Carotenoid Crystal Formation in Arabidopsis and Carrot Roots Caused by Increased Phytoene Synthase Protein Levels

Background As the first pathway-specific enzyme in carotenoid biosynthesis, phytoene synthase (PSY) is a prime regulatory target. This includes a number of biotechnological approaches that have successfully increased the carotenoid content in agronomically relevant non-green plant tissues through tissue-specific PSY overexpression. We investigated the differential effects of constitutive AtPSY overexpression in green and non-green cells of transgenic Arabidopsis lines. This revealed striking similarities to the situation found in orange carrot roots with respect to carotenoid amounts and sequestration mechanism. Methology/Principal Findings In Arabidopsis seedlings, carotenoid content remained unaffected by increased AtPSY levels although the protein was almost quantitatively imported into plastids, as shown by western blot analyses. In contrast, non-photosynthetic calli and roots overexpressing AtPSY accumulated carotenoids 10 and 100-fold above the corresponding wild-type tissues and contained 1800 and 500 µg carotenoids per g dry weight, respectively. This increase coincided with a change of the pattern of accumulated carotenoids, as xanthophylls decreased relative to β-carotene and carotene intermediates accumulated. As shown by polarization microscopy, carotenoids were found deposited in crystals, similar to crystalline-type chromoplasts of non-green tissues present in several other taxa. In fact, orange-colored carrots showed a similar situation with increased PSY protein as well as carotenoid levels and accumulation patterns whereas wild white-rooted carrots were similar to Arabidopsis wild type roots in this respect. Initiation of carotenoid crystal formation by increased PSY protein amounts was further confirmed by overexpressing crtB, a bacterial PSY gene, in white carrots, resulting in increased carotenoid amounts deposited in crystals. Conclusions The sequestration of carotenoids into crystals can be driven by the functional overexpression of one biosynthetic enzyme in non-green plastids not requiring a chromoplast developmental program as this does not exist in Arabidopsis. Thus, PSY expression plays a major, rate-limiting role in the transition from white to orange-colored carrots.


Introduction
Beta-carotene was first isolated from orange carrot roots, which are very high in this carotenoid [1]. The discovery was followed by the elucidation of numerous other carotenoids, and their number still continues to grow [2]. Carotenoids are isoprenoids synthesized by all photosynthetic organisms and some non-photosynthetic bacteria and fungi. In plants, carotenoids are synthesized in plastids, where they are constituents of light-harvesting complexes and photosynthetic reaction centers. They absorb energy from light and thus contribute to the dissipation of excess energy [3][4][5]. Carotenoids are also accumulated in chromoplasts where they act as visual attractants for pollinating insects and seed-distributing animals. In this plastid type, carotenoids are sequestered into suborganellar structures, which are classified as globular, crystalline, membranous, fibrillar and tubular types according to their architecture (reviewed in [6]).
In many cases the flux of carbon through the carotenoid biosynthetic pathway appears to be controlled by phytoene synthase (PSY), which catalyzes the head-to-head condensation of two geranylgeranyl diphosphate (GGPP) molecules to yield phytoene, the first committed reaction in carotenogenesis. GGPP is a substrate to several other enzymes and can thus enter a number of pathways, leading to the synthesis of tocopherols, chlorophylls, plastoquinones, phylloquinones, and gibberellins.
In chloroplasts, carotenoids and chlorophylls are needed in a defined stoichiometric ratio. The biosynthesis of both pigments, which share GGPP as a common substrate, is thought to occur in a strongly regulated and interdependent fashion. The regulation of PSY occurs at multiple levels: transcription of the PSY gene is lightregulated, while the presence of enzymatically inactive PSY within the prolamellar body of etiolated seedlings also indicates posttranslational control [11][12][13].
Most plants seem to contain more than one PSY gene, although Arabidopsis thaliana only has one. PSY paralogs can play specialized roles in chromoplast development, as found in developing tomato fruits [14][15][16][17]. Functional diversification of PSY was also found in rice and maize, where one of three PSYs is specifically involved in stress-induced ABA formation [18,19]. Furthermore, two PSY genes are found in carrots (Daucus carota; [20]), cassava (unpublished), and are also found annotated for poplar and grapevine.
Because of PSY's key role at the entry point into carotenogenesis, it is not surprising that biotechnological approaches to increase carotenoid levels aim at increasing its expression. However, overexpression of PSY can affect photosynthetically active and non-active tissues in different ways. Tomato plants overexpressing tomato PSY1 constitutively in all tissues, for instance, showed only slightly increased leaf carotenoid levels [21], while accumulation of total carotenoids during fruit ripening was much more pronounced [22]. Most recent transformation experiments to increase total carotenoid content in non-green tissues used tissue-specific promoters to circumvent possible adverse affects, such as a dwarf phenotype, as has been observed in some transgenic lines when using the constitutive, non-tissuespecific CaMV35S promoter [21]. Tissue-specific promoters were used to raise the carotenoid content in Arabidopsis seeds [23], canola seeds [24], potato tubers [25], and rice endosperm [26]. In the latter the choice of the PSY gene was crucial to attain high carotenoid levels [27], while increased expression of the phytoene desaturase CrtI was not effective [28]. However, PSY may not be rate-limiting in all tissues, as exemplified in transgenic potato tubers, where overexpressed lycopene b-cyclase had a dominating effect [25].
In the present work, we set out to systematically investigate the impact of PSY on carotenoid formation in green and non-green tissues, using transgenic A. thaliana seedlings, roots and calli, as well as transgenic D. carota storage roots. Analysis of RNA and protein expression, combined with quantitative carotenoid analysis, revealed profound differences in the responsiveness of carotenogenesis to increased PSY protein levels. PSY turned out to be ratelimiting in non-green tissues but not in green tissues. We propose that this difference is due to the capability of non-green plastids but not chloroplasts to form carotenoid crystals. In non-green plastids crystal formation is simply driven by enhanced carotenoid biosynthesis, and does not require the activation of intrinsic developmental programs to drive chromoplast development.

