The pathway of biosynthesis of abscisic acid in vascular plants: a review of the present state of knowledge of ABA biosynthesis

Abstract

The pathway of biosynthesis of abscisic acid (ABA) can be considered to comprise three stages: (i) early reactions in which small phosphorylated intermediates are assembled as precursors of (ii) intermediate reactions which begin with the formation of the uncyclized C₄₀ carotenoid phytoene and end with the cleavage of 9′‐cis‐neoxanthin (iii) to form xanthoxal, the C₁₅ skeleton of ABA. The final phase comprising C₁₅ intermediates is not yet completely defined, but the evidence suggests that xanthoxal is first oxidized to xanthoxic acid by a molybdenum‐containing aldehyde oxidase and this is defective in the aba3 mutant of Arabidopsis and present in a 1‐fold acetone precipitate of bean leaf proteins. This oxidation precludes the involvement of AB‐aldehyde as an intermediate. The oxidation of the 4′‐hydroxyl group to the ketone and the isomerization of the 1′,2′‐epoxy group to the 1′‐hydroxy‐2′‐ene may be brought about by one enzyme which is defective in the aba2 mutant and is present in the 3‐fold acetone fraction of bean leaves. Isopentenyl diphosphate (IPP) is now known to be derived by the pyruvate‐triose (Methyl Erythritol Phosphate, MEP) pathway in chloroplasts. (¹⁴C)IPP is incorporated into ABA by washed, intact chloroplasts of spinach leaves, but (¹⁴C)mevalonate is not, consequently, all three phases of biosynthesis of ABA occur within chloroplasts. The incorporation of labelled mevalonate into ABA by avocado fruit and orange peel is interpreted as uptake of IPP made in the cytoplasm, where it is the normal precursor of sterols, and incorporated into carotenoids after uptake by a carrier in the chloroplast envelope. An alternative bypass pathway becomes more important in aldehyde oxidase mutants, which may explain why so many wilty mutants have been found with this defect. The C‐1 alcohol group is oxidized, possibly by a mono‐oxygenase, to give the C‐1 carboxyl of ABA. The 2‐cis double bond of ABA is essential for its biological activity but it is not known how the relevant trans bond in neoxanthin is isomerized.

Biosynthesis pathway, abscisic acid, vascular plants.

Introduction

The compound that is now known as ‘abscisic acid’ (Fig. 1) was isolated as an abscission factor named ‘abscisin II’ (Ohkuma et al., 1963). At the same time, what was believed to be a dormancy hormone was isolated from sycamore leaves (Cornforth et al., 1965) and when the infrared spectra were compared they were seen to be identical (Addicott et al., 1968). Shortly after this it was discovered that the amounts of an inhibitory material increased considerably when plants wilt (Fig. 2) (Wright and Hiron, 1969) and that ABA caused stomata to close (Mittelheuser and Van Steveninck, 1969). These two discoveries highlighted the major role of ABA in plants: its anti‐stress effects.

A new line of evidence supporting the role of endogenous ABA in closing stomata was obtained (Artsaenko et al., 1995) and these authors found that an artificial anti‐ABA fV antibody in transgenic tobacco plants targeted to be retained by the endoplasmic reticulum caused the leaves to become wilty as their stomata could not close. Extracts of the transgenic leaves contained 120–1050 ng ABA g⁻¹ FW while control wild‐type leaves contained 30–100 ng g⁻¹. The ABA, therefore had penetrated into the lumen of the endoplasmic reticulum where it had been held and was unable to move to the stomata.

Subsequent work has increased the list of unfavourable conditions that can induce the rapid synthesis of ‘stress’ ABA and whose deleterious effects can be combated by ABA. These include heat, salinity, chilling and, recently, herbicide toxicity (Grossmann et al., 1996). So dramatic is the rise in the amount of ABA in stressed leaves (often 40‐fold) that an earlier, smaller effect of wilting has gone almost unnoticed: a release and redistribution of the ABA within the cells of the leaf (Cornish and Zeevaart, 1985a, b; Harris and Outlaw, 1991; Loveys, 1977). The effects of environmental factors on the content of ABA in various organs will not be considered in detail here.

Some pathogenic fungi make ABA but the pathway appears to be quite different from that by which higher plants make the compound (Neill et al., 1982). The intermediates in fungi contain no more than 15 carbon atoms and different intermediates occur in different taxa. This ‘direct’ pathway will not be considered in this review.

Molecular genetic techniques are being used increasingly to investigate pathways of biosynthesis but it is necessary to base these investigations on a clear understanding of the biochemical and physiological constraints, otherwise adventitious reactions can be misidentified as true pathways.

Changes in two aspects of biochemistry have been emphasized as a result of recent molecular biological advances. These are (i) that many peptides carry two or more different enzymic activities; (ii) a biochemical genetic lesion can induce the action of other artefactual enzymic reactions (see Section on Abscisic alcohol). An example of (i) has been reported previously (Helliwell et al., 1999). They isolated, characterized and cloned an enzyme from Arabidopsis into yeast whose single peptide chain contains three quite different enzyme activities. The enzyme has monooxygenase activity and hydroxylates ent‐kaurene to ent‐kaurenol. Dehydrogenase activity then oxidizes the first product to ent‐kaurenal. It also functions as an aldehyde oxidase to form ent‐kaurenoic acid (Fig. 3).

It is not known how mutational impairment of one of these steps would affect the others. It is even possible that the sensitivity of the aldehyde oxidase to CN‐ reported earlier (Tan et al., 1997), could be the result of its effect on a cytochrome cofactor required by one enzymic activity having an effect on another catalytic site of a complex aldehyde oxidase enzyme similar to the kaurene enzyme.

The two important new developments in ABA biosynthesis are (i) it was found that terpenoid biosynthesis in chloroplasts uses IPP which is produced from pyruvate and glyceraldehyde phosphate whereas IPP is formed from mevalonate in the cytosol for synthesis of sterols (Rohmer et al., 1993; Rohmer, 1999) (Figs 4, 5). (ii) The second development has been the exploitation of biosynthetically labelled carotenoids to obtain direct evidence that cell‐free systems were able to convert carotenoids into ABA (Cowan and Richardson, 1993; Lee and Milborrow, 1997a; Richardson and Cowan, 1996).

The formation of carotenoids

The pathway and mechanisms of biosynthesis of carotenoids were worked out by Goodwin and colleagues using labelled mevalonates to follow the interconversions of intermediates, usually in cell‐free systems (Goodwin, 1971). Although it now appears that mevalonate is normally a minor precursor of carotenoids in chloroplasts, the labelling results are valid because the labelled mevalonates supplied were converted into an intermediate (probably isopentenyl diphosphate (IPP)) which entered the chloroplasts and was incorporated into carotenoids. IPP is an intermediate in both pathways (carotenoids and steroids).

Fig. 1.

Structures.

Fig. 2.

The amount of ABA in excised wheat leaves. Measurements (μg equivalents of (±)‐ABA kg⁻¹ FW) were made 330 min after the start of the treatment to induce wilting began (data from Wright, 1977).

Fig. 3.

Conversion of the hydrocarbon ent‐kaurene into ent‐kaurenoic acid by one peptide with three enzymic activities: hydroxylation–monooxygenase; alcohol‐aldehyde oxido‐reductase; aldehyde oxidase (Helliwell et al., 1999).

Fig. 4.

The MEP biosynthetic scheme (adapted from Schwender et al., 1996; Kuzuyama et al., 1998; Rohdich et al., 1999) for the biosynthesis of isopentenyl diphosphate (IPP) from pyruvate and glyceraldehyde 3‐phosphate in chloroplasts. Carbon atoms from the precursors can be followed by their labels: C‐1 of pyruvate lost as CO₂=□; C‐2=○; C‐3=▵. Glyceraldehyde phosphate: C‐1=▪; C‐2=•; C‐3=▴. The labelling of ABA is according to the incorporation of labelled mevalonates via IPP (Milborrow, 1972, 1981; Willows and Milborrow, 1989).

Fig. 5.

The biosynthetic pathway for the formation of isopentenyl diphosphate (IPP) from mevalonate, in the cytosol. Compare with Fig. 4.

Evidence for the ‘indirect’ carotenoid pathway

Kinetics of labelling

The kinetic analysis of biosynthesis of ABA and the mechanisms that control it is beset with several major difficulties at present because some of the ‘early reactions’ between pyruvate and glyceraldehyde phosphate (to give deoxyxylulose) and on to IPP have not been fully characterized (Fig. 4). The ‘intermediate reactions’, which may be considered as those involving C₄₀ carotenoids, introduce yet another set of potential confusions because most of these compounds are virtually insoluble in water and may occur in two or more, quite separate, metabolic pools where they may, or may not, be available for further enzymic transformations (Li and Walton, 1987). Their complicated structures preclude radiochemical synthesis. β‐carotene, for example, occurs in solution in oil droplets within chloroplasts, as a light harvesting pigment within photosynthetic antennal protein molecules and as a membrane constituent (Kreuz et al., 1982) as well as a precursor of the β,β‐xanthophylls, some of which give rise to ABA. The ‘late reactions’ comprising C₁₅ intermediates formed from carotenoids are still uncertain and some may occur with the future ABA residue held by a jig or template while it is elaborated.

