A complete enzymatic capacity for biosynthesis of docosahexaenoic acid (DHA, 22 : 6n–3) exists in the marine Harpacticoida copepod Tigriopus californicus

The long-standing paradigm establishing that global production of Omega-3 (n-3) long-chain polyunsaturated fatty acids (LC-PUFA) derived almost exclusively from marine single-cell organisms, was recently challenged by the discovery that multiple invertebrates possess methyl-end (or ωx) desaturases, critical enzymes enabling the biosynthesis of n-3 LC-PUFA. However, the question of whether animals with ωx desaturases have complete n-3 LC-PUFA biosynthetic pathways and hence can contribute to the production of these compounds in marine ecosystems remained unanswered. In the present study, we investigated the complete enzymatic complement involved in the n-3 LC-PUFA biosynthesis in Tigriopus californicus, an intertidal harpacticoid copepod. A total of two ωx desaturases, five front-end desaturases and six fatty acyl elongases were successfully isolated and functionally characterized. The T. californicus ωx desaturases enable the de novo biosynthesis of C18 PUFA such as linoleic and α-linolenic acids, as well as several n-3 LC-PUFA from n-6 substrates. Functions demonstrated in front-end desaturases and fatty acyl elongases unveiled various routes through which T. californicus can biosynthesize the physiologically important arachidonic and eicosapentaenoic acids. Moreover, T. californicus possess a Δ4 desaturase, enabling the biosynthesis of docosahexaenoic acid via the 'Δ4 pathway'. In conclusion, harpacticoid copepods such as T. californicus have complete n-3 LC-PUFA biosynthetic pathways and such capacity illustrates major roles of these invertebrates in the provision of essential fatty acids to upper trophic levels.


Introduction
(PKS) complex is involved in the anaerobic pathway existing in prokaryotes and some eukaryotic microorganisms [6]. However, most eukaryotes operate the aerobic pathway of n-3 LC-PUFA biosynthesis that entails two critical components; one is the biosynthesis of n-3 C 18 polyunsaturated fatty acid (PUFA), namely α-linolenic acid (ALA, 18 : 3n-3), from saturated fatty acids (SFA) such as stearic acid (18 : 0); another is the biosynthesis of n-3 LC-PUFA including EPA and DHA from ALA (figure 1). Along with the aerobic n-3 LC-PUFA biosynthetic pathways, a varied range of enzymes catalyses multi-step desaturation (introduction of new double bonds or unsaturations) and elongation (extension of the carbon acyl chain) reactions. Initially, palmitic acid (16 : 0) is synthesized via the fatty acid synthase (FAS) system and then elongated to 18 : 0. Subsequently, a Δ9 desaturase (e.g. stearoyl-CoA desaturase) introduces the first double bond to produce the monounsaturated fatty acid (MUFA) oleic acid (18 : 1n-9) [7]. These enzymatic capacities enabling the biosynthesis of SFA and MUFA are virtually present in all eukaryotes [8,9]. However, enzymes enabling the de novo biosynthesis of PUFA (i.e. introducing a second double bond into oleic acid) have a more restricted distribution and, as discussed below, were believed to be largely absent from animals. Specifically, the de novo biosynthesis of PUFA typically requires the action of a particular type of desaturase enzymes called methyl-end (or ωx) desaturases since they introduce a new double bond between a pre-existing one and the methyl-terminus of the carbon chain [10]. Thus, there exist ωx desaturases that introduce a new double bond into oleic acid at Δ12 position producing the PUFA linoleic acid (LA, 18 : 2n-6). Others introduce a further double bond into LA at Δ15 position and thus produce the n-3 C 18 PUFA ALA (figure 1). Both LA and ALA become substrates from which n-6 and n-3 LC-PUFA, respectively, can be biosynthesized through sequential reactions catalysed by frontend desaturases, which introduce a new double bond between the pre-existing one and the carboxyl-terminus of the fatty acid, and fatty acyl elongases, which catalyse the initial condensation step of the fatty acid elongation pathway (figure 1) [11,12]. Animals possess a varied complement of front-end desaturases and fatty acyl elongases enabling them to produce LC-PUFA from the C 18 PUFA precursors LA and ALA [10,13].