PSY expression is rate-limiting in calli but not in green seedlings
We investigated the effects of PSY overexpression in green and non-green tissues, seedlings and calli of Arabidopsis lines expressing the PSY ORF from Arabidopsis (AtPSY) under the control of the constitutive, non-tissue-specific CaMV35S promoter (35S::AtPSY). Seedlings of three 35S::AtPSY lines grown for two weeks were phenotypically indistinguishable from wild-type seedlings. Caroten-oids as well as AtPSY transcript levels were determined by HPLC and Real-Time RT-PCR, respectively. Although AtPSY transcripts increased 14 to 17-fold, carotenoid levels and accumulation patterns as well as chlorophylls remained unchanged as compared to wildtype seedlings ( Figure 1A and D, left).
Calli were induced from germinating seeds, leaves, stems and roots of 35S::AtPSY lines as well as from the wild type. Since all the transgenic calli responded almost equally with respect to carotenoid accumulation (data not shown), seed-derived calli (SDC) were chosen for all further experimentation. These calli were generated by germinating seeds on callus-inducing medium for 5 days under long-day conditions, followed by 16 days in darkness. During etiolation, the transgenic lines developed intensely orange-colored calli, while those of the wild type were much less colored (Figure 2A). Quantitative HPLC revealed an approximately 10-fold increase in carotenoid levels compared to the wild type ( Figure 1A, right), together with changes in carotenoid accumulation patterns. While wild-type SDC accumulated mainly xanthophylls and only trace amounts of b-carotene, the latter was the main constituent in 35S::AtPSY SDC, followed by xanthophylls and carotene intermediates, like phytoene and phytofluene.
PSY protein levels correlate with increased carotenoid (bcarotene) levels in seed-derived calli but not in leaves We compared the response of leaves and SDC to AtPSY overexpression by determining the AtPSY protein levels in both tissues using affinity-purified anti-AtPSY antibodies.
Three AtPSY bands were observed on western blots with seedlings as well as with SDC ( Figure 1B and 3A). The 42 kD band corresponds to the mature protein after import into plastids. It co-migrates with the radiolabelled in-vitro translation product lacking the transit peptide on PAGE gels [29]. Similar internal standards also allowed relating the uppermost band (48 kD) to the unprocessed cytoplasmatic protein (data not shown). An intermediate form (ca. 45 kD) was observed in most cases but was not further investigated.
Both tissue types imported the overexpressed protein almost quantitatively into plastids; the concentration of unprocessed precursors did not rise. Therefore, the translocation step is not rate limiting and is thus not the cause for the differential response.
In fact, both transcript and protein levels were found to be substantially higher in green leaves of the transgenic lines as compared to the wild type ( Figure 1B). This means that the high carotenoid content in seedlings prevents the overexpressed AtPSY from raising carotenoid levels further. In the wild type, even low levels of AtPSY 'saturate' the system. This contrasts with the situation in calli, where AtPSY overexpression is capable of very substantially increasing the low carotenoid levels present in the wild type. Therefore, in green tissues, a relatively small amount of AtPSY is sufficient to produce high carotenoid levels while in SDC a much higher amount of AtPSY is needed to drive carotenoid levels up to seedling levels. AtPSY expression is thus rate-limiting in photosynthetically inactive calli but not in green seedlings.
Rerouting of the carotenoid biosynthetic pathway in 35S::AtPSY SDC leads to b-carotene accumulation We investigated carotenoid and transcript levels of carotenoid pathway genes during the process of SDC formation from wildtype plants and siblings of one 35S::AtPSY line ( Figure 3B, C). Data were collected after growth in the light and during subsequent etiolation for 7, 12 and 16 days. This revealed that during the light period, SDC of the 35S::AtPSY line had accumulated more than double as much carotenoids as wild-type SDC. Although bcarotene content increased about 3-fold compared with wild-type SDC at this point, it represented only 20% of total carotenoids, while xanthophylls contributed about 66%. Phytoene and phytofluene, which were absent in wild-type calli, were already present at this stage. During the subsequent 16 days of growth in the dark, xanthophylls decreased to about 25% of total carotenoid content, while b-carotene and to a minor extent also a-carotene, increased strongly, contributing to ca. 50% (or 650 mg/g dry weight) of total carotenoid content after 16 days. Remarkably, this pattern shift did not affect total carotenoid content, which remained as high as before. In addition to small bead-like, orange colored structures (C), the formation of large bar-shaped structures was observed in SDC protoplasts by light microscopy (D). As confirmed by polarization microscopy, these structures represent carotenoid crystals (E); an overlay of D and E is shown in F. A protoplast from wild-type SDC is shown in B. c, carotenoid crystals; v, vacuole. doi:10.1371/journal.pone.0006373.g002 In contrast, under the same regime, total carotenoids in wildtype SDC decreased by about 75% due to losses across all carotenoids determined. With the disappearance of chlorophylls ( Figure 3B) carotenoids became visible ( Figure 2A).
Hydroxylation of b-carotene is strongly downregulated under dark conditions in both wild type and transgenics, as revealed by Real-Time RT-PCR ( Figure 3C). Similarly, the transcripts of e-cyclase fell to about 10% of the light levels, while b-cyclase levels were less affected (ca. 50%), and phytoene desaturase, f-carotene desaturase and carotenoid isomerase decreased to about 30% (data not shown). In 35S::AtPSY SDC, expression of the transgene was 20 times higher than in the wild type, while transcript levels of the other carotenoid pathway genes decreased to the same low levels as wild-type SDC ( Figure 3C, right). This indicates that the strong b-carotene accumulation results from AtPSY overexpression accompanied by rerouting of the pathway towards the b-branch. The shift is caused by expression changes occurring during the transition to non-green callus cells and not by changes induced by AtPSY overexpression.
The decrease of carotenoid levels reflects carotenoid turnover. One interpretation is that decreasing levels of AtPSY and other carotenoid biosynthetic enzymes in wild-type SDC can no longer compensate for carotenoid turnover unless the rate of biosynthesis is increased by AtPSY overexpression.
In SDC expressing AtPSY constitutively b-carotene is deposited in crystals In chromoplasts, carotenoids are deposited in specialized structures such as plastoglobules, fibrils, crystals or proliferated membranes. Since Arabidopsis cannot form chromoplasts, we set out to determine the mode of carotenoid deposition in plastids of dark-grown 35S::AtPSY SDC accumulating ca. 1800 mg/g dry weight of these compounds, which is well in the range of crystal formation in chromoplasts. Carrots, for instance, contain 2000 to 4000 mg carotenoids/g dry weight [30] and tomatoes accumulate ca. 2000 mg/g dry weight [22].
Microscopy revealed that wild-type protoplasts contained unpigmented proplastids ( Figure 2B) while orange-colored particles were present in small numbers (1-5 per cell) in 35S::AtPSY protoplasts ( Figure 2C). As shown in Figure 2 (D-F), polarization microscopy revealed that the carotenoid particles in 35S::AtPSY SDC were birefringent, which is distinctive of their crystal nature.

