Diversification of substrate specificities in teleostei Fads2: characterization of Δ4 and Δ6Δ5 desaturases of Chirostoma estor.

Currently existing data show that the capability for long-chain PUFA (LC-PUFA) biosynthesis in teleost fish is more diverse than in other vertebrates. Such diversity has been primarily linked to the subfunctionalization that teleostei fatty acyl desaturase (Fads)2 desaturases have undergone during evolution. We previously showed that Chirostoma estor, one of the few representatives of freshwater atherinopsids, had the ability for LC-PUFA biosynthesis from C18 PUFA precursors, in agreement with this species having unusually high contents of DHA. The particular ancestry and pattern of LC-PUFA biosynthesis activity of C. estor make this species an excellent model for study to gain further insight into LC-PUFA biosynthetic abilities among teleosts. The present study aimed to characterize cDNA sequences encoding fatty acyl elongases and desaturases, key genes involved in the LC-PUFA biosynthesis. Results show that C. estor expresses an elongase of very long-chain FA (Elovl)5 elongase and two Fads2 desaturases displaying Δ4 and Δ6/Δ5 specificities, thus allowing us to conclude that these three genes cover all the enzymatic abilities required for LC-PUFA biosynthesis from C18 PUFA. In addition, the specificities of the C. estor Fads2 enabled us to propose potential evolutionary patterns and mechanisms for subfunctionalization of Fads2 among fish lineages.

The ancestry of C. estor , pattern of LC-PUFA biosynthesis activity, and effects of salinity, which confl ict with the existing paradigm, made this species an ideal candidate for studies investigating the molecular basis of LC-PUFA biosynthesis to gain further understanding of the diversity that these pathways show among teleost lineages. The objective of the present study was the functional characterization of cDNAs encoding desaturases and elongases involved in the LC-PUFA biosynthetic pathways in C. estor . The results indicated that C. estor has all the enzymatic capabilities for endogenous biosynthesis of LC-PUFAs. In addition to the presence of an Elovl5 elongase, C. estor expresses at least two distinct Fads2 enzymes displaying ⌬ 4 and ⌬ 6/ ⌬ 5 specifi cities. The fi ndings on C. estor desaturases provided further insight into the diversifi cation of substrate specifi cities in teleostei Fads2s and enabled potential evolutionary patterns of subfunctionalized Fads2s among teleost lineages to be proposed.

Tissue samples
Pike silverside ( ‫ف‬ 40 g), maintained at the facilities of the Instituto de Investigaciones Agropecuarias y Forestales (UMSNH ), Mexico, were anesthetized and euthanized with an overdose of MS-222 (Sigma-Aldrich, Alcobendas, Spain). Tissues including liver, intestine, brain, and muscle were collected in RNAlater (Ambion, Applied Biosystems, Warrington, UK) according to manufacturer's instructions. Fish maintenance and euthanasia procedures were carried out in compliance with the Offi cial Mexican Standard NOM-062-ZOO-1999 on the technical specifi cations for the production, care, and use of laboratory animals. This study was also reviewed and approved by the departmental ethics committees of Institute of Aquaculture Torre de la Sal (CSIC , Spain) and Institute of Aquaculture (University of Stirling, UK).

