Carotenoid biosynthesis during tomato fruit development: regulatory role of 1-deoxy-D-xylulose 5-phosphate synthase


For correspondence (fax +34 93 402 1219; e-mail


Plant isoprenoids represent a heterogeneous group of compounds which play essential roles not only in growth and development, but also in the interaction of plants with their environment. Higher plants contain two pathways for the biosynthesis of isoprenoids: the mevalonate pathway, located in the cytosol/endoplasmic reticulum, and the recently discovered mevalonate-independent pathway (Rohmer pathway), located in the plastids. In order to evaluate the function of the Rohmer pathway in the regulation of the synthesis of plastidial isoprenoids, we have isolated a tomato cDNA encoding 1-deoxy-D-xylulose 5-phosphate synthase (DXS), the first enzyme of the pathway. We demonstrate in vivo activity and plastid targeting of plant DXS. Expression analysis of the tomato DXS gene indicates developmental and organ-specific regulation of mRNA accumulation and a strong correlation with carotenoid synthesis during fruit development. 1-Deoxy-D-xylulose feeding experiments, together with expression analysis of DXS and PSY1 (encoding the fruit-specific isoform of phytoene synthase) in wild-type and yellow flesh mutant fruits, indicate that DXS catalyses the first potentially regulatory step in carotenoid biosynthesis during early fruit ripening. Our results change the current view that PSY1 is the only regulatory enzyme in tomato fruit carotenogenesis, and point towards a coordinated role of both DXS and PSY1 in the control of fruit carotenoid synthesis.


Isoprenoids are a large group of compounds which derive from a common building unit, isopentenyl pyrophosphate (IPP). Following the discovery of the mevalonate pathway in the 1950s, it was accepted that isoprenoids were synthesized from acetyl-CoA via mevalonic acid and IPP in all organisms, including plants ( McGarvey & Croteau 1995). In many cases, however, experimental data on the biosynthesis of specific isoprenoids in plants and microorganisms could not be explained from the exclusive operation of the mevalonate pathway (reviewed by Lichtenthaler et al. 1997 ). A few years ago, an alternative mevalonate-independent pathway for IPP biosynthesis was initially identified in bacteria by Rohmer and collaborators ( Rohmer et al. 1993 ). Evidence has subsequently emerged indicating that both the mevalonate and the Rohmer pathways operate in higher plants ( Lichtenthaler 1999; Rohmer 1999). Plant isoprenoids synthesized in the cytosol/endoplasmic reticulum (including hormones such as cytokinins and brassinosteroids, phytosterols for membrane biogenesis, phytoalexins for defence against pathogens, and prenyl groups for post-translational modification of proteins), and in the mitochondria (ubiquinone), are formed from mevalonate-derived IPP. Plastid isoprenoids (including hormones such as gibberellins and abscisic acid, photosynthesis-related pigments such as carotenoids and the phytol moiety of chlorophylls, and the side chain of electron carriers such as plastoquinone, phylloquinone K1, α-tocoquinone, and α-tocopherol) derive from IPP synthesized from the Rohmer pathway ( Fig. 1).

Figure 1.

Plastid isoprenoid biosynthesis pathway.

G3P, d-glyceraldehyde 3-phosphate; DXS, 1-deoxy- d-xylulose 5-phos- phate synthase; DX5P, 1-deoxy- d-xylulose 5-phosphate; DXR, 1-deoxy- d-xylulose 5-phosphate reductoisomerase; ME4P, 2-C-methyl- d-erythritol 4-phosphate; IPP, isopentenyl pyrophosphate; DMAPP, dimethylallyl pyrophosphate; GPS, geranylgeranyl pyrophosphate synthase; GGPP, geranylgeranyl pyrophosphate; PSY, phytoene synthase; PDS, phytoene desaturase; ZDS, ζ-carotene desaturase; LCYB, lycopene β-cyclase; LCYE, lycopene ε-cyclase; ABA, abscisic acid.

The initial reaction of the Rohmer pathway involves the formation of 1-deoxy- d-xylulose 5-phosphate (DX5P) by condensation of (hydroxyethyl)thiamin derived from the decarboxylation of pyruvate with the C1 aldehyde group of d-glyceraldehyde 3-phosphate ( Arigoni et al. 1997 ; Rohmer et al. 1996 ). This reaction is catalysed by 1-deoxy- d-xylulose 5-phosphate synthase (DXS), a novel type of transketolase ( Lange et al. 1998 ; Lois et al. 1998 ; Sprenger et al. 1997 ). In the second step, DX5P is converted to 2-C-methyl- d-erythritol 4-phosphate (ME4P) by the enzyme DX5P reductoisomerase ( Lange & Croteau 1999; Schwender et al. 1999 ; Takahashi et al. 1998 ). The remaining enzymatic reactions leading to the synthesis of IPP have not yet been fully elucidated.

