Δ12-Fatty Acid Desaturase from Candida parapsilosis Is a Multifunctional Desaturase Producing a Range of Polyunsaturated and Hydroxylated Fatty Acids

Numerous Δ12-, Δ15- and multifunctional membrane fatty acid desaturases (FADs) have been identified in fungi, revealing great variability in the enzymatic specificities of FADs involved in biosynthesis of polyunsaturated fatty acids (PUFAs). Here, we report gene isolation and characterization of novel Δ12/Δ15- and Δ15-FADs named CpFad2 and CpFad3, respectively, from the opportunistic pathogenic yeast Candida parapsilosis. Overexpression of CpFad3 in Saccharomyces cerevisiae strains supplemented with linoleic acid (Δ9,Δ12-18:2) and hexadecadienoic acid (Δ9,Δ12-16:2) leads to accumulation of Δ15-PUFAs, i.e., α-linolenic acid (Δ9,Δ12,Δ15-18:3) and hexadecatrienoic acid with an unusual terminal double bond (Δ9,Δ12,Δ15-16:3). CpFad2 produces a range of Δ12- and Δ15-PUFAs. The major products of CpFad2 are linoleic and hexadecadienoic acid (Δ9,Δ12-16:2), accompanied by α-linolenic acid and hexadecatrienoic acid (Δ9,Δ12,Δ15-16:3). Using GC/MS analysis of trimethylsilyl derivatives, we identified ricinoleic acid (12-hydroxy-9-octadecenoic acid) as an additional product of CpFad2. These results demonstrate that CpFAD2 is a multifunctional FAD and indicate that detailed analysis of fatty acid derivatives might uncover a range of enzymatic selectivities in other Δ12-FADs from budding yeasts (Ascomycota: Saccharomycotina).


Introduction
Unsaturated fatty acids (UFAs) play a key role in maintenance of optimal physical and biological properties of cell membranes [1]. UFAs are biosynthesized from saturated fatty acids by two evolutionary unrelated classes of fatty acid desaturases: soluble fatty acid desaturases, which are expressed exclusively in plant plastids, and membrane-bound fatty acid desaturases (subsequently referred to as FADs), which are widespread in eukaryotes and some prokaryotes [2].
FADs are highly specific towards their fatty acyl substrates (either acyl-CoA or acyl-lipid) and towards the position and geometric configuration of the newly introduced double bond. A range of desaturase regioselectivities dependent on the reference point used by FADs to position the introduced double bond has been described. The main regioselective modes are: (1) the double bond is introduced between specific carbon atoms counted from the carboxy terminus (DX) or (2) methyl terminus (vX) of the fatty acyl substrate, and (3) a subsequent double bond is introduced a specific number of carbon atoms (usually three) counted from the pre-existing double bond (n+3) [3]. However, these FAD regioselectivities are not mutually exclusive, and Meesapyodsuk et al. [4] suggested assigning them primary and secondary modes to more precisely describe FAD regioselectivity. For example, multifunctional FADs preferentially producing D9,D12-18:2 and D9,D12-16:2 from D9-UFAs but also capable of producing minor amounts of D9,D12,D15-18:3 and D9,D12,D15-16:3 might be termed as v+3 (D12), indicating that the primary desaturase regioselectivity is v+3 and the preferred regioselectivity is D12. However, this nomenclature may require detailed information on FAD diagnostic substrates and products, which are often not available in the literature. Therefore, we will adhere to nomenclature designating FADs as D12and/or D15-FADs.
The role of D9-FADs in pathogenic Candida sp. has been demonstrated in C. albicans, in which partial repression of Ole1p D9-FAD blocks hyphae and chlamydospore formation [25]. In C. parapsilosis, gene deletion of OLE1 impaired invasive growth, pseudohyphae formation and virulence in mice and increased susceptibility to macrophages and various stress factors such as SDS, salts and H 2 O 2 [26].
The role of PUFAs is more elusive than that of D9-UFAs, and PUFAs are probably not essential under all growth conditions. Murayama et al. [27] demonstrated that the loss of PUFA production in C. albicans via targeted disruptions of the D12and D15-FAD genes CaFAD2 and CaFAD3, respectively, did not lead to any changes in growth rate under low temperature, chlamydospore formation, hyphal formation or virulence under the conditions tested. In contrast, the hyphal form of C. albicans is enriched in PUFAs compared to the yeast form which suggests a specific role for PUFAs in C. albicans morphogenesis [28].
In this study, we isolated the coding regions CpFAD2 and CpFAD3, homologs of fungal D12and D15-FADs, respectively, from C. parapsilosis, an emerging fungal pathogen [29]. We expressed these FADs in a Saccharomyces cerevisiae expression system and identified a broad range of novel polyunsaturated and hydroxylated FA products, including an unusual D9,D12,D15-16:3 with a terminal double bond.

