Functional Characterization of the Xanthophyllomyces dendrorhous Farnesyl Pyrophosphate Synthase and Geranylgeranyl Pyrophosphate Synthase Encoding Genes That Are Involved in the Synthesis of Isoprenoid Precursors

The yeast Xanthophyllomyces dendrorhous synthesizes the carotenoid astaxanthin, which has applications in biotechnology because of its antioxidant and pigmentation properties. However, wild-type strains produce too low amounts of carotenoids to be industrially competitive. Considering this background, it is indispensable to understand how the synthesis of astaxanthin is controlled and regulated in this yeast. In this work, the steps leading to the synthesis of the carotenoid precursor geranylgeranyl pyrophosphate (GGPP, C20) in X. dendrorhous from isopentenyl pyrophosphate (IPP, C5) and dimethylallyl pyrophosphate (DMAPP, C5) was characterized. Two prenyl transferase encoding genes, FPS and crtE, were expressed in E. coli. The enzymatic assays using recombinant E. coli protein extracts demonstrated that FPS and crtE encode a farnesyl pyrophosphate (FPP, C15) synthase and a GGPP-synthase, respectively. X. dendrorhous FPP-synthase produces geranyl pyrophosphate (GPP, C10) from IPP and DMAPP and FPP from IPP and GPP, while the X. dendrorhous GGPP-synthase utilizes only FPP and IPP as substrates to produce GGPP. Additionally, the FPS and crtE genes were over-expressed in X. dendrorhous, resulting in an increase of the total carotenoid production. Because the parental strain is diploid, the deletion of one of the alleles of these genes did not affect the total carotenoid production, but the composition was significantly altered. These results suggest that the over-expression of these genes might provoke a higher carbon flux towards carotenogenesis, most likely involving an earlier formation of a carotenogenic enzyme complex. Conversely, the lower carbon flux towards carotenogenesis in the deletion mutants might delay or lead to a partial formation of a carotenogenic enzyme complex, which could explain the accumulation of astaxanthin carotenoid precursors in these mutants. In conclusion, the FPS and the crtE genes represent good candidates to manipulate to favor carotenoid biosynthesis in X. dendrorhous.


Introduction
Isoprenoids, also known as terpenoids, are one of the largest and most diverse families of natural products, consisting of over 40,000 structurally different compounds isolated from animals, plants and microorganisms [1]. Many biological functions have been attributed to them, including their roles as defensive and photoprotective agents, important components of cell membranes and both pigments and reproductive hormones, making them very attractive biotechnological metabolites [2]. These roles and structural diversity have attracted the attention of researchers from various fields of science to study isoprenoids function, biosynthesis and synthesis regulation.
Despite the enormous structural diversity of isoprenoids, their biosynthesis shares the same initial steps. These steps correspond to the sequential union of a C 5 -isoprene unit by prenyl transferase enzymes [3]. The basic isoprene building blocks are the isopentenyl pyrophosphate (IPP) and its isomer dimethylallyl pyrophosphate (DMAPP). Generally, in eukaryotes and archaea, IPP derives from the mevalonate pathway [4,5], while in most bacteria and plant plastids IPP is synthesized via the 2-C-methyl-D-erythritol 4-phosphate, DOXP/MEP, non-mevalonate pathway [6]. In the first step of isoprenoid biosynthesis (Fig. 1), IPP is converted into DMAPP, a reaction that is catalyzed by the isopentenyl pyrophosphate isomerase (encoded by idi). Then, a molecule of IPP is condensed with a molecule of DMAPP to form geranyl pyrophosphate (GPP; C 10 ), the precursor of monoterpenes. The addition of a second IPP unit to GPP generates farnesyl pyrophosphate (FPP; C 15 ), which is precursor of sesquiterpenes. The further addition of IPP to FPP results in geranylgeranyl pyrophosphate (GGPP; C 20 ), the precursor of diterpenes. The later condensation of two molecules of GGPP yields phytoene, and that of two units of FPP gives squalene, which are the precursors of carotenoids and triterpenes (including sterols), respectively. In the synthesis of GGPP from DMAPP (DMAPP to GPP to FPP to GGPP), three different systems have been described, depending on the number of enzymes involved. In the first system, only one enzyme, a GGPP-synthase encoded by the crtE gene, produces GGPP from DMAPP [7]. In the second system, two enzymes are involved: a FPP-synthase (FPS gene) that forms FPP, followed by the GGPP-synthase to form GGPP [8].
The third system is a hybrid, where the first two systems act in parallel to give GGPP [9].
Carotenoids are natural pigments, fulfilling important physiological roles in a wide range of organisms. More than 750 chemical structures have been described to date [10], and plants, algae and some fungi and bacteria synthesize carotenoids. In general, animals are unable to biosynthesize carotenoids de novo, so they must obtain them through their diet [11]. Carotenoids are attractive from an industrial point of view because they are widely used as food colorants, antioxidants and nutraceutical agents [12].
Among the carotenoids, astaxanthin (3,39-dihydroxy-b,b-carotene-4-49-dione) stands out for its commercial potential, as it is currently being used as an antioxidant and as feed additive in aquaculture for salmon fish pigmentation [13,14]. The biosynthesis of astaxanthin is limited to only a few organisms, including the basidiomycete yeast Xanthophyllomyces dendrorhous [15].
The astaxanthin biosynthesis in X. dendrorhous has been investigated, and the genes involved in the synthesis of astaxanthin from phytoene have been described. However, the early steps of carotenogenesis in this yeast are less known.
The X. dendrorhous idi [16] and crtE [17] genes have been described, but no FPS gene has been formally reported. This is interesting, as several metabolic engineering strategies have been attempted to increase carotenoid production in X. dendrorhous. To increase carotenoid precursors, the idi gene was over-expressed. However, that over-expression decreased the total amount of carotenoids [18]. Conversely, when the crtE gene was overexpressed, there was only a slight increase in the total carotenoid content [19]. An explanation for this result could be that the GGPP-synthase activity in X. dendrorhous is limited by the amount of FPP, if this is its only substrate.
In this work, we report the isolation of the FPS gene from X. dendrorhous. Through functional studies, we conclude that FPS is involved in the synthesis of FPP and that its gene product uses IPP and DMAPP, or IPP and GPP, as substrates, while the GGPPsynthase utilizes exclusively IPP and FPP as substrates to produce GGPP.