Cloning of fatty acyl desaturases and elongase from C. estor
Total RNA (2 g) extracted from C. estor tissues (TRI reagent, Sigma-Aldrich) was reverse transcribed using a GoScript TM reverse transcription system (Promega, Madison, WI) primed with random primers. The primers UNIDF (5 ′ -GGAGAGGAYGC CACG-GAGG-3 ′ ) and UNIDR (5 ′ -GTCCRCTGAACCAGTCGTTGAA-3 ′ ) for fads2 and UNIE5F (5 ′ -CATGGATGGGYCCMAGAGATC-3 ′ ) and UNIE5R (5 ′ -GTCTGAATGTAGAAGTTTGAGAAAAG-3 ′ ) for elovl5 were used to amplify a partial fragment of these genes from C. estor by PCR with GoTaq® Green Master Mix (Promega) as described previously ( 30 ). A mixture of cDNA from liver and brain was used as a template for PCR, which consisted of an initial denaturing step at 95°C for 2 min, followed by 40 cycles of denaturation at 95°C for 30 s, annealing at 50°C for 30 s, extension at 72°C for 1 min 30 s, followed by a fi nal extension at 72°C for 5 min. The PCR fragments were purifi ed and sequenced at the DNA Sequencing Service IBMCP-UPV (Valencia, Spain). While one single version of the elovl5 -like cDNA sequence was detected, sequencing results of the desaturase-targeted PCR revealed the existence of at least two distinct fads2 -like transcripts. Their individual sequences, hereafter referred to as Fads2a and Fads2b, were determined after ligation of the PCR product into pGEM-T Easy vector (Promega) and sequencing as above. The partial sequences within the open reading frame (ORF) of the two desaturases and the elongase were many marine fi sh for dietary LC-PUFAs was caused by deficiency in key enzymatic activities required for their biosynthesis from C 18 PUFAs ( 7,8 ). This was hypothesized to be a consequence of marine fi sh having evolved in an LC-PUFA-rich environment; and thus, there was low evolutionary pressure to retain the ability to desaturate and elongate C 18 PUFA . In contrast, lower levels of LC-PUFAs in the food chain meant freshwater species retained the ability to biosynthesize LC-PUFAs from C 18 PUFAs ( 8,32,33 ). This hypothesis was based upon fi sh that were largely carnivorous in the case of marine species and detritivorous/herbivorous in freshwater species ( 34 ). However, trophic level, the position of an organism within the food web, was also investigated, and it was demonstrated that the herbivorous marine fi sh Siganus canaliculatus (rabbitfi sh) had the ability to endogenously synthesize LC-PUFAs because two desaturases with ⌬ 4 and ⌬ 6/ ⌬ 5 specifi cities and two elongases (Elovl4 and Elovl5) enabled this species to perform all the enzymatic reactions required in the pathway ( 4,28 ). More recently, other confounding factors, including "trophic ecology" and diadromy, have been proposed ( 5,30 ). Currently, data indicate that the capability for LC-PUFA biosynthesis in teleost fi sh is more diverse than in other vertebrate groups, and is possibly the result of a combination of factors that interact throughout the evolutionary history of each particular group or species. Such plasticity has been primarily associated with the substrate specifi cities exhibited by desaturase Fads2, a protein that has subfunctionalized during the evolution of teleosts.
Pike silverside Chirostoma estor (previously Menidia estor ) from Lake Pátzcuaro is a highly valued freshwater fi sh in Mexico (36)(37)(38)(39). Although C. estor is a freshwater species, it has a common ancestry with marine atherinopsids ( 40 ) and shares many biological and physiological characteristics of marine species ( 41 ). Also, unusually for a freshwater fi sh, tissues of C. estor show high levels of DHA (20-32% of total FAs) and only low levels of EPA (1-3%) in contrast to the FA profi le of its zooplankton diet ( ‫ف‬ 12% DHA, 13% EPA) ( 42 ). This suggested that C. estor either selectively accumulates DHA, or has the capacity to convert EPA or other n-3 PUFAs to DHA ( 7 ). Interestingly, the activity of the LC-PUFA synthesis pathway in C. estor was very low in freshwater (0 parts per thousand salinity) in comparison to that in higher salinity (5 or 15 parts per thousand ), and this was unexpected for a freshwater fi sh, as they generally show appreciable LC-PUFA synthesis activity ( 43 ). kit, GE Healthcare, Little Chalfont, UK), digested with the corresponding restriction enzymes (Promega) and ligated into a similarly restricted pYES2 yeast expression vector (Invitrogen, Paisley, UK). The plasmid constructs prepared were designated as pYES2-fads2a, pYES2-fads2b, and pYES2-elovl5.