Carotenoid crystal formation in roots
Given that SDC reflect the situation in non-green plant cells, roots should respond similarly to PSY overexpression. In fact, as shown in Figure 4C, wild-type roots accumulated only trace amounts of carotenoids while those of 35S::AtPSY lines accumulated ca. 100 times more. In addition, the carotenoid composition pattern was very similar to SDC, consisting mainly of b-carotene and the intermediates phytoene and phytofluene ( Figure 4A). Just like in callus tissues, carotenoid increase in roots correlated with increased AtPSY protein, which was almost undetectable in wildtype roots ( Figure 4D). Real-Time RT-PCR showed that carotenoid accumulation correlated with increased expression levels of AtPSY ( Figure 4E) but did not affect expression levels of other carotenoid genes (AtPDS, AtZDS, AtbLCY, AteLCY, AtbHYD1, AtbHYD2; Figure S1). Polarization microscopy of protoplasts prepared from roots of 35S::AtPSY lines revealed an identical mode of carotenoid sequestration into birefringent carotenoid particles ( Figure S2). Thus, roots mirror the situation in SDC almost perfectly.
Carotenoid composition patterns and PSY expression in roots of cultivated carrots and Arabidopsis 35S::AtPSY lines are similar The storage root of orange carrots (Daucus carota) represents the most popular b-carotene accumulating tissue. The molecular basis of the massive carotenoid accumulation in carrots is still unclear. Path analysis on the offspring of crosses between white and orange-rooted cultivars suggested that in white roots the capability for phytoene formation was the limiting factor in carotenoid accumulation [31].
To test the hypothesis we analyzed a white-rooted wild carrot (D. carota subsp. carota, Queen Anne's Lace, QAL) and a cultivated variety with white roots (D. carota subsp. sativus var. Weisse Küttiger, KUT), as well as two orange-rooted cultivars (D. carota subsp. sativus, var. Chatenay Red Cored, CRC, and Nantaise 2, NAN). As shown in Figure 5 (A, B), white-rooted carrots and Arabidopsis wild-type roots were similarly low in carotenoids (6 mg/g DW for Arabidopsis wild type, 5 for QAL, and 14 for KUT), which consisted mostly of xanthophylls and only a minor proportion of bcarotene. Just like in Arabidopsis roots, carotenoid intermediates and a-carotene were virtually absent. The carotenoid levels in cultivated carrot roots (CRC, NAN) of 9-week old plants were as high as in SDC or roots overexpressing AtPSY ( Figure 5B; compare Figure 4C and 1A). They accumulated around 800 mg/g total carotenoids with about 42% b-carotene, and were characterized by the presence of the pathway intermediates phytoene and phytofluene. The only difference was a-carotene, which accounted for about 20% of total carotenoids in carrots but only 5% in roots of Arabidopsis 35S::AtPSY lines.
The western blot analyses again mirrored the situation met in roots of Arabidopsis 35S::AtPSY lines. While in white-rooted carrots PSY was below the detection limit, roots of the orangerooted varieties showed a strong PSY signal ( Figure 5C). We cloned the PSY cDNAs from QAL and CRC to make sure that antibody specificity and enzymatic activity were not affecting the interpretation of the results. Two PSY genes are present in the carrot genome (DcPSY1, DcPSY2; [20]), and they were identical in QAL and CRC.
We investigated whether the increase of PSY protein correlated with PSY transcript levels by Real-Time RT-PCR ( Figure 5D, E). While transcript levels of DcPSY2 were almost equivalent in the white-rooted carrots QAL and KUT, DcPSY1 transcript levels were about three times higher in KUT than in QAL. In the orange-rooted cultivars CRC and NAN, DcPSY1 transcript levels were 17 and 14-fold higher than in QAL roots, while DcPSY2 transcript levels were 8 and 5-fold higher, respectively.