Strains, media and growth conditions
Saccharomyces cerevisiae strain BY4741 (MATa his3D1 leu2D met15D ura3D; EUROSCARF, Germany) was cultivated in liquid medium lacking uracil (YNBglc-U: 0.67% yeast nitrogen base without amino acids, 2% glucose, supplemented with Brent supplement mix without uracil according to the manufacturer's instructions). In cultivation media, 2% glucose was used as the carbon source. Heterologous protein expression was induced in YNB-U media containing 2% galactose as the sole carbon source (YNBgal-U). When indicated, media were supplemented with 0.5 mM linoleic acid (Sigma-Aldrich), 0.25 mM hexadecadienoic acid (Z9,Z12-16:2; Larodan), 1% tergitol and 0.65 M NaCl. C. parapsilosis (clinical isolate P-69 obtained from the mycological collection of the Faculty of Medicine, Palacky University, Olomouc, Czech Republic) was cultivated prior to lipid extraction in YNBglc medium at 37uC until the culture reached a stationary growth phase.

Sequence and phylogenetic analysis
The assembly of shotgun reads of C. parapsilosis genome publicly available in Candida Genome database was searched using the BLAST tool (http://www.candidagenome.org/cgi-bin/compute/ blast_clade.pl) with previously characterized D12and D15-FADs as a query.
Amino acid sequences of FADs were aligned using the Muscle algorithm [30], and the phylogeny of FADs was reconstructed with MEGA5 software [31] by neighbor-joining method with calculated bootstrap support from 1,000 bootstrap replicates as a measure of statistical reliability. Desaturase topology was predicted using the programs TMHMM 2.0 [32] and HMMTOP [33]. Pairwise sequence alignment was performed using the EMBOSS Needle web-based tool (http://www.ebi.ac.uk/Tools/psa/ emboss_needle/).
Known FAD sequences were retrieved from GenBank. The FAD sequences reported here were deposited into GenBank under accession numbers FN386265 and FN386266.

Heterologous expression in Saccharomyces cerevisiae
Total cellular DNA was extracted from C. parapsilosis strain CP-69 and served as a template for PCR amplification of the CpFAD2 coding region using the following primers: CpFAD2-59 (CTG GAG CTC ATG TCT TCA GCC ACA ACT TC) and CpFAD2-39 (CTG CTC GAG TTA TTT CTC TTT CTT TGG GT) (SacI and XhoI restriction sites are underlined; start and stop codons are italicized). The CpFAD3 coding region was amplified using the primers CpFAD3-59 (CAG GAG CTC ATG AGT ACG GTC CAT GCA TC) and CpFAD3-39 (CTG CTC GAG CTA GTT TCT TGG TTT GAC AC). The resulting fragments were cloned into the vector pYES2 (Invitrogen) under control of the galactoseinducible GAL1 promoter, generating plasmids CpFAD2-pYES and CpFAD3-pYES. The constructs were verified by sequencing.
S. cerevisiae strain BY4741 was transformed with CpFAD2-pYES, CpFAD3-pYES and empty pYES2 plasmid by electroporation. The resulting CpFAD2, CpFAD3 and Empty yeast strains were selected on YNBglc-U agar plates. Single colonies were inoculated into 20 ml of YNBglc-U and cultivated in a rotary shaker (30uC, 220 RPM). At the early stationary phase, yeast cells were harvested by centrifugation (3 min, 1000 g, 20uC), washed with 20 ml of YNBgal-U and transferred into 20 ml of fresh YNBgal-U (optionally supplemented with linoleic acid, hexadecadienoic acid and tergitol). Yeast strains were cultivated for 72 h.

Lipid extraction and fatty acid methyl ester preparation
Yeast cells were harvested and total cellular lipids were extracted as described by Buček et al. [34], and fatty acid methyl esters (FAMEs) were prepared according to Matoušková et al. [35]. The resulting FAMEs were extracted with hexane (600 ml), and the extracts were analyzed by gas chromatography coupled to mass spectrometry (GC/MS) using the conditions described below. The relative abundances of individual fatty acids were calculated from the peak areas in GC/MS total ion current chromatograms.