In plants, it has been shown that 1-deoxy- d-xylulose (DX) can be incorporated into carotenoids, plastoquinone and the phytol moiety of chlorophylls ( Arigoni et al. 1997 ; Lichtenthaler et al. 1997 ; Schwender et al. 1997 ), probably after conversion to DX5P. In addition, cDNAs encoding DXS have been cloned from several plant species, including peppermint ( Lange et al. 1998 ) and pepper ( Bouvier et al. 1998 ). After the identification of the Escherichia coli dxs gene sequence, it was found that the gene disrupted in the previously reported Arabidopsis thaliana CLA1 mutant ( Mandel et al. 1996 ) encoded DXS ( Lois et al. 1998 ). CLA1 mutant plants show a very early arrest of chloroplast development, lack of chlorophyll and carotenoid pigments, and an albino phenotype ( Mandel et al. 1996 ). Therefore DXS appears to be required for the biosynthesis of isoprenoids essential for plastid function. DX5P is a biosynthetic intermediate not only for IPP synthesis, but also for thiamine and pyridoxol in plastids ( Julliard 1992; Julliard & Douce 1991). Thus it has been proposed that the first regulatory step in IPP and isoprenoid formation might be the synthesis of ME4P from DX5P ( Lange & Croteau 1999; Takahashi et al. 1998 ). However, this possibility is yet to be tested.

Because of its novelty, the contribution of the Rohmer pathway to the control of plastid isoprenoid biosynthesis has not yet been evaluated. Fruit development in tomato constitutes a good model system to investigate the regulation of plastid isoprenoid biosynthesis. Young developing fruits contain chloroplasts that synthesize isoprenoids not only for photosynthesis-related processes, but also for gibberellin and abscisic acid biosynthesis ( Gillaspy et al. 1993 ). During ripening, however, chloroplasts differentiate into chromoplasts in a process that involves degradation of chlorophylls and a massive accumulation of carotenoids (particularly lycopene), causing the fruit colour to change from green to red ( Bartley & Scolnik 1995; Gillaspy et al. 1993 ; Khudairi 1972). In order to characterize the role of the first enzyme of the Rohmer pathway in plastid isoprenoid biosynthesis, we have cloned a cDNA corresponding to the single tomato DXS gene and demonstrated that it encodes a functional DXS protein which is targeted to plastids in vivo. Analysis of tomato DXS expression indicated developmental and organ-specific regulation of mRNA levels, and a strong correlation with carotenoid accumulation in tomato fruit. In addition, our data indicate that DXS is a potentially key regulatory enzyme for carotenoid biosynthesis.


Cloning of a tomato cDNA encoding DXS

An Arabidopsis 1.5 kb cDNA fragment corresponding to the CLA1 gene ( Mandel et al. 1996 ) was used as a probe to screen a tomato leaf cDNA library. Two positive clones were isolated which, after DNA sequencing, were shown to correspond to overlapping cDNAs. The longer, LeDXS1, contains a 2568 bp insert with a 2160 bp open reading frame flanked by a 156 bp 5′ UTR and a 252 bp 3′ UTR. The protein encoded by this cDNA has 719 amino-acid residues and a predicted molecular mass of 77.6 kDa ( Fig. 2). The predicted polypeptide shows high similarity to published DXS proteins from pepper (96% identity), Arabidopsis (83%) and peppermint (64%), and it is also similar to bacterial DXS (i.e. 60% similarity with the E. coli enzyme). Plant DXS proteins contain an N-terminal domain which is not present in the microbial enzyme ( Fig. 2). This domain is poorly conserved but shows the general features of plastidial targeting sequences, including an abundance of the hydroxylated residues serine and threonine and a shortage of the acidic residues aspartic acid and glutamic acid ( Von Heijne et al. 1991 ). Genomic DNA blot hybridization experiments, using as a probe the complete LeDXS1 cDNA sequence, indicated the presence of a single copy gene in tomato (data not shown).

Figure 2.

Multiple alignment of DXS amino acid sequences.

Sequences from the following organisms are included: tomato (Lycopersicon esculentum; AF143812); pepper (Capsicum annuum CapTKT2, Bouvier et al. 1998 ); Arabidopsis thaliana (CLA1, Mandel et al. 1996 ); peppermint (Mentha × piperita, Lange et al. 1998 ); and Escherichia coli ( Lois et al. 1998 ). Identical residues are shown as white inside black boxes (when they are conserved in the five sequences); white inside grey boxes (conserved in four of the sequences); and black inside grey boxes (conserved in three sequences).

To confirm the biochemical function of the protein encoded by the cloned tomato cDNA, a complementation assay was carried out using the E. coli strain MC4100 dxs::CAT, in which the dxs gene had been disrupted by insertion of the CAT gene, encoding chloramphenicol acetyltransferase ( Charon et al. 2000 ). The disruption of the dxs gene is lethal due to the block in the biosynthesis of essential DX5P-derived compounds (isoprenoids, thiamin and pyridoxol). However, the mutant cells can grow in media supplemented with DX ( Charon et al. 2000 ). For the complementation assay, the protein encoded by the tomato LeDXS1 cDNA was expressed in E. coli dxs::CAT under the control of the PBAD promoter, which can be activated with arabinose and repressed with glucose. Induction with arabinose overcame DX auxotrophy of the transformed dxs::CAT mutant cells ( Fig. 3). However, repression with glucose prevented growth of the transformed mutant cells. These results demonstrate that the cloned tomato cDNA encoded a functional DXS.

Figure 3.

Complementation assay of E. coli dxs mutant.