Microorganisms, growth conditions and enzymes
All strains used and/or created in this work are listed in Table 1. The wild-type UCD 67-385 (ATCC 24230) X. dendrorhous strain was used for the genomic and cDNA library construction [17] and as the parental strain for genetic modifications. The yeast strains were grown at 22uC with constant agitation in YM rich medium (1% glucose, 0.3% yeast extract, 0.3% malt extract and 0.5% peptone). Yeast transformants were selected on YM-agar plates (1.5% agar) that were supplemented with 12.5 mg/mL hygromycin-B.
E. coli DH-5alpha was used for plasmid propagation and was grown at 37uC in Luria-Bertani (LB) medium, supplemented with 100 mg/mL ampicillin for plasmid selection and 20 mg/mL of Xgal (5-bromo-4chloro-3-indolyl-b-D-galactopyranoside) for recombinant clone selection by blue-white screening [20]. When necessary, recombinant clones were selected by direct colony PCR with a comprehensive set of primers. For the E. coli heterologous expression of the FPS and crtE genes, the BL21 strain was used and grown at 37uC in LB medium supplemented with 50 mg/mL kanamycin.
Taq DNA polymerase (pol), restriction enzymes, Klenow polymerase and M-MLV reverse transcriptase were purchased from Promega, and the Pfu DNA polymerase was purchased from Invitrogen.

Nucleic acids extraction
DNA extraction was performed from protoplasts of X. dendrorhous according to [21], and total RNA extraction was performed according to a modified protocol of Chomczynski and Sacchi [22,23]. Briefly, cell pellets were suspended in 200 mL of lysis buffer (0.5 M Sodium acetate pH 5.5, 5% SDS, 1 mM EDTA) and broken through mechanical rupture with 0.5 mm glass beads (BioSpec) and shaking in a mini bead beater-16 (BioSpec) for 1 min. Then, 800 mL of Tri-Reagent (Ambion) was . The three proposed model systems to achieve the synthesis of GGPP are shown: 1) first system, a FPP-and a GGPP-synthase are involved and act sequentially, 2) second system, only one GGPP synthase is involved and 3) third system, participation of the first and second system simultaneously. doi:10.1371/journal.pone.0096626.g001 added, followed by shaking in the bead beater for 1 min and incubation for 10 min at room temperature. Then, 200 mL of chloroform was added, followed by mixing and incubation for 6 min and centrifugation for 10 min at 14,0006g. RNA was extracted from the recovered aqueous phase by precipitation with 1 volume of isopropanol and 0.5 volume of precipitation buffer (1.2 M NaCl, 0.8 M sodium citrate) for 1 h at 220uC. The RNA was washed with 70% ethanol, suspended in RNase-free H 2 O and quantified spectrophotometrically at 260 nm according to [20] in a V-630 UV-Vis Spectrophotometer from JASCO.

DNA amplification and sequence analyses
The oligonucleotides that were designed and used in this study were synthesized by Integrated DNA Technologies and are listed in Table S1. The DNA amplification reactions were performed in a final volume of 25 mL containing 2 U of Taq DNA pol, 2.5 mL of 10X Taq buffer, 0.5 mL of 10 mM dNTPs, 1 mL of 50 mM MgCl 2 , 1 mL of 25 mM of each primer and 10-20 ng of template DNA. In general, the PCR reactions were performed in a 2720 Applied Biosystems thermal cycler with the following program: initial denaturation at 95uC for 3 min; 35 cycles of denaturation at 94uC for 30 s, annealing at 55uC for 30 s, synthesis at 72uC for 3 min and a final extension step at 72uC for 10 min. The samples were kept at 4uC until analyzed by 0.8% agarose gel electrophoresis in TAE buffer containing 0.5 mg/mL ethidium bromide [20]. DNA for sequencing or plasmid construction was purified from gels using the glass milk method [24]. The nucleotide sequences were obtained from an ABI 3100 Avant genetic analyzer using the BigDye terminator v3.1 kit (Applied Biosystems). DNA sequences were analyzed using Geneious 6.0.4, CLUSTAL W 1.8 and programs that are available at the NCBI web site.

