Molecular characterization of the Arabidopsis 9-cis epoxycarotenoid dioxygenase gene family


For correspondence (fax +1 352 392 6479; e-mail


A key regulated step in abscisic acid (ABA) biosynthesis in plants is catalyzed by 9-cis epoxycarotenoid dioxygenase (NCED), which cleaves 9-cis xanthophylls to xanthoxin, a precursor of ABA. In Arabidopsis, ABA biosynthesis is controlled by a small family of NCED genes. Nine carotenoid cleavage dioxygenase (CCD) genes have been identified in the complete genome sequence. Of these, five AtNCEDs (2, 3, 5, 6, and 9) have been cloned and studied for expression and subcellular localization. Although all five AtNCEDs are targeted to plastids, they differ in binding activity of the thylakoid membrane. AtNCED2, AtNCED3, and AtNCED6 are found in both stroma and thylakoid membrane-bound compartments. AtNCED5 is exclusively bound to thylakoids, whereas AtNCED9 remains soluble in stroma. A quantitative real-time PCR analysis and histochemical staining of promoter::GUS activity in transgenic Arabidopsis revealed a complex pattern of localized NCED expression in well-watered plants during development. AtNCED2 and AtNCED3 account for the NCED activity in roots, with localized expression in root tips, pericycle, and cortex cells at the base of lateral roots. Localized AtNCED2 and AtNCED3 expression in pericycle cells is an early marker of lateral initiation sites. AtNCED5, AtNCED6, AtNCED3, and AtNCED2 are expressed in flowers with very high AtNCED6::GUS activity occurring in pollen. AtNCED5::GUS, and to lesser degrees, AtNCED2::GUS and AtNCED3::GUS are expressed in developing anthers. AtNCED5, AtNCED6, AtNCED9, and AtNCED3 contribute to expression in developing seeds with high levels of AtNCED6 present at an early stage. GUS analysis indicates that AtNCED3 expression is confined to the base of the seed, whereas AtNCED5 and AtNCED6 are expressed throughout the seed. Consistent with the studies conducted by Iuchi and his colleagues in 2001, AtNCED3 is the major stress-induced NCED in leaves. Our results indicate that developmental control of ABA synthesis involves localized patterns of AtNCED gene expression. In addition, differential membrane-binding capacity of AtNCEDs is a potential means of post-translational regulation of NCED activity.


Abscisic acid (ABA) is a key regulator of seed development, root growth, stomatal aperture, and plant responses to water stress (Zeevaart and Creelman, 1988). ABA is an apo-carotenoid compound derived from oxidative cleavage of the 11,12 double bond of 9-cis epoxy carotenoids (neoxanthin and/or violaxanthin) (Zeevaart, 1999). The 9-cis-epoxycarotenoid dioxygenase (NCED) enzyme was identified by analysis of the maize viviparous14 (vp14) mutant (Schwartz et al., 1997; Tan et al., 1997). NCED genes have, subsequently, been identified in several species including tomato (Burbidge et al., 1999), bean (Qin and Zeevaart, 1999), avocado (Chernys and Zeevaart, 2000), cowpea (Iuchi et al., 2000), and Arabidopsis (Iuchi et al., 2001).

Apo-carotenoid compounds derived from carotenoid cleavage are widely distributed in nature. The NCED genes of plants identify a large family of double-bond-cleaving dioxygenases found in genomes of plants, animals, and bacteria (refer to Figure 1). Carotenoid dioxygenases are distinguished by the specificity of double bond cleavage. The carotenoid dioxygenases of animals include 15,15′β-carotene dioxygenase responsible for vitamin A synthesis (von Lintig and Vogt, 2000). Enzymes catalyzing specific cleavage of the 9,10 and 9′,10′ double bonds have been identified in plants and animals. The 9,10(9′,10′) cleavage products include C13 compounds related to β-ionone and compounds derived from the C14 central fragment of the carotenoid backbone (Bouvier et al., 2003; Schwartz et al., 2001). Compounds in the latter group accumulate in mycorrhiza-infected roots of maize (Walter et al., 2000). A 7,8(7′,8′) cleavage enzyme has been described recently (Bouvier et al., 2003). The Arabidopsis genome contains at least one 9,10(9′,10′) cleavage enzyme, AtCCD1, and several additional carotenoid cleavage dioxygenase (CCD)-like genes of unknown function. We generically designate genes in this family as AtCCDs. Genes closely related to AtCCD1 are found in leaf-derived ESTs from a variety of species, indicating that mRNA for this enzyme is expressed at significant levels in normal leaves of most species.

Figure 1.

clustalw tree analysis of VP14-related dioxygenase proteins in different organisms.

The nine carotenoid dioxygenases of Arabidopsis are highlighted in bold, and the NCED branch is highlighted in thick lines. Other dioxygenases include: VP14 of Zea mays (Tan et al., 1997); PaCCD2, PaNCED1, and PaNCED3 of Persea americana (Chernys and Zeevaart, 2000); PvNCED1 of Phaseolus vulgaris (Qin and Zeevaart, 1999); LeNCED1 of Lycopersicon esculentum (Burbidge et al., 1999); OsNCED1 (AC120531), OsNCED2 (AL731885), OsNCED3 (AP005632), OsCCD4a (AP005825), OsCCD4b (AL731749), OsCCD1 (AL732378), OsCCD7 (AL663000), and OsCCD8 (AP003376) of Oryza sativa; SYN1 and SYN2, two sequences in the complete Synechocystis cyanobacterium genome (Kaneko et al., 1996); HsRPE65 of (AF039855) and HsVitA of Homo sapiens, also named as β,β-carotene-15,15′-dioxygenase (BCDO) (Yan et al., 2001); DmVitA of Drosophila melanogaster (von Lintig and Vogt, 2000); LSD of Pseudomonas paucimobilis (Kamoda and Saburi, 1993); and MYC1 and MYC2 open-reading frames of mycobacteria (NC_002974).

