Metabolic Engineering of Saccharomyces cerevisiae for Production of Eicosapentaenoic Acid, Using a Novel Δ5-Desaturase from Paramecium tetraurelia

Abstract

Very-long-chain polyunsaturated fatty acids, such as arachidonic acid (ARA), eicosapentaenoic acid (EPA), and docosahexaenoic acid (DHA), have well-documented importance in human health and nutrition. Sustainable production in robust host organisms that do not synthesize them naturally requires the coordinated expression of several heterologous desaturases and elongases. In the present study we show production of EPA in Saccharomyces cerevisiae using glucose as the sole carbon source through expression of five heterologous fatty acid desaturases and an elongase. Novel Δ5-desaturases from the ciliate protozoan Paramecium tetraurelia and from the microalgae Ostreococcus tauri and Ostreococcus lucimarinus were identified via a BLAST search, and their substrate preferences and desaturation efficiencies were assayed in a yeast strain producing the ω6 and ω3 fatty acid substrates for Δ5-desaturation. The Δ5-desaturase from P. tetraurelia was up-to-2-fold more efficient than the microalgal desaturases and was also more efficient than Δ5-desaturases from Mortierella alpina and Leishmania major. In vivo investigation of acyl carrier substrate specificities showed that the Δ5-desaturases from P. tetraurelia, O. lucimarinus, O. tauri, and M. alpina are promiscuous toward the acyl carrier substrate but prefer phospholipid-bound substrates. In contrast, the Δ5-desaturase from L. major showed no activity on phospholipid-bound substrate and thus appears to be an exclusively acyl coenzyme A-dependent desaturase.

During the past two decades, very-long-chain polyunsaturated fatty acids (VLC-PUFAs) like arachidonic acid (ARA; 20:4ω6), eicosapentaenoic acid (EPA; 20:5ω3), and docosahexaenoic acid (DHA; 22:6ω3) have attracted the attention of the scientific community as well as the dietary supplement and food industries due to their proven health benefits.

VLC-PUFAs are constituents of biological membranes, participate in cellular processes like cell signaling and hormone receptor binding (29), and act as biosynthesis precursors for eicosanoids and other anti-inflammatory mediators. Consumption of VLC-PUFAs is associated with cardioprotective effects, prevention of diseases like Alzheimer's disease, Parkinson's disease, and cancer, and is essential for correct development of brain and visual function in infants (3, 18, 29). Since the human body cannot synthesize VLC-PUFAs de novo in adequate quantities, dietary intake of these compounds is important. Fish and fish oils are the major nutritional dietary sources of VLC-PUFAs, but concerns over the continual depletion of wild fish stocks and contamination of the oceans lead to the obvious conclusion that alternative sustainable sources are urgently needed (24). VLC-PUFAs are present in a wide range of different organisms, from mammals, fungi, mosses, and bacteria to lower plants, but it is within the microalgae that the most efficient producers of VLC-PUFA are found, especially for the ω3 fatty acids. Curiously, saprophytic and pathogenic forms of life, like the Oomycetes (e.g., species from Pythium [37] and Saprolegnia [22]), the Entomophthorales (Entomophthora [6] and Conidiobolus [13]), the Labyrinthulids (30), and the protozoa (Leishmania and Trypanosoma [39]), also produce VLC-PUFAs and apparently contain the complete pathway for PUFA synthesis.

Some industrial processes exploit these efficient PUFA producers as cell factories for more sustainable production of these compounds (25), although the culture of nonadapted native organisms can often prove to be problematic and costly. Alternatively, the heterologous production of VLC-PUFAs in dedicated hosts is also seen as a potential solution. The increasing number of reports describing the isolation of desaturases and elongases from a variety of organisms and their successful expression in several hosts like Saccharomyces cerevisiae and plants (1, 16, 20, 28, 31, 36) confirm the availability of a genetic tool kit with which to metabolically engineer the synthesis of these important fatty acids.

Synthesis of ARA and EPA can be achieved through a sequence of desaturations and elongations, the final steps being the introduction of a double bond between carbons 5 and 6 in dihomo-γ-linolenic acid (20:3ω6) and eicosatetraenoic acid (20:4ω3), respectively, by a Δ5-desaturase (Fig. 1). The Δ5-desaturases belong to the “front-end” desaturases family, since they introduce a double bond between an existing double bond and the carboxyl terminal of the fatty acid chain.

FIG. 1.

Simplified representation of PUFA biosynthetic pathways from stearic acid (18:0) to ARA, EPA, and DHA.