Root-specific PSY expression in white carrots leads to yellow-colored roots
Seemingly, the correlation of PSY protein abundance and carotenoid levels in non-photosynthetic tissues of Arabidopsis and carrot roots represents a common theme. Differential expression of other carotenogenic genes does not seem to be required to drive carotenoid overproduction, as can be concluded from comparative transcriptional analysis conducted with several carrot cultivars differing in root color [32].
Consequently, increased PSY in a white-rooted carrot background should produce colored carrot roots. To prove the point we generated transgenic QAL lines overexpressing the bacterial PSY crtB under the control of a storage root-specific promoter cloned from yam (Dioscorea spec.; Arango et al., unpublished; yam::crtB). In fact, roots of about 8-week old heterozygous T1 transformants exhibited intense yellow color (see Figure 6A). However, the carotenoid composition pattern showed some differences compared to orange carrot roots. b-carotene accounted for only 10% of total carotenoids in most lines, with only one line (yam::crtB#3) containing about 30% b-carotene ( Figure 6B, C). Carotene intermediates, mainly phytoene, but also phytofluene, fcarotene, and lycopene predominated in roots of all lines, and acarotene was absent. The pattern remained constant during prolonged growth while carotenoid levels increased from about 180 mg/g DW at 8 weeks to 400 mg/g DW at 16 weeks.
Even though carotenoid content was lower in roots of yam::crtB lines than in orange carrot roots, carotenoid sequestration into crystals followed the same pattern as in roots of cultivated carrots ( Figure 6D). From this we conclude that only increased PSY protein levels are required to attain both high carotenoid levels as well as crystal formation, while carotenoid composition patterns may vary.