A further complication has recently been discovered (Vittone and Milborrow, 1999; BV Milborrow, unpublished results) in that most of the ABA in leaves is not free in solution but is held by some proteinaceous material; this is referred to as ‘fettered’ ABA. Loveys found that the majority of the ABA in broad bean leaves was in the chloroplasts, but when they were sonicated and then centrifuged the ABA was in the pellet (Loveys, 1977). Wilting caused release of free ABA and most of the solvent extraction procedures that have been used in the past remove the ABA from the material that binds it. Exogenously applied ABA does not mix with this ‘fettered’ ABA so measurements of turnover of the endogenous ABA using labelled ABA applied to the tissues are of uncertain validity.

There may even be a small separate pool of carotenoids destined to be precursors of ABA because it has been found that successive wilt/recovery cycles produced smaller and smaller pulses of ‘stress’ ABA (Li and Walton, 1987). Thus any attempts to analyse the flux through the pathway are confused by the complications of the different pools. An experiment which may show an example of this has been reported (Nonhebel and Milborrow, 1987). Tomato plants supplied with ²H₂O for some days were extracted and the specific ²H enrichment of the endogenous ABA measured by GC/MS. When the leaves were wilted so that large quantities of ‘stress’ ABA were formed, its specific enrichment was less. The interpretation of this result was that the ‘stress’ ABA was produced from a larger pool of unlabelled precursors formed before ²H₂O was supplied. Even this explanation is subject to criticism as the distribution of ²H₂O could not be controlled and young and old leaves have been shown to form ‘stress’ ABA at different rates (Cornish and Zeevaart, 1984). Some parts of the shoots containing more heavily labelled precursors may have contributed a disproportionate amount of the original ABA.

Quantitative measurements of ABA and carotenoids

The amounts of each of the xanthophylls that could give rise to ABA, violaxanthin, anthera‐ and neoxanthin, exceed the amount of ABA in a leaf by 3–100‐fold so it is difficult to relate the rise in ABA to a very small percentage fall in the amount of any carotenoid present. There is also the difficulty of a continuing synthesis of carotenoids. This was overcome by growing barley in darkness and treating the leaves with fluridone, an inhibitor of the formation of carotenoids (Gamble and Mullet, 1986). When the leaves were wilted the amounts of viola‐ and neoxanthin fell by 7.7 nmol (41%) and 1.2 nmol (57%) while ABA rose by 3.9 nmol g⁻¹ DW (7.8 times). Lutein and β‐carotene increased from 41 to 43 and 2.2 to 2.5 nmol. Similar results were obtained using etiolated bean leaves (Li and Walton, 1990): these authors found that the rise in ABA plus its metabolites phaseic acid [2] and dihydrophaseic acid [3] closely matched the decrease in amounts of neoxanthin plus violaxanthin.

Octadiene dioic acid

Octadiene dioic acid (ODA) [4] was discovered in ABA‐deficient flacca mutant tomato plants (Linforth et al., 1987a, b). Its carbon skeleton suggested that it was the residual, central, C₁₀, part of a carotenoid from which the two terminal rings had been removed as the carbon skeletons of ABA.

An attempt was made to find if ODA was derived from the same carotenoid molecule as ABA by growing normal tomato plants in 40% ²H₂O (Milborrow et al., 1988). After 6 d the ABA molecules were found to contain between three and ten ²H atoms. The central part of the carotenoid molecule would be expected to carry eight ²H atoms if it were derived from the same carotenoid molecule whose end group had given rise to the ABA. The ODA contained just one ²H atom which was attributed to the reduction of one of three double bonds in the central residue by NADHH⁺ where the H⁺ would have been ²H derived from the water.

This result showed that the ABA could not have been formed from the same carotenoid molecule as that which gave rise to the ODA. Later work has established that ABA is derived from 9′‐cis‐neoxanthin (Fig. 1) and so only one ABA residue per carotenoid is likely. The dioxygenase cleavage produces an apo‐carotenoid (C₂₅) with an allenic terminal ring. (Schwartz et al., 1997b). Why ODA is abundant in ABA‐deficient mutants is a mystery.

¹⁸O in the carboxyl group of ABA

The ‘indirect’ or ‘carotenoid’ pathway has generally been supposed to occur by the cleavage of the C₁₅ carbon skeleton of ABA from a C₄₀ carotenoid. The demonstration that an ¹⁸O atom from ¹⁸O₂ is incorporated into the carboxyl group of ABA made in leaves of several species was a strong indication that some precursor larger than the 15 carbon atoms of ABA underwent oxidative cleavage (Creelman et al., 1987).

¹⁸O labelling

One of the difficulties with using ¹⁸O for detecting labelling patterns in ABA is that the 4′‐keto oxygen of ABA (Fig. 1) exchanges with the medium (Zeevaart and Milborrow, 1976). The complexity of using ¹⁸O is highlighted by the rigorous stability of the 1′‐hydroxyl oxygen atom of ABA. In contrast, when the 4′‐ketone group of ABA is reduced to form the diols then the 1′‐oxygen becomes labile and exchanges and even epimerizes (Vaughan and Milborrow, 1987) and so also does the oxygen atom of the 4′‐hydroxyl group. Rock and Zeevaart (1990) found that the C‐1 carbonyl oxygen of AB‐aldehyde (Fig. 1) also exchanges rapidly in water. The less than expected amount of ¹⁸O at C‐1 of ABA was taken to imply that AB‐aldehyde was a precursor of ABA but the C‐1 carbonyl oxygen of xanthoxal (Fig. 1) may be equally susceptible to exchange (Rock and Zeevaart, 1990). It is not known if the oxygen atom of the 4′‐hydroxyl of xanthoxal or of other precursor molecules is capable of exchange.

Pathways

Mevalonate pathway

The discovery of mevalonate as the precursor of terpenoids (Fig. 5) and the establishment of their pathway of biosynthesis was mostly worked out in animals. The stereospecifically labelled mevalonates were then used to probe the mechanisms of biosynthesis of carotenoids and other terpenoids in plants (Wright, 1961). Gradually observations accumulated which hinted that all was not quite right with the hypothesis that the pathway was the same in plants and animals: it was found that pyruvate was incorporated into carotenoids much more efficiently than acetate (Shah and Rogers, 1969) and that mevinolin (a potent inhibitor of mevalonate biosynthesis) hardly affected carotenoid formation (Alberts et al., 1980; Bach and Lichtenthaler, 1982). Recently, other authors found ¹³CO₂ was incorporated rapidly in the light into leaf terpenoids but only slowly into leaf steroids, which are made in the cytoplasm (McCaskill and Croteau, 1995; Loreto et al., 1996).

The methyl erythrose phosphate (MEP) pathway

The paradox was resolved by Rohmer and colleagues who found that IPP was produced in the chloroplasts of plants, and in some bacteria, by a novel pathway where two carbon atoms of pyruvate were condensed with glyceraldehyde phosphate to produce 1‐deoxyxylulose‐5‐P (Fig. 4) (Rohmer et al., 1993; Schwender et al., 1996; Lichtenthaler et al., 1997a, b; Rohmer, 1999). Kuzuyama et al. showed that a reductive rearrangement gave 2‐C‐methyl erythritol‐4‐phosphate (MEP) (Kuzuyama et al., 1998). Rodich et al. (1999) found that MEP was conjugated in a cytidine triphosphate (CTP) dependent reaction to give the sugar phosphate in Fig. 8A. This was accomplished by the daringly novel means of scanning DNA and protein databases for CTP‐dependent reactions, which revealed a candidate enzyme in Haemophilus influenzae. They expressed the gene in E. coli and found that the enzyme catalysed the formation of the phosphoribosyl cytidine derivative of MEP. The remaining enzymes and intermediates to give IPP are unknown. Perhaps CTP should be added to cell‐free systems producing IPP from pyruvate as it may increase yields.

Chloroplasts take up IPP and incorporate it into ABA (Milborrow and Lee, 1998a) so it is probably cytosolic IPP produced from added, labelled mevalonate that can enter chloroplasts and be incorporated into carotenoids. This provides an explanation for the mechanism by which mevalonate is incorporated into ABA by intact cells of avocado fruit. Photosynthetically fixed ¹³CO₂ was found in sterols made in the cytoplasm (McCaskill and Croteau, 1995) but it has not been established if this is caused by leakage of chloroplastic IPP itself or leakage of glucose, glycerate or other photosynthetic intermediates into the cytoplasm where they can be converted into acetyl residues and then into IPP via mevalonate. The separation of cytoplasmic sterol biosynthesis from chloroplastic terpenoid biosynthesis has been well documented (Cvejic and Rohmer, 2000).