Single-strand DNA synthesis and quantitative RT-PCR (RT-qPCR)
The cDNA was synthesized using M-MLV reverse transcriptase (Invitrogen) with 5 mg of total RNA in a final volume of 20 mL, according to the manufacturer's protocol. The relative transcript level analyses for each gene were performed in an Mx3000P quantitative PCR system (Stratagene) using 1 mL of the reverse transcription reaction, 0.25 mM of each primer (Table S1) and 10 mL of the SensiMix SYBR Green I (Quantace) kit in a final volume of 20 mL. The pairs of primers used had efficiencies greater than 95%, as determined by standard curves with a correlation coefficient of R2$0.996. The Ct values obtained were normalized to the respective value of the X. dendrorhous actin gene [Genbank: X89898.1] [25] and were later expressed as a function of the control conditions using the DDCt algorithm [26].
Obtaining the genes controlling the GGPP synthesis from IPP in X. dendrorhous According to the Kyoto Encyclopedia of Genes and Genomes [27], three enzymatic activities, EC 2.5.1.1 (dimethylallyl transtransferase), EC 2.5.1.10 (geranyl transtransferase) and EC 2.5.1.29 (GGPP-synthase) are involved in the synthesis of GGPP from IPP and DMAPP. In S. cerevisiae, the dimethylallyl-and the geranyl-transtransferase activities are performed by the FPPsynthase encoded by the ERG20 gene [28], and the BTS1 gene, which is homologous to the crtE gene described in other organisms, encodes the GGPP-synthase activity [29]. Only the X. dendrorhous crtE gene has been previously described [17]. To find the X. dendrorhous FPS gene, a BLASTp analysis was performed using the deduced amino acid sequences from ERG20 as a query. By this way, the homologous genes from the basidiomycetes Coprinopis cinnerea, Cryptococcus neoformans, Laccaria bicolor and Ustilago maydis were obtained [Genbank: EAU81564.1, XP_571137, DS547151 and XP_757593, respectively], and the deduced amino acid sequences were aligned. From the conserved regions, primers were designed to amplify by PCR the X. dendrorhous FPS gene. By this way, a X. dendrorhous genomic PCR amplification product of approximately 800 bp (using primers FPSF and FPSR) was obtained, which was completely sequenced. BLAST analyses of the obtained nucleotide sequence against the patent database available at the National Center of Biotechnology Information (NCBI), showed 100% identity with the X. dendrorhous FPS gene sequence from patent EP0955 363A2 [30]. This patent described the X. dendrorhous FPS gene isolation and reported its sequence [Genbank: AR349868.1], but it did not analyzed the gene functionality nor its impact on the yeast carotenogenesis. Nevertheless, primers were designed from this sequence to amplify the X. dendrorhous FPS gene.
Plasmid construction for E. coli and X. dendrorhous transformations All plasmids used and constructed in this work are listed in Table 2.
A 3,312 bp EcoRI/DraI DNA fragment containing the X. dendrorhous crtE gene was sub-cloned from plasmid pXD-I10 [17] into the pBluescript SK-vector, giving rise to pXd-gcrtE. The plasmid pXd-gFPS was constructed by inserting a 2,627 bp PCRamplified DNA fragment encoding the FPS gene into the EcoRV site of the pBluescript SK-. This last DNA fragment was amplified using primers FPSORF1F and FPSORF1R (Table S1) and genomic DNA of the UCD 67-385 wild-type strain as template. The genomic versions of both genes from the wild-type strain UCD 67-385 were uploaded at the Genbank database as follows: Next, from the genomic crtE and FPS gene versions, primers were designed to amplify their cDNA obtained by reverse transcription (RT) of total RNA from the wild-type strain UCD 67-385. The amplified products were also cloned into the pBluescript SK-vector, resulting in the plasmids pXd-ccrtE and pXd-cFPS, whose inserts were completely sequenced. The plasmids pET28-ccrtE and pET28-cFPS for heterologous expression in E. coli were constructed by PCR amplification of the cDNA of each gene from plasmids pXd-ccrtE and pXd-cFPS, respectively, and using primers with restriction sites (Table S1) for an in-frame ligation into the NcoI+NdeI digested plasmid pET28a.
The plasmids pXd-gcrtE::hph and pXd-gFPS::hph were constructed as shown in Figure S1. For pXd-gcrtE::hph, a 1,984 bp EcoRV DNA fragment in pXD-gcrtE was replaced by the hygromycin B resistance cassette obtained from the digestion of pMN-hph with EcoRV; in the case of pXd-gFPS::hph, a 2,038 bp EcoRV/BglII DNA fragment in pXd-gFPS was replaced by the hygromycin B cassette.
To over-express the crtE and FPS genes in X. dendrorhous, plasmids pXdVexp2-ccrtE and pXdVexp2-cFPS were constructed as shown in Figure S1. Briefly, the crtE and the FPS cDNAs were amplified from plasmids pXd-ccrtE and pXd-cFPS, respectively, and were then independently inserted at the HpaI site of pXdVexp2. In this approach, after transformation, the gene to be over-expressed is integrated into the yeast genome at locus int [Genbank: KJ140286], which represents a non-coding DNA region that does not reveal an evident new phenotype when it is interrupted. Each cDNA is regulated by the promoter of the ubiquitin gene [Genbank: KJ140285] and the terminator from the glyceraldehyde-3-phosphate dehydrogenase [Genbank: Y08366] gene of X. dendrorhous.
Transformation of X. dendrorhous X. dendrorhous was transformed by electroporation [31] using a BioRad gene pulser X cell with PC and CE cassettes under the following conditions: 125 mF, 600 V and 0.45 kV [32]. The electrocompetent cells were prepared from an exponential culture at OD 600nm = 1.2, grown in YM medium and transformed with 1 to 5 mg of linear donor DNA. The resulting transformant strains were confirmed as X. dendrorhous through an analysis of the ITS1, 5.8 rRNA gene and ITS2 DNA sequences [33].