Much attention has been focused on the regulation of NCED genes in response to stress. A variety of studies indicate that the carotenoid cleavage reaction is a key regulated step in the pathway controlling stress-induced ABA synthesis (Qin and Zeevaart, 1999). Notably, induction of maize Vp14 (Tan et al., 1997) and NCED genes of other species by water stress in leaves is well correlated with stress-induced ABA synthesis. Mutations in NCED genes reduce stress-induced ABA synthesis in leaves (Iuchi et al., 2001). Overexpression of AtNCED3 in transgenic Arabidopsis and PvNCED1 in tobacco plants results in increased ABA accumulation and resistance to water stress (Iuchi et al., 2001; Qin and Zeevaart, 2002).

Developmental control of ABA synthesis in seeds, roots, and other plant organs where ABA has essential developmental functions is less well understood. While AtNCED3 evidently plays a major role in stress-induced ABA synthesis in leaves, the Arabidopsis genome includes at least four other NCED genes that have a lesser or no demonstrated role in stress-induced synthesis (Iuchi et al., 2001).

In this paper, we evaluate the developmental and stress regulation of the AtNCED genes using sensitive gene-specific RT-PCR probes and β-glucuronidase (GUS) reporter gene::transgene constructs. In addition, we show evidence that NCEDs differ from other AtCCD proteins in subcellular localization, suggesting alternative modes of post-transcriptional regulation of apo-carotenoid pathways. Moreover, the plastid-targeted NCEDs show differences in their association with chloroplast membranes. Our results indicate a potential for multiple layers of regulation that affect developmental control of ABA synthesis.


Molecular cloning and analysis of AtCCDs in Arabidopsis

The clustalw tree shown in Figure 1 summarizes the family of carotenoid cleavage dioxygenase genes (AtCCDs) identified in the Arabidopsis genome with closely related dioxygenases in human beings, Drosophila, yeast, cyanobacteria, mycobacteria, and other plant species. Five AtNCEDs that are closely related to VP14 of maize (AtNCED2, AtNCED3, AtNCED5, AtNCED6, and AtNCED9) have been directly implicated in ABA synthesis, and four of them are reported to have 11,12(11′,12′) 9-cis epoxycarotenoid cleavage activity (Iuchi et al., 2001). Although enzymatic activity has not been reported, AtNCED5 shares high homology to other NCEDs, suggesting that it may encode very likely an 11,12(11′,12′)-NCED. By comparison, the AtCCD1 enzyme has been shown to specifically cleave the 9,10 and 9′,10′ double bonds of a variety of carotenoid substrates (Schwartz et al., 2001). The primary products of these reactions are various β-ionone-related C13 fragments and the C14 central fragment. A related CCD of avocado, PaNCED2 as initially named (Chernys and Zeevaart, 2000), is closely related to AtCCD1. Thus, most likely, it encodes a 9,10(9′,10′) cleavage dioxygenase. The biochemical functions of the AtCCD4, AtCCD7, and AtCCD8 proteins are unknown. Interestingly, in the tree, AtCCD7 and AtCCD8 proteins align more closely to mammalian 15,15′-carotenoid dioxygenases (HsRPE65 and HsVitA) than to the 11,12 double-bond-cleaving NCEDs of plants. All the plant NCEDs are clustered together in one branch. AtCCD4, positioned between AtCCD1 and the plant NCED cluster in this tree, is apparently more closely related to AtNCEDs than the CCDs. However, the C-terminal amino acid sequence of AtCCD4 showed high homology (56% identity, 72% similarity) to CsZCD of Crocus, a 7,8(7′,8′) zeaxanthin cleavage dioxygenase (Bouvier et al., 2003). Intriguingly, the predicted AtCCD4 (595 aa) is a much larger protein than CsZCD (369 aa).

Analysis of the recently completed rice genome draft sequence assemblies revealed that the CCD gene family of rice differs from that of Arabidopsis. blast analysis of public Japonica and Indica databases revealed three OsNCED genes for each subspecies. OsNCED1 (AC120531), OsNCED2 (AL731885), and OsNCED3 (AP005632) have intronless gene structures and 86, 78, and 66% amino acid identity to VP14, respectively. In the phylogenetic tree, OsNCED1 is most closely related to VP14, suggesting that it may be the ortholog of VP14. While the rice genome sequence assembly is not completely finished, the fact that we obtain the same three genes from independent sampling of the Japonica and Indica genomes indicates that rice has fewer NCED genes than Arabidopsis. Moreover, the gene phylogeny suggests that significant diversification of the NCED genes has occurred since the divergence of rice and Arabidopsis. Hence, there are not clear orthology relationships between rice and Arabidopsis NCED genes. In contrast, putative orthologs for each of the non-NCED AtCCD genes are found in rice. Close homologs of AtCCD1, AtCCD7, and AtCCD8 are found as single-copy genes (OsCCD1, OsCCD7, and OsCCD8, respectively), whereas there are two AtCCD4-like genes in rice (OsCCD4a and OsCCD4b). These results indicate that genomes of diverse angiosperms carry very similar complements of carotenoid dioxygenases, suggesting that the corresponding apo-carotenoid synthetic pathways are broadly conserved.

All nine Arabidopsis CCDs contain four highly conserved histidine residues that may be involved in coordinating a non-heme iron required for activity (Schwartz et al., 1997; Tan et al., 1997). In addition to sequence relationships, members of the CCD gene family are distinguished by their gene structures and subcellular localization signals (Figure 2). The five members of the AtNCED subfamily and the AtCCD4 gene lack introns, whereas AtCCD1, AtCCD7, and AtCCD8 genes have a more typical eukaryotic gene structure including multiple introns. While the regulatory significance of the intronless structure, if any, is unknown, this property is conserved in NCED genes of other species including maize Vp14 (Tan et al., 1997), Notabilis of tomato (Burbidge et al., 1999), and bean PvNCED1 (Qin and Zeevaart, 1999). The avocado PaNCED1 and PaNCED3 have not been characterized with regard to their gene structure (Chernys and Zeevaart, 2000). There is some evidence that the efficiency of intron splicing may be degraded under severe stress in plants (Bournay et al., 1996; Marrs and Walbot, 1997). In that case, the absence of introns may be a mechanism for enhancing NCED expression and ABA synthesis under stress.