Challenging the widely accepted dogma establishing that, with few exceptions (e.g. [14][15][16][17]), animals lack the ability to biosynthesize PUFA de novo, Kabeya et al. [18] demonstrated that a plethora of invertebrates including cnidarians, nematodes, lophotrochozoans and arthropods possess ωx desaturases enabling them to biosynthesize PUFA de novo. Importantly, it was further established that invertebrates also have ωx desaturases commonly known as 'ω3 desaturases', which enable conversions of multiple n-6 fatty acids into the corresponding n-3 desaturated products including LC-PUFA such as EPA and DHA [18][19][20]. These findings have obvious ecological implications associated with the contribution that these invertebrates can have on the primary production of the essential compounds n-3 LC-PUFA at a global scale. This is particularly true for groups that are abundant and widely distributed in marine ecosystems.
Copepods play critical roles in the trophic ecology of marine ecosystems not only because they are the most royalsocietypublishing.org/journal/rsob Open Biol. 11: 200402 abundant zooplanktonic crustaceans but also because they represent key prey items ensuring the transfer of essential nutrients to upper trophic level organisms. Indeed, copepods contain high levels of n-3 LC-PUFA, particularly EPA and DHA [21], and thus these crustaceans become critically important to guarantee the provision of these essential nutrients to high trophic level organisms such as fish [22]. While the abovementioned abundance of LC-PUFA in primary producers suggests that the high LC-PUFA content of copepods' lipids can have a dietary origin, it is unclear to what extent copepods can modulate their own fatty acid profiles and contribute to their characteristic high LC-PUFA lipid profiles. Early studies using 14 C-labelled fatty acids suggested some copepods showed the ability to bioconvert PUFA into LC-PUFA [23,24]. Studies involving feeding trials using LC-PUFA-deficient diets or stable isotope labelled fatty acids provided further evidence that Harpacticoida and Cyclopoida copepods possess some capacity to produce n-3 LC-PUFA endogenously [25][26][27][28][29][30][31][32][33]. However, these studies could not unequivocally establish that the abovementioned metabolic activities observed were indeed due to the copepod's enzymatic complement, and hence it is difficult to completely rule out that LC-PUFA are rather synthesized by microbes that coexist within the copepod. More robust evidence has been recently collected when genes encoding ωx desaturases were found in several species from groups of copepods including Cyclopoida, Harpacticoida and Siphonostomatoida [18]. However, with the sole exception of Lepeophtheirus salmonis (Siphonostomatoida), none of copepod ωx desaturases has been functionally characterized yet, and therefore, it remains unknown whether the activities of the copepod ωx desaturases are conserved across the group or, on the contrary, have diversified during evolution as occurred for other fatty acyl desaturases [13]. In addition, it is unclear whether, beyond ωx desaturases, copepods have further enzymes such as front-end desaturases and fatty acyl elongases that, along ωx desaturases, enable complete enzymatic activities allowing the production of a variety of LC-PUFA including DHA (figure 1). Recent studies involving genomic and transcriptomic analyses identified sequences of desaturase and elongase genes with putative roles in the LC-PUFA biosynthesis in Harpacticoida and Cyclopoida copepods [31][32][33][34]. However, no functional evidence demonstrating the actual role of these enzymes in the LC-PUFA biosynthetic pathways was reported.
The marine intertidal Harpacticoida copepod Tigriopus californicus has been used as a model organism for the study in several disciplines including ecology, evolution and genetics [35][36][37]. In this study, we investigated the LC-PUFA biosynthetic capability of T. californicus by performing a comprehensive search for genes encoding putative ωx desaturases, front-end desaturases and fatty acyl elongases in T. californicus transcriptomic and genomic datasets. A total of 13 genes encoding two ωx desaturases, five front-end desaturases and six fatty acyl elongases were successfully identified. We further cloned and functionally characterized the 13 T. californicus desaturases and elongases by heterologous expression in yeast demonstrating that T. californicus has a comprehensive enzymatic set enabling the endogenous production of LC-PUFA including EPA and DHA. These results suggest that harpacticoid copepods such as T. californicus can be a net producer of the physiologically important n-3 LC-PUFA in marine supralittoral ecosystems.