AtPSY expression in green and non-green tissues: Effects on carotenoid formation
The expression of AtPSY results -as we show-in a differential response in green and non-green plant cells with respect to carotenoid formation. While seedlings were essentially unaffected, carotenoid content in non-green SDC and roots increased by 10fold and ca. 100-fold over wild-type levels, respectively (Figure 1,  4). In absolute terms, chloroplasts can synthesize very large amounts of carotenoids (2000 mg/g) at low levels of AtPSY expression. Non-green plastids accumulate carotenoids less effectively, requiring strongly elevated AtPSY levels to reach levels of 500 (roots) to 1800 (SDC) mg/g dry weight.
This differential response is not due to limitations in the import machinery, since AtPSY pre-proteins did not accumulate in any of the tissues examined. This finding points towards fundamental differences in carotenoid storage mechanisms. In chloroplasts, carotenoid is generally found in protein-bound form, mainly in light-harvesting complexes accommodating carotenoids and chlorophylls [33,34]. Protein binding of carotenoids may impact on biosynthetic capacity directly or indirectly and provide a sequestering site for carotenoids, allowing highest levels of accumulation in plastid membranes without interfering with their physicochemical properties. Sequestration into light-harvesting complexes most probably also affects carotenoid biosynthesis by shifting the balance of pathway products while at the same time protecting the bound carotenoids from being catabolized. Moreover, PSY is only active when bound to membranes, as shown for daffodil chromoplasts [29,35]. Accordingly, PSY was essentially inactive in prolamellar bodies of etioplasts while its enzymatic activity was elicited upon the formation of thylakoid membranes [11].
Carotenoid catabolism, which is initiated by carotenoid oxygenases, can influence carotenoid accumulation profoundly [36][37][38], and may be comparatively higher in non-green cells. Here, increased PSY levels or enzymatic activity might be able to override carotenoid turnover, leading to elevated steady-state carotenoid concentrations. It is conceivable that carotenoid stability may be very significantly increased when carotenoids accumulate in suborganellar structures, such as fibrils [39] or crystals, in the absence of functional carotenoid-binding photosynthetic proteins.

Increased PSY levels lead to carotenoid crystal formation in non-green cells
Carotenoid crystals form in SDC and roots from AtPSY overexpressing lines. During the formation of SDC, crystal buildup is supported by reduced carotenoid diversity through the concerted downregulation of b-hydroxylases and e-cyclase expression, and to a much lesser extent b-cyclase (Figure 3). This process favors b-carotene formation until enough is enriched at the site of formation, causing the b-carotene to leave the lipophilic phase of the membrane and form crystals. Roots of transgenic lines behaved very similarly. In both cases, carotenoid accumulation is due to increased levels of the rate-limiting enzyme PSY although without impacting on the expression of other carotenoid genes, which remained unchanged in SDC and roots of 35S::AtPSY lines relative to the corresponding wild-type tissues.  A similar mode of carotenoid deposition occurs during the conversion of chloroplasts to chromoplasts in tomato fruits, for instance, which is accompanied by the fruit-specific upregulation of tomato PSY1. Here, due to concurrent downregulation of e and b-cyclases [15,40], lycopene accumulates in crystalline bodies within the chromoplast [41,42]; they possibly show up as small, bead-like structures at an early stage of formation [43]. Similar structures were also observed in SDC and root protoplasts of AtPSY overexpressing lines (Figure 2, Figure S2). The formation of b-carotene crystals in these cells seems to be solely driven by the overexpression of AtPSY, not requiring additional components of a chloroplast or chromoplast developmental program that does not exist in Arabidopsis.