Figure 2.

Schematic structure of the carotenoid dioxygenase family in Arabidopsis.

From N- to C-terminal, the dotted box represents a putative plastid-targeting transit peptide; bar box, a putative amphipathic α-helix domain; and open box, the catalytic domain containing four highly conserved histidine residues (H) that are required for iron sequestration (Schwartz et al., 1997).

Differential localization and targeting of AtCCD proteins

Based on transit peptide prediction analysis, all five AtNCEDs carry a putative transit peptide. AtCCD4 shares further affinity with the NCED class proteins in having a putative transit peptide. Maize VP14 and NCEDs isolated from other species are known to be plastid localized (Qin and Zeevaart, 1999; Tan et al., 2001). Chloroplast import assays shown in Figure 3(a) confirm plastid localization of all five AtNCED proteins. In contrast, in vitro translated AtCCD1, which lacks an evident transit peptide sequence, was not imported into pea chloroplasts, suggesting that plastid localization is not a universal property of carotenoid dioxygenases. Cytosol-localized CCDs, conceivably, may access carotenoids contained in the outer envelope membrane of plastids (Parry and Horgan, 1991). In import reactions, small amounts of AtCCD1 protein were found associated with washed chloroplasts in a protease-accessible compartment, indicating localization on the exterior surface of the plastids. The implication that at least some CCDs are cytosolic suggests that co-localization of CCDs with substrates may be a major control point for the reactions catalyzed by these enzymes. For cytosolic CCDs, substrate availability would, presumably, be in part determined by the carotenoids exposed on the outer face of the envelope membrane in intact cells. The approximately 65-kDa size, typical of CCDs, would likely preclude diffusion through the outer envelope.

Figure 3.

Chloroplast targeting of CCDs by chloroplast protein import assay.

(a) Protein import into pea chloroplasts. Precursor protein (Tp) was translated in vitro in the presence of 3H-Leu and incubated with pea chloroplasts (Cp) under import condition. After import, the chloroplasts were treated with or without thermolysin (Th).

(b) Sub-localization of five NCEDs inside chloroplasts. After import, purified chloroplasts were lysed and separated into subfractions of stroma (S), thylakoids (T), and envelopes (E). A pea light-harvesting chlorophyll a/b-binding protein (LHCP) and a Rubisco small subunit (ssRBS) protein were imported and subfractionated as a control. The imported proteins were resolved on an SDS–PAGE and visualized by fluorography. All samples were loaded on the basis of equal chloroplasts.

Targeting of the soluble dioxygenase proteins to plastid membranes that contain carotenoids is a second potential control point affecting substrate availability of carotenoid dioxygenases. We have shown that the N-terminal domain of the mature VP14 protein is necessary and sufficient for specific binding of VP14 to thylakoid membranes (Tan et al., 2001). This domain includes an amphipathic sequence that is required for membrane binding. As shown in Figure 4, similar amphipathic sequences are found at comparable positions in all five AtNCEDs, as well as in the AtCCD1 and AtCCD4 proteins, suggesting a conserved mechanism of membrane targeting for plastid-localized and extra-plastidic CCDs. Amphipathic structures are less evident in the AtCCD7 and AtCCD8 sequences, suggesting that these proteins may be targeted to their substrates by a different mechanism. Mammalian and bacterial CCDs also lack amphipathic domains.

Figure 4.

Analysis of the putative amphipathic α-helical regions in AtCCD proteins.

(a) The helix wheel plot of the putative amphipathic region of two representative carotenoid dioxygenases, AtNCED2 and AtNCED5. Charged residues are marked with circles.

(b) Aligned sequences of the amphipathic region in AtCCDs and VP14.

We have previously shown that NCEDs isolated from different species differ significantly in the partitioning of protein between soluble (stromal) and membrane-bound compartments of chloroplasts. Roughly 40% of the VP14 in chloroplasts extracted either from maize leaves or from chloroplasts fractionated following import reactions performed in vitro exists in a thylakoid-bound form, whereas PvNCED1 isolated from bean is found exclusively bound to thylakoid membranes following import into pea chloroplasts (Qin and Zeevaart, 1999). To address the partitioning of Arabidopsis CCDs in chloroplasts, we fractionated pea chloroplasts following import of five AtNCEDs. As shown in Figure 3(b), the AtNCED proteins differed in their partitioning between thylakoid and stromal fractions. AtNCED2, AtNCED3, and AtNCED6 were partially localized in the stroma, whereas AtNCED5 was exclusively partitioned into the thylakoid-bound fraction, and AtNCED9 exclusively in stroma. In this respect, AtNCED2, AtNCED3, and AtNCED6 are similar to VP14 of maize (Tan et al., 2001), while AtNCED5 is similar to PvNCED1 of bean (Qin and Zeevaart, 1999). The predominant stromal form of imported AtNCED2 had an apparent molecular weight lower than that of the thylakoid-bound form, suggesting partial proteolysis or other post-translational modification. Interestingly, this phenomenon was shared with AtNCED3 and AtNCED6, both of which partition between stroma and thylakoid fractions. The detected thylakoid-bound form is, clearly, not a contamination of incompletely digested precursor protein because it does not appear in the envelope fraction. The cause of the apparent upward mobility shift of imported AtNCED5 is also unknown. It is also possible that assembly of the iron cofactor contributes to size shifts observed following import into chloroplasts. In any case, the presence of two processed AtNCED2 bands of different sizes is consistent with the import experiments shown in Figure 3. The larger sized form of the protein was roughly equally partitioned between stroma and thylakoid fractions. These results indicate that different NCEDs from a single species differ with respect to membrane targeting.

The stroma localization of AtNCED9 indicates that presence of the N-terminal amphipathic structure is an imperfect predictor of membrane binding. It is possible that the amphipathic domain of AtNCED9 mediates interactions with other soluble proteins rather than with a membrane component, or that a necessary membrane component is missing in pea chloroplasts. Alternatively, membrane targeting of the this protein may be strictly regulated.