E. coli MC4100 dxs::CAT cells transformed with plasmid pBAD-DXS, in which the expression of the protein encoded by the tomato LeDXS1 cDNA is under the control of the PBAD promoter, were grown on 2 × TY plates supplemented with either 0.02% arabinose or 0.2% glucose (right side of plates). Wild-type E. coli strain MC4100 was grown on the same plates as a control (left side).

Tomato DXS is targeted to plastids

The N-terminal region of plant DXS has been suggested to be a plastid targeting sequence ( Lange et al. 1998 ; Mandel et al. 1996 ). However, the only reported experimental evidence on the plastidial localization of this enzyme is based on the cross-reaction of plastid polypeptides of the predicted size with an anti-DXS antibody in pepper fruit ( Bouvier et al. 1998 ). To determine whether DXS is targeted to plastids in vivo, we fused the N-terminal 430 residues of tomato DXS to sGFP, a soluble codon-optimized green fluorescent protein ( Rodríguez-Concepción et al. 1999 ), for expression of the corresponding fusion protein (DXS-GFP) in living cells. Tomato leaves were microbombarded with constructs to transiently express either GFP or DXS-GFP under the control of the CaMV 35S promoter. As expected, green fluorescence corresponding to GFP was localized in the cytoplasm and the nucleus ( Fig. 4a). Fluorescence from DXS-GFP, however, colocalized with chlorophyll autofluorescence ( Fig. 4b,c). This result confirms that tomato DXS is targeted to chloroplasts in vivo, in agreement with the proposed role of this enzyme in the biosynthesis of plastidic isoprenoids.

Figure 4.

Subcellular localization of DXS-GFP in tomato.

Leaves were microbombarded with constructs to express either GFP (a) or the fusion protein DXS-GFP (b, c). Green fluorescence from the GFP proteins (a, b) and red fluorescence from chlorophyll (c) was detected by confocal microscopy. Bars, 10 μm.

Organ-specific and developmental expression of tomato DXS

To gain more insights into the function of DXS in tomato plants, the expression pattern of the DXS gene was analysed in several tissues by RNA blot hybridization. DXS transcripts are abundant in young, developing, and fully expanded leaves, inflorescences, and stems ( Fig. 5a). DXS transcripts in roots, however, were hardly detectable, even after long exposure. The level of DXS transcripts in young and mature green fruit was similar to that in the other photosynthetically active shoot tissues, although a slight decrease in DXS mRNA accumulation was detected in mature green fruit as compared to young fruit ( Fig. 5b). However, the level of DXS transcripts increased greatly during fruit ripening. The highest level of transcripts was detected in orange fruit, and decreased during the latest stages of ripening ( Fig. 5b). These results show that DXS mRNA accumulation in tomato is controlled by organ-specific and developmental signals. They also show that the increase in DXS mRNA accumulation correlates with the transition from mature green to orange fruit, suggesting that DXS induction is associated with the activation of carotenoid biosynthesis at the onset of ripening. Therefore it could be expected that cells actively synthesizing carotenoids would have the highest level of DXS expression. To study this possibility, in situ hybridization experiments were carried out on tomato orange fruit sections. DXS transcripts were predominantly localized in the outer layers of cells of the fruit pericarp ( Fig. 5c). Hand-made sections of orange and red fruit pericarps also showed that most of the orange and red carotenoid pigments were distributed in the same peripheral area of the pericarp ( Fig. 5c). These results further support a role for DXS in carotenoid biosynthesis during tomato fruit ripening.

Figure 5.

Analysis of tomato DXS expression pattern.

(a) Blot with RNA samples from tomato leaves (L1, young; L2, expanding; L3, fully expanded); inflorescences (I); stems (S); and roots (X) hybridized with a tomato DXS probe. The size of the transcript detected is shown. Exposure was carried out for 4 days.

(b) Blots with RNA samples from tomato Ailsa Craig wild-type young (Y); mature green (M); orange (O); and red ripe (R) fruit; and mutant yellow flesh (r) young (Y); mature green (M); orange-like (O); and yellow ripe (R) fruit were hybridized with probes for DXS and PSY1, and exposed for 2 days. The blots described in (a) and (b) were also hybridized with a 25S rRNA probe to compare the RNA amounts loaded in each lane.

(c) Localization of DXS mRNA and carotenoid pigments on tomato fruit sections. I, In situ hybridization of an orange fruit section (8 μm) with a tomato DXS sense riboprobe; II, in situ hybridization of a similar section (8 μm) with a tomato DXS antisense riboprobe; III, red fruit hand-made section. Bars, 0.5 mm.