FA analysis of yeast
Total lipids were extracted from yeast samples and fatty acyl methyl esters (FAMEs) were prepared as described previously ( 9,30 ). FAMEs were quantifi ed using a Thermo gas chromatograph (Thermo Trace GC Ultra, Thermo Electron Corporation, Waltham, MA) fi tted with an on-column injection system and a fl ame ionization detector (FID). Additionally, FAMEs were identifi ed using an Agilent 6850 gas chromatograph system coupled to a 5975 series MSD (Agilent Technologies, Santa Clara, CA). The desaturation or elongation conversion effi ciencies from exogenously added PUFA substrates were calculated by the proportion of substrate FA converted to desaturated or elongated products as [individual product area/(all products areas + substrate area)] × 100. For the elongase, some of the initial elongation products were further elongated, and thus the accumulated conversions were calculated by summing all elongated products ( 28 ). Similarly, the desaturase Fads2b exhibited multifunctional abilities, and thus the conversions on ⌬ 8 substrates (20:3n-3 and 20:2n-6) include stepwise reactions.

Tissue distribution of fads2 and elovl5 transcripts
Expression of the target genes ( fads2a , fads2b , and elovl5 ) was measured by quantitative real-time PCR (qPCR). Total RNA from liver, brain, intestine, and muscle was extracted from three C. estor adults, as described above, and 2 g of total RNA were reverse transcribed into cDNA (M-MLV reverse transcriptase, Promega) using oligo-dT primer. The qPCR was performed using primers shown in supplementary Table I. Copy numbers of target genes were normalized with copy number of the reference gene ef-1 ␣ further extended by 5 ′ and 3 ′ rapid amplifi cation of cDNA ends (RACE) PCR (FirstChoice® RLM-RACE kit, Ambion) through a two-round (nested) PCR approach as detailed below (see supplementary Table I for primer details).
For desaturase fads2a , fi rst round PCR was performed combining the adaptor-specifi c 5 ′ RACE outer primer with the gene-specifi c reverse primer CEFaR1, whose 3 ′ end sequence contained two nucleotides that differed from the fads2b sequence. A second round PCR with 5 ′ RACE inner primer and CEFaR2 produced a positive band expanding out the putative end of the 5 ′ untranslated region . For the fads2b cDNA sequence, a specifi c 5 ′ RACE product was not obtained, but the ORF of fads2b was obtained to enable functional characterization as detailed below.
Positive 3 ′ RACE PCR products were obtained using isoformspecifi c primers and 3 ′ adaptor primers. First round PCR involved the use of forward primers CEFaF1 (isoform a) and CEFbF1 (isoform b) with the 3 ′ RACE outer primer. First round products were subsequently used as a template for nested PCR with forward primers CEFaF2 (isoform a) and CEFbF2 (isoform b), and with reverse primers consisting of the adaptor primer 3 ′ RACE inner.
A similar approach was followed to obtain the full-length cDNA of C. estor elovl5 . The gene-specifi c primers CEE5R1 and CEE5R2 (5 ′ RACE) and CEE5F1 and CEE5F2 (3 ′ RACE) used for RACE PCR are listed in supplementary Table I. General RACE PCR conditions consisted of an initial denaturing step at 95°C for 2 min, followed by 32-35 cycles of denaturation at 95°C for 30 s, annealing at 55-60°C for 30 s, extension at 72°C for 2 min 30 s, followed by a fi nal extension at 72°C for 5 min (GoTaq® Green Master Mix, Promega). PCR products were cloned into pGEM-T Easy vector (Promega) and sequenced as above.

Sequence and phylogenetic analyses
The amino acid (aa) sequences of the C. estor desaturases Fads2a and Fads2b, and elongase Elovl5 proteins were compared with those of orthologs from other fi sh and tetrapods (mammals, amphibians, and birds) and sequence identity scores were obtained using the EMBOSS Needle pairwise sequence alignment tool . For phylogenetic analysis of the C. estor deduced aa sequences of fads2a , fads2b , and elovl5 , two trees were constructed using the neighbor-joining method ( 44 ), with confi dence in the resulting tree branch topology measured by bootstrapping through 10,000 iterations. Desaturase and elongase C. estor sequences were compared with homologous proteins from a variety of vertebrate lineages. The ⌬ 6 desaturase and the PUFA elongase sequences from the oleaginous fungus Mortierella alpina were used as outgroup sequences to construct both rooted trees.