Developmental regulation and stress induction of AtNCED genes

Iuchi et al. (2001) have shown that AtNCED3 is the dominant stress-induced NCED expressed in Arabidopsis leaves. However, the previous study provided relatively little insight into the function and developmental regulation of the other Arabidopsis AtNCED genes. Expression analysis by Northern blotting hybridization is hindered by probe specificity when applied to closely related members of a gene family. In addition, it lacks quantitative information on relative levels of each member of the family. In order to quantify mRNAs of each AtNCED gene in organs of developing Arabidopsis plants, we have developed an accurate and specific transcript quantification assay based on fluorescent real-time RT-PCR (TaqMan). We designed and tested gene-specific TaqMan probes and primers for each of the five NCED genes. The cross-specificity of the primers and probes is summarized in Table 1. In pairwise tests with other cloned AtNCED genes, each probe provided a minimum of 104-fold discrimination in detection of DNA.

Table 1.  Specificity of each set of AtNCED TaqMan real-time RT-PCR primers and probe in detecting related AtNCEDs
  • *

    One picogram of each cloned AtNCED gene was added per reaction in a pairwise test against all sets of the primers and probes. The amount given was determined by this quantitative real-time RT-PCR using each set of primers and probe. A standard curve was made by the corresponding cloned gene. The data represent the equivalent amount of signal produced by 1 pg of each specific gene.

AtNCED62 × 10−4<10−73 × 10−61<10−7

Using the Taqman real-time PCR tool, we quantified expression of each AtNCED in response to stress. Mature leaves of Arabidopsis were detached and allowed to lose 15% FW through transpiration, and then sealed in plastid bags for 6 h at room temperature to allow the stress response to develop. The non-stressed control was sealed in a bag without water loss. Figure 5 shows the mRNA levels of each AtNCED in leaves. Consistent with a previous report (Iuchi et al., 2001), we found that AtNCED3 is the dominant NCED transcript in either stressed leaves or non-stressed leaves. However, stress increased mRNA levels of all NCEDs, particularly AtNCED5 and AtNCED9. To investigate how quickly these AtNCEDs express in response to a water loss, we quantified mRNA changes of each AtNCED in the first 35 min following leaf detachment. As shown in Figure 5(b), AtNCED3 mRNA dramatically increased over 35 min, and a transcriptional response was detected within 10 min after detachment. AtNCED2 also showed a several-fold induction; however, the overall mRNA levels were much less than those for AtNCED3.

Figure 5.

TaqMan real-time RT-PCR analysis on stress induction of AtNCED expression in leaves.

(a) Detached leaves were allowed to lose 15% of water and sealed in a bag for 6 h (ST). Non-stressed (NS) materials were sealed directly and kept for 6 h at the same condition.

(b) Time course of AtNCED expression during first 35 min of dehydration.

Members of the AtNCED gene family showed overlapping patterns of expression in other major organs. As shown in Figure 6, AtNCED3 and AtNCED2 are the major transcripts found in roots. In flowers, AtNCED5, AtNCED6, and AtNCED2 accounted for approximately 85% of the NCED mRNAs, whereas lower levels of AtNCED3 and AtNCED9 were detected. While previous studies have been focused on ABA functions in plant responses to environmental cues and its role in dormancy, little is known about the possible functions of ABA in flower development, pollen formation, and pollination processes. The desiccation tolerance of Arabidopsis pollen is in reminiscence of the ABA function in developing seed desiccation tolerance. In developing siliques, we found that AtNCED6 is the major mRNA with significant levels of transcripts of AtNCED5, AtNCED9, and AtNCED3 also present. In maturing siliques, AtNCED6 mRNA level decreased to the level roughly equivalent to that for the other three, whereas AtNCED3 and AtNCED5 increased, indicating that these genes regulate ABA synthesis during seed development. AtNCED2 is less likely to play a significant role in siliques based on its very low expression. Intriguingly, we detected a high level of transcripts, particularly for AtNCED3, in secondary inflorescent stems. It has not been known that ABA is synthesized in the stem, although it contains green cells containing carotenoids. It is possible that this effect may be caused by the normalization on a total RNA basis if stem cells have a substantially lower level of ribosomal RNA than other cells. However, recent studies indicate that the abscisic aldehyde oxidase (AAO1 and AAO3) of Arabidopsis, which catalyzes the final step in ABA biosynthesis, is also strongly expressed in stems (Seo et al., 2000).

Figure 6.

Expression of the AtNCED gene family in major organs of Arabidopsis during development as determined by TaqMan real-time RT-PCR.

AtNCED promoter-directed GUS expression pattern

To further localize the expression of each AtNCED gene at the cellular level, promoter:: β-glucuronidase (GUS) fusions were created for each gene and transgenic Arabidopsis was generated. Each construct contained approximately 1.5 kb of genomic DNA upstream of the translation start with the exception of AtNCED2::GUS, which contained 2.3 kb of upstream sequence. blast analysis of the putative promoter regions confirmed the absence of sequences from neighboring genes. Between 20 and 50 independent transgenic lines were generated for each AtNCED::GUS construct, and at least 15 lines were analyzed by histochemical GUS staining. Three developmental stages were analyzed: vegetative tissues at 15–16 days; developing flowers and developing seeds.

In roots, AtNCED2::GUS activity is localized in a narrow ring of cells at the tip of lateral roots, in the pericycle, and in the cortex cells surrounding new lateral roots (Figure 7a–c). Pericycle cells are a single layer of cells that surround vascular tissues, and these cells are organized longitudinally along the length of the roots. Lateral roots are initiated from pericycle cells that are located adjacent to a xylem pole; thus, there are two columns of lateral roots 180° apart. AtNCED3::GUS is expressed in a similar pattern in cells surrounding lateral root initiation sites and tips of both primary and lateral roots (Figure 7g–i). In contrast to AtNCED2, AtNCED3 is expressed in cells of the entire root tip including the root cap, root meristematic zone, and vasculature. Only AtNCED3::GUS was observed in the root vascular tissues. GUS staining for both AtNCED2 and AtNCED3 was absent from the emerging radicle of germinating seedlings but was readily detectable in fully developed primary roots bearing lateral roots. In the zone of lateral root initiation, GUS expression was observed in a punctuate pattern at regular intervals along the primary root axis apparently corresponding to the lateral root initiation sites. Consistent with that association, treatment of roots with IAA, which induces lateral root initiation, enhanced GUS staining in this region (data not shown).