Fruit DXS mRNA accumulation pattern is affected by PSY1 activity

It has been proposed that PSY1, the fruit-specific isoform of phytoene synthase (which catalyses the formation of phytoene from two molecules of GGPP; Fig. 1), is the limiting enzyme for carotenoid biosynthesis during tomato fruit ripening ( Bartley & Scolnik 1995). To compare the expression patterns of PSY1 and DXS genes during fruit development, the same filter used with the DXS probe ( Fig. 5b) was stripped and rehybridized with a PSY1 probe. Similarly to DXS, a significant accumulation of PSY1 mRNA was detected in ripening fruit, reaching the highest levels at the orange stage and slightly decreasing in red, ripe fruit ( Fig. 5b). The similar pattern of expression of DXS and PSY1 suggests a coordinated role during carotenoid biosynthesis in ripening fruit. However, the kinetics of DXS and PSY1 transcript accumulation were clearly different during the transition from growing to ripening fruit. Consistent with previous results ( Giuliano et al. 1993 ; Ronen et al. 1999 ), the PSY1 mRNA level began to increase in mature green fruits before the first colour changes were seen ( Fig. 5b). The observation that the induction of PSY1 transcript accumulation precedes that of DXS suggests that changes in the level of carotenoid intermediates and/or end-products associated with higher PSY1 activity in mature green fruit might be involved in the regulation of DXS expression during fruit ripening. This possibility was tested by analysing DXS expression in the yellow flesh (r) mutant. The r mutation is known to affect the PSY1 gene, resulting in the expression of a shorter mRNA that encodes a non-functional enzyme ( Fray & Grierson 1993). Therefore, the r fruit contains virtually no carotenoids, although it shows a yellow colour that is primarily due to the accumulation of the flavonoid chalconaringenin.

The results of the RNA blot analysis of DXS expression in fruits from the r mutant are shown in Fig. 5(b). DXS transcript levels in young and mature green fruits of the r mutant were similar to those in wild-type fruits. Fruit ripening stages in the r mutant were defined as orange-like (when chlorophyll degradation and pigment accumulation changed fruit colour from green to yellow) and yellow ripe (yellow soft fruit). Although this colour- and softening-based staging system for ripening may not be accurate, the pattern of accumulation of PSY1 transcripts in wild-type fruits was identical to that of the shorter PSY1 mRNAs in the selected r fruits ( Fig. 5b; Ronen et al. 1999 ), suggesting that the fruits from r and wild-type tomato plants used for the experiment were at similar ripening stages. The onset of ripening in r orange-like fruits was also associated with a high increase in DXS mRNA accumulation. However, DXS transcript level did not decrease in later stages of ripening, leading to mRNA levels that were much higher than those detected in ripe wild-type fruits ( Fig. 5b). This result shows that PSY1 activity and carotenoid synthesis are not required for the induction of DXS transcript accumulation at the beginning of ripening, but they seem to be involved in the down-regulation of DXS expression in the last stages of ripening.

DX induces expression of PSY1 and DXS, and carotenoid accumulation

The strong correlation between the temporal and spatial patterns of DXS expression and carotenoid accumulation suggested that DXS activity might be limiting for carotenoid biosynthesis during tomato fruit ripening. To mimic the effect of increased DXS activity on carotenoid accumulation, DX was directly injected into mature green tomato fruit. DX was used instead of DX5P to overcome problems associated with uptake of the phosphorylated compound into cells. Previous reports have shown that DX can readily enter the plastid isoprenoid pathway when externally supplied to plant cells ( Arigoni et al. 1997 ; Lichtenthaler et al. 1997 ; Schwender et al. 1997 ).

A total of 14 mature green fruits were used in three independent experiments. Each fruit was longitudinally cut in two halves. A syringe with a fine needle was used to inject 50 μl of a 100 m m DX solution to one of the halves. The same volume of a control solution without DX was injected to the other half. After 3 days at room temperature, the fruits were classified into three different groups according to the colour of the corresponding halves ( Fig. 6). In two out of the 14 fruits (group 1), both halves remained green. In contrast, in seven fruits the DX-injected half of each fruit became red, whereas the control half remained green (group 2). In the remaining five fruits, carotenoid pigments accumulated in both halves, but the one treated with DX typically showed a stronger red colour (group 3).

Figure 6.

Effect of DX injection on tomato fruit development.

Mature green fruits were longitudinally cut in two halves and either a DX (+) or a control (–) solution were injected into each half. After 3 days, fruits were classified into groups 1, 2, or 3 according to the visible accumulation of carotenoid pigments.

(a) Representative fruits of each group.

(b) Measurement of chlorophyll a (green) and total carotenoids (orange) in the fruit halves shown in (a). Columns represent means and bars represent standard errors of three replicas.

(c) Blot with RNA samples from the fruit halves shown in (a) hybridized with DXS, PSY1 and 25S rRNA probes.

Representative fruits from the three groups were selected ( Fig. 6a), and their corresponding halves were frozen and ground in liquid nitrogen to a fine powder which was used for measurement of both chlorophyll and carotenoid contents ( Fig. 6b) and RNA analysis ( Fig. 6c). Although no visible colour difference was observed between the two halves of group 1 fruit, chlorophyll a content in the DX-treated half was lower than in the control half, whereas the carotenoid level was similar ( Fig. 6b). The DX-treated half of group 1 fruit showed a higher level of PSY1 mRNA, although the DXS transcript level remained unchanged ( Fig. 6c). In the fruits from groups 2 and 3, DX injection induced carotenoid accumulation ( Fig. 6b). In group 2 fruit, the DX-injected half showed significantly less chlorophyll a and more carotenoids than the control half. In these fruits, DX induced not only PSY1 but also DXS gene expression ( Fig. 6c). The control half from group 3 fruit had low chlorophyll a and high carotenoid contents ( Fig. 6b), and it showed higher levels of DXS and PSY1 transcripts than those of control fruits from groups 1 and 2 ( Fig. 6c). The DX-injected half of group 3 fruit, however, again accumulated more carotenoids and showed higher levels of DXS and PSY1 transcripts than the control. The different level of chlorophyll a, carotenoids, and DXS and PSY1 transcripts in the control halves from group 1, 2 and 3 fruits shows that their starting metabolic status was different. This was expected as ‘mature green’ is a relatively broad developmental definition that can span a few days. However, our experimental design suggests that DX injection activates accumulation of PSY1 transcripts and eventually carotenoids independently of the fruit stage. Taken together, our data suggest that synthesis of DX5P by DXS is probably a key regulatory step for carotenoid biosynthesis.