Terpenoid synthesis: the pyruvate‐triose (MEP) pathway

Milborrow and Lee found that an avocado cell‐free preparation that routinely incorporated [¹⁴C] mevalonate into ABA gave much higher yields when supplied with [¹⁴C]pyruvate (Milborrow and Lee, 1998a). As intact cells of avocado fruit incorporate labelled mevalonates into ABA it is now assumed that IPP, or a similar precursor produced from mevalonate in the cytosol, is taken into the plastids and incorporated into carotenoids and ABA. A chloroplastic fraction from avocado fruit was described that was able to convert [¹⁴C]mevalonate into ABA, but the chloroplasts were contaminated with cytosol (Milborrow, 1974). When spinach leaf mesophyll protoplasts were carefully lysed and the clean, intact chloroplasts collected by centrifugation, washed by gentle resuspension and recentrifugation (Milborrow and Lee, 1998a), [¹⁴C]pyruvate (2700 dpm) and [¹⁴C]IPP (1575 dpm) were incorporated into ABA in high yield, but mevalonate was hardly incorporated (85 dpm) (Table 1 ). This result explains why small amounts of [¹⁴C]mevalonate were incorporated into ABA by chloroplasts from avocado—they are notoriously sticky and contaminated with an almost equal volume of cytosol. It was also found that ABA was not synthesized by chloroplasts (Hartung et al., 1981), but [¹⁴C]mevalonate was fed, which is now known not to be assimilated by chloroplasts unless the preparation also contains some cytoplasmic enzymes. The avocado fruit cell‐free system that was used by Lee and Milborrow to demonstrate the conversion of mevalonate into carotenoids and ABA was deliberately prepared from total cell homogenates so that enzymes from all organelles would be present. Other authors prepared an acetone powder from ripening orange peel, again, enzymes from all parts of the tissue would have been present (Cowan and Richardson, 1993; Richardson and Cowan, 1996). The rate of incorporation of [¹⁴C]mevalonate into ABA by the orange peel extract was several hundred times greater than that in the avocado cell‐free system. An explanation of this can be proposed based on present knowledge of the pyruvate/ triose pathway and the physiology of the ripening process in orange peel. Orange peel flavedo synthesizes large quantities of carotenoids in chromoplasts as the chlorophyll disappears, so the major source of IPP probably has to come from mevalonate via the cytosolic pathway. Hence the rapid assimilation and incorporation of mevalonate into carotenoids and ABA. One might predict that the orange peel powder would be relatively less effective than the avocado system at forming ABA from [¹⁴C]pyruvate.

Table 1.

Formation of labelled ABA by isolated, washed intact chloroplasts from spinach leaves and by lysed protoplasts (data from Milborrow and Lee, 1998a)

	Compound added	dpm in ABA
Experiment 1
Disrupted protoplasts	R‐[14C]mevalonate	575
(intact chloroplasts+cytosol)	[14C]IPP	1025
Washed, intact chloroplasts	[14C]IPP	1575
Experiment 2
Isolated, washed, intact   chloroplasts	R‐[14C]mevalonate	85
chloroplasts
	[14C]pyruvate	2700

Location of biosynthesis in plant parts

The ability of a number of plant organs to make ABA has been tested by wilting isolated leaves, stems, fruits, and seeds and measuring if an increase in ABA occurs within several hours. There is also an indication from work on ABA‐deficient mutants that the regulation of formation of ABA in the seed is different from that in leaves (Ellis et al., 1999).

ABA is synthesized in several different parts of plants (Milborrow and Robinson, 1973) and the regulatory controls appear to be different. For example, the wilting of leaves induces a sudden and rapid increase in ABA content to 40 times higher than the normal value (Milborrow, 1981) (‘stress’ ABA) (Fig. 2) while the amounts of ABA in roots increase progressively as the tissues lose water (Cornish and Zeevaart, 1985). Avocado fruit discs synthesized ABA rapidly during the ripening period and were unaffected by wilting. Seeds of cereals make ABA during certain phases of growth and it stimulates accumulation of storage reserves and induces seed dormancy, although excess ABA imported from wilting leaves during early seed development appears to cause abortion of cereal seeds. The source of these diverse patterns of response may lie with the occurrence of eight copies of the dioxygenase gene detected by low stringency Southern blotting of the cloned VP14 gene with genomic DNA from maize (Tan et al., 1997). The different genes may have different initiation sequences which respond in particular ways in different tissues.

Support for this suggestion comes from the results of Phillips et al. where the DNA for an anti‐ABA antibody under the control of the seed‐specific USP promotor from Vicia faba was used to transform tobacco (Phillips et al., 1997). Seeds of the transformed tobacco plants had near zero ABA contents 21 d after pollination. The seeds could be germinated precociously and could not tolerate drying. Seed‐specific immunomodulation blocked normal seed maturation and permitted germination.

Although guard cells contain chloroplasts, it was found that epidermal strips from broad bean leaves did not increase their content of ABA when wilted (Loveys, 1977) nor did isolated epidermal strips of Commelina communis (Doerffling et al., 1980). It was found that stomata were unable to convert xanthoxal into ABA (Parry et al., 1988) (2‐cis‐xanthoxin is referred to as xanthoxal (XAN) according to the agreed new nomenclature (Milborrow et al. 1997a)). As stomatal guard cells contain chloroplasts they would be expected to contain the apparatus necessary to make ABA. If stomata cannot synthesize ABA then they must be entirely dependent on ABA coming through the apoplast as guard cells lack plasmodesmata (Sanchez, 1977). These organ‐specific differences are all the more surprising when ABA has been shown to move round the plant quite readily in the xylem and phloem (Zeevaart, 1977; Hoad 1973; Munns and Sharp, 1993).

After wilted leaves regain turgor there is rapid inactivation of ABA by oxidation to phaseic acid (Grantz et al., 1985; Windsor and Zeevaart, 1997; Harris and Outlaw, 1991; Cornish and Zeevaart, 1984) and some is converted into the glucose ester (Carrington et al., 1988) and the glucoside (Milborrow and Vaughan, 1982) of the reduction product: dihydrophaseic acid (Fig. 1).

The main rise in ABA caused by water loss occurs some 2–3 h after wilting and involves the synthesis of new enzymes. The rise in ‘stress’ ABA can be blocked by cycloheximide so it involves synthesis of enzymes in cytoplasmic ribosomes (Guerrero and Mullet, 1986; Li and Walton, 1990; Quarrie and Lister, 1984a, b). Tan et al. have sequenced the gene for the dioxygenase enzyme that cleaves 9′‐cis‐neoxanthin to give xanthoxal and found the first 100 amino acid residues comprise a chloroplastic transit signal sequence (Tan et al., 1997). The amounts of its m‐RNA showed a clear response to wilting conditions so it is probably the rate‐limiting step for the synthesis of ‘stress’ ABA: its genetics and sensitivity to cycloheximide establish that it is produced in the cytosol and imported into the chloroplasts.

Location of biosynthesis within cells

Sindhu and Walton claimed that xanthoxal was not converted by a chloroplast fraction into ABA, but a cytoplasmic fraction was able to form ABA from it (Sindhu and Walton, 1987). Subsequent commentators have assumed that xanthoxal is converted into ABA in the cytoplasm after being released from a carotenoid in a plastid. In contrast to this Milborrow and Lee isolated washed, intact spinach leaf chloroplasts from protoplasts and found that they incorporated pyruvate and isopentenyl diphosphate into ABA in high yield, mevalonate in very low yield (Milborrow and Lee, 1998a). This established that the whole pathway of biosynthesis from five carbon precursors onwards to ABA can occur within chloroplasts; if any damage had occurred to the chloroplasts during washing they would have lost stromal enzymes and as they were able to synthesize ABA their biosynthetic machinery was not seriously damaged. The washing would have removed the cytosolic enzymes, as demonstrated by the very low incorporation of [¹⁴C]mevalonate (Table 1). The penetration of isopentenyl diphosphate through the plastid envelope has been shown to be mediated by a carrier (Soler et al. 1993) in Vitis vinifera plastids. Scrutiny of Sindhu and Walton's methods revealed that their ‘chloroplasts’ were prepared by homogenization of the leaves in a Waring Blendor at maximum speed followed by centrifugation at 1000 g, resuspension and dialysis (Sindhu and Walton, 1987). The chloroplast fragments could be expected to have lost a large proportion of their stromal enzymes so their inability to convert xanthoxal into ABA may reflect their treatment rather than an inherent lack of an ability to make ABA. The ‘cytosol’ fraction was the supernatant fraction from a homogenate prepared by grinding a leaf with a pestle and mortar and then centrifuging at 100 000 g to remove all other membranes and particles. It can be expected, therefore, to contain a proportion of soluble stromal enzymes which have subsequently been shown to be able to convert xanthoxal into ABA (Sindhu et al., 1990).

Possible artefactual reactions in cell‐free biosynthesis

As can be seen from Fig. 6, it is possible for labelled ABA to be synthesized in cell‐free systems from labelled carotenoids by spurious, unnatural reactions. The demonstration of the conversion of a putative precursor into ABA by a cellular extract, therefore, does not establish that the reaction occurs in vivo.

It was considered essential to devise a protocol that would exclude an adventitious path to ABA. ‘Cold traps’ are inadequate as they could exhibit all the characteristic features of a pathway in spite of its being unnatural. The procedure adopted by Lee and Milborrow was to add a number of specific metabolic inhibitors to the cell‐free system forming [¹⁴C]ABA from [¹⁴C]mevalonic acid and to compare their potency with their effect against incorporation of [¹⁴C]mevalonate into ABA by avocado fruit discs (Lee and Milborrow, 1997a). If the inhibitors had similar effects on [¹⁴C]ABA production in the cell‐free system as in intact cells in vivo then one can confidently expect that ABA is being formed by the same processes in the two situations. In addition, the ability of ‘cold pools’ or ‘cold traps’ of possible carotenoid intermediates to dilute the ¹⁴C of mevalonate (and thereby diminish the amount of ¹⁴C present in ABA) was also examined.