Material and methods
2.1. RNA extraction and cDNA synthesis Live T. californicus were purchased from Reefphyto Ltd. (Newport, UK). Total RNA was extracted from pooled (approx. 50) whole individuals using TRI Reagent (Sigma-Aldrich, Dorset, UK) following the manufacturer's recommendations. The extracted RNA was treated with RQ1 RNase-Free DNase (Promega, Madison, WI, USA) to remove potential genomic DNA contamination. Subsequently, the complementary DNA (cDNA) was synthesized from 2 μg of total RNA using Super-Script III First-Strand Synthesis System for RT-PCR (Thermo Fisher Scientific, Waltham, MA, USA). The obtained cDNA was stored at −20°C until further use. All experiments were carried out in compliance with the guidelines for the care and use of laboratory animals of the Tokyo University of Marine Science and Technology.

Isolation of putative desaturase and elongase genes from Tigriopus californicus
In addition to the two ωx desaturase sequences identified by Kabeya et al. [18], termed herein as ωx1 (NCBI acc. No. JW524768) and ωx2 (GBTC01004759), further genes encoding PUFA biosynthesizing enzymes including putative front-end desaturases and fatty acyl elongases were retrieved from the T. californicus transcriptome shotgun assembly (NCBI BioProject No. PRJNA158547 and PRJNA263967) by DELTA-BLAST. Several functionally characterized front-end desaturases and fatty acyl elongases from invertebrates including molluscs and echinoderms were used as queries [38][39][40][41][42][43]. The DELTA-BLAST search allowed the identification of a total of 21 front-end desaturase-like and 31 fatty acyl elongase-like sequences that were subsequently assembled to generate several consensus sequences. As a result, full-length open reading frames (ORF) of five putative front-end desaturases and five putative elongases were obtained, which were named as Fed1 to Fed5 and Elo1 to Elo5, respectively. After this search, a new transcriptomic dataset became available at NCBI (No. PRJNA504307). Thus, the same search against the new database was repeated and an additional elongaselike sequence was successfully obtained, which was named as Elo6. We then performed blastn search against T. californicus genome assembly (https://i5k.nal.usda.gov/Tigriopus_ californicus) and confirmed all five front-end desaturase and six fatty acyl elongase sequences were successfully found in the T. californicus genome. Apart from the T. californicus ωx1 (JW524768) and ωx2 (GBTC01004759) found by Kabeya et al. [18], no further ωx desaturases were identified in the T. californicus genome.

Phylogenetic analysis
Phylogenetic trees comparing the deduced amino acid (aa) sequences of front-end desaturases and fatty acyl elongases from a wide range of eukaryotic organisms were constructed using the maximum-likelihood (ML) method [44]. A comprehensive phylogenetic analysis of ωx desaturases including the two of T. californicus was carried out by Kabeya et al. [18] and hence, it was not performed in the present study. The sequence dataset for the ML phylogenetic analysis of front-end royalsocietypublishing.org/journal/rsob Open Biol. 11: 200402 desaturases was built through the following steps: (i) retrieval of all eukaryotic aa sequences from RefSeq-specific proteincontaining 'Delta6-FADS-like' domain (cd03506); (ii) selection of representative species from each taxonomic group; (iii) addition of functionally characterized genes and copepod genes and (iv) generation of multiple sequence alignment (MSA) using MAFFT v. 7 [45] with FFT-NS-i method (-reorder -auto), and removal of all duplicated sequences and incomplete gappy sequences. A similar strategy was applied for elongases as follows: (i) retrieval of all metazoan aa sequences from RefSeq-specific protein-containing 'ELO' domain (pfam01151); (ii) and (iii) were the same as above and (iv) generation of cleaned MSA as described above and removal of potential non-PUFA elongase sequences. Each dataset was then aligned using MAFFT v. 7 [45] with E-INS-i method (-genafpair -maxiterate 1000) and filtered through using GUIDANCE v. 2.0 [46] with 100 bootstrap replicates, sequence and column masking cut-off threshold less than 0.5. Subsequently, the resulting MSA were filtered to delete columns containing gaps that were greater than 95% of the sequences by TrimAl [47]. The final MSA contained 254 columns and 222 sequences for front-end desaturases, and 229 columns and 184 sequences for elongases. The ML phylogenetic inference was carried out using RAxML-NG with automatic bootstrapping (-all) [48]. The protein substitution model was selected as LG + I + G4 for both gene types by ModelTest-NG (-model LG + I + G4) [49]. The resulting trees were visualized using Interactive Tree of Life (iTOL v. 5, https://itol.embl.de).

