants is presumably the cleavage of 9‐cis‐epoxycarotenoids, the first committed step of ABA biosynthesis’. It is therefore difficult to overstate the importance of cloning this gene for studies on the control of ABA biosynthesis.

A programme of non‐targeted transposon mutagenesis was carried out in maize using the highly active transposon (MuI) in the Robertson's Mutator strain. Complete or partial precocious germination of seeds on the cob was used as a screen for new viviparous mutants; a phenotype which can be caused by genetic lesions in ABA biosynthesis. Two new viviparous mutants showed some signs of precocious germination but did not emerge fully on the cob. Both had weak penetrance and neither of the mutants appeared to be prone to water‐stress under field conditions. Complementation testing showed that the two mutant lines were allelic with each other, but not with any other known viviparous mutants. The new mutant alleles were termed vp14–2274 and vp14–3250.

Genomic DNA flanking one copy of the Mu1 element that co‐segregated with vp14–2274 allele was cloned into λgt10 and used as a probe to isolate cDNA and genomic clones of the wild‐type Vp14 gene. Further molecular analyses showed that in vp14–2274 the Mu1 element had integrated into the coding sequence, completely disrupting the ORF, resulting in a null allele. In the other allele, vp14–3250, the Mu1 transposon had inserted in the promoter region at a position 1 kb upstream from the start of the coding sequence (Tan et al., 1997).

The ORF of the Vp14 gene contains no introns and has a deduced sequence of 604 amino acids (Tan et al., 1997). The motif FDGDGM in VP14 is very similar to the FNGDGM motif of the two subunits of lignostilbene‐α,β‐dioxygenase (LSD) from the bacterium Pseudomonas paucimobilis (Kamoda and Saburi, 1995). A second VP14 motif, MIHDFAI, is also similar to that of the LSD α subunit MMHDFGV (Kamoda and Saburi, 1995). These and many other amino acid sequence homologies are consistent with VP14 encoding a dioxygenase (Schwartz et al., 1997a). More direct proof that VP14 was specifically involved in the oxidative cleavage reaction in ABA biosynthesis was obtained by producing a recombinant VP14 protein in E. coli. This was shown to cleave specifically xanthophyll substrates provided that they had the 9‐cis configuration. The putative ABA precursors, 9‐cis‐violaxanthin and 9′‐cis‐neoxanthin were both cleaved by recombinant VP14 to form xanthoxin in vitro (Schwartz et al., 1997a).

The ABA‐deficient mutant of tomato, notabilis (not), had been shown to metabolise xanthoxin to ABA readily (Parry et al., 1988). Subsequent analysis of leaf xanthophyll levels revealed no abnormalities in not mutant homozygotes (Parry et al., 1990). From this it could be inferred that not mutants were probably impaired in the same step of the pathway as the vp14 mutant of maize. Degenerate primers were designed (D McCarty, personal communication) based on the two conserved motifs between VP14 and LSD which were discussed in the previous paragraph. These were used to generate a 540 bp PCR product from tomato genomic DNA, which was used as a probe to screen a cDNA library generated from wilted tomato leaf tissue. A full‐length wild‐type cDNA clone was isolated (Burbidge et al., 1997b) which shared 63% identity with VP14 and encoded a protein of 605 amino acids (Burbidge et al., 1999). Primers designed to the extreme 5′ and 3′ ends of the coding region were used to PCR‐amplify the equivalent sequence from genomic DNA extracted from not mutant plants. DNA sequence analysis revealed only one consistent difference, a single base‐pair deletion in the not allele which corrupted a restriction enzyme cleavage site (TspRI). Southern analysis confirmed that this deletion was real and thus demonstrated that this wilty mutant had a frameshift mutation disrupting the ORF encoding a 9‐cis‐epoxycarotenoid dioxygenase of tomato (Burbidge et al., 1999). This is now referred to as an LeNCED gene and Vp14 is referred to as a ZmNCED gene (Liontenberg et al., 1999). The deduced amino acid sequences encoded by the wild‐type allele of the not gene locus (now known as LeNCED1), VP14 from maize and similar sequences from Arabidopsis and bean (Phaseolus vulgaris) have recently been compared (Qin and Zeevaart, 1999).

9‐cis‐epoxycarotenoid dioxygenase gene expression

Information on the expression of an NCED gene was first published quite recently (Tan et al., 1997). Since then reports have appeared describing NCED gene expression in tomato, Arabidopsis and in bean (Phaseolus vulgaris). A number of general conclusions can be drawn from these studies.