Carotenoid and sterol extraction and RP-HPLC analyses
The acetone extraction method [34] was used for carotenoid extraction from cellular pellets. Total carotenoids were quantified spectrophotometrically at 465 nm using an absorption coefficient of A1% = 2,100 and normalized to the dry weight of the yeast. The extracted carotenoids were separated by RP-HPLC using a RP-18 Lichrocart125-4 (Merck) column with acetonitrile:methanol:isopropanol (85:10:5, v/v) as the mobile phase, with a 1 mL/ min flux under isocratic conditions, and the elution spectra were recovered using a diode array detector. Carotenoids were identified according to their spectra and retention time in comparison to standards.
Sterols were extracted according to [35]. For this extraction, cell pellets were mixed with 4 g of KOH and 16 mL of 60% (v/v) ethanol/water and incubated at 8062uC for 2 h. Non-saponificable sterols were extracted with 10 mL of petroleum ether, quantified spectrophotometrically at 280 nm using an absorption coefficient of A1% = 11.500 and normalized to the dry weight of the yeast. Sterols were separated by RP-HPLC with a C-18 column, using methanol:water (97:3, v/v) as the mobile phase at 1 mL/min under isocratic conditions. The elution spectra were recovered using a diode array detector, and sterols were visualized in the 280 nm channel and compared to ergosterol standard purchased from Sigma-Aldrich.

Protein extraction
Soluble proteins from E. coli were obtained by sonication of cell pellets (5 mL from a culture with a DO 600 0.5-0.7) suspended in 1 mL of extraction buffer, 50 mM MOPS pH 7.5 and protease inhibitor cocktail (Complete, Roche Biomedicals), using a Cole Parmer 4710 ultrasonic homogenizer. Twelve sonication pulses of 10 s at 40% duty, followed by 30 s of incubation in ice, showed the best protein yield from the soluble fraction. The lysate was centrifuged at 14,0006g, and the supernatant was recovered and stored at 280uC until its use.
Soluble proteins from X. dendrorhous were also obtained by sonication of cell pellets from 10 mL of a stationary phase culture. Each pellet was suspended in 5 mL of extraction buffer (50 mM MOPS, 5 mM EDTA, 0.5 M sorbitol and protease inhibitor cocktail) and ruptured by 15 pulses of sonication of 30 s at 90% duty followed by 1.5 min of incubation in ice. The lysate was sequentially centrifuged at 4,0006g for 5 min, 10,0006g for 20 min and 100,0006g for 1 h, recovering the supernatant from each step. The recovered supernatants were purified from endogenous metabolites through a PD-10 exclusion column (GE Healthcare). The recovered proteins were stored at 280uC.

Prenyl transferase activity assays
Prenyl transferase activity assays were adapted from [36]. Two protein extract samples were obtained from two replicate cultures of each of the X. dendrorhous or E. coli strains that were assayed. The enzymatic assays were performed in duplicates in a final volume of 500 mL containing 12 mg of protein extract (14,0006g or 100,0006g supernatants), 50 mM MOPS, 5 mM MgCl 2 , 6 mM DTT, 100,000 dpm 14  Incubations were performed at 22uC for 1 h. 14 C-GPP, 14 C-FPP and/or 14 C-GGPP were expected as products, as was 14 C-DMAPP possibly formed by an IPP isomerase present in the protein extracts. The corresponding 14 C-prenyl alcohols, which are formed by phosphatase activities on the prenyl transferase products, may also be obtained, though the amount produced in most cases was negligible and thus not considered. Reactions were stopped by adding 500 mL of hexane to extract the 14 C-prenyl alcohol phosphatase products. After extraction, the organic phase was separated and removed. The remaining aqueous phase was acidified with HCl to a final concentration of 0.66 N and incubated for 1 h at 37uC to hydrolyze allylic 14 C-prenyl phosphates formed by prenyl transferases. The resulting labeled, rearranged prenyl alcohols were extracted with 500 mL of hexane and the organic phase (allylic hexane phase) was stored at 220uC for quantification and further thin layer chromatography (TLC) analysis. One enzyme unit (U) was defined as 1 mmol of product formed per min under the assay conditions. Prenyl transferase specific activity was expressed in U/mg of protein. A representative chromatogram prenyl product analysis by TLC is shown in Figure S2.

Product quantification and analysis
14 C-labelled products, as well as the remaining substrate, were quantified in the organic and aqueous phases, respectively, by liquid scintillation counting in a Perkin Elmer Tri-Carb 2810TR instrument. 14 C-Prenyl alcohols in the allylic hexane phase (C 5 , C 10 , C 15 and/or C 20 ), which are formed from allylic prenyl phosphate products, were analyzed by TLC in 60 mm silica gel plates (TLC Silica gel 60 RP-18, Merck) developed with acetone/ H 2 O (9:1, v/v) as mobile phase. Standard dimethyl allyl alcohol (C 5 ), linalool (C 10 ), trans-nerolidol (C 15 ) and geranylgeraniol (C 20 ) (4 mg, 8 mg, 2 mg and 0.5 mg, respectively) were added to the samples that were concentrated under N 2 up to 10-15 mL and loaded into the plates. The plates were sprayed with Libermann-Burchard solution (H 2 SO 4 :acetic anhydride:ethanol (1:1:8, v/v) and heated gently on a heating plate to reveal the standards (Rf values were 0.9, 0.8, 0.71 or 0.62 for the C 5 , C 10 , C 15 or C 20 standards, respectively). Segments of 2.5 mm were cut from the origin to the solvent front, and radioactivity was quantified in each fraction by liquid scintillation counting.