When the two sets of results were compared (Fig. 7) they were so similar that there can be little doubt that the reactions in the cell‐free system were the same as those by which ABA was made in vivo. For example, AMO1618, an inhibitor of sterol cyclization, increased the [¹⁴C]mevalonate incorporated into ABA in both instances. It is presumed that the blocking of cyclization of squalene epoxide to give steroids caused the earlier intermediates to feed back and accumulate and so allow [¹⁴C]IPP to be diverted into carotenoid formation. The lowering by carbon monoxide of the amount of ¹⁴C passing into ABA was not as great as expected in both the cell‐free system and fruit discs, while the monooxygenase inhibitor piperonyl butoxide was considerably more potent. A similar response to CO was noted by Creelman et al. (Creelman et al., 1992).

Fig. 6.

A sequence of artefactual reactions that could lead to the formation of ABA from violaxanthin by spurious reactions.

Fig. 7.

Comparison of the effectiveness of some specific metabolic inhibitors on the incorporation of [¹⁴C]mevalonate into ABA in an avocado cell‐free system compared with their effectiveness in avocado fruit discs. The similarity of the effectiveness in the two situations, in spite of the possibility of imperfect distribution in the fruit, is taken as evidence that the biosynthetic pathway in the cell‐free system is the same as that by which ABA is made in vivo.

Fig. 8.

(A) Carbon skeleton of a hypothetical precursor of abscisic acid showing the pro‐chirality of the H atoms in mevalonic acid before any are lost during the formation of the double bonds. The numbers refer to the positions of the carbon atoms in mevalonic acid. (B) The stereochemical origin in mevalonic acid of the H atoms which are retained in ABA. The double bond at C‐4 in the side chain is formed by a trans elimination of the C‐2‐pro‐S and 5‐pro‐R H atoms of mevalonate from the C‐4 and C‐5 of ABA, respectively. C‐5′ of the ring of ABA (derived from C‐4 of mevalonate) carries the 4‐pro‐R‐H atom of mevalonate in the 4′‐pro‐R position and an H atom derived from the medium during cyclization (shown as H⁺) in the 5′‐pro‐S position. The H atom retained at C‐3′ of the ring of ABA is the C‐2‐pro‐R of mevalonate.

Inhibition by tungstate

Molybdate has been shown to be a necessary cofactor for ABA biosynthesis (Walker‐Simmons et al., 1989) while in the aldehyde oxidase of a deep sea bacterium (Chan et al., 1995) it is replaced by tungstate—a metal below Mo in group VIB of the Periodic Table. Sodium tungstate, therefore, was added to the cell‐free system in the expectation that it would replace molybdate as the metal ion cofactor in the aldehyde oxidase and block the formation of ABA. Not only did tungstate act as a potent inhibitor of ABA formation, but it forms a strong, insoluble complex with an alkaloid, cinchonine, which, when added to the cell‐free system, restored enzyme activity. Chromate ions at the same concentration had virtually no effect on the amount of [¹⁴C]ABA formed (Cr is above Mo in Group VIB).

When tungstate was added to the cell‐free system it reduced the amount of ¹⁴C incorporated into ABA but [¹⁴C]xanthoxal accumulated and addition of the alkaloid cinchonine (to remove the tungstate) allowed the [¹⁴C]xanthoxal to be converted into [¹⁴C]ABA. The ready restoration of biosynthesis by removal of tungstate suggests that the molybdate is weakly held by the enzyme and that xanthoxal is its substrate. If oxidation of an aldehyde later in the sequence of reactions were inhibited there would not be a backing up of a series of intermediates sufficient to block the oxidation of xanthoxal by product inhibition at the nanomolar concentrations of the 15 carbon substrates. If the C‐1 aldehyde group of xanthoxal is the first part of the molecule to be oxidized then AB‐aldehyde cannot be an intermediate.

Xanthoxal, therefore, is the substrate and the product is xanthoxic acid (Fig. 1). Xanthoxic acid has been found in a cell‐free system from orange peel (Cowan and Richardson, 1997) and (+)‐[¹⁴C]xanthoxic acid was converted rapidly into [¹⁴C]ABA by avocado fruit (Milborrow et al., 1997b). Consequently AB‐aldehyde cannot be the main natural precursor of ABA and although it was detected in extremely small quantities in apple fruit (Rock and Zeevaart, 1990), the majority (c. 5:1 2‐trans:cis) was the 2‐trans‐isomer which gives 2‐trans‐ABA.

Bisulphite and dimedone form complexes with aldehydes and they inhibited the production of ABA in the cell‐free system (Lee and Milborrow, 1997a). The dimedone complex of xanthoxal was identified, but this particular result does not exclude the participation of AB‐ald as the sequestering of the first aldehyde (XAN) would automatically preclude the formation of the second (AB‐ald) if it were formed from XAN.

Mutants

Wilting mutants

ABA‐deficient mutants are known in tomato, pea, maize, tobacco, barley, Arabidopsis, and potato, but they have provided less help in defining the pathway of biosynthesis than expected. The exception to this generalization has been the maize transposon mutant VP14 which led to the isolation of the gene for the dioxygenase enzyme which cleaves 9′‐cis‐neoxanthin to give xanthoxal (Schwartz et al., 1997b).

Tomato mutants flacca and sitiens have been found to be unable to oxidize AB‐aldehyde to ABA so the genetic lesion was concluded to be in the aldehyde oxidase gene. These and other mutants unable to make ABA (Leydecker et al., 1995) were believed to lack aldehyde oxidase because they failed to oxidize heptaldehyde (Fig. 1). The wild type was able to oxidize AB‐aldehyde and heptaldehyde. This established that the oxidase is somewhat unspecific, however, the other possible aldehydic precursor, xanthoxal, was tested by Parry et al. and flacca and sitiens failed to oxidize it (Parry et al., 1988). Thus xanthoxal has equal claim as AB‐aldehyde to be the substrate of the oxidase which is defective in the mutants. Recent research (Sagi et al., 1999) has complicated the role of the aldehyde oxidase in flacca even more because leaves contain about 23% of the normal amount of ABA (Neill and Horgan, 1985; Rock et al., 1991) while roots contain almost as much as those of the wild type. An example of an ABA‐deficient mutant barley was found in which a cofactor deficiency made the plants unable to synthesize ABA and also impaired the mutant's capacity to reduce nitrate (Walker‐Simmons et al., 1989) (see also aba3 mutant in the next section).

Identification of the genetic lesions

Two Arabidopsis mutants that are deficient in ABA have been characterized (Schwartz et al., 1997a) and both were found to convert very little xanthoxal into ABA. One (aba3) was unable to oxidize AB‐aldehyde to ABA, while the other (aba2) did have aldehyde oxidase activity. Schwartz et al. proposed that aba2 is blocked between XAN and AB‐ald while aba3 cannot convert AB‐ald into ABA.

If one does not accept that AB‐aldehyde is the immediate precursor of ABA then an alternative explanation of the data has to be proposed. This is possible given that xanthoxal is converted readily into ABA equally by turgid and wilted plants (Parry et al., 1988; Sindhu and Walton, 1987; Tan et al., 1997) so the enzymes for the steps between XAN and ABA are constitutive and have adequate capacity for rapid synthesis of ABA. aba3 lacks the ability to oxidize hypoxanthin and heptaldehyde but this ability was restored on treating the extracts with dithionite and Na₂S—a reaction that is known to thiolate the molybdate cofactor of the aldehyde oxidase and the extract was then able to form ABA.

Thus aba3 is a mutant that cannot produce a functional molybdate cofactor required by the aldehyde oxidase. If these results are compatible with those obtained in the orange and avocado fruit cell‐free systems (Lee and Milborrow, 1997b; Cowan and Richardson, 1997), then this was the enzyme blocked by tungstate and which oxidized the aldehyde group of xanthoxal to give xanthoxic acid. The restoration of its activity with dithionite would enable it to produce substrate for subsequent reaction(s) which are constitutive and so this ‘mended’ mutant would be expected to be able to form ABA (as found by Schwartz et al., 1997a).

The aba2 mutant could not convert xanthoxal into ABA but it did have the ability to oxidize AB‐aldehyde to ABA so this establishes that its aldehyde oxidase is functional. The enzyme deficiency of the aba2 mutant, therefore, must lie in its inability to oxidize the 4′‐hydroxyl to a ketone or to isomerize the 1′,2′‐epoxy group to a 1′‐hydroxy‐2‐ene. It would be interesting to have plants carrying both mutations as it would be predicted from the above that aba2 plants should be able to oxidize xanthoxal to xanthoxic acid and then the enzymes of aba3 should be able to form ABA.

The Arabidopsis aba3 mutant appears to have a very similar defect to the flacca tomato which, on wilting, produces large amounts of AB‐alcohol and 2‐trans‐AB‐alcohol which are roughly equivalent to the amount of ‘stress’ AB formed in the wild type (Linforth et al., 1987a, b) but no report was made on whether the aba3 mutant also accumulated AB‐alcohol. The mechanism for the formation of the AB‐ol is unknown but it would be interesting to discover if AB‐ol is produced when plants are wilted in the presence of tungstate and CN‐ so that the aldehyde oxidase enzyme is inhibited. Perhaps the defect in the mutant enzyme causes it to reduce the C‐1 aldehyde to a primary alcohol when unable to oxidize it to a carboxylic acid.