Figure 7.

AtNCED promoter::GUS expression in vegetative tissues of transgenic Arabidopsis.

AtNCED2::GUS staining in:

(a) A narrow ring of cells in root tips.

(b) Base of lateral roots.

(c) Pericycle and cortex cells surrounding new lateral initials.

(d) Apical shoot meristem.

(e) Primordia of young flowers and the abscision zone of mature flowers.

(f) Stomatal guard cells in senescent leaves.

AtNCED3::GUS staining in:

(g) Root tips and vascular tissues.

(h, i) Base of lateral roots and cortex cells.

(j, k, l) Stomatal guard cells of cotyledon leaves and petiole.

(m) Cortex region of hypocotyl.

(n) Inflorescence stem.

(o) Pedicel.

AtNCED5::GUS staining in:

(p, q) Shoot meristem and at the point of attachment of petiole to stem.

In aerial plant parts, weak staining of AtNCED2::GUS was found in shoot meristems, flower primordia, abscission of mature flowers, and stomata guard cells of senescing leaves (Figure 7d–f). AtNCED3::GUS staining was found in guard cells of cotyledonous leaves, hypocotyl, and petioles (Figure 7j–l). However, GUS activity was not observed in the guard cells of rosette and cauline leaves. AtNCED3::GUS activity was found in the cortex vascular tissue of hypocotyls, green inflorescence stems, and pedicles of the developing siliques (Figure 7m–o). Faint blue staining was observed along the vascular tissue in petioles of cotyledonous and first two to four leaves, but not in the petioles of leaves emerging thereafter. AtNCED5::GUS showed strong expression at the center of leaf whorls (Figure 7p) corresponding to the axillary and apical inflorescence meristems, and eventually in fully developed shoot buds. AtNCED5::GUS staining was visible in cells at the basal portion of leaf petioles at the site of attachment to stem in a region corresponding to the leaf abscission zone (Figure 7q).

In developing flowers, AtNCED3 promoter activity was observed in floral buds, localized mainly in the anthers (Figure 8a). GUS expression in mature flowers was limited to the anther connective zone, pedicel, and the transition zones between pedicel and developing siliques. AtNCED5::GUS showed intense expression in floral buds and open flowers throughout petals and sepals, and, more prominently, in the anthers (Figure 8b,c). In anthers of open mature flowers, GUS expression was localized in anther filaments and anther connective zones. AtNCED6::GUS was expressed strongly in mature pollen grains inside anthers and pollen adhering to the stigma, as well as in germinating pollens (Figure 8d,g). No AtNCED6::GUS activity was observed in developing anthers of floral buds, indicating that AtNCED6 promoter is active in the late stages of pollen development. Faint GUS staining was observed in immature ovules in transgenic lines of AtNCED3::GUS (Figure 8e). All AtNCED::GUS genes except AtNCED9::GUS were expressed during flower development, suggesting that ABA plays a key role in this process. AtNCED6 is uniquely expressed in pollen, suggesting a possible role in pollen desiccation tolerance.

Figure 8.

Expression of AtNCED promoter–GUS fusions in flowers and developing seeds of transgenic Arabidopsis.

(a) AtNCED3::GUS activity in anthers of floral buds.

(b, c) AtNCED5::GUS staining in developing anthers of floral buds, petals, anther filament, and anther connectives.

(d) AtNCED6::GUS staining in pollens in anther sac and germinating pollens on stigma.

(e) AtNCED3::GUS in immature ovule.

(f) AtNCED5::GUS in early stages of developing seeds.

(g, h) AtNCED6::GUS in pollens and early stages of developing seeds.

(i) AtNCED3::GUS expression in developing seeds.

(j) AtNCED5::GUS expression in developing seeds.

(k) AtNCED6::GUS expression in developing seeds.

(l) AtNCED3::GUS in the funiculus of mature seeds.

Following fertilization, AtNCED6::GUS and AtNCED5::GUS were highly active through early stages of embryo and endosperm development in seeds (Figure 8f–l) and remained active in embryos until maturity (Figure 8j,k). No staining was evident in the seeds after a late maturation stage, even where mechanical rupture of the seed coat was used to ensure substrate infiltration (data not shown). In contrast, AtNCED3::GUS was expressed in a localized basal region of the seed (Figure 8i). In mature seeds, AtNCED3::GUS expression is restricted to the funiculus (Figure 8l).

Overall, the expression patterns of the AtNCED promoter::GUS genes are in good agreement with the real-time RT-PCR analysis. One exception is that we failed to detect any visible GUS staining in either vegetative or reproductive tissues in any of the AtNCED9::GUS transformants, suggesting that the 1.5 kb upstream flanking sequences are not sufficient for appropriate regulation of AtNCED9.


Our results reveal striking patterns of differential expression and localization of Arabidopsis CCD genes at the organ, subcellular, and suborganelle levels. Although our knowledge of the biochemical functions and substrate specificities of this family is still incomplete, at least five of the nine genes encode NCEDs involved in ABA synthesis (Iuchi et al., 2001).

The differential regulation of the five AtNCED genes provides insight into the overlapping programs of developmental and environmental signals that regulate ABA biosynthesis in plants. Developmental regulation of ABA synthesis is particularly important in seed development, whereas stress induction of ABA synthesis in leaves is a key response to water stress. In agreement with Iuchi et al. (2001), we find that AtNCED3 is the major stress-induced gene in leaves, with minor contributions coming from other AtNCEDs. While AtNCED5, AtNCED9, and AtNCED6 showed little evidence of strong stress regulation in the organs tested, there is no clear division between developmentally regulated and stress-regulated members of the family. Our real-time RT-PCR results indicate that AtNCED6, AtNCED5, and AtNCED9 are the dominant contributors to developmentally regulated ABA synthesis in seeds with a quantitatively smaller contribution made by AtNCED3. The hypothesis that patterns of expression correlate with relative contributions to ABA synthesis in these organs will be tested by evaluating a comprehensive set of insertional knockouts of these genes.