Functional characterization of the Tigriopus californicus desaturases and elongases
The full-length ORF of the T. californicus ωx desaturases, frontend desaturases and fatty acyl elongases were amplified by polymerase chain reactions (PCR) using a high-fidelity DNA polymerase (PrimeSTAR Max DNA polymerase, Takara Bio Inc., Shiga, Japan). All PCR runs were carried out following the manufacturer's default recommendations as follows: 10 s at 98°C for the initial denaturation, 35 cycles of 10 s at 98°C, 5 s at 55°C and 20 s at 72°C. The primer sequences containing restriction enzyme sites to enable further cloning into the yeast expression vector pYES2 (Thermo Fisher Scientific) are given in table 1. The resulting PCR products were purified from an agarose gel (GenElute Gel Extraction Kit, Sigma-Aldrich Japan K.K., Tokyo, Japan) and subsequently digested with the corresponding restriction enzymes (table 1). Next, the purified and digested full-length ORF were ligated into a similarly restricted pYES2 using T4 DNA ligase (Promega) and transformed into DH5α competent E. coli (Nippon Gene Co., Ltd., Tokyo, Japan). Positive transformant colonies were grown overnight in LB broth, and plasmid preparations (GenElute Plasmid Miniprep Kit, Sigma-Aldrich Japan K.K.) were sent to the DNA sequencing service (Eurofin Genomics K.K., Tokyo, Japan) to confirm their sequences. Plasmid constructs of the corresponding T. californicus two ωx desaturases, five front-end desaturases and six fatty acyl elongases were individually transformed into INVSc1 yeast Saccharomyces cerevisiae (Life Technologies Japan, Tokyo, Japan) using the S.c. EasyComp yeast transformation kit (Life Technologies Japan). After growing the transformed yeast on S. cerevisiae minimal medium minus uracil (SCMM −ura ) plates for 3 days at 30°C, one of the successful transformants from each gene was selected and used for the functional characterization assay as described below.

Fatty acid analysis by gas chromatography
FAME samples were injected on a GC-2025 (Shimadzu Corporation, Kyoto, Japan) gas chromatograph equipped with a capillary column (Supelcowax 10, 30 m × 0.32 mm i.d. × 0.25 µm, Sigma-Aldrich) and a flame ionization detector (FID). The temperature conditions consisted of an initial ramp from 50°C to 180°C at a rate of 40°C min −1 , 180°C to 230°C at a rate of 1°C min −1 and then 230°C for 10 min. The injection port and FID temperature were 250°C and 280°C, respectively. Helium was used as a carrier gas in constant velocity mode (30 cm sec −1 ). FAME was identified by the comparison of their retention times with those from commercial FAME standards. The conversion efficiency of three types of assayed enzymes towards the exogenously supplied PUFA substrates was calculated with the formula (all product areas/(all product areas + substrate area)) × 100 [20].

Statistical analysis
Comparisons of means of FA composition between control yeast (n = 3) and transgenic yeast expressing the T. californicus royalsocietypublishing.org/journal/rsob Open Biol. 11: 200402

Phylogeny of the Tigriopus californicus front-end desaturases and fatty acyl elongases
The ML phylogenetic tree of the front-end desaturases retrieved from T. californicus and other organisms is shown in figure 2 (a complete view is shown in electronic supplementary material, figure S1). All five T. californicus front-end desaturases (termed Fed1-5, figure 2) clustered together (bootstrap = 98%) with other copepod sequences from Tigriopus japonicus (Harpacticoida), Caligus rogercresseyi (Siphonostomatoida) and Paracyclopina nana (Cyclopoida). A direct sister group of the copepod clade could not be clearly identified due to insufficient resolution of the tree. However, the copepod clade clustered (bootstrap = 82%) with several taxonomically unrelated species belonging to the classes Kinetoplastea (e.g. Bodo saltans, Trypanosoma cruzi), Choanoflagellata (Salpingoeca rosetta) and Haptophyta (Rebecca salina and Pavlova lutheri).
Interestingly, front-end desaturases from other metazoan species including Vertebrata, Echinodermata and Mollusca grouped together in a clearly separated clade (bootstrap = 100%) from that containing the copepod sequences (figure 2).