Results similar to those obtained with the aba2 and aba3 mutants, and probably involving the same enzymes, were reported by Sindhu et al. who were able to extract and separate two enzymic activities: an aldehyde oxidase and a 4′‐dehydrogenase‐1′,2′‐ isomerase (Sindhu et al., 1990). These enzymic activities appear to be the two points where the genetic lesions occur in, respectively, the aba3 and aba2 genes of Arabidopsis (Schwartz et al., 1997a) and the same arguments can be applied. These results (Sindhu et al., 1990) show that a 1‐fold acetone precipitate readily converted AB‐ald into ABA but was considerably less (one‐fifth) as active at converting XAN into ABA. A 4‐fold acetone (total) precipitate was able to convert XAN into ABA but the 1‐fold and 3‐fold acetone fractions could not form ABA from XAN unless they were combined and NADP was present. NADP was not required by the extracts which oxidized AB‐ald to ABA (405 ng ABA formed from 1 μg) but no ABA was detected when XAN was supplied. Extracts of the barley mutant Az34 also failed to form ABA when presented with AB‐ald, confirming that it too lacked aldehyde oxidizing capacity.

These results can be interpreted without involving AB‐ald as an intermediate. It is clear that the 1‐fold acetone precipitate fraction contains aldehyde oxidase activity because it converts AB‐ald into ABA (519 ng ABA from 1 μg AB‐ald). The 3‐fold acetone precipitate was unable to produce much ABA from XAN (59 ng) and unable even to make significant quantities of ABA (87 ng) when supplied with AB‐ald. It therefore lacked aldehyde oxidase activity. When the 1‐fold and 3‐fold acetone fractions were recombined the mixture was able to convert XAN into ABA (417 μg). The 3‐fold fraction, therefore, contains 4′‐dehydrogenase‐1′,2′‐epoxide into 1′‐hydroxy‐2‐ene isomerase activity. There is no evidence to show that the enzymes operated in vivo in the order (3‐fold then 1‐fold acetone extracts) to produce first AB‐aldehyde then ABA. The authors did not report looking for xanthoxic acid in the whole extract, 3‐fold fraction, nor in the recombined extract.

Abscisic alcohol

Zeevaart and colleagues (Rock et al., 1991) found that the mutant tomatoes flacca and sitiens are deficient in aldehyde oxidase and produce 2‐trans‐AB‐alcohol labelled with ¹⁸O in the C‐1 hydroxyl group. They also found that traces of ABA with two ¹⁸O atoms in the carboxyl group were produced in the mutant plants and, as the reaction was inhibited by carbon monoxide, they concluded that a P‐450 monooxygenase was capable of adding a second oxygen atom to the C‐1 of AB alcohol, which carries a hydroxyl group at C‐1, thereby forming the carboxyl of ABA. This effectively bypassed the defective aldehyde oxidase and formed traces of ABA containing two ¹⁸O atoms in the carboxyl group. Small amounts of such molecules were found in all species examined (tomato, barley, potato). This reaction also occurs in normal plants as a minor pathway. It is physiologically important in mutants impaired in aldehyde oxidation. It may be significant that so many mutations of this enzyme are known (Arabidopsis, Schwartz et al., 1997a; tomato and tobacco, Leydecker et al., 1995). If small quantities of ABA are essential for growth (as suggested by Herde et al., 1999a, b), then a complete mutant would die but ‘leaky’ mutants, or those for which there is a potential shunt reaction, would survive. Rock and Zeevaart found minute amounts of epoxy carotenoids even in the most severely ABA‐deficient Arabidopsis mutant (aba4) which is unable to form epoxy carotenoids, so even this mutation of the epoxidase enzyme allows traces of viola‐ and neoxanthin to be formed, or another reaction can supply the intermediate (Rock and Zeevaart, 1991). Similar results were found in an Arabidopsis mutant, which failed to epoxidize zeaxanthin to violaxanthin (Marin et al., 1996). It is not known if traces of zeaxanthin can be epoxidized by physico‐chemical reactions in the plant to give a 1′,2′ epoxide (ABA numbering). The synthetic, unnatural C₁₅ acid, β‐ionylidene acetate (the C₁₅ equivalent of β‐carotene) is epoxidized to the racemic 1′,2′‐epoxide by light and O₂ (Milborrow and Noddle, 1970) and the ring is hydroxylated enzymically and stereospecifically at C‐4′ in the plant but only the (S,R)1′,2′ isomer, equivalent to natural xanthoxic acid, is converted into (+)‐ABA (Milborrow and Garmston, 1973). It is possible, therefore, that traces of β‐carotene or zeaxanthin can be oxidized without enzyme activity into 1′,2′ epoxides to give (+)‐ABA sufficient to allow the plant's survival.

Intermediates in the biosynthesis of ABA

Xanthoxal is an intermediate in the biosynthesis of ABA

Xanthoxal was isolated as a breakdown product comprising about 1% of the cleavage products formed from violaxanthin by light and air (Taylor and Burden, 1972). It was found to be converted into ABA by bean leaves and Taylor and Burden were able to oxidize it chemically to (+)‐ABA. They also found it was present in leaves of dwarf bean (Taylor and Burden, 1970). Unfortunately, a similar range of carotenoid photolytic cleavage products were also present in extracts of the leaves. Firn and Friend found that the soybean leaf lipoxygenase cleaved violaxanthin into the same range of products, so the finding of xanthoxal in leaves is not firm evidence for its being a natural precursor of ABA (Firn and Friend, 1972). Li and Walton claimed that xanthoxal was not converted into ABA by chloroplasts but it was by the cytosol (Li and Walton, 1987).

In contrast to this Milborrow and Lee found washed, isolated, intact chloroplasts from spinach leaves converted [¹⁴C]isopentenyl diphosphate into ABA so the whole pathway can occur within the chloroplasts (Milborrow and Lee, 1998a).

The determination of the stereospecificity of some of the reactions of ABA biosynthesis using labelled mevalonate (Robinson and Ryback, 1969; Milborrow, 1972) (Fig. 7) was possible because of the fortuitous accident that a derivative of mevalonate, probably isopentenyl diphosphate (IPP), formed from chirally labelled mevalonate in the cytosol, is able to pass into the chloroplasts and mix with the intrachloroplastic pool and become incorporated into carotenoids and ABA.

Xanthoxal is considered to be an intermediate in the biosynthesis of ABA because it accumulates in tomato mutants flacca and sitiens, and in cell‐free systems when the aldehyde oxidase is blocked by tungstate ions. It is labelled with ¹⁴C when [¹⁴C]mevalonate is converted into [¹⁴C]ABA in a cell‐free preparation (Lee and Milborrow, 1997b) and the addition of dimedone, an aldehyde complexing reagent, to avocado slices reduced the amount of labelled ABA formed and the labelled xanthoxal–dimedone complex accumulated.

Xanthoxal has now been shown to be the cleavage product formed from 9′‐cis‐neoxanthin by the dioxygenase enzyme (VP14) (Schwartz et al., 1997b) which, when present as a transposon‐mutated form in maize, rendered the plants unable to form ABA. Xanthoxal is labelled with ¹⁴C in cell‐free systems converting [¹⁴C]mevalonate into ABA (Lee and Milborrow, 1997a).

Xanthoxal contents were unaffected by wilting the leaves of bean (Phaseolus vulgaris) (Parry et al., 1988; Sindhu and Walton, 1987) so presumably the large amounts of xanthoxal formed by the dioxygenase can be processed adequately by the constitutive enzyme activities which convert xanthoxal into ABA. Consequently, the formation of xanthoxal by the dioxygenase is probably the rate‐limiting step in the production of ‘stress’ ABA. It is a widely observed biochemical phenomenon that enzymes which catalyse the rate‐limiting steps of metabolic pathways are themselves highly regulated allosterically (at least in animals, and are usually multimeric, e.g. glutamate dehydrogenase). So far there is no report of the regulation of VP14 under normal conditions in turgid cells. When cells wilt the amounts of m‐RNA that code for the VP14 increase considerably and the synthesis of ‘stress’ ABA commences. Yet under ambient conditions cells produce ABA and it appears to ‘turn over’, albeit slowly, according to ²H₂O and ¹⁸O₂ labelling experiments.

ABA is made, presumably by a small amount of VP14 under normal, unstressed conditions and its content is regulated by mechanisms that are unknown at present. The amount of ABA present in leaves is maintained at a constant value until a critical water potential deficit of about −9 atmospheres is reached whereupon synthesis of ‘stress’ ABA begins. Wright's results show that after 330 min the amount of ABA in the leaf has reached a plateau value proportional to the severity of the wilt (Fig. 4) (Wright, 1977). When experiments have been carried out with varying intensities of a stress, such as wilting, the contents of ABA appear to reflect the severity of the stimulus.

Effects of inhibitors

The inhibitor of sterol cyclases (AMO 1618) increased the ¹⁴C of mevalonate incorporated into ABA. This is interpreted as the inhibition of the sterol cyclase causes a build‐up of labelled, small, early intermediates in the pathway leading to sterols in the cytoplasm and so diverts labelled precursors into carotenoids Fig. 6).

The lowering by carbon monoxide of the amount of ¹⁴C passing into ABA was not as great as expected in both the cell‐free system and fruit discs. Naproxen is an inhibitor of lipoxygenases and dioxygenases and so would be expected to inhibit the formation of xanthoxal.