In well-watered plants, pronounced patterns of highly localized NCED expression are evident in roots and reproductive organs. ABA has been implicated recently in the regulation of lateral root growth and initiation (De Smet et al., 2003; Signora et al., 2001). In roots, the punctuate pattern of AtNCED2 and AtNCED3 expression in the pericyle predicts sites of lateral root initiation, indicating that these genes are expressed very early in lateral root initiation. Both genes remain active in cells at the base of mature lateral roots, suggesting a capacity for continued influence on lateral root growth. More recently, an inhibitory effect of lateral root development in Arabidopsis by exogenously applied ABA has been reported (De Smet et al., 2003). However, it is also known that primary root elongation is stimulated by accumulation of ABA in maize roots under drought conditions (Sharp et al., 1994). The highly localized expression of AtNCED2 and AtNCED3 in roots hints that ABA may play dual roles, i.e. promoting or inhibiting lateral root development. In that case, a threshold of ABA exists in these cells that provides the switch for the two opposite functions.

Abscisic acid has multiple functions in seed development including regulation of seed establishment and abortion, embryo maturation, and dormancy. Through genetic analysis of the aba1 mutant, Karssen et al. (1983) resolved distinct maternal and zygotic sources of ABA synthesis in developing seed of Arabidopsis. An early peak of ABA is maternally derived, whereas a later peak involved in establishment of dormancy is primarily derived from synthesis in the embryo. AtNCED5 and AtNCED6 show the strongest expression in mid- to late-stage embryos, suggesting that these genes are likely to contribute to the latter embryo source, ABA. At least three AtNCED genes are expressed in fertilized ovules and very early stages of seed development, suggesting complex control of maternal ABA synthesis in tissues surrounding the nascent embryo and endosperm. At their present resolution, the data do not clearly distinguish AtNCED5 and AtNCED6 expression in the female gametophyte/endosperm and the maternal nucellar tissues of newly fertilized ovules. AtNCED3 expression is, predominantly, in the maternal tissues located in the basal region of the seed and the funiculus throughout seed formation. This extends a general pattern of AtNCED2 and AtNCED3 expression at the points of organ attachment and the abscission zones in the plant.

Consistent with previous studies (Iuchi et al., 2001; Qin and Zeevaart, 1999; Tan et al., 2001), five AtNCED isoforms have transit peptides, indicating that these proteins are co-localized with their carotenoid substrates in the plastid. We have confirmed that all five AtNCEDs are imported into the chloroplasts. The implied localization of AtCCD1 in the cytosol is, on the other hand, intriguing. In standard chloroplast import reactions, we detected significant binding of in vitro translated AtCCD1 on the protease-accessible surface of intact pea chloroplasts; however, we have not determined whether this interaction is mediated by the putative amphipathic domain located in the N-terminus of AtCCD1. AtCCD4 carries a putative transit peptide; thus, it is likely targeted into the plastids. The localization of AtCCD7 and AtCCD8 is unknown, although these proteins also lack unambiguous transit peptide sequences. The lack of membrane-binding activity of AtNCED9 is also interesting. As epoxy-carotenoid substrates are exclusively found in the membranes, the membrane affinity is a potential layer of regulation of NCED activity (Tan et al., 2001). It is possible that the membrane binding of AtNCED9 is regulated in some way. Membrane binding may be mediated by an interaction with another membrane-binding protein or heterodimer formation with other AtNCEDs such as AtNCED5 (Biggin and Sansom, 1999). In that case, AtNCED9 would only contribute to ABA synthesis in cells that express another AtNCED.

The AtNCEDs, as well as NCEDs, isolated from other species (Qin and Zeevaart, 1999; Tan et al., 2001) differ substantially in their degree of association with the carotenoid-bearing thylakoid membranes. The functional significance of partitioning between stroma and thylakoid compartments is not clear. One possibility is that different NCEDs interact specifically with different components of the thylakoid membrane, and that these components may vary in abundance. This notion is consistent with the evidence that VP14 binding to thylakoid membranes is saturable, indicating that thylakoid membranes have a limited capacity to bind VP14. In this respect, AtNCED3 resembles VP14 of maize. Other similarities including their expression in roots and stress induction in leaves are consistent with the hypothesis that AtNCED3 and VP14 are orthologs. One apparent difference, however, is that AtNCED3 appears not to be the dominant NCED expressed in developing embryos. In maize, VP14 accounts for about 70% of the developmentally regulated ABA synthesis in seeds and 35% of the stress-induced ABA synthesis in leaves. Although the estimated size of the NCED gene family is similar in maize (Tan et al., 1997), exact orthology relationships between members of these gene families do not necessarily exist between these species.

AtCCD1 has been shown to catalyze symmetric cleavage of diverse carotenoids at the 9,10 and 9′,10′ double bonds (Schwartz et al., 2001). A related CsCCD has been reported recently in Crocus that cleaves the same double bonds of carotenoids (Bouvier et al., 2003). Based on the distribution of related ESTs and Northern blot analyses (Schwartz et al., 2001), the 9,10 cleavage enzyme is, evidently, normally expressed in non-stressed leaves of a broad range of species including maize. Nevertheless, constitutive production of β-ionone and related compounds is generally very low from plant leaves. One reason may be that, in intact cells, AtCCD1 access to substrates may be limited to the outer face of the outer envelope membrane of plastids. Although AtCCD1 has a broad substrate specificity in vitro, the outer envelop membrane has a unique carotenoid composition (Parry and Horgan, 1991) relative to other plastid membranes that is likely to influence the spectrum of available substrates in vivo. Importantly, localization in the cytosol would also place the AtCCD1 in a unique regulatory environment relative to the plastid-localized NCEDs subject to an independent regime of protein turnover and post-translational modification.