Functional characterization of the C. estor Fads2 and Elovl5 by heterologous expression in yeast
PCR fragments corresponding to the ORF of pike silverside desaturases fads2a and fads2b and elongase elovl5 were amplifi ed from a mixture of cDNA (liver and brain) by PCR using the high fi delity Pfu DNA polymerase (Promega) with primers containing Hin dIII and Sac I restriction sites (underlined in supplementary Table I) as follows. For fads2a , the primer pair CEFVF-CEFaVR was used. For fads2b , the same forward primer CEFVF and the antisense primer CEFbVR were used. While CEFVF was specifi c for both isoforms, its use in combination with the fads2b -specifi c primer CEFbVR enabled us to successfully isolate fads2b . Finally, the ORF of the elovl5 was isolated using the primers CEE5VF and CEE5VR (supplementary Table I). PCR consisted of an initial denaturing step at 95°C for 2 min, followed by 35 cycles of denaturation at 95°C for 30 s, annealing at 57°C for 30 s, extension at 72°C for 3 min (desaturase) or 2 min (elongase), followed by a fi nal extension at 72°C for 5 min. The PCR products were subsequently purifi ed (Illustra GFX PCR DNA/gel band purifi cation striata , L. calcarifer , and S. aurata ), and 62.5% to that from H. sapiens , 62.7% to that from Xenopus tropicalis , and 64.4% to that of Gallus gallus . Lower identities (38.0-52.3%) were obtained when the C. estor Elovl5 was compared with other elongase families such as Elovl2 and Elovl4 from different vertebrate lineages including teleosts, amphibians, birds, and mammals. The C. estor Elovl5 had a diagnostic histidine box ( H VY HH ), and lysine (K) and arginine (R) residues at the carboxyl terminus ( KK L R VD). The sequence of the elovl5 cDNA from C. estor was deposited in GenBank with accession number KJ417837.
Phylogenetic trees constructed on the basis of aa sequence comparisons of the C. estor cDNAs with homologous proteins from fi sh and other vertebrate species refl ected the identity scores shown above. For desaturases, the C. estor Fads2s clustered with Fads2s from fi sh and, more distantly, with Fads2s from mammals, birds, and amphibians ( Fig. 1 ). Additionally, Fads1-like proteins, "front-end" desaturases not present in teleost fi sh ( 32 ), grouped separately from the Fads2 cluster that included the C. estor desaturases ( Fig. 1 ). On the other hand, the C. estor Elovl5 grouped together with other teleost and tetrapod (mammals, birds, and amphibians) orthologs, and separately from members of Elovl2 and Elovl4 families from fi sh and other vertebrates ( Fig. 2 ).
In spite of its homology with the Fads2a aa sequence, the C. estor Fads2b showed a remarkably different pattern of FA substrate specifi city. Whereas no ⌬ 4 desaturation was detected, Fads2b conferred on yeast the ability to desaturate both ⌬ 6 substrates 18:3n-3 and 18:2n-6 to 18:4n-3 and 18:3n-6, respectively, and ⌬ 5 substrates 20:4n-3 and 20:3n-6 to 20:5n-3 and 20:4n-6, respectively ( Fig. 3 ; Table 1 ), which indicated the C. estor Fads2b had dual ⌬ 6 ⌬ 5 activity. Conversions obtained from the yeast assays suggested (GenBank accession number KJ439615). PCR amplicons of each gene were cloned into pGEM-T Easy vector (Promega) that was then linearized, quantifi ed spectrophotometrically (NanoDrop 2000c, Thermo Scientifi c, Wilmington, DE), and serial-diluted to generate a standard curve of known copy numbers. The qPCR amplifi cations were carried out in duplicate using a CFX96 Bio-Rad machine (Alcobendas, Spain) in a fi nal volume of 20 l containing 5 l diluted (1/20) cDNA, 0.25 M of each primer, and 4 l PyroTaq EvaGreen® mix (Cultek Molecular Bioline, Madrid, Spain). Amplifi cations were carried out with a systematic negative control (no template control) containing no cDNA. The qPCR profi les consisted of an initial activation step at 95°C for 15 min, followed by 40 cycles of 15 s at 95°C, 20 s at the specifi c primer pair annealing melting temperature (supplementary Table I), and 15 s at 72°C. After the amplifi cation phase, a dissociation curve of 0.5°C increments from 60 to 90°C was performed, enabling confi rmation of the amplifi cation of a single product in each reaction. No primer-dimer formation occurred in the no template control. All reactions were carried out in duplicate, a linear standard curve was drawn, and the absolute copy number of the targeted gene in each sample was calculated.