The “front-end” desaturases contain a fused N-terminal cytochrome b₅-like domain, which has been claimed to be essential for catalytic activity, and three conserved histidine boxes (17, 33). Genes encoding Δ5-desaturase activities were first isolated from the ARA-producing fungus Mortierella alpina (9, 12) by using a PCR-based approach with primers designed for those conserved histidine boxes, which had been previously observed in Δ6-desaturases (14, 15, 26, 34). The same method has been successfully used to isolate most of the desaturases described in the literature. An alternative is the use of the conserved histidine boxes as well as other typical motifs in bioinformatics searches, which, combined with the increasing number of genome sequencing projects, provides a faster and more direct approach toward identification and expression of novel genes. In addition, the facile and robust procedure of whole-gene synthesis and codon optimization opens the door to a limited synthetic biology approach to pathway engineering, meaning that multiple nonnative sequences encoding the VLC-PUFA biosynthetic pathway can be expressed in a suitable “chassis” (a host such as yeast) to generate a transgenic organism with novel functionality. Despite the benefits of bioinformatics, our understanding of the underlying biochemistry and regulation of this pathway is still incomplete.

Although detailed biochemical studies on purified front-end desaturases and PUFA elongases are still missing, several research groups have studied their acyl carrier substrate specificities in detail using in vivo systems (4, 5). In lower eukaryotes, it is suggested that the elongation step in PUFA biosynthesis takes place in the acyl coenzyme A (acyl-CoA) pool, whereas Δ5- and Δ6-desaturations occur mainly on phospholipid (PL)-linked substrates (4). This creates an acyl exchange bottleneck (so-called “substrate dichotomy”) in the pathway, since the Δ6-desaturated products do not enter the acyl-CoA pool, and therefore elongation and subsequent Δ5-desaturation are not efficiently reconstituted. However, recent studies have identified microalgal Δ6- and Δ5-desaturases which accept CoA-bound substrates, and these findings suggest that an acyl-CoA-dependent pathway might overcome this endogenous metabolic blockade (8). Similarly, it has been suggested that use of the alternative Δ9-elongation/Δ8-desaturation pathway (1) (which requires only a single acyl exchange) represents an additional solution to this problem.

In the present study we describe the identification of novel Δ5-desaturases from a ciliate protozoan, Paramecium tetraurelia, and two microalgae, Ostreococcus tauri and Ostreococcus lucimarinus. The Δ5-desaturases were functionally expressed in an engineered yeast strain capable of producing the substrates 20:3ω6 and 20:4ω3, resulting in the production of ARA and EPA. To our knowledge, this represents the first report of an EPA/ARA-producing Saccharomyces cerevisiae strain without the need of fatty acid supplementation. In order to fully evaluate the acyl carrier substrate specificities of these novel desaturases, we studied the distribution over time of the fatty acid substrate and product of Δ5-desaturation in different lipid fractions and compared the results with previously known Δ5-desaturases from Mortierella alpina and Leishmania major.

MATERIALS AND METHODS

Materials.

Restriction enzymes were obtained from New England BioLabs (United Kingdom). Polymerases were from Finnzymes (Finland), and the fatty acids used as substrates were purchased from Larodan Fine Chemicals (Sweden). Fatty acid methyl ester (FAME) standards were from Nu-Check-Prep. Fatty acyl-CoAs (16:0, 17:0, 18:0, 18:1, 18:2, 20:0, 22:0, and 24:0) were obtained from Avanti Polar Lipids and Sigma-Aldrich. Unless otherwise stated, all other chemicals were from Sigma. Synthetic genes were obtained from Genscript and codon optimized for expression in S. cerevisiae.

Identification and cloning of Δ5-desaturases.

A list of Δ5-desaturase (D5D) protein sequences with verified activities was used as query for a BLAST search in the GenBank database. The outputs were combined into a unique list in which redundant sequences were removed. In the subsequent analysis the sequences were ranked based on typical features of front-end desaturases, i.e., the presence of the HPGG motif in the N terminus of the amino acid sequence, responsible for heme binding and catalytic activity; the existence and order of three conserved histidine boxes (HXXXH, HXXXHH, and QXXHHLFP); the presence of the motif DPDI (and variations) between the second and third HIS boxes; the percent identity with other Δ5-desaturases; and the length of the sequence. Moreover, the best hits were assessed based on the fatty acid composition of the original organisms in order to confirm the presence of ARA, EPA, and the respective precursors in the fatty acid profiles. The selected sequences were acquired as synthetic genes codon optimized for S. cerevisiae in vector pUC57 (Table 1). The open reading frames (ORFs) were reamplified with primers containing NotI/EcoRI restriction sites and introducing the translation initiation sequence ACC immediately upstream of the start codon, and they were subcloned into a replicative pESC-URA-based expression vector under the control of the constitutive S. cerevisiae TEF1 promoter. The constructs were designated pSF28-OtD5D, pSF28-OlD5D, pSF28-Ptet1D5D, and pSF28-Ptet2D5D. In addition to the putative desaturases, a described Δ5-desaturase from Leishmania major (accession number XM_001680969) (10) was also acquired as a codon-optimized gene and expressed in the same system as the putative genes. The construct was designated pSF28-LmD5D.