Experimental procedures

Plant materials and growth conditions

All the AtNCEDs were cloned from Arabidopsis ecotype Columbia (Col-0). The expression analysis by TaqMan real-time RT-PCR was performed with Arabidopsis ecotype Wassilewskija (WS). For germination, Arabidopsis seeds were surface-sterilized with 50% (v/v) household bleach (active ingredient 6% sodium hypochlorite) containing 0.1% (v/v) Tween 20 for 5 min, and then rinsed several times with sterile water. Seeds in sterile water were kept at 4°C for 3 days before spreading on Petri dishes containing 0.5× Murashige and Skoog medium (Murashige and Skoog, 1962) supplied with 1% sucrose, 0.05% MES (2-[N-morpholino]ethanesulfonic acid), and 0.2% phytogel. Plates were kept at 22°C with continuous light in a growth chamber for germination. After 10 days, seedlings were transplanted in sterile vermiculite/soil mix in pots and kept in a controlled environment cabinet at 22°C under continuous illumination.

Stress treatment

The detached leaves and roots were harvested from germination plates after 14 days and left on the bench to allow the tissues to lose 15% of their fresh weight. The materials were then sealed in plastic bags and kept in the dark for 6 h. Non-stressed leaves and roots were sealed directly in bags and kept in the dark for 6 h.

Molecular cloning of AtNCEDs

AtCCD1: The AtCCD1 gene was identified from an Arabidopsis EST database homology search with the maize VP14. The genomic clone of AtCCD1 was isolated from an Arabidopsis EMBL3 genomic library (Clontech, Palo Alto, CA, USA). The nearly full-length AtCCD1 cDNA was obtained from the Arabidopsis stock center.

AtNCED2: Molecular cloning of AtNCEDs was started when only small portions of Arabidopsis sequences were known. Thus, homologous cloning was first used. A pair of degenerate primers (forward TTC/TGAC/TGGIGAC/TGGIATGGT; reverse ACIG/CCA/GAAA/GTCA/GTGCATCAT, where I is inosine) were designed based on the conserved sequences between VP14 and its homologous protein in bacteria, lignostilbene dioxygenase (LSD). PCR conditions were 95°C for 5 min and 35 cycles of 94°C for 1 min/50°C for 2 min/72°C for 1 min. This PCR amplified a 500-bp fragment that has a sequence homology to VP14, and thus was used as a probe to screen an Arabidopsis thaliana genomic EMBL3 library (Clontech). Ten positive clones were isolated from the library. Sequence and restriction map analysis indicated that these clones were identical. This clone was designated as AtNCED2.

AtNCED3: AtNCED3 was identified from a blast search of VP14-related genes in Arabidopsis in the database. Then, two primers (forward 5′-GAAACAGAAGAAGCCACAAAAAAAGACAA; reverse 5′-AAACCGCACCCCAAAAGAAACAACAAACA) were designed to amplify the gene using Arabidopsis genomic DNA (Columbia) as template. The PCR conditions were 95°C for 5 min and 35 cycles of 94°C for 1 min, 58°C for 1 min, and 72°C for 3 min. The resulting fragment was cloned into pCR4®-TOPO (Invitrogen, CA, USA) and sequenced for confirmation.

AtNCED5: AtNCED5 was identified from the end sequence of an Arabidopsis BAC (bacterial artificial chromosome) clone T31A21. The BAC clone T31A21 was obtained from Arabidopsis stock center and was sequenced from the end containing AtNCED5. The BAC clone contained full-length AtNCED5, and the gene was amplified by PCR (forward primer TCTCATGGCTTGTTCTTACATA; reverse primer TCTGGCTCTGCACTTTAAAC) and cloned into pCR4-TOPO vector (Invitrogen).

AtNCED6 and AtNCED9: AtNCED6 and AtNCED9 were identified in a manner similar to that for AtNCED3. Primers for AtNCED6 were (forward 5′-GCTCAAACCAACGTATCCAAACTTAAACC; reverse 5′-AAACGTAAACCTAAAAAGAGGAGATAGCA); and for AtNCED9 (forward 5′-GATAAATTGTGGGAGATAAACAAGATGAG; reverse 5′-CTCAGGCAACAATCCTATGTCACTAATCA).

Plasmid constructs for chloroplast import assay

All the constructs for in vitro transcription were made under SP6 promoter in pGEM-3Z vector. To avoid the start codon in the SphI restriction site (GCATGC), pGEM-3Z vector (Promega, Madison, WI, USA) was digested with SphI, blunted with Klenow DNA polymerase, and self-ligated. To enhance in vitro transcription, translation start ATG of the target gene was put close to SP6 promoter. For AtCCD1, a construct containing the full-length AtNCED1 cDNA sequence was digested with SalI and SmaI, and the insert was cloned into pGEM-3Z vector. For AtNCED2, a construct containing the full-length AtNCED2 genomic sequence (intronless) was digested with XhoI and SmaI, and ligated into SalI and SmaI sites. For AtNCED5, the full-length sequence was amplified by PCR using BAC clone T31A21 as a template, with forward primer 5′-TCTCATGGCTTGTTCTTACATA and reverse primer 5′-TCTGGCTCTGCACTTTAAAC. The PCR fragment was cloned into pCR4-TOPO vector (Invitrogen). The fragment of HindIII and XbaI was cloned into pGEM-3Z vector (Promega). Similar strategy was used to make the constructs for AtNCED3 (forward 5′-CAAAGCTTAAAAATGGCTTCTTTCACGG; reverse 5′-TCCCACTGGTAAATCTCGCT), AtNCED6 (forward 5′-AGAAGCTTCACCACCATGCAACACTCTC; reverse 5′-TTTCTAGACCCAAAAATATCAACACATTAACC), and ATNCED9 (forward 5′-CACTGCAGAAAAGATGAGAAAATAGTAGCGTGAA; reverse 5′-TAAAGCTTATTCAAGAAACTCGTTGCCG).

In vitro transcription, translation, and import assay

All the plasmids were linearized at 3′ end before being used as templates for in vitro transcription. In vitro transcription, translation, and import were performed according to Cline et al. (1993) and were described previously (Tan et al., 2001). Fresh chloroplasts isolated from pea were used.