Statistical analysis
Tissue expression (qPCR) results are expressed as mean normalized values (± SE) corresponding to the ratio between the copy numbers of fads2a , fads2b , and elovl5 transcripts and the copy numbers of the reference gene ef-1 ␣ . A one-way ANOVA followed by Tukey HSD test ( P < 0.05) was performed to compare the expression level among the selected tissues (SPSS, Chicago, IL).
The inherent capability of vertebrate Fads2 enzymes for ⌬ 8 desaturation was investigated in both isoforms. Only Fads2b exhibited the ability to desaturate 20:3n-3 and 20:2n-6 to the corresponding ⌬ 8-desaturated products 20:4n-3 and 20:3n-6, respectively ( Table 1 ). The relative ⌬ 6/ ⌬ 8 activity Fig. 1. Phylogenetic tree comparing the deduced aa sequence of the newly cloned C. estor Fads2-like with Fads1-and Fads2-like desaturases from a variety of vertebrates. The tree was constructed using the neighbor-joining method ( 44 ) with MEGA4. The horizontal branch length is proportional to aa substitution per site. The numbers represent the frequencies with which the tree topology presented was replicated after 10,000 bootstrap iterations. All accession numbers are from GenBank database.

Fig. 2.
Phylogenetic tree comparing the deduced aa sequence of the newly cloned C. estor elongase of very long-chain FAs (Elovl) with elongases Elovl2, Elovl4, and Elovl5 from a variety of vertebrates. The tree was constructed using the neighbor-joining method ( 44 ) with MEGA4. The horizontal branch length is proportional to aa substitution per site. The numbers represent the frequencies with which the tree topology presented was replicated after 10,000 bootstrap iterations. All accession numbers are from GenBank database. 20:4n-3, and 20:3n-6, respectively ( Table 2 ). Further elongations to C 22 secondary products could be detected in yeast incubated with 18:3n-3, 18:2n-6, and 18:4n-3. For C 20 PUFA substrates, almost 90% of total EPA was elongated to 22:5n-3, whereas ARA was elongated to 22:4n-6 to a lower extent (54.8%). C 22 PUFA substrates, including 22:5n-3 yeast expressing the coding sequence of the pike silverside elovl5 showed activity toward most of the PUFA substrates assayed, with particularly high conversions for most C 18 and C 20 substrates. Among C 18 substrates, high elongations were obtained with 18:3n-3, 18:4n-3, and 18:3n-6, which were converted to the corresponding C 20 products 20:3n-3,   Tissue distribution of fads2 and elovl5 transcripts Tissue distribution of the C. estor fads2 (a and b isoforms) and elovl5 mRNA in adult specimens was analyzed by qPCR. Both fads2 transcripts were detected in all four tissues analyzed, with signifi cantly higher expression signals found in liver compared with brain, intestine, and muscle ( Fig. 6 ). With regard to elovl5 , liver also showed the highest expression rates, but signifi cant differences could only be established with brain and muscle signals ( Fig. 6 ).

DISCUSSION
A previous study suggested that C. estor had an active LC-PUFA biosynthesis pathway that enabled this species to endogenously synthesize DHA from PUFA precursors ( 43 ). Here, we provide robust data supporting a likely molecular mechanism demonstrating that C. estor expresses genes encoding enzymatic activities that would enable the synthesis of DHA.