DMOX derivatization
Fatty acid 4,4-dimethyloxazoline (DMOX) derivatives were prepared from FAME extracts according to Fay and Richli [36]. Briefly, hexane solvent was evaporated under a stream of nitrogen, and FAMEs were heated at 180uC overnight with 0.5 g 2-amino-2-methylpropanol. After cooling, the DMOX derivatives were dissolved in 5 ml dichloromethane and washed three times with 2 ml distilled water. The dichloromethane phase was dried over anhydrous MgSO 4 , evaporated under a stream of nitrogen and redissolved in hexane. DMOX derivatives were analyzed using GC/MS under the conditions described below and identified by comparing their mass spectra to previously described spectra and using inferred empirical rules [37].

TMS derivatization
Hydroxy FAs were analyzed in the form of their trimethylsilyl (TMS) derivatives by treating the FAME extracts with excess N,Obis(trimethylsilyl)acetamide (Sigma-Aldrich) in acetonitrile (10 min, 40uC), evaporated under a gentle stream of N 2 and redissolved in chloroform [24]. Freshly prepared TMS derivatives were analyzed using GC/MS under the conditions described below, and their mass spectra and retention behavior were compared to the previously published mass spectra of FAME TMS derivatives [24] and to that of a TMS derivative prepared from 1 mg methyl ricinoleate (methyl 12-hydroxyoleate, Sigma-Aldrich).

GC/MS analysis
The FAME extracts and respective DMOX and TMS derivatives were analyzed with a 7890A gas chromatograph coupled to a 5975C mass spectrometer, equipped with electron ionisation and quadruple analyser (Agilent Technologies, Santa Clara, CA, USA) using DB-5MS or DB-WAX capillary columns (both J&W Scientific, Folsom, CA, USA; 30 m60.25 mm, film thickness 0.25 mm); the electron ionisation was set at 70 eV.
Conditions for the analysis of FAMEs and DMOX derivatives were as follows: carrier gas, He: 1 ml/min; 10:1 split ratio, injection volume 2 ml; injector temperature 220uC; thermal gradient 140uC to 245uC at 3uC/min, then at 8uC/min to 280uC and temperature held for 5 min. The temperature program was terminated at 245uC and held at this temperature for 10 min when the DB-WAX column was used. The TMS derivatives were analyzed on a DB-5MS column using the following thermal gradient: 50uC to 140uC at 10uC/min, then 1uC/min to 198uC, 3uC/min to 320uC and temperature held for 3 min.
In the reconstructed phylogenetic tree of functionally characterized fungal D12and D15-FADs, all FADs cluster into clades according to the class or subphylum of the source organism, with the exception of FADs from the filamentous yeasts Aspergillus nidulans, Fusarium moniliforme and Magnaporthe grisea, which create a group of multifunctional D15/D12 FADs [9]. Within the clades, FADs cluster generally into ''D12-FAD'' subclades and ''D15-FAD'' subclades. CpFad2 clusters within a subclade of strictly monofunctional D12-FADs, whereas CpFad3 belongs to a subclade of D15-FADs (Fig. 1.).

Functional characterization of CpFAD2 and CpFAD3 in S. cerevisiae
We designed two pairs of specific primers to amplify the CpFAD2 and CpFAD3 ORFs, and we obtained a single amplicon for each primer pair from genomic DNA isolated from C. parapsilosis. Sequencing of CpFAD2 and CpFAD3 confirmed that the nucleotide sequences of the amplified coding regions are identical to those obtained from the C. parapsilosis genome database.
The sequence comparison of CpFad2 and CpFad3 with previously characterized FADs strongly suggests that CpFad2 is a D12-FAD, whereas CpFad3 is a D15-FAD. However, the homology-based functional annotation of FADs may fail to predict FAD substrate specificity [39] or regioselectivity [14]. To experimentally determine the specificities and regioselectivities of CpFad2 and CpFad3, we cloned their coding regions into the pYES2 vector under control of a galactose-inducible promoter. We transformed the resulting plasmids into S. cerevisiae, generating the CpFAD2 and CpFAD3 strains.
The total lipidic extracts of galactose-induced CpFAD2 and CpFAD3 strains were transesterified, and the resulting fatty acid methyl esters (FAMEs) were compared to the FAME extract from a control strain transformed with an empty plasmid (Empty strain).
To determine the distribution of novel PUFAs in individual lipid classes, the total lipidic extract was analyzed using liquid chromatography with mass spectrometric analysis. The preliminary data indicates that PUFAs are distributed in phospholipids, i.e., phosphatidylserine, phosphatidylethanolamine, phosphatidylcholine, and triacylglycerols (data not shown). We have also investigated the effect of accumulation of PUFAs in the CpFAD2 and CpFAD3 strains on their growth rate and tolerance to alkali-  metal cations. When compared to the Empty strain, we observed a decreased growth rate for yeasts expressing CpFAD2 on YNBgal-U agar plates (Fig. S4B). The decreased growth rate of the CpFAD2 strain became even more prominent on YNBgal-U containing 0.65 M NaCl (Fig. S4C). The growth rate of the CpFAD3 strain on YNBgal-U or YNBgal-U supplemented with linoleic acid was unaltered compared to that of the Empty strain (Fig. S4B, D).