Results and Discussion
Sequence analysis of genes controlling the GGPP synthesis from IPP in X. dendrorhous The X. dendrorhous crtE gene has been previously isolated and described [17], which encodes a 376 amino acid protein with a predicted molecular weight of 42. Because the crtE and the FPS genes encode prenyl transferase enzymes, their deduced amino acid sequences were compared, including the amino acid sequences from C. neoformans (crtE [Genbank: XP_572774.1] and FPS [Genbank: XP_571137.1]) to increase the robustness of the analysis (Fig. 2). The amino acid identity between the homologous enzymes was 47% and 63% for the crtE and FPS gene products, respectively. As expected, the identity among the four enzymes decreased dramatically. Even though the sequence identity among the prenyl transferase enzymes encoded by the analyzed crtE and FPS genes is low, the previously reported conserved regions in these enzymes [37,38] could be distinguished (Fig. 2). This was the case for the twoaspartic acid-rich motifs (DDxxD) that were recognized: i) the First Aspartic acid-Rich Motif (FARM; DDxxDxxxxRRG), and ii) the Second Aspartic acid-Rich Motif (SARM; GxxFQxxDDxxD). Moreover, the chain length domain (CLD) proposed by Ohnuma and co-workers [39] could be distinguished in the GGPP-and FPP-synthase deduced sequences. The X. dendrorhous GGPPsynthase contains a small amino acid and a serine at the fourth and fifth positions before the FARM motif, while FPP-synthase has a tyrosine and phenylalanine at these positions, which is consistent with the chain length determination of these two types of prenyl transferases [40,41]. Although the predicted X. dendrorhous FPPsynthase shows the seven conserved regions (I to VII) described for this type of enzyme, the predicted GGPP-synthase from X. dendrorhous lacks any distinctive aminoacids from this last conserved region.
Heterologous expression of FPS and crtE genes from X. dendrorhous in E. coli To study the enzymatic activities encoded by the FPS and crtE genes from X. dendrorhous, the genes were independently expressed in E. coli. The synthesis of FPP-synthase and GGPP-synthase in the transformant strains (BL21+FPS and BL21+crtE) was confirmed by SDS-PAGE analyses (Fig. S3), which showed protein bands of the expected size (40.24 and 42.16 kDa, respectively) only when cultures were induced by IPTG. Moreover, the protein bands were excised from gels and analyzed by MALDI-TOF to confirm the identity of the proteins. Prenyl transferase assays were performed with protein extracts from transformant strains BL21+pET28, BL21+FPS and BL21+crtE (Table 1), using induced and noninduced cultures as controls. The results are summarized in Table 3. Prenyl transferase activity was not detected with any of the substrates in the extracts of non-induced cultures or in extracts from BL21+pET28. In contrast, 14 C-labeled products were found in the allylic fraction from incubations with induced BL21+FPS and BL21+crtE extracts. The extracts from BL21+FPS formed C 10 and C 15 14 C-allylic pyrophosphates from DMAPP plus 14 C-IPP, as evidenced by the C 10 and C 15 14 C-alcohols detected in the allylic hexanic phase (Table 3), while GPP was converted by BL21+FPS extracts into a C 15 allyl pyrophosphate that was detected as a C 15 alcohol after acid hydrolysis. Interestingly, no products were found in the allylic fraction when FPP plus 14 C-IPP were utilized as substrates. In contrast, the extracts from BL21+ crtE did not utilize DMAPP or GPP as substrates but were able to convert FPP in the presence of 14 C-IPP to give C 20 14 C-alcohols in the allylic fraction, which is evidence of 14 C-GGPP synthesis. These results demonstrate that the prenyl transferases encoded by FPS or crtE differ strongly in their substrate specificity, in agreement with their sequence homology to the FPP-and GGPP-synthase encoding genes from other organisms. Thus, GGPP production for astaxanthin biosynthesis by X. dendrorhous would result from the combined action of both prenyl transferases, i.e., the FPP-and GGPP-synthases, with the latter providing a link to the central metabolism essential for astaxanthin production. This was demonstrated by the conversion of DMAPP plus 14 C-IPP into C 20 14 C-products by the combined BL21+FPS and BL21+crtE protein extracts, in addition to the C 15 and C 10 labeled products found in the hexane phase after acid hydrolysis (Table 3). However, the C 20 labeled products fraction was lower than the other fractions, which would result from a low geranylgeranylpyrophosphate synthase (crtE gene) activity compared to farnesylpyrophosphate synthase (FPS gene) activity in the protein extracts from E. coli (Table 3). Thus, FPP accumulates and is only partially converted into GGPP. Product distribution in E. coli combined extracts is consistent with results found in the X. dendrorhous enzymatic assays with DMAPP as substrate in which 86% of the product corresponded to FPP and only 12% to GGPP (see below, Table 4). The geranylgeranyl-pyrophosphate synthase activity in X. dendrorhous is much lower than the farnesyl-pyrophosphate synthase activity (Table 4), thus FPP conversion to GGPP is a slow reaction.