TABLE 1.

Δ5-Desaturases used in this study

Source organism	GenBank accession no.
Paramecium tetraurelia 1	HQ678517
Paramecium tetraurelia 2	HQ678518
Ostreococcus lucimarinus	HQ678519
Ostreococcus tauri	HQ678520
Leishmania major	HQ678521
Mortierella alpina	GU593328

The Δ5-desaturase sequence MaD5D (GenBank accession number GU593328) was isolated from Mortierella alpina CBS 608.70 cDNA by PCR using the primers 5′-ATGGGTACGGACCAAGGAAAAACC-3′ and 5′-CTACTCTTCCTTGGGACGGAGTCC-3′, designed to match the Δ5-desaturase described previously in reference 12. The ORF was reamplified with primers containing restriction sites and was subcloned for sequencing. Sequence verification of two clones showed that the isolated Δ5-desaturase gene contained slight variations compared to the MaD5D reported in reference 12 (96% identity of the protein sequence). The verified sequence was then subcloned into a pESC-URA-based expression vector, under the control of the strong S. cerevisiae TEF1 promoter, and designated pSF28-MaD5D. The Δ5-desaturase sequences used in this study are summarized in Table 1.

Strains.

For activity assays with substrate supplementation and for time course experiments, S. cerevisiae strain CEN.PK 113-5D (MATa ura3-52) was transformed with plasmid DNA using the polyethylene glycol-lithium acetate method. Transformants were selected on 2% glucose agar plates without uracil (35).

In order to verify Δ5-desaturation activity in a yeast strain producing 20:3ω6 and 20:4ω3, plasmid DNA was transformed into the proprietary S. cerevisiae strain FS01699 (Fluxome A/S), which contains chromosomal integrations of genes encoding a Δ9-desaturase (D9D), a Δ12-desaturase (D12D), a Δ6-desaturase (D6D), a Δ6-elongase (D6E), and an ω3-desaturase (FAD3) under the control of strong yeast promoters (Table 2). Briefly, FS01699 was constructed as follows: the MaD9D, MaD12D, and MaD6E genes were isolated from M. alpina cDNA as described above for the Δ5-desaturase, using amplification primers designed to match described M. alpina genes (references 41, 31, and 21, respectively). The SkFAD3 gene was isolated from Saccharomyces kluyveri genomic DNA using primers matching the ω3-desaturase described in reference 20. The PCR primers used for gene amplification were 5′-ATGGCAACTCCTCTTCCCCCCTCC-3′- and 5′-CTATTCGGCCTTGACGTGGTCAGTGC-3′ for the Δ9-desaturase, 5′-AACCCTTTTTCAGGATGGCACC-3′ and 5′-AAAGTTGTGTCCGGTAAATGCTTC-3′ for the Δ12-desaturase, 5′-ATGGAGTCGATTGCGCCATTCC-3′ and 5′-TTACTGCAACTTCCTTGCCTTCTCC-3′ for the Δ6-elongase, and 5′-GGTCTCGAGCCACCATGTCTATTGAAACAGTCGG-3′ and 5′-GGCCGCGGATCATTGACTGGAACCATCTT-3′ for the ω3-desaturase. The Ostreococcus tauri Δ6-desaturase (OtD6D) (5) was codon optimized for expression in S. cerevisiae (Backtranslation tool; Entelechon) and was assembled from synthetic oligonucleotides by PCR. Following subcloning and sequencing of the genes, OtD6D, MaD9D, MaD12D, and SkFAD3 were sequentially integrated together with constitutive yeast promoters at selected chromosomal sites through a PCR-based gene-targeting approach modified from that described in reference 7. The gene-targeting substrates were designed such that a direct repeat of the promoter sequence on either side of the insert enabled looping out of the Kluyveromyces lactis URA3 selection marker through homologous recombination between repeats; this event was selected for on 5-fluoroorotic acid plates. MaD6E under the control of the S. cerevisiae PYK1 promoter was integrated at the trp1-289 marker. Strain FS01699 additionally contains a deletion of the POT1 gene and a replacement of the FAS1 promoter with the ADH1 promoter.