The other three inhibitors were compounds that form a complex with aldehydes (dimedone, sodium sulphite) or that block aldehyde oxidase (tungstate) had similar potency in the cell‐free preparation and in intact cells.

Occurrence of AB‐aldehyde

Most of the evidence for AB‐aldehyde's being the immediate biosynthetic precursor of ABA stems from its rapid oxidation to ABA when fed to shoots or cell‐free systems and the failure of flacca and sitiens mutants to effect the conversion. Labelled xanthoxal has been used less frequently, probably because of its difficult synthesis (Parry et al., 1988). Flacca tomato plants were also found to be unable to oxidize xanthoxal to ABA (Parry et al., 1988).

The inability of flacca to oxidize AB‐aldehyde and xanthoxal to ABA suggests that the mutants lack a functional aldehyde oxidase, the data do not establish that AB‐aldehyde is the intermediate. The ABA‐deficient potato mutant ‘droopy’ was unable to convert (±)‐[²H₁]ABA‐ald into ABA whereas wild‐type plants readily oxidized it to [²H₁]ABA (stereochemistry not specified) (Duckham et al., 1989). The mutant plants were found also to have converted the AB‐ald into AB‐alcohol and 2‐trans‐AB‐alcohol. There was a very slight incorporation of the [²H₁]AB‐ald into ABA, but it is not known whether this can be attributed to a leaky mutant enzyme or to the oxidation of C‐1 to a carboxylic acid by a monooxygenase or peroxidase type of enzyme reaction as reported previously (Rock et al., 1991) or even spontaneous oxidation in air. It emphasizes once again that metabolic lesions may be bypassed by other, unnatural reactions coming into play in mutants.

ABA aldehyde

The main interest of the tungstate experiments was that [¹⁴C]xanthoxal increased when the formation of [¹⁴C]ABA was blocked by tungstate. At the nanomolar or lower concentrations of the substrates present in the cell‐free preparations it is extremely unlikely that inhibition by tungstate of the aldehyde oxidase, if its function is to convert AB‐aldehyde into ABA, would cause previous intermediates in the pathway to accumulate by product inhibition as far back as xanthoxal. Consequently, it was proposed that the substrate for the aldehyde oxidase, blocked by tungstate and non‐functional in the ABA‐deficient mutants flacca and sitiens, was xanthoxal. Removal of tungstate with cinchonine allowed the xanthoxal to be oxidized to ABA.

AB‐aldehyde is prone to oxidation in air (Willows and Milborrow, 1992) and because of this its role as an intermediate in the biosynthesis of ABA has been doubted although, of course, it could occur as a Schiff base or be complexed in some way in vivo. The inability of ABA‐deficient tomato mutants flacca and sitiens to oxidize AB‐aldehyde revived its claims to be an intermediate. Willows and Milborrow fed racemic [²H₄]AB‐aldehyde to tomato shoots and found that both (+) and (−) enantiomers were converted into ABA, but in a ratio of 2:1 (Willows and Milborrow, 1992). While this lack of stereospecificity does not establish that the oxidation is artefactual (the chiral centre is five bonds away from the aldehyde group) it was, nevertheless, cause for disquiet. Shortly afterwards Lee and Milborrow found evidence that xanthoxal was the substrate for the Mo‐requiring aldehyde oxidase in ABA biosynthesis and so xanthoxic acid was the next intermediate (Lee and Milborrow, 1997b). Xanthoxic acid was formed, labelled, in cell‐free systems that converted [¹⁴C]mevalonate into [¹⁴C]ABA (Cowan and Richardson, 1997). Labelled AB‐ald was not detected in the cell‐free system (nor by Sindhu and Walton, 1987). If xanthoxal is the substrate of the aldehyde oxidase then xanthoxic acid must be the next intermediate and AB‐aldehyde could not be a major precursor of ABA. Re‐examination of the flacca, sitiens data showed that they lacked ‘aldehyde oxidase activity’ but the enzymic activity was assayed with heptaldehyde (Leydecker et al., 1995; Akaba et al., 1998). The enzyme must have quite broad structural requirements to accept an aliphatic as well as an α‐β unsaturated aldehyde. However, this non‐specificity has not been considered and the AB‐aldehyde has continued to be assumed to be the immediate precursor of ABA. Cowan has recently questioned the status of AB‐ald as a precursor of ABA (Cowan, 2000).

Although the specific enrichment of ¹⁸O of the AB‐aldehyde was found to be greater than that of the 2‐trans‐isomer this can not be attributed to a stronger cleavage reaction of 9′‐cis‐neoxanthin compared with cleavage of 9′‐trans‐neoxanthin because it has been demonstrated that the VP14 dioxygense enzyme could cleave 9‐cis‐violaxanthin and 9′‐cis‐neoxanthin, but not their all‐trans isomers (Schwartz et al., 1997b). The mechanism responsible could be the slower conversion of 2‐trans‐AB‐aldehyde and 2‐trans‐xanthoxal into 2‐trans‐ABA compared with the rate of conversion of 2‐cis‐aldehyde and xanthoxal into ABA; this would give time for more ¹⁸O to exchange into the medium from C‐1 of the 2‐trans‐isomer. All this assumes, of course, that the rate of exchange of the oxygen atom of the aldehyde is the same in the cis and trans isomers.

Rock and Zeevaart detected very small amounts of AB‐aldehyde in apple fruit, but it comprised about a 5:1 mixture of the 2‐trans to 2‐cis isomer (Rock and Zeevaart, 1990). Taylor and Burden had shown that 2‐trans‐xanthoxal was converted into 2‐trans‐ABA while the 2‐cis isomer gave ABA (Taylor and Burden, 1973). Thus, once the 2‐trans bond of xanthoxal has been isomerized to trans, it can not be re‐isomerized. At present (1999) the only known mechanism to form the cis double bond of ABA operates in C₄₀ carotenoids and it is not certain if this is spontaneous or is catalysed enzymically.

AB‐ald could not be found in bean leaves (Li and Walton, 1990) and no [¹⁴C]AB‐ald was found in cell‐free systems biosynthesizing ABA from [¹⁴C]mevalonate (Cowan and Richardson, 1997) but xanthoxic acid was detected. Although it was stated that xanthoxic acid is not converted into ABA (Sindhu and Walton, 1988), these authors’ data do show a 10% conversion (i.e. 10% of the rate at which xanthoxal was converted into ABA by tomato leaves. More recently, the formation of [¹⁴C]ABA from [¹⁴C]xanthoxic acid in avocado fruit has been reported (Milborrow et al., 1997b).

The 1′,4′‐trans‐diol is not an intermediate in the biosynthesis of ABA

The trans‐diol is a precursor of ABA in fungi, but the data on its being a precursor in leaves is ambiguous. Cowan and Richardson found that it was formed from [¹⁴C]mevalonate in tissue homogenates (Cowan and Richardson, 1993) while in other experiments (Sindhu and Walton, 1988; Richardson and Cowan, 1996) the 1′,4′‐trans‐diol was converted into ABA. A majority of the trans‐diol formed from (±)‐ABA was the unnatural (−) epimer (Vaughan and Milborrow, 1987). A ‘cold pool’ of trans‐diol added to a cell‐free system from orange peel which synthesized ABA vigorously from [¹⁴C]mevalonate not only trapped ¹⁴C in the diol (Cowan and Richardson, 1997) but the amount of radioactivity present in ABA was less. The formation of the diol from ABA by a reduction reaction and the ready oxidation of the diol to ABA suggested that a secondary alcohol dehydrogenase was operating close to its equilibrium.

A definitive difference between the trans‐diol's being a precursor and a metabolite formed from ABA lies in the C‐4′ hydrogen atom. The hydrogen atoms at C‐4′ of ABA and the diol would be derived from those at C‐5 of [5‐³H₂]mevalonate (Fig. 7). Both ³H atoms at C‐4′ would be lost from ABA when the ketone was formed. On the other hand, one tritium atom from C‐5 of mevalonate would be retained at C‐4′‐ in the trans‐diol if it were a precursor of ABA. If [¹⁴C; 5‐³H₂]mevalonate were incorporated into ABA both the ABA and the trans‐diol would carry a tritium atom from C‐5 of mevalonate at C‐5 of their side chain (Milborrow, 1972). Thus, if the diol were the precursor of ABA, its ³H:¹⁴C ratio would be 2:3. If the trans‐diol were formed by reduction of ABA, then both ABA and the diol would be expected to retain just one ³H atom at C‐5 of their side chains and, consequently, the diol would have a ³H:¹⁴C ratio of 1:3. The ABA isolated from the avocado had a ratio of 0.915:3, close to the 1:3 expected, while that of the trans‐diol was 0.905:3. Oxidation of the C‐4′ hydroxyl of the diol gave ABA which had the same ³H:¹⁴C ratio and so established that there was no ³H at C‐4′ (Milborrow and Lee, 1998b). The trans‐diol, therefore, is not a precursor of ABA and was formed from ABA. This agrees with the results of Rock and Zeevaart (Rock and Zeevaart, 1990).