Diurnal variation

Northern analyses were carried out using an LeNCED1 probe in parallel to those discussed earlier in relation to steady‐state levels of LeZEP mRNA(Thompson et al., 2000a). It was a surprise to find that LeNCED1 mRNA levels also showed a strong diurnal rhythm in leaves, but with a clearly different phase from that of LeZEP mRNA. It appears that in tomato leaves NCED1 mRNA abundance reaches a peak at the end of the light period, rather than in the middle. Unlike ZEP mRNA, there was apparently no anticipation of the fact that the lights were about to go out. The LeNCED1 mRNA levels then fell by between 3‐ and 5‐fold with a minimum recorded value after 5 h of darkness. The steady‐state mRNA level was still at this low level when measured 1 h before the lights were due to come on, showing no sign that this event was being anticipated. This again was in contrast to LeZEP mRNA. During 48 h of continuous darkness, both LeZEP and LHCII mRNAs displayed one additional complete cycle and began to rise for a second cycle. This was not the case with LeNCED1 mRNA which remained at a low level throughout the extended dark period, showing no sign that it was being controlled by a circadian oscillator. It appears that this gene is subject to control by a mechanism that is sensitive to light and dark, but is otherwise insensitive to the time of day (Thompson et al., 2000a). Care must be taken to ensure that experiments monitoring the response of NCED mRNAs to other environmental stimuli do not span a light/dark transition. Further experiments should also be carried out to test for possible diurnal fluctuations in NCED mRNAs at high relative humidity, when fluctuations in water potential have been prevented. If such studies confirm that there is a genuine light effect, then the phenomenon requires an explanation. It is conceivable that the plant prepares for a possible emergency requirement for ABA biosynthesis by turning on this gene at dawn. The plant may anticipate that water stress would most likely occur during the day. Additional control mechanisms may determine whether this diurnal rise in NCED mRNA eventually results in elevated ABA levels.

Drought‐induced expression

In maize, dehydrated detached leaves accumulated higher levels of ZmNCED mRNA after a 5 h incubation period than corresponding controls which were not water‐stressed (Tan et al., 1997). In tomato, a 3‐fold increase in LeNCED1 mRNA was observed in leaves taken from whole plants which were not watered. The levels were relatively low at 12.00 h on day 2 of the drought and showed the full 3‐fold rise by the next time point at 16.00 h on the same day (Burbidge et al., 1997b). More recently, steady‐state LeNCED1 mRNA levels were determined at earlier time points after dehydration in detached tomato leaflets (Thompson et al., 2000a). At the first time point, only 15 min after detachment, a small increase in LeNCED1 mRNA abundance was detected. The increase was complete 45 min after detachment. In bean leaves a similarly rapid response was detected with PvNCED1 mRNA being induced strongly within 30 min of the start of the stress treatment (Qin and Zeevaart, 1999). These authors also reported that the mature form of the PvNCED1 protein could be detected after 1 h of stress. In both the tomato and the bean, these changes preceded a rise in ABA levels in the leaves. This is consistent with a cause and effect relationship; although it is necessary to upregulate steady‐state NCED mRNA levels in the absence of water stress in order to demonstrate this directly. The rapidity of the responses in both tomato and bean leaves, suggests that an initial reduction in turgor almost instantly triggers expression of this gene; which therefore must be extremely pressure sensitive.

In roots, the steady‐state LeNCED1 mRNA levels were almost 5‐fold lower than in the leaves on non‐stressed tomato plants (Thompson et al., 2000a). After 54 h without water being supplied, an 8‐fold rise in LeNCED1 mRNA was detected in the leaves and an increase of over 60‐fold was observed in the roots (Thompson et al., 2000a). Bean roots responded similarly to those of tomato, when they were detached and subjected to osmotic stress (Qin and Zeevaart, 1999). The bean roots accumulated PvNCED1 mRNA rapidly, within 1 h of the start of the stress treatment. As with leaves, rises in PvNCED1 mRNA and protein levels preceded the subsequent increase in ABA (Qin and Zeevaart, 1999). In both the leaves and the roots PvNCED1 mRNA was less abundant following re‐hydration (Qin and Zeevaart, 1999). This is consistent with an important role for NCED mRNA in the control of ABA biosynthesis in both leaves and roots. In the roots, at least, it seems probable that the increase in steady‐state ZEP mRNA is also required to maintain the supply of epoxycarotenoids for continued ABA biosynthesis (see earlier sections).