TABLE 2.

Desaturase- and elongase-encoding genes integrated in strain FS01699

Gene name	Activity	GenBank accession no.	Integration site	Promoter
OtD6D	Δ6-Desaturase (Ostreococcus tauri)	HQ678522	DCI1 (−2,893)	ADH1
MaD9D	Δ9-Desaturase (Mortierella alpina)	GU593324	POX1 (−112,263)	TDH3
MaD12D	Δ12-Desaturase (Mortierella alpina)	GU593325	FOX2 (−12,709)	TDH3
MaD6E	Δ6-Elongase (Mortierella alpina)	GU593327	trp1-289	PYK1
SkFAD3	ω3-Desaturase (Saccharomyces kluyveri)	GU593329	GPP1 (−70,791)	HXT7

Expression of Δ5-desaturases in S. cerevisiae.

For substrate feeding experiments, precultures were grown in 2% glucose minimal medium (40) at 30°C until an optical density at 600 nm (OD₆₀₀) of approximately 2 to 3 was reached. Glass inserts (Supelco) containing 0.5 ml of 2% glucose minimal medium with 1% (wt/vol) Tergitol solution (type NP-40; 70% in H₂O) and a 0.5 mM concentration of the appropriate fatty acid substrate dissolved in absolute ethanol were inoculated to an initial OD₆₀₀ of 0.2. The glass inserts (triplicates) were placed into a deep multiwell plate, covered with sterile sealing tape, and incubated for 48 h at 30°C, tilted, with shaking at 230 rpm. For expression of Δ5-desaturases in 20:3ω6- and 20:4ω3-producing S. cerevisiae, precultures grown as described above were inoculated in 100 ml 2% glucose minimal medium to an initial OD₆₀₀ of 0.1 and cultivated for 48 h at 30°C with shaking at 150 rpm.

For the time course experiments, preinocula were grown for approximately 24 h in 2% glucose minimal medium at 30°C. Strains were then reinoculated in 100 ml 2% glucose minimal medium with 1% (wt/vol) Tergitol (type NP-40; 70% in H₂O) to an initial OD₆₀₀ of 0.1 and grown at 30°C until reaching late exponential phase (OD₆₀₀, ∼3) in order to guarantee expression of the desaturases and a sufficient amount of biomass for analysis. At this point (time 0 min), samples of approximately 30 mg (dry weight) of cells were harvested by centrifugation at 1,730 × g for 5 min and washed with an equal volume of water, and the pellet was used for fatty acid analysis. For acyl-CoA analysis, samples of 1 OD unit were harvested by a short-spin centrifugation of 20 s and dropped into liquid nitrogen, with the exception of the 24-h samples, in which 8 OD units were harvested. The kinetic experiment was started immediately after taking time zero samples, by addition of 0.5 mM 20:3ω6 to the cultures. Samples were taken after 5 min, 1 h, 4 h, and 24 h in the presence of the exogenous substrate as described above for time zero.

Fatty acid and lipid analyses.

Total fatty acid analysis of yeast cultures was performed according to the methods described in reference 19. FAMEs were analyzed on a gas chromatograph (GC; Agilent 7890A) coupled to a flame ionization detector (FID). Samples were injected at 190°C into a DB-Wax column (10-m by 0.1-mm inner diameter; 0.1-μm film thickness; J&W Scientifics), and immediately after injection the temperature was increased to 260°C at 12°C/min. Free fatty acid 23:0 was used as the internal standard. The FAMEs were identified based on their relative retention times by comparison with standards of LC- and VLC-PUFAs. Analysis of the different lipid classes (neutral lipids and phospholipids) was performed according to the methods described in reference 23. The lipid fractions were transmethylated, and FAMEs were analyzed as described previously. Fatty acid yield (in mg FA/OD unit) was determined by the amount of each fatty acid and the OD₆₀₀ of the initial cell suspension.

Acyl-CoA analysis.

Yeast cells harvested from suspension and snap-frozen in liquid nitrogen were extracted according to methods described in refererence 11 for subsequent quantitative analysis of fluorescent acyl-etheno-CoA derivatives by high-performance liquid chromatography. Analysis of acyl-CoA was performed using an Agilent 1100 LC system with a Phenomenex LUNA 150-mm by 2-mm C₁₈(2) column. The methodology and gradient conditions were described previously (10, 32). The synthesis of acyl-CoA 18:4 was performed according to methods described in reference 38.

RESULTS

Identification of putative Δ5-desaturases.