Since the discovery of the mevalonate-independent pathway for the synthesis of plastid isoprenoids (Rohmer pathway), the full elucidation of the pathway and its regulatory role in the synthesis of specific plastidic isoprenoid end-products represents a novel major challenge in plant cell metabolism. Here we present the molecular characterization of tomato DXS, the first enzyme of the Rohmer pathway, and propose its regulatory role in the biosynthesis of carotenoids during tomato fruit ripening.

Using in vivo systems, we have demonstrated that the protein encoded by the cloned tomato LeDXS1 cDNA is a functional DXS that is targeted to chloroplasts, where it probably synthesizes DX5P for plastid isoprenoid biosynthesis. In agreement, analysis of the DXS expression pattern showed a clear correlation between DXS mRNA accumulation and plastid isoprenoid requirements of the tissue. The level of DXS mRNA was similar in all the photosynthetic tissues tested, whereas it was very low in roots ( Fig. 5). Most of the plastid isoprenoid products are related to photosynthesis ( Fig. 1), and therefore a more active DXS expression in shoot tissues compared to the root can be expected. Consistently, the CLA1 gene encoding Arabidopsis DXS is positively regulated by light ( Mandel et al. 1996 ). Nevertheless, very low levels of expression of DXS in tomato roots and in dark-grown Arabidopsis seedlings can be detected, suggesting that DXS activity is required, at least at a basal level, in all cell types.

DXS expression also correlates with the production of specific isoprenoids in other plants. Peppermint leaves contain oil gland secretory cells highly specialized for isoprenoid (monoterpene) accumulation. In these cells, DXS mRNA level is highest during the first stages of leaf development, preceding the peak of monoterpene biosynthesis ( Lange et al. 1998 ). In tomato, DXS expression in young to fully expanded leaves remained constant, consistent with the fact that no specific isoprenoid product is accumulated at high levels during leaf development. Unlike leaves, however, tomato fruit showed a developmentally regulated pattern of DXS expression. Young and mature green fruit had DXS transcript levels similar to those detected in other photosynthetic tissues, although a slight decrease in DXS transcript level was observed when young fruit developed into mature green fruit. This DXS down-regulation correlated with a lower accumulation of photosynthesis-related isoprenoids such as carotenoids and chlorophylls ( Fraser et al. 1994 ; Giuliano et al. 1993 ), and with a decreased requirement for other plastid isoprenoid-derived products such as gibberellins and abscisic acid in mature green fruit ( Gillaspy et al. 1993 ). In contrast, a strong induction of DXS expression was observed during fruit ripening at the stages of high rate of carotenoid accumulation, although the DXS transcript level decreased in ripe fruit ( Fig. 5). DXS expression is also up-regulated in correlation with carotenoid accumulation in pepper fruit ( Bouvier et al. 1998 ). Furthermore, the accumulation of DXS transcripts in cells from the outer pericarp tissue of tomato orange fruit closely paralleled the localization of carotenoid pigments. In summary, the pattern of DXS expression was almost identical to that reported for the accumulation of carotenoids throughout fruit development ( Fraser et al. 1994 ; Giuliano et al. 1993 ), except in the last stages of ripening. Carotenoid biosynthesis is a highly controlled process which differs in chloroplast- and chromoplast-containing tissues, suggesting a specific regulation of genes and enzymes of the carotenoid pathway at each plastid developmental stage ( Fraser et al. 1994 ; Giuliano et al. 1993 ). In the transition from green to orange fruit, the carotenoid profile changes dramatically. In green fruits, the main carotenoid products are lutein and β-carotene ( Fraser et al. 1994 ; Ronen et al. 1999 ). During ripening, however, both lycopene β- and ε-cyclase activities (LCYB and LCYE, respectively; Fig. 1) decrease to undetectable levels and lycopene accumulates in chromoplasts, resulting in the characteristic red colour of tomato ripe fruit ( Fraser et al. 1994 ; Ronen et al. 1999 ). Our expression studies in the r mutant suggest that the qualitative and/or quantitative changes in ripe fruit carotenoids contribute to DXS down-regulation ( Fig. 5). The fact that DXS mRNA levels do not decrease (but increase) as a result of the accumulation of higher levels of orange-red carotenoids after injection of DX to mature green fruit ( Fig. 6) confirm that the carotenoid-induced DXS down-regulation is specific to the last stages of ripening. End-product regulation of other tomato genes encoding enzymes of the carotenoid pathway, such as phytoene desaturase (PDS; Fig. 1), has also been reported ( Corona et al. 1996 ). The complex regulation of tomato DXS gene expression is consistent with the current model of control of carotenoid biosynthesis by modulation of the transcriptional activities of genes encoding enzymes of the pathway ( Corona et al. 1996 ; Fraser et al. 1994 ; Ronen et al. 1999 ).