The ready reversibility of the 4′‐oxidoreductase was established by adding NAD NADH, NADP or NADPH to a mung bean tissue homogenate with [¹⁴C]ABA. The reduced coenzymes increased the amount of [¹⁴C]ABA converted into the diol compared to the oxidized forms. Reduced glutathione had the same effect, this was attributed to the widespread occurrence of NADP‐glutathione reductase (Grant et al., 1996) able to form NADPHH⁺. Dithiothreitol also increased the amount of diol formed but the mechanism for the transfer of reducing potential is not known in this case. It should be noted that DTT (10 mM) was present in cell‐free systems (Cowan and Richardson, 1997; Sindhu and Walton, 1988) although the pH of 7.4 would tend to favour oxidation of the substrate by neutralization of the H⁺ removed as NADPHH⁺. This result highlights the care necessary for the interpretation of biosynthetic results obtained with cell‐free systems.

Radiochemical purity of compounds formed biosynthetically

The demonstration that a labelled precursor has been converted into ABA depends on establishing that the isotopic label is present in the ABA molecule and in no other in a purified sample. This has often been achieved with other compounds by a sequence of recrystallizations to constant specific activity. This method is less appropriate for ABA because of the difficulty of extracting sufficient natural (+)‐ ABA for crystallization and, until recently, the inability to resolve the synthetic, racemic material which cannot be used as it has a different crystallization behaviour. The preferred method for preparing radiochemically pure ABA is to make use of several unusual chemical features to convert the compound into derivatives which have different chromatographic characteristics. These are, usually, methylation of the free acid with diazomethane, reduction of the 4′‐ketone group to give equal quantities of the isomeric 1′,4′‐diols, oxidation of the diols, separately, with MnO₂ back to ABA (a reaction that requires an α,β‐unsaturated hydroxyl group). Alternatively, the MeABA‐containing fraction can be acetylated with acetic anhydride which decreases the polarity of impurities containing hydroxyl groups, but does not affect ABA as its tertiary 1′‐hydroxyl group cannot be acetylated. Any labelled material which still co‐chromatographs with ABA after these treatments can be safely assumed to be ABA.

Stereochemistry of biosynthetic reactions

The stereochemistry of several of the reactions of biosynthesis of ABA has been probed by the use of stereospecifically tritiated mevalonic acids (Milborrow, 1972) which have been incorporated into ABA by ripening avocado fruit. Although the biosynthesis of ABA occurs within chloroplasts by the pyruvate‐glyceraldehyde pathway (Milborrow and Lee, 1998a) [¹⁴C]isopentenyl diphosphate has been shown to be converted into ABA in intact, washed spinach chloroplasts so presumably IPP or a closely related metabolite made in the cytoplasm can enter the plastids of avocado fruit after it has been formed from mevalonate.

The early experiments of Robinson and Ryback showed that the C‐2 double bond of ABA was formed in the trans configuration because the C‐4‐pro‐R‐³H atom of mevalonate was retained at C‐2 of the cis double bond of ABA (Robinson and Ryback, 1969). The homologous atom is retained during the biosynthesis of carotenoids while the cis‐double bond in rubber is formed with the retention of the C‐4‐pro‐S‐³H atom of mevalonate. It follows from this that the C‐2 double bond is formed trans and is subsequently isomerized to cis, with retention of the ³H atom.

The C‐4 double bond of ABA is formed with the loss of the C‐2‐pro‐³H atom of mevalonate from C‐4 of ABA and the loss of the C‐5‐pro‐S‐³H atom of mevalonate from what becomes the C‐5 of ABA (Milborrow, 1972) (Fig. 8A).

Cyclization occurs with the formation of a C‐1′, C‐2′ β double bond (and a concomitant loss of a C‐4‐pro‐R‐³H atom of mevalonate). The ε double bond (C‐2′) is formed by the isomerization of the 1′,2′‐epoxide to a 1′‐hydroxyl group and the loss of the C‐2‐pro‐S‐³H atom of mevalonate from C‐3′ of ABA (Fig. 8B).

The arrangement of the precursor during cyclization places the methyl group derived from the C‐2 of mevalonate in the C‐8′ position of ABA and the methyl group from the C‐6 Me of mevalonate as C‐9′ (Milborrow, 1984). Protonation of the future C‐5′ of ABA gives the H derived from the medium at the C‐5′‐pro‐S‐position and the ³H from C‐4‐pro‐R of mevalonate as the C‐5′‐pro‐R H atom (Willows and Milborrow, 1989).

Free ABA in leaf cells

The finding that exogenous [²H₄]ABA passed through mung bean hypocotyl segments without mixing with the endogenous [¹H]ABA (Warganegara and Milborrow, 1997; BV Milborrow, unpublished results) led to the discovery that there was very little free ABA in leaf cells (Vittone and Milborrow, 1999; BV Milborrow, unpublished results). Most of the ABA is held by a proteinaceous material, but is extracted by solvents such as ethanol. The bound form is referred to as ‘fettered’ ABA as the form of binding is unknown. At first sight it might appear that this could account for the apparently lower specific enrichment with deuterium of the ABA in wilted tomato (noted by Nonhebel and Milborrow, 1987). The lesser enrichment in the ‘stress’ ABA was attributed to the mobilization of a larger, pre‐formed pool of carotenoids. Although wilting would release preformed, and therefore unlabelled ‘fettered’ ABA this could not account for the lower specific enrichment because the solvent extraction procedure used would have extracted all the ‘fettered’‐ABA fraction in both instances.

Reactions forming ABA

Dioxygenase

Transposon mutations that gave a maize plant unable to produce ABA, enabled McCarty and colleagues to isolate and clone the dioxygenase enzyme responsible for the release of the C₁₅ residue that is elaborated into ABA (Schwartz et al., 1997b). The enzyme in vitro cleaves 9′‐cis‐neoxanthin and 9‐cis‐violaxanthin so it has dioxygenase activity, but the failure of the enzyme to cleave the all‐trans isomers establishes that it lacks 9‐trans‐ to cis‐isomerase activity. At the time of writing it is not known if the 9‐bond of neoxanthin is isomerized enzymically or spontaneously. Chemical cleavage of epoxy‐carotenoids produces 2‐trans‐xanthoxal and xanthoxal in a ratio of about 2:1. The trans isomer is converted into 2‐trans‐ABA, the 2‐trans bond is not isomerized to 2‐cis after the C₁₅ residue has been separated.

Parry et al. found that the amounts of carotenoids in tomato leaves exceeded the amounts of ABA by between 20‐ and 100‐fold so any reciprocal changes caused by synthesis of stress‐induced ABA would be difficult to detect (Parry et al., 1990). The exception was 9′‐cis‐violaxanthin which occurred at 0.62 nmol g⁻¹ FW, but was even higher in stressed leaves whose ABA content had risen from 1.34 to 4.23 nmol g⁻¹. A similar result was obtained by Li and Walton who found that the amount of 9‐cis‐violaxanthin in bean leaves treated with fluoridone was not noticeably affected by wilting (0.8 nmol decrease) when the content of ABA rose from 0.9 to 6.8 nmol g⁻¹ over 150 min (Li and Walton, 1990). These results indicate that 9‐cis‐violaxanthin is not a close precursor of ABA. The content of 9′‐cis‐neoxanthin decreased by 3.3 nmol g⁻¹. It is interesting to note that the content of all‐trans‐violaxanthin decreased by 12.9 nmol. The dioxygenase enzyme VP14 failed to cleave all‐trans‐violaxanthin (Schwartz et al., 1997b) so this suggests that violaxanthin is converted rapidly into 9′‐cis‐neoxanthin to provide for the synthesis of ‘stress’ ABA and comprises a major store of precursor.

Li and Walton found that excised bean leaves supplied with water increased their content of ABA to 1070 ng g⁻¹ FW when wilted (Li and Walton, 1987). Leaves supplied with 30 μm fluridone to block synthesis of carotenoids, and then wilted, increased their ‘stress’ ABA to 911 ng g⁻¹ on the first wilt. After recovery of turgor and the imposition of a second wilt the ABA increased to 424 ng g⁻¹ while 190 ng g⁻¹ and 238 ng g⁻¹ were present after the third and fourth wilts, respectively. These data showing progressively smaller amounts of ‘stress’ ABA formed over successive wilts suggest that there is a small, dedicated pool of carotenoids which is available to form ‘stress’ ABA. Bonora et al. have recently detected a possible biosynthetic pathway in ripening fruits of Arum italicum where carotenoids with a 9‐cis‐double bond are elaborated (Bonora et al., 2000). There is, as yet, no evidence to show that it occurs in leaves or in colourless tissues able to make ABA, nor how the trans double bond is isomerized to cis. In view of Li and Walton's results referred to above and those of Parry et al. who also reported that the amounts of 9‐cis‐ violaxanthin were unaffected by wilting, it seems unlikely that these compounds are important biosynthetic precursors (Li and Walton, 1990; Parry et al., 1990).

It has been shown that pea and bean plants treated with cycloheximide, an inhibitor of cytosolic protein synthesis, did not produce ‘stress ABA’ when wilted (Guerrero and Mullet, 1986; Li and Walton, 1990). So some of the enzymes responsible were synthesized in the cytosol. It is now known that ABA and carotenoids are produced in chloroplasts so some enzymes, at least, are made in the cytosol in response to unfavourable conditions and enter the chloroplasts to produce ‘stress ABA’.