A BLAST search was performed in GenBank using characterized Δ5-desaturase protein sequences as queries. Putative Δ5-desaturase sequences from the unicellular ciliate Paramecium tetraurelia (Ptet1D5D) and from the microalgae Ostreococcus tauri (OtD5D) and Ostreococcus lucimarinus (OlD5D) were chosen for further analysis (Table 1). The Ptet1D5D protein shared 30% identity with M. alpina D5D (accession number AF054824), 26% with Phaeodactylum tricornutum D5D (accession number AY082392), and 32% identity with Mantoniella squamata D5D (accession number AM949596). O. tauri and O. lucimarinus sequences shared 82% identity between each other, 31% with M. alpina D5D, and 66% with M. squamata D5D.

Several of the Δ5-desaturases identified in our search were previously annotated and functionally characterized as such, contributing to validating the scoring method. The Δ5-desaturases from L. major (LmD5D) (39) and M. alpina (MaD5D) (9, 12) were also included in the further analysis, the latter being previously described as a phospholipid-dependent desaturase (4).

Functional characterization of Δ5-desaturases in S. cerevisiae.

Synthetic genes encoding codon-optimized versions of the putative desaturases were expressed in S. cerevisiae in the presence of exogenously supplied fatty acid substrates for Δ5-desaturation. Analysis of the fatty acid compositions of the strains expressing Ptet1D5D, OtD5D, OlD5D, MaD5D, and LmD5D in the presence of 20:3ω6 and 20:4ω3 showed the appearance of new peaks, identified as ARA and EPA, respectively, confirming that the novel genes encode Δ5-desaturases (Fig. 2 shows representative chromatograms). The analysis showed that P. tetraurelia D5D had the highest desaturation efficiency in the system used, followed by M. alpina D5D and O. tauri D5D, while O. lucimarinus D5D and L. major D5D were less active (Fig. 3).

FIG. 2.

Δ5-Desaturase activities of Ptet1D5D. GC-FID chromatograms of fatty acids from yeast strains expressing Ptet1D5D (a and c) or harboring the empty plasmid pSF28 (b and d), cultivated with supplementation of 0.5 mM 20:3ω6 (a and b) or 20:4ω3 (c and d) fatty acids, respectively.

FIG. 3.

Desaturation efficiencies of Ptet1D5D, OlD5D, OtD5D, LmD5D, and MaD5D. Yeast strains expressing Ptet1D5D, OlD5D, OtD5D, LmD5D, and MaD5D genes were cultivated in the presence of 0.5 mM 20:3ω6 or 20:4ω3 fatty acids. Desaturation efficiencies {[(product)/(substrate + product)] × 100} were calculated using the percentages of substrates and products in total fatty acids. Each value represents the mean ± standard deviation of five biological replicates.

Functional characterization of Δ5-desaturases in a strain producing 20:3ω6 and 20:4ω3.

To evaluate activity in a completely reconstituted ARA/EPA pathway, the Δ5-desaturases were expressed in an engineered S. cerevisiae strain producing 20:3ω6 and 20:4ω3. The background strain coexpressed an acyl-CoA-dependent Δ6-desaturase from the microalga O. tauri; a Δ9-desaturase, a Δ12-desaturase, and a Δ6-elongase from M. alpina; and an ω3-desaturase from S. kluyveri. Since in this strain an acyl-CoA-dependent Δ6-desaturase delivers the direct substrate for Δ6-elongation (avoiding the step of transferring the Δ6-desaturated fatty acid from phospholipid to acyl-CoA), Δ6-elongation is not expected to be a bottleneck (5). Fatty acid analysis of the background strain expressing the various Δ5-desaturases (Table 3) confirmed that the elongation step is highly efficient in this strain (substrate conversions were in the range of 85 to 95% for both the ω6 and the ω3 substrates) and showed that all the Δ5-desaturases were able to convert the endogenously produced 20:3ω6 and 20:4ω3 into ARA and EPA, respectively. Analysis of the desaturation efficiencies for the Δ5-desaturases (Fig. 4) showed that Ptet1D5D was the most efficient enzyme in both substrates, followed closely by MaD5D. OlD5D, OtD5D, and LmD5D showed very similar conversions of 20:3ω6, and OlD5D and OtD5D were more efficient toward the ω3 substrate than LmD5D.

TABLE 3.