The pattern of DXS mRNA accumulation is similar to the pattern of expression of PSY1, the gene reported to encode the committed enzyme in carotenoid biosynthesis during tomato fruit development ( Bartley & Scolnik 1995). But unlike DXS, PSY1 expression ( Fig. 5b) and activity ( Fraser et al. 1994 ) are induced before the onset of ripening, in the mature green stage. Induction of PSY1 activity, however, does not result in increased carotenoid levels (which actually decrease at this stage of development; Fraser et al. 1994 ; Giuliano et al. 1993 ), as would be expected from the current view of PSY1 as the regulatory enzyme in fruit carotenoid biosynthesis ( Bartley & Scolnik 1995). The activation of PSY1 expression preceding the increase in DXS transcript levels was also observed after DX injection ( Fig. 6). This induction of PSY1 mRNA accumulation by DX feeding and the modulation of DXS expression by PSY1 activity during the last stages of fruit development ( Fig. 5b) suggest a significant cross-talk between PSY1 and DXS activities that may contribute to the fine regulation of carotenoid accumulation in tomato fruit. Assuming that the changes in DXS mRNA levels result in similar changes in DXS enzyme activity, we propose an alternative model for the regulation of carotenoid synthesis during tomato fruit development ( Fig. 7). In this model, DXS rather than PSY1 would control the start of massive carotenoid accumulation when tomato fruit enters the ripening phase. It is possible that PSY1 activity precedes the increase in DX5P synthesis, so the bulk of GGPP derived from DX5P in the first stages of ripening may be immediately channelled to phytoene and readily enter the carotenoid biosynthesis pathway. Thus PSY1 activity would be limiting for tomato fruit carotenoid biosynthesis only if non-limiting levels of the GGPP substrate were available.

Figure 7.

Proposed model of the role of DXS and PSY1 in the regulation of carotenoid biosynthesis in tomato fruit.

The model is based on the expression pattern of DXS and PSY1 during fruit development and the DX injection experiments in mature green fruit. DXS and PSY1 activities are represented as gates that control flux (vertical arrows) from substrates to products. In young fruit, the coordinated activities of the two enzymes result in the production of carotenoids for photosynthesis-related processes. In mature green fruit, DXS transcript level does not significantly change, but there is an increase in PSY1 expression and activity. Despite the proposed regulatory role of PSY1 for carotenoid biosynthesis, the higher level of PSY1 activity in mature green fruit does not result in more carotenoids, probably because DXS activity does not provide enough DX5P to be channelled to GGPP, the substrate of PSY1. In orange and red fruit, the high increase in both DXS and PSY1 expression leads to massive accumulation of carotenoids.

Since DX5P is a precursor not only of isoprenoids, but also of thiamine and pyridoxol ( Julliard 1992; Julliard & Douce 1991), it has been suggested that the conversion of DX5P into ME4P catalysed by DX5P reductoisomerase (DXR; Fig. 1) could be the first regulatory step in the Rohmer pathway ( Lange & Croteau 1999; Takahashi et al. 1998 ). However, the temporal and spatial expression patterns of the tomato DXR gene during fruit development do not support the hypothesis that the reaction catalysed by DXR could be such a regulatory step (M.R.C., L.M.L. and A.B., unpublished results). Our model ( Fig. 7) proposes that DXS activity (and therefore DX5P availability) is limiting for carotenoid synthesis during the first stages of ripening. In agreement, injection experiments ( Fig. 6) showed that an increase in the levels of DX (or most probably its metabolically active form, DX5P, which mimics a specific increase in DXS activity) results in up-regulation of the expression of PSY1 and eventually of DXS, and increased carotenoid content in the fruit. These observations indicate that DXS activity is indeed the first potentially regulatory step in the biosynthesis of fruit carotenoids, at least during early stages of ripening. Furthermore, in addition to induction of PSY1 expression, DX injection activated other processes typically associated with the onset of ripening such as chlorophyll a degradation ( Fig. 6) and down-regulation of the rbcS2 gene (M.R.C., L.M.L. and A.B., unpublished results), both of which precede carotenoid accumulation ( Fraser et al. 1994 ; Gillaspy et al. 1993 ; Khudairi 1972). We speculate that increased synthesis of DX5P might contribute to the induction of the developmental process of fruit ripening as a whole. Since ripening is an aspect of development that is unique to fruit, it is possible that the proposed regulatory character of DXS is fruit-specific. However, overexpression of DXS in E. coli cells carrying a plasmid harbouring lycopene biosynthesis genes leads to higher accumulation of lycopene and ubiquinone ( Harker & Bramley 1999), suggesting that DXS activity may also be limiting for isoprenoid biosynthesis in microorganisms.

In conclusion, our results support previous studies showing that carotenogenesis in tomato fruit is independent of the mevalonate pathway ( Rodríguez-Concepción & Gruissem 1999), and provide further evidence that the isoprenoid intermediates for carotenoid biosynthesis derive from the Rohmer pathway. In addition, we show several lines of experimental evidence that synthesis of DX5P is a key rate-limiting step for carotenoid biosynthesis in tomato fruit. Our results change the current view that PSY1 is the only regulatory enzyme in tomato fruit carotenogenesis, and point towards a coordinated role of both DXS and PSY1 in the control of carotenoid synthesis during ripening. This opens up a new perspective to the study and eventual manipulation of carotenoid accumulation in tomato fruit.