FPS and crtE gene mutations and over-expression in X. dendrorhous
To study the involvement of the FPS and crtE genes in the astaxanthin and ergosterol biosynthetic pathways of X. dendrorhous, FPS and crtE deletion and over-expressing mutant strains were constructed.
To generate deletion mutants, the wild-type strain UCD 67-385 was independently transformed using the linearized plasmids pXd-gcrtE::hph and pXd-gFPS::hph. In this way, the wild-type genes (FPS or crtE) were replaced by the DNA fragment containing the resistance marker for transformant strain selection, flanked by DNA sequences that allowed homologous recombination between the transforming DNA and the corresponding target locus. Several transformants were selected from each transformation event, and the modified genotype was confirmed by PCR reactions using comprehensive sets of primers (Table S1). Figure S4 shows the genotype analysis of one transformant of each kind, indicating that they are heterozygotes for the FPS or the crtE genes because they have a mutant allele and still have a wild-type allele. For this reason, these transformants were named 385-FPS (+/2) and 385-crtE (+/2) , respectively. This result was expected because the experimental evidence suggests that the UCD 67-385 strain is diploid [42]. Although several attempts at transformant homozygotization using the Double Recombinant Method (DRM) [43] were made, it was not possible to obtain crtE [17] and FPS homozygous mutants, suggesting that these genes are essential and that their homozygote mutant condition could be lethal. Nevertheless, a different color phenotype is apparent to the naked eye between the parental and the heterozygous strains (Fig. 3). This observation suggests that under substrate limiting conditions, the carbon flux may be preferentially directed towards vital functions instead of carotenogenesis, as the growth curves of both heterozygous strains were not significantly altered (Fig. S5). Additionally, the heterozygous condition supports the gene dose effect in the production of carotenoids in X. dendrorhous, which has been described in other carotenogenic heterozygous mutants [17,44].
Similarly, the FPS and crtE genes were over-expressed in the wild-type strain UCD 67-385 by transforming with linearized plasmids pXdVexp2-cFPS and pXdVexp2-ccrtE; a transformant strain of each type was selected and named 385-FPS (+/+, +1) and 385-crtE (+/+, +1) , respectively. In the same way as the heterozygous mutant strains (Fig. S4), PCR-based genotype analyses of the transformant strains confirmed that the over-expression cassettes were integrated at one of the int locus alleles. Under visual inspection, the colonies of both over-expressing mutants have a more intense red color than the parental strain (Fig. 3). Additionally, the 385-FPS (+/+, +1) colonies have a glossy appearance, in contrast to the X. dendrorhous wild-type colonies, which are usually opaque. The crtE over-expression affected the biomass production, as 385-crtE (+/+, +1) cultures reached a lower OD 600 than did the parental strain (8.7660.62 versus 12.860.93) after 120 h of culture at 22uC in YM media with constant agitation (Fig. S5). This fact suggests that crtE over-expression favors the carbon flux towards carotenogenesis at the expense of other metabolic pathways, negatively impacting the strain growth. Conversely, FPS over-expression does not affect strain growth, most likely because this condition would favor FPP production, which is also a sterol biosynthetic pathway precursor. Finally, in both over-expressing strains, pigmentation is apparent to the naked eye earlier than in the parental strain and in both heterozygote mutant strains, suggesting that carotenogenesis begins earlier in the FPS and crtE over-expressing strains (Fig. 3).

Carotenoids and sterols in X. dendrorhous wild-type, FPS and crtE mutant strains
To confirm the visual inspection observations, total carotenoids and sterols were extracted from the wild-type and 385-FPS (+/2) , 385-crtE (+/2) , 385-FPS (+/+, +1) and 385-crtE (+/+, +1) mutant strains after 72 and 120 h of culture in YM media with constant agitation ( Table 5). As expected, the total carotenoid content and/or composition were different in the mutant strains with respect to the wild-type strain. Statistical analysis (Student's t test, p,0.01) confirmed that the FPS and crtE over-expressing strains have a higher carotenoid content and a slight increase in the astaxanthin fraction compared with the parental strain after 72 h of incubation. Previously, a similar carotenoid content increment (approximately 1.4-fold) was reported by over-expressing the crtE gene in X. dendrorhous [19]. Conversely, the total carotenoid content in the heterozygote FPS and crtE mutant strains was not Enzymatic assays were performed as described in the Materials and Methods Section. Two protein extract samples from two replicate cultures were obtained for each strain, and then two enzymatic assays were performed with each protein extract. The table values are the mean values of prenyl transferase activity and percentage of product 6 standard error of the mean. *Percentage of each product considering 100% the total radioactivity found in the hexanic phase. ND: Product not detected in TLC analysis. C 10 , C 15 and C 20 correspond to products containing 10, 15 and 20 carbon atoms, respectively. **The assays conducted using FPP as a substrate yielded low amounts of products, which could not be detected under the experimental conditions. doi:10.1371/journal.pone.0096626.t004 significantly different from the wild-type strain (Student's t test, p, 0.01), but the composition was altered, mainly by an increase of beta-carotene, echinenone, OH-echinenone, cantaxanthin and phoenicoxanthin at the expense of the astaxanthin fraction. As mentioned above, the fundamental phenotypic differences between the five strains analyzed in this work can most likely be explained by carbon flux distribution. Thus, the FPS and crtE overexpressing strains produce more of the carotenogenic pathway precursors, such as FPP and GGPP, while their production should be reduced in the heterozygous mutants. In X. dendrorhous, carotenogenesis is induced when glucose in the medium is depleted [25,45,46], which occurs close to the stationary phase of growth. Thus, carotenogenesis begins when cultures stop growing and the carbon flux is directed towards carotenoid biosynthesis. However, the FPS and crtE over-expressing strains might provide an excess of substrate, which might favor an early start for carotenogenesis. This result contrasts with the heterozygous FPS and crtE mutant strains, in which the growth is not altered (Fig. S5) and the limited carotenogenic substrates may be directed towards biomass formation, thus delaying carotenogenesis (Fig. 3).
Conversely, the altered carotenoid composition in the heterozygous FPS and crtE mutants may support the hypothesis of the formation of a carotenogenic complex in X. dendrorhous [47]. Under this hypothesis, a varying level of one of the protein components of the complex might affect the sequence of the reactions, thus affecting the end products that are formed [47]. In this sense, the carotenogenic complex assembly may possibly be driven, at least in part, by the availability of carotenogenic substrates. Then, under limited substrate concentrations (this would be the case of the heterozygous FPS and crtE mutants), an incomplete carotenogenic complex might be assembled, favoring the formation of intermediary carotenoids other than astaxanthin. This would not be the case in the FPS and crtE over-expressing strains, in which a higher amount of carotenogenic substrates might favor an earlier assembly of a carotenogenic complex, leading to an earlier and a higher carotenoid production.
Finally, in contrast to what we expected, no significant differences regarding the ergosterol production among the X. dendrorhous transformants and the wild-type parental strain were observed. However, after 120 h of cultivation, there was a small reduction of the ergosterol content in the wild-type strain, which was only statistically significant when compared with the strain 385-crtE (+/+, +1) that had an ergosterol content increment (about 36% more).