Fatty acid composition of strains endogenously producing 20:3ω6 and 20:4ω3 and additionally expressing Ptet1D5D, OlD5D, OtD5D, LmD5D, or MaD5D^a

Fatty acid	% of total FA (mean ± SD)
	Ptet1D5D	OlD5D	OtD5D	LmD5D	MaD5D	Control
16:0	15.77 ± 1.78	19.57 ± 1.88	23.25 ± 3.26	17.12 ± 2.28	18.38 ± 0.66	15.36 ± 4.34
16:1ω9	36.32 ± 2.91	29.18 ± 2.73	29.52 ± 3.7	29.85 ± 2.82	32.44 ± 0.99	34,35 ± 2,71
18:0	4.14 ± 0.52	3.72 ± 0.51	3.56 ± 0.17	3.91 ± 0.57	3.8 ± 0.48	4.41 ± 0.48
18:1ω9/ω7	30.35 ± 2.36	24.69 ± 2.03	19.38 ± 2.17	20.17 ± 2.98	21.6 ± 0.86	24.04 ± 2.75
18:2ω6	6.46 ± 1.98	7.66 ± 1.25	8.21 ± 0.94	13.98 ± 3.31	10.12 ± 1.25	8.27 ± 3.71
18:3ω6	0.08 ± 0.01	0.1 ± 0.09	0.18 ± 0.02	0.09 ± 0.08	0,16 ± 0,02	0.06 ± 0.07
20:3ω6	0.57 ± 0.08	1.45 ± 0.26	1.48 ± 0.22	1.17 ± 0.31	1.32 ± 0.38	1.35 ± 0.21
20:4ω6	0.21 ± 0.05	0.17 ± 0	0.19 ± 0	0.17 ± 0.02	0.46 ± 0.26	0.03 ± 0.03
18:3ω3	0.54 ± 0.72	2.5 ± 0.32	3.01 ± 0.44	2.57 ± 0.5	1.59 ± 0.64	1.61 ± 0.77
18:4ω3	0.07 ± 0	0.18 ± 0.19	0.2 ± 0.06	0.1 ± 0.09	0.13 ± 0.12	0.08 ± 0.11
20:4ω3	0.62 ± 0.04	1.96 ± 0.62	1.86 ± 0.1	1.65 ± 0.67	1.22 ± 0.45	1.41 ± 0.88
20:5ω3	0.34 ± 0.04	0.42 ± 0.1	0.49 ± 0.05	0.3 ± 0.15	0.5 ± 0.29	0 ± 0.01

^aControls were the same background strain harboring the empty plasmid pSF28. The values represent the means ± SD of three independent experiments.

FIG. 4.

Desaturation efficiencies of Ptet1D5D, OtD5D, OlD5D, LmD5D, and MaD5D in a strain endogenously producing 20:3ω6 and 20:4ω3. The Δ5-desaturases were coexpressed with MaD9D, MaD12D, OtD6D, and MaD6E genes. Desaturation efficiencies were calculated as described in the legend of Fig. 3. Each value represents the mean ± standard deviation of three independent experiments.

A second Δ5-desaturase from Paramecium tetraurelia.

In the initial BLAST search, several putative front-end desaturase sequences from P. tetraurelia were found. Interestingly, each sequence was accompanied by a nearly identical (ca. 95% amino acid sequence identity) second allele in the genome of P. tetraurelia. The unicellular eukaryotic ciliate P. tetraurelia has in fact experienced a whole-genome duplication event (2), and this may explain the presence of a second, closely related sequence. Another particularity of P. tetraurelia is that TAA and TAG, stop codons in the standard genetic code, encode glutamine (27), making it necessary to codon optimize genes from this organism before expression in hosts using the universal genetic code. After verifying Δ5-desaturase activity for one of the alleles (Ptet1D5D), we sought to investigate whether the second, nearly identical allele (accession number XM_001436681; Ptet2D5D) was also functional. The codon-optimized Ptet2D5D exhibited similar enzymatic activity as Ptet1D5D, both when expressed in a 20:3ω6/20:4ω3-producing strain (Fig. 5) and when tested by exogenous supplementation of 20:3ω6 and 20:4ω3 fatty acids (data not shown). No activity was measured when 18:2ω6, 18:3ω3, or 20:2ω6 was supplied to the cultivation medium (data not shown).

FIG. 5.

Desaturation efficiencies of Ptet1D5D and Ptet2D5D in a strain endogenously producing 20:3ω6 and 20:4ω3. The Δ5-desaturases were coexpressed with MaD9D, MaD12D, OtD6D, and MaD6E genes. Desaturation efficiencies were calculated as described in the legend of Fig. 3. Each value represents the mean ± standard deviation of three independent experiments.