The six elongases (Elo1-6) from T. californicus were separated to three different clades in the ML phylogenetic tree (figure 3; electronic supplementary material, figure S2). The Elo1, Elo2, Elo3 and Elo5 were branched to a Pancrustacea (=Crustacea + Hexapoda)-specific elongase clade with a relatively high support value (90%). Despite its direct sister group could not be clearly established, this Pancrustaceaspecific clade grouped itself with the vertebrate Elovl1 and Elovl7 cluster (bootstrap = 98%) ( figure 3). Within the Pancrustacea clade, the relationship among each taxonomical group was not well-resolved but, along Copepoda (T. californicus, T. japonicus, P. nana, C. rogercresseyi, Caligus clemensi, Lepeophtheirus salmonis and Eurytemora affinis), it included several insects (e.g. Drosophila melanogaster, Bombyx mori), a collembola (Folsomia candida), and crustaceans such as Decapoda (Scylla olivacea, Eriocheir sinensis and Penaeus vannamei), Branchiopoda (Daphnia pulex) and Amphipoda (Hyalella azteca) ( figure 3). Interestingly, the T. californicus Elo1, Elo2 and Elo5 grouped with sequences from other copepods, particularly T. japonicus, while the relationship of the T. californicus Elo3 with other copepod genes could not be established ( figure 3). Away from the Pancrustacea clade described above, the T. californicus Elo6 clustered within another well-supported clade (89%) containing several Arthropoda sequences including functionally characterized Elovl4-like sequences from two decapods (S. olivacea and Portunus trituberculatus) [52,53] (figure 3). The T. californicus Elo4 was grouped in a completely distinct clade that out-grouped all other sequences (figure 3). This clade was comprised a wide range of metazoan sequences except bony vertebrates. The T. californicus Elo4 was the only gene from Ecdysozoa species identified in our searches.

Discussion
Copepods, one of the most abundant groups within zooplankton, contain high levels of n-3 LC-PUFA, particularly EPA and DHA [21,55], prompting interest to elucidate to which extent biosynthesis, along diet can contribute to the abundance of these essential nutrients. While several studies provided evidence suggesting that some copepod species can indeed produce LC-PUFA endogenously, the potential contribution of microbial endosymbionts for such metabolic ability could not be completely ruled out [31][32][33]. In the present study, we addressed this methodological drawback by performing a comprehensive retrieval for all genes encoding desaturase and elongase enzymes with potential roles in the PUFA and LC-PUFA biosynthesis in a representative species within Harpacticoida, T. californicus. We selected this species not only because of its importance as a dominant zooplanktonic component in the marine tidal area, but also because it has been recently shown to possess putative ωx desaturases, enzymes limiting the ability of animals for de novo biosynthesis of PUFA, as well as being major components enabling bioconversions of n-6 substrates into n-3 LC-PUFA [18]. We herein demonstrate that T. californicus has multiple genes encoding ωx desaturases, front-end desaturases and fatty acyl elongases whose functions enable this species to carry out all reactions required for de novo biosynthesis of PUFA and, from them, LC-PUFA up to DHA. A comprehensive phylogenetic analysis of ωx desaturases performed by Kabeya et al. [18] depicted an overall distribution of these gene families in animals. The T. californicus ωx1 and ωx2 grouped with other copepod ωx desaturases to form a single clade, itself clustering along with sequences from lophotrochozoans including molluscs, annelids and rotifers [18]. In the present study, the phylogenetic analysis of front-end desaturase showed that the T. californicus Fed1-5 formed a well-supported branch including the other two 18 : 2n-6 18 : 3n-3 18 : 1n-9   non-Harpacticoida copepods, namely C. rogercresseyi (Siphonostomatoida) and P. nana (Cyclopoida). These results suggest that the copepod front-end desaturase genes evolved from their common ancestor. Interestingly, the front-end desaturase sequences retrieved from T. californicus and other copepods are phylogenetically unrelated with decapod crustacean desaturases hypothesized to be Δ6 front-end desaturases [56][57][58][59][60]. Unlike copepod sequences, the decapod putative front-end desaturase-like sequences do not contain the conserved Delta6-FADS-like domain (cd03506) in the NCBI-specific protein database and hence did not pass the domain-specific threshold (bit-score = 113.891) and were excluded from the ML phylogenetic inference. These findings strongly suggest that the decapod desaturases are not frontend desaturases and probably explain why no functional data supporting their annotation as 'Δ6 desaturases' were reported [56][57][58][59][60]. Indeed, the absence of front-end desaturases appears to the extent to other major groups within Malacostraca crustaceans such as isopods and amphipods, according to our search strategy that did not identify any putative front-end desaturases in these groups. It is interesting to note that the copepod front-end desaturase clade is clearly separated from the well-supported branch comprising sequences from other metazoans including vertebrates, echinoderms and lophotrochozoans (see [10,13]). This result illustrates the diversity of front-end desaturase gene families that play pivotal roles in PUFA biosynthesis in metazoans. Such diversity can be partly accounted for phenomena such as horizontal gene transfer (HGT), an evolutionary mechanism that has been proposed to explain the presence of ωx desaturases [18,61] and other desaturases [62] in animals. Consistently, the copepod sequences retrieved in the present study clustered closely with front-end desaturase sequences from several protists and algae including Leishmania major and P. lutheri, both characterized as Δ4 desaturases and thus playing key roles in DHA synthesis from 22 : 5n-3 [63,64]. The insufficient resolution of the tree does not allow us to conclude whether the copepod front-end desaturases were acquired via HGT. However, the ever-increasing availability of genomic sequences will probably enable us to clarify their overall evolutionary history in the near future.