FPS and crtE relative transcript levels and prenyl transferase activity in X. dendrorhous deletion and over-expressing transformants
To assess whether the decrease or increase of the FPS and crtE gene dose effectively changes mRNA levels, RT-qPCR analyses were performed for wild-type, heterozygous and over-expressing strains grown under the same conditions. Total RNA was extracted after 72 h of culture to determine the relative FPS and crtE transcript levels, which were normalized to the transcript levels of the actin gene [25]. The RT-qPCR analyses revealed that the relative transcript levels of the FPS gene in the 385-FPS (+/+, +1) strain was significantly increased with respect to the wild-type and it was reduced to about half of the wild-type value under the heterozygous condition. However, the relative transcript levels of the crtE gene in the 385-crtE (+/+, +1) strain was slightly increased and it was decreased in the 385-crtE (+/2) strain (Fig. 4).
Unlike other astaxanthin-producing organisms, the synthesis of astaxanthin from b-carotene in X. dendrorhous is catalyzed by a cytochrome P450 system [48] composed of the cytochrome P450 enzyme (astaxanthin synthase, encoded by the crtS gene [49,50]) and the cytochrome P450 reductase (CPR, encoded by the crtR gene [44]). Given that carotenoid composition in both heterozygous mutant strains, 385-FPS (+/2) and 385-crtE (+/2) , was different with respect to the wild-type strain, mainly by the reduction of the astaxanthin fraction and an increase of intermediary carotenoids between b-carotene and astaxanthin, the expression of the genes involved at this step, was analyzed (Fig. 4). The relative crtS transcript levels showed only a slight reduction in the 385-FPS (+/2) strain. However, the relative crtR transcript levels were significantly reduced in both heterozygous strains, while the FPS and crtE overexpressing strains showed no significant changes. Previous studies have shown that mutation of one of the crtR alleles in the diploid UCD 67-385 strain affects the carotenoid composition, decreasing the astaxanthin fraction in favor of b-carotene and other intermediary carotenoids [44]. Thus, our results suggest that the relative crtR transcript levels reduction in strains 385-FPS (+/2) and 385-crtE (+/2) could also be responsible for their altered phenotypes. Considering the vast number of reactions in which CPR enzymes are involved, the expression and activity regulation of CPR is generally a complex process that involves several mechanisms [51] that act at different levels. The mechanism by which the crtE and FPS gene deletions affect the crtR transcript levels requires further investigation. Finally, prenyl transferase activities were determined in soluble protein extracts from the five analyzed X. dendrorhous strains, which were obtained after 72 h of cultivation under the conditions described in Materials and Methods (Table 4). Prenyl transferase activity, mainly FPP-synthase, was increased by 85% in the FPS over-expressing strain while prenyl transferase activity, measured with FPP, remained unchanged compared to the wild-type strain UCD 67-385. Conversely, the prenyl transferase activity with FPP was increased in 144% in the crtE over-expressing mutant, while the activity measured with DMAPP remained unchanged. However, C 20 14 C-products are a minor fraction in all cases. Also, C 10 and C 15 14 C-products can be formed in the presence of FPP as a consequence of IPP isomerase activity that achieves the isomerization of IPP to DMAPP. In enzymatic assays with UCD 67-385 protein extracts and only including IPP as substrate, C 10 , C 15 and C 20 were obtained, which evidences IPP isomerase activity in these extracts. However, this endogenous IPP isomerase activity would only affect the enzymatic assays including FPP as substrate, since no DMAPP is added to the assay. On the other hand, the enzymatic assays including DMAPP as substrate, should not be affected as substrates were used at saturated concentrations, so the IPP isomerase activity would be insignificant.
The prenyl transferase activity in heterozygous strain extracts using DMAPP or FPP as a substrate showed no significant differences compared to the wild-type strain (Table 4). Probably, the methodology that we used might not be sensitive enough to detect variations in the prenyl transferase activity in the protein extracts from the strains that we assayed. Also, a transcript level reduction, which we observed for the respective gene in the heterozygous strains, does not necessarily lead to a reduction of the enzymatic activity with the same magnitude.
The results reported here demonstrate that in X. dendrorhous, two prenyl transferases act sequentially to synthesize GGPP from DMAPP and IPP precursors: an FPP-synthase (encoded by FPS) that produces FPP and an GGPP-synthase (encoded by crtE) that transforms the C 15 prenyl pyrophosphate into GGPP. The accumulation of 14 C-FPP (detected as C 15 alcohol in the allylic fraction) in the above assays indicates that the reaction catalyzed by the crtE gene product would be rate-limiting in the biosynthetic pathway, which thus represents a possible target for astaxanthin production improvement. Considering that the DNA assembler methodology, which allows integration in the genome of several DNA fragments during one transformation event, was proven to be effective and successful in X. dendrorhous [52], crtE could be simultaneously over-expressed together with other rate-limiting astaxanthin production genes to favor the synthesis of carotenoids in X. dendrorhous.