In addition to Ptet1D5D and Ptet2D5D, other sequences similar to front-end desaturases but either shorter or longer than the average Δ5-desaturase were identified in the P. tetraurelia genome (accession numbers XM_001450326, XM_001453029, XM_001425061, XM_001461896, XM_001434078, and XM_001425226). These sequences were also expressed in yeast and tested by feeding with exogenous fatty acids, but no Δ5-, Δ6-, or Δ8-desaturase activity was detected in the presence of the appropriate substrates (data not shown).

Acyl carrier specificities of Δ5-desaturases.

In order to investigate the acyl carrier specificity of the Δ5-desaturases from this study, distribution of fatty acids in the main lipid fractions was followed over a 24-h period after addition of exogenous 20:3ω6 to the cultures. Incorporation of the fatty acid substrate 20:3ω6 reflected an identical pattern in all the strains analyzed (Fig. 6 a for Ptet1D5D). Already at 5 min after supplementation, 20:3ω6 could be detected in the total fatty acids (TFA), steryl esters (SE), triacylglycerols (TAG) (Fig. 6a), and acyl-CoA pool (Fig. 7). The amount of 20:3ω6 in TFA, TAG, and SE pools increased during the 4 h after supplementation and remained approximately constant for up to 24 h in the first two pools, while in SE the amount of 20:3ω6 decreased drastically from the 4-h to the 24-h sample. In contrast to the immediate incorporation into TAG and SE, 20:3ω6 was detected in the PL and diacylglycerols (DAG) pools only 4 h after substrate addition.

FIG. 6.

Distribution of 20:3ω6 and total fatty acids in the lipid fractions of a Ptet1D5D-expressing yeast strain during a time course feeding experiment. Exogenous fatty acid substrate (0.5 mM 20:3ω6) was added to the yeast culture at time zero, and samples for lipid analysis were collected at the indicated time points. (a) Absolute amount of 20:3ω6 in each lipid fraction, expressed in mg/OD unit. (b) Absolute amount of total fatty acids in each lipid fraction, expressed in mg/OD unit.

FIG. 7.

Fatty acid profile of the acyl-CoA pool of a Ptet1D5D-expressing yeast strain during a time course feeding experiment. The experiment was carried out as described in the legend of Fig. 6. Samples of 1 OD unit were collected at the indicated time points (for the 24-h samples, aliquots of 8 OD units were collected), and the acyl-CoA fraction was analyzed. *, ARA peak in the 24-h sample.

The desaturation product ARA was first observed after 1 h in the TFA of Ptet1D5D- and LmD5D-expressing strains, as well as in the TAG pool of the latter, although in smaller amounts. ARA was not detected in the 5-min samples of any strain, nor in the 1-h samples of OlD5D, OtD5D, or MaD5D (Fig. 8). After 4 h, the desaturation product was present in the TFA and TAG pools of all strains. Also at this time point, ARA was first detected in the PL fraction, but only in the Ptet1D5D-expressing strains. This appearance of ARA in the TAG prior to its detection in the phospholipids suggests that desaturation takes place in the acyl-CoA pool and that the Δ5-desaturated product is transferred to the TAG via diacylglycerol acyltransferases. After 24 h the desaturation product was present in the PL of all strains except the strain expressing LmD5D (Fig. 8). In fact, ARA was not detected in the phospholipids of this strain at any time point, indicating that LmD5D does not accept the phospholipid-bound substrate. In the strains where ARA was found in the phospholipids, its percentage relative to other fatty acids was always higher in that pool than other lipids, indicating a preference for phospholipid-bound substrate. This was particularly pronounced for Ptet1D5D, where already in the 4-h sample, ARA constituted 2.3% of fatty acids in PL and 0.6% of fatty acids in TAG (corresponding to 27% and 1.9% conversion efficiencies, respectively).

FIG. 8.

Absolute amounts of ARA in the lipid fractions of Δ5-desaturase-expressing strains during time course feeding experiments. Yeast strains expressing Ptet1D5D, OtD5D, OlD5D, LmD5D, or MaD5D were cultivated as described in the legend of Fig. 6, and samples were taken at the indicated time points. No ARA was detected in any of the lipid fractions at time points 0 or 5 min.

ARA could not be detected in the acyl-CoA pool of any strain with the method used, except in the 24-h samples (Fig. 7). It is, however, likely that ARA was present at levels below the detection limit in the acyl-CoA pool, where it was efficiently taken up and accumulated as TAG and SE, thus just passing through the acyl-CoA pool at a high turnover rate. Interestingly, in contrast to 20:3ω6, ARA was never detected in the SE pool of any strain, which indicates that Are1p and Are2p acyltransferases cannot use that fatty acid as substrate. ARA also could not be detected in the DAG pool, indicating that it is not used for DAG synthesis by Gat1p, Gat2p, and Slc1p but is efficiently channeled to TAG by Dga1p activity. This is furthermore supported by the fact that ARA appears as TAG prior to other lipid pools.