Hydroxylation activity of CpFad2 and CpFad3
To determine whether CpFad2 and CpFad3 are capable of FA hydroxylation, we treated the FAME extracts with N,O-bis(trimethylsilyl)acetamide to convert hydroxy FAMEs to their corresponding trimethylsilyl-FAMEs, which exhibit better chromatographic properties and provide MS fragment ions characteristic of particular TMS group locations [42]. Trace amounts of TMS derivatives of methyl 9-hydroxypalmitic acid (MS fragment ions at m/z 201 and 259) and methyl 9-hydroxystearic acid (MS fragment ions at m/z 229 and 259) were detected in all transformed yeast strains (Fig. 5, S5 and Table 2), providing evidence of the intrinsic D9-fatty acid hydroxylase activity of yeast D9-FAD Ole1p [22,24]. Notably, in the CpFAD2 strain, an additional TMS derivative with low abundance (0.10% 6 0.02%) was detected ( Table 2). This was  determined to be a TMS derivative of methyl ricinoleate (methyl 12-hydroxyoleate) based on its retention time, which was identical to that of TMS derivative of methyl ricinoleate standard, and on the presence of characteristic fragment ions, namely (M-15) ion at m/z 369 and ions at m/z 187, 270 and 299 (Fig. 5A). This observation indicates that CpFAD2 exhibits 12-hydroxylation activity.