The ML phylogenetic analysis of elongases denoted that four out of six genes (Elo1, Elo2, Elo3 and Elo5) isolated from T. californicus formed a well-supported clade with other Pancrustacea sequences. This Pancrustacea-specific clade includes the Elovl7-like elongase isolated from royalsocietypublishing.org/journal/rsob Open Biol. 11: 200402 S. olivacea, which showed elongation capability on a range of PUFA [65]. The Pancrustacea elongase clade is closely related to that including the vertebrate Elovl1 and Elovl7 but, unexpectedly, not Elovl5, Elovl2 and Elovl4, major enzymes of PUFA elongation in vertebrates [13,66,67]. Moreover, the Pancrustacea elongase clade contained multiple insect sequences that, rather than roles in PUFA elongation, have been shown to participate in the biosynthesis of sex pheromones [68] and cuticular hydrocarbons [69]. Elo6 belonged to another wellsupported Arthropoda clade, which includes functionally characterised Elovl4-like genes isolated from two decapod crustaceans, namely S. olivacea and P. trituberculatus [52,53] ( figure 3). Unlike the other five elongase genes from T. californicus, one elongase gene (Elo4) clustered within the most diverged clade, which was an out-group of all other elongase sequences. Since Elo4 showed very low activity towards all PUFA tested in the present study, it is reasonable to speculate that Elo4 does not play a prominent role in PUFA elongation in T. californicus.
In marine ecosystems, high trophic level organisms such as fish have some capacity to biosynthesize n-3 LC-PUFA from the C 18 n-3 PUFA ALA (18 : 3n-3), a phenomenon often referred to as 'trophic up-grading' [70]. Vital to this process is the provision of precursor ALA by organisms occupying lower trophic levels that serve as a food item for fish and other top predators. The functional assays of the two T. californicus ωx desaturases studied here showed that harpacticoid copepods can efficiently produce ALA and, hence, play such a pivotal role in marine ecosystems. Indeed, the T. californicus ωx desaturases contain key enzymatic activities including Δ12 (ωx1) and Δ15 (ωx1 and ωx2) required to convert oleic acid (18 : 1n-9) into LA (18 : 2n-6) and ALA (18 : 3n-3), respectively (figure 1). Interestingly, the T. californicus ωx1, along with Δ12 and Δ15 desaturase activities, further showed Δ17 activity towards C 20 n-6 PUFA substrates, while ωx2, in addition to Δ15, also had Δ17 and Δ19 activities towards C 20 and C 22 n-6 PUFA substrates, respectively. These results allow us to conclude that both ωx desaturases can be categorized as 'ω3 desaturases' enabling desaturation of multiple n-6 substrates into the corresponding n-3 metabolic products (figure 1). Similar functions were observed in one of the ωx desaturases functionally characterized from the Siphonostomatoida copepod L. salmonis, although this species possesses a further enzyme that exclusively exhibited Δ12 desaturase activity [18]. Beyond the phylogenetic diversification of desaturase gene/protein families pointed out above, these results further illustrate that diversification also expands to function since the enzymes from relatively closely related species such as T. californicus and L. salmonis have different substrate specificities. Along with the abovementioned ability to provide C 18 PUFA precursors for trophic upgrading to LC-PUFA by fish and other predators, our results demonstrated that harpacticoid copepods such as T. californicus can further produce LC-PUFA by themselves. They can do so by the combined action of the herein characterized front-end desaturases and elongases, which enable the biosynthesis of LC-PUFA through any of the routes described in animals to date (figure 1). For the T. californicus front-end desaturases, an apparent separation of regioselectivities was observed, with Δ6 desaturase activity being contained mostly in Fed1, Δ8 in Fed3, Δ5 in Fed5 and Δ4 in Fed2. Using the same set of enzymes in both the n-6 and n-3 routes, T. californicus can use two distinct pathways to biosynthesize both arachidonic acid (ARA) and