Experimental procedures

Plant material and treatments

Seeds from tomato (Lycopersicon esculentum Mill. cv. Ailsa Craig) wild type and yellow flesh mutant (r) were a gift from the laboratories of Dr P.M. Bramley and Dr W. Gruissem. Plants were grown under controlled greenhouse conditions. Green fruit development stages were defined as described ( Gillaspy et al. 1993 ) and are referred to as young (up to 1.5 cm in diameter) and mature green. Orange and ripe fruit stages in the wild type were also defined as described ( Gillaspy et al. 1993 ). Ripening stages in the r mutant were defined as orange-like (when chlorophyll degradation and pigment accumulation changed fruit colour from green to yellow), and yellow ripe (yellow soft fruit). 1-Deoxy- d-xylulose (DX) was prepared basically as described ( Lois et al. 1998 ). For injection experiments, a total of 14 mature green fruits were collected from the plant and longitudinally cut in two halves. Immediately, the open sections of both halves were protected from desiccation with Saran Wrap. A 0.5 ml syringe with a fine needle was used to inject 50 μl of 100 m m DX in 0.1% (v/v) Tween 20 through the pericarp to one of the halves. The same volume of 0.1% (v/v) Tween 20 was injected into the other half as a control. After 3 days at room temperature, both halves were photographed, frozen in liquid nitrogen, and stored at −80°C until used for experiments.

cDNA cloning and analysis

Plasmid pH 2A12, containing a 2.6 kb cDNA which encodes the full-length CLA1 (DXS) protein from Arabidopsis thaliana, was obtained from the DNA Stock Center at ABRC, Ohio State University (EST 22939 CD4-16, GenBank accession number W43562). A 1.5 kb XbaI-EcoRI fragment from pH 2A12 was used as a probe to screen a tomato leaf cDNA library (kindly provided by Dr Salomé Prat). After screening 90 000 λgt11 plaques under low-stringency conditions ( Sambrook et al. 1989 ), two positive clones (LeDXS1 and LeDXS2) were identified and used for further analysis. The cloned cDNAs were excised from the λ vector by digestion with BamHI and the fragments released were cloned into pBluescript (Stratagene). Both strands were sequenced with the ABI PRISM DNA Sequencing kit (Amersham) using vector primers and specific primers. Sequencing showed that both clones corresponded to overlapping cDNAs. The plasmid containing the longer clone, LeDXS1 (encoding a full-length protein with homology to DXS), was named pLeDXS. Sequence analyses were performed using the GCG 9.0 software package (Genetics Computer Group Inc.) and the NCBI blast program available on the web ( Multiple alignment was performed using the pileup program (GCG) with the default options.

E. coli complementation

An NdeI restriction site was introduced at the ATG translation start codon of the tomato cDNA in plasmid pLeDXS with the QuickChange site-directed mutagenesis kit (Stratagene) using Pfu DNA polymerase (Gibco-BRL) and the complementary primers PTom-Nde (5′-CAGTTGAATTGACTACATATGGCTTTGTGTGC-3′) and PTom-Nde-C (5′-GCACACAAAGCCATATGTAGTCAATTCAA- CTG-3′), as specified by the supplier. The resulting plasmid was designated pLeDXSmut. A modified version of plasmid pBAD-GFPuv (Clontech) was created by removing the NdeI site located at position 4926 by site-directed mutagenesis with the oligonucleotide PBAD-mut1 (5′-CTGAGAGTGCACCATCTGCGGTGT- GAAATACC-3′) as mutagenic primer. The resulting plasmid was designated pBAD-M1. Expression plasmid pBAD-DXS was constructed by cloning the 2.4 kb NdeI-KpnI fragment generated by partial digestion of pLeDXSmut with NdeI into the corresponding sites of plasmid pBAD-M1. Plasmid pBAD-DXS was used to transform E. coli MC4100 dxs::CAT cells ( Charon et al. 2000 ). One isolated colony was selected and streaked on 2 × TY plates supplemented with either 0.02% arabinose (to induce expression from the PBAD promoter) or 0.2% glucose (to repress expression from the PBAD promoter).

DNA and RNA analysis

Genomic DNA and RNA from several tomato tissues, except from fruit, were extracted according to Dean et al. (1985) . Genomic DNA (10 μg) extracted from young leaves was digested with several restriction enzymes, size-fractioned by electrophoresis in a 0.8% (w/v) agarose gel, and blotted onto a Hybond-C nitrocellulose membrane (Amersham). Hybridization was carried out in 5 × SSC, 5 × Denhardt's, 1% (w/v) SDS and 500 μg ml−1 denatured salmon sperm DNA at 68°C for 18 h. The entire 2.6 kb tomato DXS cDNA excised from plasmid pLeDXS was 32P-labelled with the Ready to Go kit (Pharmacia), and used as a probe. Washes were performed for 20 min at room temperature twice in 2 × SSC, 0.1% (w/v) SDS, twice in 0.2 × SSC, 0.1% (w/v) SDS, and at 42°C twice in 0.2 × SSC, 0.1% (w/v) SDS.