Quarrie and Lister had shown that chloromycin and streptomycin inhibited ABA biosynthesis, these compounds block chloroplastic peptide synthesis so some steps of ABA biosynthesis probably occurred within chloroplasts (Quarrie and Lister, 1984a, b). However, the antibiotics so damaged the chloroplasts that the conclusion could not be considered well established (Oelmueller et al., 1986).

Isomerization

The C‐2 double bond of natural ABA is in the cis‐Z‐configuration, but in solution in the light this bond isomerizes to a 1:1 mixture with the 2‐trans‐(E) isomer which is biologically inactive. The 2‐trans‐ABA is also isomerized by light to the 2‐cis. ABA is biosynthesized in the 2‐cis form (which is the only isomer detected in roots), but a few per cent of the ABA in leaves is found in the 2‐t rans form, possibly by sunlight‐activated isomerization in vivo (Milborrow, 1970).

The 2‐cis double bond of xanthoxal is readily isomerized to 2‐trans but there appears to be little or no reversal to give 2‐cis (Parry et al., 1990). 2‐trans‐xanthoxal is converted into 2‐trans‐xanthoxic acid which, in turn, is converted into 2‐trans‐ABA. Thus is appears that only the 2‐cis‐ isomer of xanthoxal is a precursor of ABA.

The C‐2‐double bond of ABA appears to be far enough away from the ring to have little or no effect on the biosynthetic reactions which take place on the ring in the pathway between xanthoxal (or 2‐trans‐xanthoxal) and ABA. This has also been found to apply to the rapid hydroxylation of 2‐trans‐ABA to give 2‐trans phaseic acid (BV Milborrow, unpublished data) and the 4′‐hydroxylation, oxidation and isomerization of 1′,2′‐epoxyionylideneacetic acid (a synthetic compound) to give xanthoxic acid and then ABA (Milborrow and Garmston, 1973).

Xanthoxic acid is converted into ABA

An experiment in which tungstate was added to a cell‐free system from avocado fruit which converted [¹⁴C]mevalonate into ABA was repeated (Lee and Milborrow, 1997b) and the [¹⁴C] present in xanthoxal was determined by adding a ‘cold scavenger’ (unlabelled material added to an extract to assist purification). Tungstate increased the ¹⁴C in xanthoxal, but when the alkaloid cinchonine was added (to complex with and precipitate the tungstate) the ¹⁴C did not accumulate in xanthoxal.

From this, it was concluded that xanthoxal was oxidized to xanthoxic acid by a molybdenum‐requiring aldehyde oxidase. The mutant flacca tomatoes lacked aldehyde oxidase (Parry et al., 1988), but were found to be able to convert very small amounts of [¹³C]xanthoxal into ABA. This possibly occurred by the bypass reaction described earlier. Xanthoxic acid has been shown to be readily converted into ABA (Milborrow et al., 1997b) so the remaining unknown reactions of the pathway from carotenoids to ABA are now: oxidation of the 4′‐hydroxyl group and isomerization of the 1′,2′‐epoxide to a 1′‐hydroxyl group and a 2′‐ene. The chemical synthesis of ABA from xanthoxal (Taylor and Burden, 1972) used an oxidation of the 4′‐hydroxyl group to give the 4′‐keto group, but this compound was unstable and the 1′,2′‐epoxy group isomerized spontaneously to a 1′‐hydroxy‐2′‐ene, thereby forming AB‐aldehyde. It is not known if 4′‐keto‐xanthoxic acid is equally unstable or if two enzymic steps are required.

The presence of a higher ratio of the 2‐trans isomer to xanthoxal in mutant tissues, compared to wild‐type tissues, may be attributable to the longer time the xanthoxal was in the cells, thereby allowing more spontaneous isomerization to the trans‐ isomer.

There are two obstacles to interpreting the data on the last few intermediates prior to ABA. (1) Previous experiments have fed xanthoxal and assessed the conversion into ABA, which requires two or three enzymic steps whose order and intermediates have not been characterized. Either xanthoxal's or AB‐aldehyde's ‐CHO group could have remained unoxidized while the ring was elaborated, or the reverse. (2) It appears that at least some of the ABA residue is bound in some way (‘ABA adduct’; Netting et al., 1992) and the last few intermediates may not be free in the cell.

ABA adduct

AB‐aldehyde has been found to autoxidize spontaneously to ABA (Willows and Milborrow, 1992) so its participation in biosynthesis as the immediate precursor into ABA was considered unlikely. It is possible, however, that the last few steps of biosynthesis of ABA are carried out with the future ABA residue held by a jig or template molecule while it undergoes the final chemical changes. The candidate for this is a compound referred to as ABA‐adduct (Netting et al., 1992). It is highly unstable and has eluded characterization as it breaks down spontaneously in strong salt solutions and at pH below 5 and above 8. It comprises two components, both polar, one neutral, the other weakly basic. When adduct is treated with mineral acid some of the ABA is released as methyl ester, but the yield of MeABA is a small fraction (10–20%) compared to the total amount of ABA released by saponification. Adduct was not labelled when μCi amounts of ±‐[¹⁴C]ABA were supplied to onion shoots. However, when tomato seedlings were grown in ²H₂O (55 atom%), 50% of the ABA extracted from them carried from 2–11 ²H atoms with a majority having 5 atoms. In contrast to this the ²H content of Me‐ABA released from the adduct isolated from these plants contained a higher enrichment of ²H: 71% carried from 2–11 ²H atoms (a majority with 5). In a second experiment 57% of the ABA carried from 2–11 ²H atoms with a majority of 7 (Netting et al., 1997).

The higher specific enrichment of ²H in the Me‐ABA from the adduct compared with the enrichment of the free ABA indicates that the adduct is a precursor of ABA, but which intermediate first binds to the adduct moiety is not known. (±)‐[²H₃]AB‐aldehyde was not assimilated into the adduct fraction in tomato shoots (Willows and Milborrow, 1992). An alternative explanation of the greater enrichment of the adduct with newly biosynthesized [²H]ABA is that the newly formed [²H]ABA in the chloroplasts is at a high concentration and able to exchange into the adduct. Newly formed [²H]ABA passing into the cytosol would be diluted into previously formed unlabelled, free ABA and so it would show a lower specific enrichment. It was reported that [²H₆]ABA passed through mung bean hypocotyl segments without mixing with the endogenous, unlabelled ABA (Warganegara and Milborrow, 1997). The site in which this ABA is held in the cells is not known. When the mungbean segments were analysed it was found that some (+)‐[²H₆]ABA was present in the adduct, but whether metabolic assimilation into the adduct occurred over the 16 h exposure or physical exchange into the molecule took place in the explants is not known.

The lack of labelling of the adduct by [¹⁴C]ABA in onion shoots over several days growth may have occurred by the labelled ABA's being metabolized and the newly formed, unlabelled ABA then exchanged into the adduct to replace the labelled material. Turnover rates are unknown. The one probable intermediate between xanthoxal and ABA that has been synthesized is 4′‐ketoxanthoxal. This compound was made by Taylor and Burden and it was found to isomerize spontaneously to give the ring of ABA (Taylor and Burden, 1972). 4′‐keto‐xanthoxic acid has not been synthesized and may be more stable. These reactions, therefore, are difficult to investigate and it is not known if the intermediates occur free or are held as adducts.

Summary

The most probable pathway of biosynthesis of ABA in plants appears to be confined to the chloroplasts where pyruvate and glyceraldehyde phosphate are combined and rearranged, via deoxyxylulosephosphate, to give isopentenyl diphosphate (IPP). In some tissues [¹⁴C]mevalonate is incorporated into IPP in the cytosol. Some of the IPP can then enter the chloroplasts where it is elaborated into carotenoids and can give rise to ABA. Eight IPP residues are combined and converted into β‐carotene, then oxidized to violaxanthin and rearranged to 9′‐cis‐neoxanthin (Fig. 9). The large increase in the amount of ABA formed when leaves wilt (‘stress’ ABA) is produced by the de novo synthesis of the carotenoid cleavage enzyme (dioxygenase VP14) in the cytosol and which enters the chloroplasts to catalyse what appears to be the rate‐limiting step in biosynthesis: the formation of xanthoxal, the 15 carbon skeleton of the future molecule of ABA. This is oxidized to xanthoxic acid and undergoes further oxidation and rearrangement to ABA. It is possible that some of these last reactions occur while the future ABA residue is loosely held as an adduct, a larger molecule of unknown structure. A minor bypass reaction occurs in several mutant plants which lack an effective aldehyde oxidase: AB‐alcohol is formed and this is slowly converted into ABA. This reaction also occurs to a small extent in wild‐type plants.

Fig. 9.

The proposed pathway of biosynthesis of ABA in chloroplasts. The ‘early reactions’ involving the production of small isoprenoid phosphate building blocks (see Fig. 4) are considered to end with the formation of the C₄₀ uncyclized carotenoid phytoene. Progressive desaturation and two cyclizations give β,β‐carotene. Then two hydroxylations (3, 3′) give zeaxanthin (shown at top) and two epoxidations (site of the aba 1 mutation in Arabidopsis) give violaxanthin. The ‘late intermediates’ are formed by cleavage of the C₁₅ carbon skeleton of ABA as xanthoxal from 9′‐cis‐neoxanthin. Some of these compounds and the newly formed ABA may be held by a larger molecule referred to as ‘adduct’.

1

Fax: [phone]. E‐mail: [email protected]