EPA from the C 18 PUFA LA and ALA, respectively (figure 1). These two pathways are the socalled 'Δ6 pathway', consisting of a Δ6 desaturation, followed by an elongation and a final Δ5 desaturation, and the 'Δ8 pathway', starting with an elongation, followed by a Δ8 desaturation and a final Δ5 desaturation (figure 1) [71,72]. In addition to front-end desaturases, we herein demonstrate that T. californicus has several elongases that can contribute to both pathways. More specifically, the T. californicus Elo2 and Elo3, with preference towards C 18 and C 20 PUFA, and Elo5 with preference towards C 20 PUFA, appear to be major enzymes involved in the biosynthesis of LC-PUFA such as ARA and EPA. Although further studies will be needed to investigate physiological significance of some 'over-lapping' activities observed in both Fed and Elo, the apparent separation of regioselectivities/substrate specificities would imply the importance of precise modulation of LC-PUFA composition in T. californicus.
Copepods including harpacticoids contain lipids with particularly high levels of DHA (approximately 10 to 20%, e.g. [30,73,74]). Functional analyses of the T. californicus LC-PUFA biosynthesising genes demonstrate that some copepods have the ability to produce DHA from EPA. DHA biosynthesis in animals takes place via two different pathways, including the 'Sprecher pathway', reported in rats [75], fish [51,76,77] and the razor clam Sinonovacula constricta [78], and the 'Δ4 pathway' that is mostly present in teleost lineages [51] (figure 1). Both pathways require the first elongation from EPA (20 : 5n-3) to 22 : 5n-3 that, in the case of T. californicus, can be achieved by different elongases, namely Elo2, Elo3 and Elo5. However, the Sprecher pathway requires a further elongation to produce 24 : 5n-3 before this compound can be Δ6 desaturated to 24 : 6n-3 and chain-shortened ( partial β-oxidation) to DHA (figure 1). None of the T. californicus desaturases characterized in this study appeared to be able to desaturate 24 : 5n-3 to 24 : 6n-3, and therefore it is unlikely that this copepod can produce DHA via the Sprecher pathway. Importantly, one of the T. californicus front-end desaturases, the herein termed 'Fed2' is a Δ4 desaturase, an enzyme enabling the direct conversion of 22 : 5n-3 to DHA within the Δ4 pathway. Possessing Δ4 desaturases has been previously reported in some microalgae [4] and few vertebrates mostly teleosts [22] but, to the best of our knowledge, this key enzyme has not been hitherto identified in invertebrates. While further investigations will help to clarify the occurrence of Δ4 desaturases among copepods, it is reasonable to speculate that endogenous production via the Δ4 pathway accounts for part of the DHA found in copepods' lipids [25][26][27][28][29][30][31][32][33].
In conclusion, we have successfully isolated and functionally characterized two ωx desaturases, five front-end desaturases and six elongases from T. californicus. The T. californicus ωx desaturases are ω3 desaturases enabling (i) the biosynthesis of LA and ALA that are precursors of LC-PUFA and (ii) the conversion of multiple n-6 PUFA into the corresponding n-3 metabolic products including LC-PUFA. Upgrading from the C 18 PUFA LA and ALA to LC-PUFA is also possible in T. californicus, since the complementary action of its front-end desaturases and elongases enable multiple routes for the biosynthesis of the physiologically active compounds ARA, EPA and DHA. Collectively, the results obtained in this study demonstrate that T. californicus has a complete enzymatic complement enabling this species to produce n-3 LC-PUFA up to DHA endogenously. Therefore, harpacticoid copepods arise as primary producers of n-3 LC-PUFA in marine ecosystems and, given their widespread distribution and abundance at a global scale, it is likely that such contribution is not negligible.
Data accessibility. All isolated sequences from T. californicus in the present study were deposited into the NCBI GenBank with the accession numbers MT757162 to MT757173 and MW246081.