Developmental changes in expression

The abundance of NCED mRNA can clearly vary in a tissue‐specific manner, in that unstressed tomato roots have almost 5‐fold lower levels of LeNCED1 mRNA than leaves of the same plants (Thompson et al., 2000a). Although NpZEP mRNA increased as flowers matured and peaked one‐third to one‐half of the way through the process of seed development in N. plumbaginifolia (Audran et al., 1998), corresponding Northern data for NpNCED are not currently available. This aspect of NCED expression has not yet been investigated as thoroughly as the dramatic response to water stress.

Artificially‐induced modifications in expression

Work on transgenic plants with altered NCED gene expression is about a year behind that described in the corresponding section on ZEP. A number of studies on this topic are likely to be published shortly. Antisense and sense constructs involving the tomato LeNCED1 gene have been produced as a result of a collaboration between the groups at Nottingham (UK) and at HRI (Wellesbourne, UK). Early work with these constructs has been mainly focused on those with the NCED1 ORF in the sense orientation. A total of 61 independent primary transformants have been obtained (Thompson et al., 2000b). The majority of these transgenic tomato plants (39 in total) had a strong tendency to wilt and are therefore likely to be showing co‐suppression effects. Another group of 19 plants appeared to be normal and have not been investigated further.

All of the primary transformants were transferred to a controlled environment set to 90% relative humidity. Under these conditions three of the transformants grew slowly and were observed to guttate, i.e. to exude droplets of liquid at the leaf margins. Because all of the other 58 plants in the same environment were not showing this symptom, these three primary transformants were referred to as ‘over‐guttating’. ABA has been shown to result in 3‐fold increases in apparent hydraulic conductivity of sunflower roots resulting in substantial increases in the volume of xylem sap exuding from the cut surface of de‐topped plants (Glinka, 1980; Glinka and Abir, 1983). Increased volumes of xylem sap flowing from the roots to the shoots, combined with stomatal closure, probably results in exudation from the hydathodes at the leaf margins. Since this could be interpreted as a symptom of excessive ABA levels, these ‘over‐guttating’ transformants were selected for further analysis.

One of the three over‐guttating transgenics was found to contain multiple inserts and this presents difficulties in establishing homozygous lines. Most of the recent experiments have involved homozygous progeny derived from the other two primary transformants. Progeny from one of these primary transformants was shown to have 17‐fold higher LeNCED1 mRNA abundance in their leaves than corresponding wild‐type plants and a substantial and significant increase in bulk leaf ABA content. Reduced stomatal conductance and increased seed dormancy were also noted (Thompson et al., 2000b). Progeny of the second primary transformant had almost 4‐fold higher LeNCED1 mRNA abundance in their leaves than corresponding wild‐type plants. This did not result in a detectable increase in bulk leaf ABA content, but in spite of this a reduction in stomatal conductance and an increase in seed dormancy was observed. The seed ABA levels were significantly higher in both of these transgenic lines (Thompson et al., 2000b). It is possible that in the second line, any potential increase in ABA, resulting from its 4‐fold higher LeNCED1 mRNA level was balanced by increased catabolism in the leaves (as suggested by Zeevaart, 1999). However, the resulting increased flux of ABA, a possible change in compartmentation, or a possible increase in root‐sourced ABA, could explain the reduced stomatal conductance. In the other line, the 17‐fold higher steady‐state LeNCED1 mRNA level appears to have been sufficient to overwhelm the capacity for increased catabolism to prevent ABA from accumulating to excessive levels in the leaves. This study provides direct evidence that increasing steady‐state NCED mRNA alone is sufficient to result in enhanced ABA accumulation. This confirms that NCED is a key rate‐limiting enzyme in the control of ABA biosynthesis.

Molecular characterization of ABA aldehyde oxidase (ABAO)