DISCUSSION

Production of very-long-chain polyunsaturated fatty acids in transgenic organisms is a process still far from optimal. Heterologous expression of desaturases and elongases usually results in low PUFA yields in the host organism compared to the levels observed in native PUFA producers. The overall efficiency of the PUFA pathway is likely to be influenced by lipid dynamics inside the cell, where interplay between desaturases, elongases, and acyltransferases may be determinants. For example, a higher flux in the pathway is to be expected when both desaturases and elongases have an identical acyl carrier preference, using CoA-activated fatty acids as substrate (4, 8). On the other hand, competition for that same substrate by some acyltransferases present in the host may limit the flux through the pathway, since such competing reactions transfer the acyl-CoA substrate to other lipids which are not substrates for elongation and desaturation. In the present study, the time-dependent incorporation of the substrate 20:3ω6 into different lipids provides some hints on the activities of these acyltransferases in S. cerevisiae. The exogenously supplied 20:3ω6 appeared in the TAG and SE lipid fractions already at 5 min after addition of this fatty acid to the medium, while it was detected in the phospholipids and DAG only after 4 h. This pattern of 20:3ω6 assimilation agrees well with previous findings that exogenous free fatty acids are CoA activated as soon as they enter yeast cells (4) and indicates that diacylglycerol and sterol acyltransferases (Dga1p, Are1p, and Are2p) are very active with the CoA-activated form of 20:3ω6 as substrate. In contrast, the later appearance of 20:3ω6 in DAG and PL compared to TAG (and in smaller amounts) may indicate that the acyltransferases involved in DAG synthesis in S. cerevisiae (i.e., Gat1p, Gat2p, and Slc1p) are less active on CoA-activated 20:3ω6 than the DAG-acyltransferases Dga1p, Are1p, and Are2p. Another possible explanation is that TAG is formed from preexisting DAG and that the replenishment of DAG is achieved through the activity of phospholipase C on the phospholipids, which initially decreased (Fig. 6b). In the context of a fully reconstituted PUFA pathway, the highly active native diacylglycerol and sterol acyltransferases may compete with Δ5-desaturases for CoA-activated 20:3ω6, and acyl-CoA-dependent Δ5-desaturases may therefore be less efficient than phospholipid-dependent versions due to limited substrate availability.

The overall most efficient desaturase in the present study, Ptet1D5D, was highly active on phospholipid-linked substrates but also seemed to have the capacity to act on the acyl-CoA substrate. In contrast, LmD5D, which exclusively preferred the acyl-CoA substrate, was the least efficient enzyme, illustrating that exclusive acyl-CoA substrate specificity is not an advantage per se. In the Ptet1D5D-expressing strain, Δ5-desaturation in the PL fraction was evident at the earliest possible sampling point, i.e., as soon as the 20:3ω6 substrate was detected in PL. At this sampling point, ARA was predominantly found in PL and to a lesser extent in TAG. At the later sampling point, increased amounts of ARA were found both in the PL and in TAG. This may indicate that Δ5-desaturation occurs on acyl-CoA substrates followed by incorporation into TAG by acyltransferases, or that PL-bound fatty acids are released and taken up into TAG. Both of these models are supported by our data.

Analyses of lipid distribution in different pools performed after 24 or 48 h of incubation have in the past been used to distinguish between lipid-linked and acyl-CoA-dependent Δ5-desaturases (4, 8). A predominant accumulation of desaturation products in the phospholipids was considered to be characteristic of lipid-linked Δ5-desaturases, whereas a roughly equal distribution between phospholipids and neutral lipids was associated with acyl-CoA-dependent Δ5-desaturases. In our results, we observed approximately equal distribution of ARA in the phospholipids and TAG of strains expressing Ptet1D5D, OlD5D, and OtD5D after 24 h of incubation, while in the MaD5D-expressing strain larger amounts of ARA accumulated in the phospholipids compared to the TAG. However, our results also show that initial desaturation in MaD5D, Ptet1D5D, OtD5D, and OlD5D occurs in the acyl-CoA until the substrate is available in the PL, subsequently taking place there with higher conversion efficiency. Therefore, assessing acyl carrier specificity based solely on measurement at a single time point does not appear to be feasible. Altogether, this leads us to conclude that acyl carrier specificity is not the only important feature of front-end desaturases and we hypothesize that certain promiscuity toward the acyl carrier used as substrate is beneficial for overall Δ5-desaturation efficiency.