Conclusions
In X. dendrorhous, GGPP is synthesized from IPP and DMAPP by two prenyl transferase enzymes that act sequentially. In the first stage, an FPP-synthase (encoded by the FPS gene) produces FPP in two steps: GPP is first produced by the condensation of IPP and DMAPP, and the addition of a second IPP molecule to GPP then produces FPP. In the second stage, a GGPP-synthase (encoded by the crtE gene) produces GGPP by adding a third molecule of IPP to FPP. Both of the genes described in this work, FPS and crtE, represent good target candidates to enhance the astaxanthin production in X. dendrorhous by metabolic engineering approaches. Figure S1 Plasmids constructed in this work. In each plasmid illustration, the relevant features such as endonuclease recognition sites and primer binding sites (thin arrows) are shown. Some elements of the original plasmid (pBluescript SK-) were kept in the figure and are shown in gray. The pBluescript SK-skeleton was kept as light gray. Plasmids: A) pXd-gFPS harbors a 2.5 kb DNA fragment carrying the X. dendrorhous FPS gene. Green, thick arrows represent the nine exons of the FPS gene. B) pXd-gFPS::hph was constructed from pXd-gFPS, which was digested with EcoRV and BglII to release a 2.0 kb DNA fragment that contained the FPS gene that was replaced by a hygromycin B resistance cassette (HygR, represented by an ochre, thick arrow). For transformation purposes, pXd-gFPS::hph was linearized with AvaI. C) pXd-gcrtE harbors a 3.3 kb DNA fragment carrying the X. dendrorhous crtE gene. Light blue, thick arrows represent the nine exons of the crtE gene. D) pXd-gcrtE::hph was constructed from pXd-gcrtE, which was digested with EcoRV to release a 2.0 kb DNA fragment that contained the crtE gene that was replaced by a hygromycin B resistance cassette (HygR, represented by an ochre, thick arrow). For transformation purposes, pXd-gcrtE::hph was linearized with KpnI and SmaI. E) The pXdVexp2 expression vector was constructed by cloning an X. dendrorhous non-coding genomic region (Int, represented by a light orange, thick arrow) to target the integration into the X. dendrorhous genome. The Int region was interrupted by inserting the X. dendrorhous ubiquitin promoter (Ubi-P, represented by a light purple, thick arrow) and the GPD terminator (GPDT, represented by a dark purple, thick arrow), with a HpaI site between them to insert the gene being expressed. The hygromycin B cassette for transformant selection (HygR, represented by a ochre, thick arrow) was also included. F) The insertion of FPS or crtE cDNA at the HpaI site of pXdVexp2, yielded pXdVexp2-cFPS and pXdVexp2-ccrtE, respectively. (TIF) Figure S2 Representative chromatogram from prenyl product analysis by TLC. The prenyl alcohol products that were obtained in the prenyl transferase activity assays were resolved by TLC. The radioactivity detected on each of the TLCplate fragments is displayed in each graph. The peaks corresponding to the C 10 , C 15 and C 20 products are indicated. Graphs correspond to the following enzymatic assay: A) protein extracts from wild-type X. dendrorhous strain with DMAPP as additional substrate, B) protein extracts from E. coli BL21+FPS strain with DMAPP as additional substrate, C) protein extracts from E. coli BL21+crtE strain with FPP as additional substrate and D) a mixture of protein extracts from E. coli strains BL21+FPS and BL21+crtE with DMAPP as additional substrate.  Figure S4 PCR-based analysis of the FPS and crtE gene deletion in X. dendrorhous. PCR analyses to confirm the insertion of the Hygromycin B resistance cassette and the replacement of the target gene of the parental wild-type UCD 67-385 strain (wt). The resulting heterozygous strains (+/2) are 385-FPS (+/2) (upper gel photographs) and 385-crtE (+/2) (lower gel photographs), including the PCR negative control without DNA (2). Between the upper and lower gel photographs, a scheme is included to represent the primer sets (shown in arrows) that were used, the expected PCR-product size and the target DNA. The primers used in the analyses depended on the gene deletion that was studied (crtE or FPS gene) and were: 1: FPSnewF or crtE_CDS_F1, 2: FPSnewR or crtE_CDS_R1, 3: FPS_Out_F7 or crtE_Out_F1, 4: TEF Antisense, 5: GPDT Sec F and 6: FPS_Out_R1 or crtER2 (Table S1). The scheme shading is in accordance with Figure