The most recent breakthrough in understanding the molecular genetics of ABA biosynthesis involves the identification of an Arabidopsis gene encoding an ABA aldehyde (ABAld) oxidase (ABAO) which is specific to ABA biosynthesis (Seo et al., 2000). At the time of writing, it is only possible to refer to this in the form of abstracts. This represents the culmination of a groundbreaking series of papers on plant AOs by Koshiba and co‐workers. Koshiba and Matsuyama investigated indole‐3‐acetaldehyde oxidase activity in maize seedlings (Koshiba and Matsuyama, 1993). The first plant AO was subsequently purified from maize seedlings and found to be a 300 kDa homodimer which was clearly a molybdoenzyme (Koshiba et al., 1996). Alternative evidence on the amino acid sequence of a plant AO was obtained by database searching with a genomic DNA sequence obtained from tomato chromosome 11, which revealed homology with human (Wright et al., 1993) and bovine (Calzi et al., 1995) AOs. This suggested a function for the tomato ORF in encoding an aldehyde oxidase which was termed TAO1 (Ori et al., 1997). The predicted amino acid sequence of TAO1 and the purified maize AO revealed many conserved regions in plant AOs (Sekimoto et al., 1998). Plant AOs are members of complex multigene families; four AO cDNAs from Arabidopsis were obtained and positioned on four of the five chromosomes in the genome, i.e. atAO‐1 on chromsome 5; atAO‐2 on chromosome 3; atAO‐3 on chromsome 2, and atAO‐4 on chromosome 1 (Sekimoto et al., 1998). Two of these Arabidopsis cDNAs were obtained following PCR using degenerate primers based on conserved regions between maize and bovine AOs (atAO‐1 and atAO‐2). The other two were obtained by searching the Arabidopsis EST (atAO‐3) and genomic (atAO‐4) databases (Sekimoto et al., 1998). It now appears certain that the Arabidopsis EST E10BIT7 (Newman et al., 1994) was derived from an mRNA encoding an AO enzyme specifically involved in ABA biosynthesis (ABAO). The evidence for this conclusion is as follows:‐

• (i)Expressing genes encoding enzymes with a MoCo binding domain has been difficult in E. coli. Expression in various yeasts has been more successful, e.g. Arabidopsis nitrate reductase (NR‐holoenzyme) was expressed in an active form in the yeast Pichia (Su et al., 1996, 1997). Consequently, a Pichia heterologous expression system was used to demonstrate that an enzyme encoded by Arabidopsis atAO‐3 had substrate preference/specificity for ABAld (Seo et al., 1999—poster 454 at the IBC at St Louis). Experiments of this type provide a means of testing the suggestion that an AO converts xanthoxin to xanthoxic acid, rather than ABAld to ABA (Lee and Milborrow, 1997).

• (ii)An Arabidopsis ABA‐deficient wilty mutant (originally termed v11) co‐mapped with atAO‐3 on chromosome 2 and subsequent DNA sequencing of the atAO‐3 region in the mutant revealed a point mutation. The wilty mutant was therefore renamed aao3 and it was concluded that the gene product of the wild‐type allele was the ABAO enzyme involved in ABA biosynthesis in leaves (Seo et al., 2000).

There has been insufficient time, compared with ZEP and NCED, to obtain a full picture of the role of ABAO in the control of ABA biosynthesis; consequently, it is not yet possible to review ABAO on the basis of the sections/subheadings previously used for ZEP and NCED. However, it is interesting to note that Seo et al. have found that in Arabidopsis the AAO3 mRNA is mainly expressed in leaves and that steady‐state mRNA levels are rapidly increased in dehydrated leaf tissue (Seo et al., 2000).

Concluding remarks

From this review it can be seen that three of the steps in the ABA biosynthetic pathway can now be investigated at the level of gene expression. This is providing a fascinating insight into the way that plants respond to environmental stimuli by turning on ABA biosynthesis. Steady‐state levels of all three mRNAs appear to increase in response to water stress in at least some plant tissues. It is important to know how expression of the three genes is co‐ordinated to ensure that ABA synthesis can be sustained efficiently. It is also necessary to remember that the balance between ABA synthesis and metabolism is vital in determining whether ABA accumulates in whole plants (see Cutler and Krochko, 1999, for a review of both aspects). ABA is metabolized by ABA 8′‐hydroxylase activity which is induced by ABA itself (Cutler et al., 1997). This effect is clearly antagonistic to attempts to increase ABA levels by over‐expressing ABA biosynthesis genes. If ABA were also to inhibit its own synthesis by some form of feedback, then a combination of these two effects would tend to frustrate attempts to produce transgenic plants with elevated ABA levels. No suppression of drought‐induced increases in steady‐state LeNCED1 (Thompson et al., 2000a) and PvNCED1 (Qin and Zeevaart, 1999) mRNA by ABA has been detected. In practice, it has proved possible to produce transgenic plants with an increased ABA content by over‐expressing ABA biosynthetic genes (Frey et al., 1999; Thompson et al., 2000b). ABA biosynthesis is now subject to human as well as environmental control.

References