Similarities between AtPSY overexpressing tissues and the cauliflower OR mutant
The accumulation of large amounts of b-carotene in the form of crystals is strikingly similar to the cauliflower OR mutant [44]. However, the OR gene is not biosynthetic but encodes a plastidlocalized protein containing a DnaJ Cys-rich domain, making the OR protein a likely constituent of the plastidic chaperone system [45]. Indicative for the functionality of the underlying principle, potatoes transformed with the mutated OR gene showed increased b-carotene accumulation in tubers [46]. Also, crystal formation is maintained in cauliflower OR calli [47,48].
It is interesting to note that a very similar effect, including altered plastid morphology, can be achieved in non-green Arabidopsis cells by increasing the carotenoid biosynthetic capacity. Carotenoid biosynthetic genes are not upregulated in the OR mutant [44], and the phytoene biosynthetic capacity is not increased, as shown using the phytoene desaturase inhibitor norflurazone [47]. In both cases b-carotene crystal formation is the common principle, although the underlying mechanisms may be different. Yet, the resulting effect, namely deposition of b-carotene in a form that is less accessible to catabolic reactions, might be the same. Alternatively, sequestration of b-carotene into crystals may shift local chemical equilibria, thus enhancing the flux of carbon through the pathway. Targeted metabolic profiling of carotenoid degradation products should allow distinguishing between these possibilities.

Why are carrots orange?
While tomato chromoplasts develop from chloroplasts during fruit maturation, chromoplasts of carrot roots emerge from proplastids with low carotenoid content [49]. Despite these differences in origin, in both cases carotenoids are sequestered as crystals [50,51].
As we were able to show, PSY overexpression can lead to carotenoid accumulation in Arabidopsis roots, matching orange carrots almost perfectly, both in qualitative and quantitative terms. The presence of high PSY levels in orange carrot roots, while undetectable in white-rooted cultivars, provides further support for its proposed key role. Furthermore -again similar to Arabidopsis roots-increased PSY seems sufficient to explain heightened carotenoid accumulation, since transcript levels of other carotenogenic genes do not increase in orange as opposed to white-rooted carrots [32]. This is in agreement with the conclusions drawn from pathway analyses performed by Santos et al. suggesting that the primary difference between white and orange carrots resides in an enzymatic change blocking carotenoid accumulation early in the pathway [31]. Our results suggest that in white-rooted carrots low PSY maintains carotenoid biosynthesis at a low level while in orange carrots high PSY is responsible for the very large amount of accumulating carotenoids.
To confirm the hypothesis, white carrots were transformed to overexpress PSY (CrtB) in the roots. As expected, their roots showed strongly increased carotenoid levels. However, in contrast to Arabidopsis roots and SDC of 35S::AtPSY lines the carotenoid composition pattern was characterized by a higher proportion of carotene intermediates like phytoene, phytofluene and f-carotene, pointing towards rate-limiting carotene desaturation. Similarly, lycopene which was absent in orange-rooted cultivars and lower levels of b-carotene accumulation indicated lower cyclization capacity. Furthermore, the high proportion of a-carotene in orange-rooted cultivars was neither found in roots of the transgenic carrots nor in non-green tissues of AtPSY overexpressing Arabidopsis lines. However, such pathway differences are known in carrot germplasm [52]. Hence, we propose that PSY protein overexpression is the main driving force behind the orange phenotype. Orange carrots are believed to have been domesticated from white cultivars in the 16 th century [53], and it seems plausible that the orange phenotype we all know originated from one single mutation event.
SDC were generated essentially as described [56]. Ten milligrams of surface-sterilized seeds were plated onto petri dishes (145 mm diameter) containing SDC medium Seedlings were obtained by plating 7.5 mg surface-sterilized seeds onto MS medium. Petri plates were sealed and the seedlings grown under long-day conditions for 14 days. Seedlings, SDC and roots were ground in liquid nitrogen immediately after harvest and stored at 270uC for further analysis.

Plant material and growth
Carrot seeds were obtained from the following sources: Queen Anne's Lace, Richters Herbs (Goodwood, Canada); Küttiger, Dreschflegel Saatgut (Witzenhausen, Germany), Chatenay Red Cored, B&T World Seeds (Aigues-Vives, France), Nantaise 2, Freya (Solingen, Germany). Carrots were grown in soil under long-day conditions. Roots of 9-week old carrot plants were removed from the soil, ground in liquid nitrogen and stored at 270uC for further analysis.