The capabilities exhibited by the two desaturase cDNAs cloned from C. estor (Fads2a and Fads2b) cover the set of desaturation requirements for DHA synthesis from LNA (18:3n-3). Heterologous expression of Fads2b showed this protein was a dual ⌬ 6 ⌬ 5 desaturase and, thus, it catalyzed the desaturation of 18:3n-3 to 18:4n-3 ( ⌬ 6) and also the ⌬ 5 desaturation required to convert 20:4n-3 to EPA. While the C. estor Fads2a can partly contribute to the ⌬ 5 desaturation leading to EPA biosynthesis as described for Fads2b, the higher conversion activities of Fads2a suggested that its major role in the overall pathway was to catalyze the ⌬ 4 desaturation involved in the direct conversion of 22:5n-3 to DHA. A similar pathway of DHA biosynthesis from EPA was postulated to operate in the rabbitfi sh, S. canaliculatus ( 4,28 ), and more recently Senegalese sole, S. senegalensis ( 5 ). In comparison with the Sprecher pathway, the socalled " ⌬ 4 pathway" is a more direct metabolic route, as it avoids translocation of PUFA intermediates (namely 24:6n-3) between endoplasmic reticulum and peroxisomes, and also the further catabolic step (partial oxidation to DHA) occurring in the latter organelle ( 3,46 ). and 22:4n-6, were only marginally or not converted to longer products ( Table 2 ; Fig. 5 ). For each pair of homologous substrates considered, conversions obtained from the yeast assays suggested that the C. estor Elovl5 more efficiently elongated n-3 than n-6 PUFAs on a consistent basis ( Table 2 ). Thus, 18:3n-3, 18:4n-3, and 20:5n-3 were elongated to a greater extent than the corresponding n-6 PUFA substrates 18:2n-6, 18:3n-6 and 20:4n-6, respectively. Particularly interesting was the difference in the conversions observed between 18:3n-3 (up to 41.1% converted to longer products) and 18:2n-6 (only 5.5% converted to longer products).  mechanism, such as the ⌬ 4 pathway. In agreement, neither hepatocytes nor enterocytes of C. estor that had been incubated with either [1-14 C]18:3n-3 or [1-14 C]20:5n-3 showed any recovery of radioactivity in Sprecher pathway intermediates, namely 24:5n-3 to 24:6n-3 ( 43 ). Moreover, the functional characterization of the C. estor Elovl5 supports such a hypothesis.
The Elovl5 of C. estor showed virtually no ability to elongate 22:5n-3 to 24:5n-3 as would be required in the Sprecher pathway. This is in agreement with the great majority of fi sh Elovl5, with the exception of orthologs from rabbitfi sh (10.6% conversion from 22:5n-3 to 24:5n-3) ( 28 ) and, to a lesser extent, cobia (6.6% conversion) ( 19 ). Also consistent with Elovl5 from fi sh, the C. estor Elovl5 showed substantial activity for the elongation of C 18 and C 20 PUFAs to the corresponding C 20 and C 22 PUFA products, including the elongations of 18:4n-3 to 20:4n-3 and 20:5n-3 to 22:5n-3, which were catalyzed in yeast at high conversions (60.2 and 89.6%, respectively). Clearly, the C. estor Elovl5 can support all the elongation reactions required for DHA biosynthesis through the ⌬ 4 pathway. Taken together, the functional characterization data for Elovl5, Fads2a, and Fads2b, and the biochemical assays with radiolabeled PUFA substrates ( 43 ), predicts a putative pathway of biosynthesis of LC-PUFAs in C. estor from dietary essential C 18 PUFAs, 18:3n-3 and 18:2n-6 ( Fig. 7 ). In addition to the DHA biosynthetic pathway described above, two possible pathways for EPA and ARA biosynthesis are shown. First, the "classical", and likely the most prominent, pathway involving ⌬ 6 desaturation → elongation → ⌬ 5 desaturation is possible through the consecutive action of Fads2b, Elovl5, and Fads2b , respectively. Second, the alternative " ⌬ 8 pathway" proceeds through an initial elongation of dietary essential C 18 PUFAs by Elovl5, followed by two consecutive desaturations catalyzed by Fads2b, fi rst as ⌬ 8 and second as ⌬ 5 ( Fig. 7 ). The C. estor Elovl5 was effective in elongating While we did not test the ability of C. estor desaturases to mediate the ⌬ 6 desaturation of 24:5n-3 to 24:6n-3 required in the Sprecher pathway, we hypothesize that it is not operative in the presence of a more direct and effi cient  particular evolutionary history of C. estor . Therefore, while C. estor shares some "marine" features with most of its atherinopsid counterparts, it also refl ects some characteristics typically found in freshwater species ( 42 ).