Isolation of total RNA and real-time quantitative RT-PCR

Total RNA was isolated from tissues by using the RNeasy Plant Mini Kit (Qiagen, Germany). RNA samples were treated with RNase-free DNase and purified using the Qiagen mini-column. Successful removal of trace amounts of contaminating DNA was confirmed by the sensitive TaqMan real-time PCR in a minus reverse transcription reaction. The concentration of RNA was determined at A260. TaqMan one-step real-time RT-PCR was performed at conditions recommended by the manufacturer (Applied Biosystems, Perkin-Elmer, Foster City, CA, USA) with 30 min at 48°C for transcription, 95°C for 10 min, and 40 cycles of 95°C for 15 sec/60°C for 1 min. All reactions contained 1× TaqMan buffer (Perkin-Elmer) and 5 mm MgCl2, 200 µm each of dATP, dCTP, and dGTP, 400 µm dUTP, 100 nm TaqMan probe, 0.625 units of AmpliTaq Gold polymerase, and 0.25 units of multiScribe RNA reverse transcriptase and RNase inhibitor in a 25-µl volume, 100–300 ng of total RNA, 800 nm of each AtNCED forward and reverse primer, and 200 nm of the gene-specific TaqMan probe. Reactions were carried out in the GeneAmp 5700 Sequence Detection (Perkin-Elmer). A plasmid containing each AtNCED was used in a standard curve assay, and all the AtNCED transcripts were normalized as copy number per nanogram of total RNA. The sequences of forward and reverse primers are listed in Table 2.

Table 2.  Primers and probes used in TaqMan real-time quantitative RT-PCR assay
GeneForward/reverse primerTaqMan probea
  • a

    All probes were labeled at 5′ with fluorescent reporter dye 6-carboxyfluorescein (FAM) and at 3′ with quencher dye 6-carboxy-N,N,N,N-tetramethylrhodamine (TAMRA). Sequences are given from 5′ to 3′.


AtNCED promoter::GUS fusions and transgenic plants

To construct the AtNCED promoter::GUS gene, a fragment approximately 1.5 kb 5′ of the translational start of each AtNCED gene except AtNCED2 was amplified by PCR and cloned into pCR4-TOPO vector (Invitrogen). For AtNCED2, a 2.3-kb promoter sequence was amplified by PCR from a genomic plasmid using forward primer T7 and reverse primer 5′-CGTGGATCCCATGGCTTTTGTTTCTTATC. For other AtNCEDs, the primers were: for AtNCED3: forward primer 5′-AGTCGACTCTCGTCAACTAAAATCCAAGTG and reverse primer 5′-GTCTAGATGTGTTCAATCAGTATTTGGTGG; for AtNCED5: forward primer 5′-ATCTAGAGGTATTTCTCCATGGCATTTT and reverse primer 5′-AGGATCCGAGAGCAAATGACTTAAGAAAAGAA; for AtNCED6: forward primer 5′-TAAGCTTCCACTACCATCTCCCCAGTG and reverse primer 5′-GGGATCCGGTGGTGACTTGTGAGCGAC; and for AtNCED9: forward primer 5′-TAAGCTTAAAATGGCTCAAACAACCGA and reverse primer 5′-TGGATCCTCACGCTACTATTTTCTCATCTTTT.

The cloned promoter sequence was sequenced to confirm the absence of PCR-caused mutation. Then, the promoter fragment was excised and subcloned into pBI101.1 binary vector (Clontech) upstream of the ATG of GUS gene. The plasmids were sequenced again to verify the junctions of the chimeric gene. The pBI101.1 clone containing different NCED promoter::GUS genes were transformed by electroporation into Agrobacterium tumefaciens strain GV1301. Plasmids were isolated from Agrobacterium and verified by restriction digests.

Arabidopsis ectotype Col-0 and WS were transformed with Agrobacterium carrying the NCED promoter::GUS gene pBI101.1 by floral dip method (Clough and Bent, 1998). Siliques were harvested separately to ensure independent transgenic lines. T1 seeds were plated on the germination medium containing 60 µg ml−1 of kanamycin and 100 µg ml−1 of carbencillin to select for transformants. Kanamycin-resistant T1 seedlings were transferred to soil and grown to maturity. The resulting T2 seed populations were screened for kanamycin resistance to use for analysis of promoter activity.

GUS assay

The T2 transgenic plants of each NCED promoter::GUS lines were grown on Murashige and Skoog media containing kanamycin and carbencillin. Some transgenic seedlings were transferred to soil and grown to adult plants for GUS staining of flowers, floral shoots, and siliques. The positive transformants were confirmed for carrying a particular NCED construct by PCR using the promoter-specific primer for the 5′ end and GUS primer, 5′-TCACGGGTTGGGGTTTCTAC, for the 3′ end. Histochemical assay of GUS activity was conducted as described by Jefferson et al. (1987) with modifications. Prior to incubation in substrate solution, plant material was fixed in 90% (v/v) acetone for 5–10 min and rinsed in a buffer containing 50 mm NaPO4, pH 7.2, 0.5 mm K3Fe(CN)6, and 0.5 mm K4Fe(CN)6. GUS staining was performed using 4 mm 5-bromo-4-chloro-3-indolyl-β-glucuronic acid (GBT) in the staining buffer (50 mm NaPO4, pH 7.2, 0.5 mm K3Fe(CN)6, 0.5 mm K4Fe(CN)6, and 4 mm X-Gluc), vacuum-infiltrated for 10 min, and incubated at 37°C in the dark for 24 h. The tissue was then cleared in several changes of 70% (v/v) ethanol. The samples were examined under Leica MZ125 stereomicroscope (Leica Microscopy Systems Ltd, Heerbrugg, Switzerland), and photographed using spot rt software v. 3.1 (SPOT Diagnostic Instruments, Inc., MI, USA) camera equipped to the microscope.


The authors thank Shan Wu and Mike McCaffery for technical assistance. This work was supported, in part, by the Florida Agricultural Experiment Station and by grants from the US Department of Energy, US Department of Agriculture, and NSF2010 awarded to D.R.M., and approved for publication as Journal Series No. R09420.