Discussion
Fungal FAD research is largely motivated by the search for novel FADs that could be used in metabolic engineering of microorganisms to produce PUFAs and other valuable UFAs on an industrial scale [43,44]. Additionally, desaturation, as a part of fungal fatty acid metabolism, has been shown to play a crucial role in the growth and morphogenesis of pathogenic yeast species in plants and humans. Therefore, FADs have been suggested as potential targets for antifungal drugs [25][26][27]45,46].
Nguyen et al. [26] described the role of D9-FAD in the pathobiology of C. parapsilosis, a fungal pathogen that forms persistent biofilms on implanted medical devices such as catheters [29]. In addition to saturated and monounsaturated fatty acids, C. parapsilosis cellular lipids contain a major fraction of PUFAs, suggesting activity of D12and D15-FADs [16]. In this study, we identified and characterized CpFad2 and CpFad3 desaturases responsible for production of PUFAs in C. parapsilosis.
Sequence analysis of CpFad2 and CpFad3 using TMHMM 2.0 and HMMTOP topology prediction algorithms indicated the presence of more than four transmembrane helices. The proposed topology, however, does not satisfy the requirement of arrangement of conserved histidine motifs on a common, cytosolic ER membrane face [47,48] (Fig. S1). Therefore, the third and fourth transmembrane helix likely represents a hydrophobic region peripherally associated with an ER membrane. This topology was previously proposed by Hoffmann et al. for D12-FAD and a bifunctional D12/D15-FAD from Aspergillus nidulans [10]. Alternatively, some of the additional TM helices might be correctly predicted and play a role in binding the acyl-lipid substrate, as hypothesized by Diaz et al. [49].
Previously, D9,D12,D15-16:3 was not detected as a product of CaFad2 heterologously expressed in S. cerevisiae under control of the GAL1 promoter. CaFad2 is a D12-FAD from C. albicans, which can produce D9,D12-16:2 and D9,D12-18:2 fatty acids [27]. However, the presence of D9,D12,D15-16:3 might be easily missed due to its overall low abundance or due to the low abundance of the molecular ion at m/z 264 and low abundance of fragments at m/z 74 and 87 characteristic of FAMEs (Fig. S3).
Production of a-linolenic, hexadecatrienoic (D9,D12,D15-16:3) and ricinoleic acid (12-hydroxyoleate) by CpFad2 suggests that CpFad2 is a bifunctional D12/D15-FAD. The D15-desaturase activity of D12-FAD from budding yeast (Saccharomycotina) has been reported only by Yan et al. [17], and the hydroxylase activity has not been investigated. Damude et al. suggested that D15-FADs independently evolved from ancestral D12-FADs multiple times [9]. The facile evolutionary shift between D12and D15-FAD regioselectivities was further supported by the effect of site-directed mutagenesis of a single amino acid residue [4] and by domainswapping experiments [4,10]. Analogously, one to four amino acid mutations were found to be sufficient to convert plant D12-FAD Table 2. Relative abundances of TMS derivatives of hydroxy FAs in FAME extracts from CpFAD2, CpFAD3 and Empty strains.  into a bifunctional D12-FAD/hydroxylase [22,50]. Based on these observations and our current results, we hypothesize that minor D15-desaturase regioselectivity and hydroxylase activity might be present in numerous functionally characterized D12-FADs from budding yeasts, including CaFad2 from C. albicans [27]. The fact that minor PUFA and hydroxylated products were rarely detected in previously described D12-FADs from budding yeast might be caused by a low level of desaturase expression under the experimental conditions, by a low level of accumulation of PUFAs and hydroxylated fatty acids or by the low sensitivity of the FAME analysis procedure. The reason underlying the absence of specific D15-hydroxylated FAs in CpFAD3 strain supplemented with linoleic acid might be the absence of CpFad3 D15-hydroxylase activity and/or the overall low enzymatic activity of CpFad3 compared to CpFad2 in our expression system. Based on the range of observed desaturase specificities of CpFad2 [its capability of introducing a double bond at the D15 (v1) position in D9,D12,D15-16:3 and D15(v3) position in D9,D12,D15-18:3 and the lack of D9,D15-18:2 products], we propose that the primary regioselective mode of CpFad2 is n+3 and the secondary mode is D12. This would imply, in accordance with the FAD regioselectivity classification described by Meesapyodsuk et al. [4], that the preexisting double bond serves as a reference point for positioning of the introduced double bond and that the D12-position is preferred over the D15-position.
Overexpression of D12and D15(v3)-FADs in S. cerevisiae provides a tool to study the influence of PUFAs on yeast physiology [19,[53][54][55]. The PUFAs were present in phospholipid fraction of the CpFAD2 and CpFAD3 strain supplemented with linoleic acid, suggesting that they might influence the membrane properties and the yeast phenotype. However, we did not observe any increase in tolerance to NaCl, which has been observed in yeast strains producing PUFAs [55]. A decreased growth rate of yeasts heterologously expressing D12-FADs was previously attributed to impairment of tryptophan uptake [54]. Although we employed the BY4741 yeast strain which is not auxotrophic for tryptophan, the modified structure and physical properties of yeast cell membranes containing PUFAs might affect the properties of numerous cell membrane proteins and result in a decreased growth rate.
In C. parapsilosis, the low amount or complete absence of C16 PUFAs and low amount of palmitoleic acid is in general agreement with the results of a previous study by Moss et al. [16] that describes the fatty acid composition of various Candida species. Together, these data suggests that in Candida species, C16 PUFAs do not accumulate. The low content of C16 PUFAs might be a consequence of a low amount of their precursor, palmitoleic acid. Alternatively, overexpression of CpFad2 under control of the GAL1 promoter might result in higher expression and therefore higher desaturase activity of CpFad2 in S. cerevisiae, as compared to the activity of CpFad2 in C. parapsilosis.
Taken together, this study provides further evidence that, despite the growing database of functionally characterized FADs, detailed characterization of FADs by mass spectrometry analysis of fatty acid derivatives can reveal surprising desaturase specificities that cannot be inferred solely from sequence comparisons. Figure S1 Amino acid sequence alignment of CpFad2 and CpFad3. Three conserved histidine-rich regions (H1-H3) are marked by boxes. Predicted transmembrane domains for CpFad2 and CpFad3 are indicated by bars above or below the sequence, respectively. The consensus transmembrane region predicted by both HMMTOP and TMHMM 2.0 algorithms are indicated by solid bars; transmembrane regions predicted by only one algorithm are indicated by dashed bars. Identical residues are indicated by a black background, similar residues by a grey background. (TIF) Figure S2 Mass spectra of DMOX derivative of D11,D14-18:2-methylester detected in FAME extract from CpFAD2 yeast strain. Characteristic fragments are highlighted, and the fragmentation pattern of the DMOX derivative is shown above the spectra.