The substrate specifi cities of C. estor desaturases described here are not entirely unique among fi sh Fads2 enzymes and, as mentioned above, similar substrate specifi cities ( ⌬ 4 and ⌬ 6 ⌬ 5) were found in two desaturases isolated from rabbitfi sh ( 4 ). Furthermore, a ⌬ 6 ⌬ 5 Fads2 and a ⌬ 4 Fads2 were previously identifi ed from zebrafi sh ( 9 ) and Senegalese sole ( 5 ), respectively, and a monofunctional ⌬ 5 Fads2 was also cloned from Atlantic salmon ( 13 ). Provided the Gnathostomata (jawed fi sh) ancestral Fads2 had ⌬ 6 desaturase activity ( 32 ), as for mammalian FADS2s ( 35 ), the expansion of Fads2s in teleosts has been accompanied by subfunctionalization in the enzyme derived from independent mutations in the primary aa sequence ( 32 ). Although a recent study identifi ed a single aa residue as determining the differential ability for 22:5n-3 elongation between ELOVL2 and ELOVL5 elongases in rat ( 48 ), identifi cation of specifi c domains/residues controlling the functionality of desaturases has been elusive and studies are restricted to nonvertebrate enzymes (49)(50)(51).
However, the above said, the existence of three ⌬ 6 ⌬ 5 desaturases and three ⌬ 4 desaturases distributed in four distinct species, allows us to explore potential evolutionary scenarios for teleostei Fads2 subfunctionalization. The recently revised tree of life of bony fi sh ( 52 ), based on 369 families, allows us to investigate whether the diversity of substrate specifi cities of C. estor desaturases can be related in an evolutionary context along with those of other desaturases with ⌬ 4 and ⌬ 6 ⌬ 5 activities, previously described from S. canaliculatus ( 4 ), S. senegalensis ( 5 ), and D. rerio ( 9 ). The three species possessing a ⌬ 4 desaturase ( S. canaliculatus , S. senegalensis , and C. estor ) belong to three different clades including Percomorpharia ( S. canaliculatus ), Carangimorphariae ( S. senegalensis ), and Ovalentariae ( C. estor ), all sharing a common ancestor ( ‫ف‬ 115 Ma). It is tempting to speculate that Fads2 enzymes in fi sh with ⌬ 4 activity are restricted to these three groups, albeit these groups contain >200 families and so the activity is likely further restricted to specifi c families, or even individual species within the groups. Thus, other species within these both 18:3n-3 and 18:2n-6 as required in the ⌬ 8 pathway, elongase abilities previously demonstrated in Elovl5 from Southern bluefi n tuna ( 20 ) and meagre ( 30 ). Moreover, the desaturase activities required for the ⌬ 8 pathway, namely ⌬ 8 and ⌬ 5, were demonstrated by Fads2b, but not Fads2a. While the activity of Fads2a as ⌬ 4 desaturase suggested a steric impediment disabling the insertion of new double bonds at the ⌬ 6 or ⌬ 8 positions, the Fads2b desaturase showed the ability to desaturate both 20:3n-3 and 20:2n-6 at the ⌬ 8 carbon.