For fruit RNA extraction, fruits at different developmental stages were collected, seeds were removed, and the pericarp tissue was immediately frozen in liquid nitrogen. RNA was isolated as described ( Rodríguez-Concepción & Gruissem 1999). For RNA blot analysis, 10 μg from each sample were fractionated by gel electrophoresis in 1% (w/v) agarose gels containing 2% formaldehyde, and blotted onto Hybond-N+ membranes (Amersham). Probes were made by labelling cDNA inserts with the Ready to Go kit (Pharmacia). A DXS probe was made from a 726 bp NdeI-SacII DXS cDNA fragment (corresponding to the predicted C-terminal region of the protein) excised from plasmid pLeDXS. The PSY1 cDNA used as a probe contained the entire coding region and most of the 3′ UTR, and was isolated from tomato fruit RNA by RT–PCR with the primers PSY1F (5′-ACCATGGTTTTCTTGCTCAG-3′) and PSY1R (5′-GTCTAGAAG- TCTCTCAAAGGAG-3′). The 25S rRNA probe was made as described ( Cunillera et al. 1997 ). Hybridization was performed at 68°C in ExpressHyb7 hybrization solution (Clontech) for 1 h. Washes were carried out twice at room temperature in 1 × SSC, 0.1% (w/v) SDS and once at 45°C in 1 × SSC, 0.1% (w/v) SDS. Biomax (Kodak) film was used for exposures.

Microbombardment assay

A 1.4 kb KpnI-NcoI fragment from the 5′ region of LeDXS1 cDNA was cloned into the same sites of a modified pGFP-MRC plasmid in which a KpnI site had been created next to the XhoI site ( Rodríguez-Concepción et al. 1999 ). In the resulting construct, the CaMV 35S promoter directed the expression of the fusion protein DXS-GFP, in which the first 430 amino-acid residues of tomato DXS were fused in frame to the N-terminus of a synthetic green fluorescent protein (sGFP). Subcellular localization of the DXS-GFP fusion protein in tomato leaf epidermis cells was carried out by microbombardment with tungsten particles (1.0 μm) coated with the corresponding plasmid DNA, using a Biolistic PDS-1000/He system (Biorad) at a pressure of 1100 psi. After 24 h at 22°C with continuous light, samples were examined directly with a Leica TCS 4D Confocal Laser Scanning Microscope (CSLM). Green fluorescence corresponding to the GFP fusion proteins was detected using a BP515-525 filter after excitation with blue light at 488 nm. Red autofluorescence from chlorophyll was detected using a LP590 filter after excitation with green light at 568 nm.

In situ hybridization

Tomato orange fruit tissue was fixed in a solution containing 2% (v/v) formaldehyde, 4% (v/v) acetic acid and 60% (v/v) ethanol at 4°C for 2 days. After an ethanol series, dehydrated samples were embedded in paraffin (Paraplast Plus). Sections (8 μm) were used for in situ hybridization with digoxigenin-labelled (Roche) DXS RNA probes as previously described ( Coen et al. 1990 ). A 399 bp 3′ClaI-BamHI LeDXS1 cDNA fragment cloned in pBluescript vector was used as a template to prepare the antisense probe with T3 RNA polymerase and 1 μg of plasmid DNA digested with ClaI. As a control, a sense probe was synthesized using T7 RNA polymerase and 1 μg of plasmid template linearized with XbaI. Specific activity of probes was checked by dot-blot analyses and working dilutions of 1 : 150 in 120 μl hybridization buffer were used. The hybridization signal was detected with NBT/BCIP (Roche) and the colorimetric reaction finally stopped with 10 m m Tris–HCl, 1 m m EDTA (pH 8) after 14 h incubation.

Chlorophyll and carotenoid analysis

Frozen powder (0.5 g) from the same fruit pericarp tissue used for RNA extraction was ground and mixed with 2 ml of methanol : water (1 : 1) followed by addition of 4 ml chloroform. After 20 min, samples were centrifuged at 3000 r.p.m. for 10 min at room temperature. The lower phase was recovered and evaporated. Dried samples were resuspended in 100 μl ethylacetate, and quantification was performed in acetone as described ( Lichtenthaler 1987).


We thank Dr S. Prat for the tomato cDNA library and Drs P. Bramley, S. Jenkins and W. Gruissem for the gift of tomato seeds. We also thank the staff in charge of the Serveis de Camps Experimentals of the Universitat de Barcelona for the care of plants. The help and advice of C. López and S. Ruíz (in situ hybridization) and S. Castel and L. Napal (microscopy), from the Serveis Cientificotècnics of the Universitat de Barcelona are greatly appreciated. We also thank the DNA Stock Center at ABRC, for the Arabidopsis H2A12 EST clone. This work was supported by grants from the European Commission DG XII Biotechnology Programme (Contract number BIO4-96 2077), Spanish Ministerio de Educación y Cultura (DGICT, PB96-0176), and Generalitat de Catalunya (CIRIT, 1997SGR-00088). L.M.L. received a Spanish MEC fellowship.


  1. Tomato DXS cDNA accession number AF143812.