Carrot transformation
Transgenic QAL lines were transformed with pCAMBIA1305.2yam::crtB as described [57]. Transformation was carried out using pieces of cotyledons and hypocotyls of one-week-old light-grown QAL seedlings and the A. tumefaciens strain GV3101. After co-cultivation, vancomycin (200 mg/l) was used instead of claforan to arrest bacterial growth. For selection of transgenic calli, hygromycin (10 mg/ml) was used instead of geneticin used in the original protocol.

Generation of Arabidopsis roots
Roots of Arabidopsis wild type and AtPSY overexpressing lines were generated as described [58]. Seed were sterilized, sown on MS medium ( [59]; MS salts, Gamborg's B5 vitamins, 1% [w/v] sucrose, and 0.25% [w/v] Phytagel), stratified for 4 days and grown for 7 days at 26uC under long-day conditions. Seedlings were transferred to B5 liquid medium (Gamborg's B5 salts, Gamborg's B5 vitamins, and 2% [w/v] glucose) and incubated for 14 days with shaking at 120 rpm under long-day conditions at 26uC. Plantlets were removed from the liquid, green plant parts were carefully removed and roots immediately ground in liquid nitrogen and stored at 270uC for further analysis.

Protoplast generation
Protoplasts from Arabidopsis SDC and roots were generated as described [60]. Light microscopy was performed using an Axioskop 2 microscope (Carl Zeiss, Jena, Germany) equipped with the Axiovision 4.6.3 software.

TaqMan Real-Time RT-PCR Assay
Total RNA was isolated using the plant RNA purification reagent (Invitrogen, Heidelberg, Germany). RNA purification, DNase I digestion and Real-Time RT-PCR assays were performed as described [19]. Primers and 6FAM-labeled probes were designed using Primer Express software (Applied Biosystems, Darmstadt, Germany) and are given in Table S1.

Carotenoid Extraction and Quantification
Carotenoids were extracted using lyophilized plant materials and analyzed by HPLC as described [19]. The following approximate dry weights were used: 30 mg SDC; 5 mg seedlings; 70 mg Arabidopsis roots; 100 mg white carrot roots; 20 mg orange carrot roots.

Generation of anti-AtPSY antibodies
AtPSY cDNA was amplified without the transit peptide coding sequence by PCR using mutagenized primers (position 574 according to accession no NM_121729). The PCR product was subcloned into the vector pCOLDI (Takara-Clontech, Heidelberg, Germany) to yield an N-terminal 6*His tag fusion. 6*His-AtPSYdTP inclusion bodies were isolated as described [19], dissolved in 10 ml 6 M Gu-HCl, 0.1 M Na 2 HPO 4 (pH 7.15), the protein purified using HIS-Select Spin Columns (Sigma, Taufkirchen, Germany) and dialyzed against water. The protein precipitate was dissolved in SDS sample buffer, subjected to SDS-PAGE and gel slices containing 6*His-AtPSYdTP were used to immunize rabbits.

Protein extraction and Western blot analysis
Proteins were extracted with phenol as described [61]. Protein concentration was determined with Bio-Rad protein assay. After SDS-PAGE, blotting onto PVDF membranes (Carl Roth, Karlsruhe, Germany) and treatment with blocking solution (TBS containing 5% [w/v] milk powder), membranes were incubated with affinity-purified anti-AtPSY antibodies in PBS containing 0.1% (v/v) Tween-20 and 1% (w/v) milk powder for 2 h. For detection, the ECL system (GE Healthcare, München, Germany) was used. Western blots were stripped and reprobed with antiactin antibodies (Sigma). Signal quantification was done using the Quantity One software (Bio-Rad). Figure S1 Expression levels of carotenogenic enzymes in 35S::AtPSY lines. Expression levels of carotenogenic enzymes were determined by Real-Time RT-PCR using total RNA isolated from roots of wild type and 35S::AtPSY lines#12 and #22. Transcript levels were normalized to 18S rRNA level of the corresponding sample and expressed relative to the content in the wild type. PDS, phytoene desaturase, ZDS, fcarotene desaturase, eLCY, lycopene e-cyclase; bLCY, lycopene bcyclase; bHYD1/2, b-carotene hydroxylase 1/2.