The ability of fi sh Fads2 to catalyze ⌬ 8 desaturation has been regarded as a characteristic primarily of marine species, and thus relatively low values of ⌬ 6/ ⌬ 8 ratio; i.e., the ratio of the conversions toward 18:3n-3 and 20:3n-3 are shown by Fads2 from species of marine origin ( 2 ). The ⌬ 6/ ⌬ 8 of the C. estor Fads2b was 4.6, slightly above the range of ⌬ 6/ ⌬ 8 ratios of marine fi sh desaturases (1.8-4.2), but well below those of desaturases from salmonid/freshwater species (12.0-91.2) ( 2 ). Moreover, the activities of the LC-PUFA biosynthetic pathways measured in hepatocyte and enterocyte primary cultures from C. estor ( 40 ) were generally lower than those obtained in Atlantic salmon ( 47 ), but higher than those obtained in the marine teleost Atlantic cod ( Gadus morhua ) ( 15 ). Interestingly, the tissue distribution of the desaturases and elongases generally refl ected a "freshwater fi sh" pattern for C. estor . Thus, liver had the highest mRNA levels for both desaturases and the elongase, and appeared as a major metabolic site for LC-PUFA biosynthesis, over intestine, brain, and muscle. Consistently with the mRNA tissue distribution data, previous studies highlighted the unusually high contents of DHA in C. estor liver, brain, and muscle, and also other tissues including gonads and adipose tissues ( 38 ). Similar tissue distribution patterns were observed in freshwater/salmonid species, including Atlantic salmon ( 14,18 ) and zebrafi sh ( 17 ), in which liver and intestine exhibited the highest expression signals of fads2 and elovl2 and elovl5 elongase genes. Marine fi sh species, including Asian sea bass, Atlantic cod, and cobia, had the highest expression levels of LC-PUFA biosynthesis genes in brain ( 15,19,21 ). The relatively low ⌬ 6/ ⌬ 8 ratio of Fads2b, which was typical of marine species on one hand, and the higher expression of desaturase/elongase mRNA in liver, which was typical of freshwater species on the other, probably refl ected the Fig. 7. The biosynthesis pathway of long-chain PUFAs from ␣ -linolenic (18:3n-3) and linoleic (18:2n-6) acids in C. estor . Enzymatic activities shown in the scheme are predicted from heterologous expression in S. cerevisiae of the ⌬ 4 and ⌬ 6/ ⌬ 5 desaturases (Fads2a and Fads2b, respectively) and the Elovl5-like elongase characterized in the present study. three groups, including A. regius , T. thynnus , L. calcarifer , D. labrax , S. aurata , R. canadum (Percomorpharia), Psetta maxima (Carangimorphariae), and Oreochromis niloticus (Ovalentariae), have Fads2 enzymes functionally characterized as ⌬ 6 desaturases ( 11,16,19,21,26,27,29,30 ). However, investigation of desaturases from representatives of other phylogenetic branches is required to confi rm this hypothesis.
Establishing a potential pattern of distribution for Fads2 enzymes in teleost fi sh lineages showing ⌬ 6 ⌬ 5 activity is more speculative, as the three species displaying these activities ( D. rerio , S. canaliculatus , and C. estor ) are more distantly related. While the groups of C. estor and S. canaliculatus are relatively close and thus the presence of ⌬ 6 ⌬ 5 Fads2s might have a similar restricted evolutionary context as described for ⌬ 4 desaturases, the existence of a dual ⌬ 6 ⌬ 5 desaturase in D. rerio , belonging to the more ancient cyprinid lineage, does not follow common evolutionary patterns, although it is possible that convergent evolution may have occurred. Assuming the ancient Fads2 was a ⌬ 6 desaturase ( 32 ), subfunctionalization of some Fads2 in distantly related lineages (cyprinids vs. atherinopsids and siganids) led to the acquisition of ⌬ 5 desaturase activity, possibly through discrete mutations at the catalytic site. According to the proposed time-calibrated tree, the emergence of the dual ⌬ 6 ⌬ 5 in cyprinids occurred ‫ف‬ 100 Ma ago ( 52 ), as the desaturase from common carp (also cyprinid) does not show dual activity.
In summary, the present study demonstrates that C. estor expresses desaturase and elongase genes encoding all the enzymatic activities required for the biosynthesis of DHA from the C 18 precursor LNA. While the C. estor Elovl5 accounted for all the elongation steps, two distinct Fads2-like desaturase enzymes displaying ⌬ 4 and ⌬ 6 ⌬ 5 specifi cities operate along the pathway. More importantly, the uncommon substrate specifi cities of Fads2s from C. estor and other species like D. rerio , S. canliculatus , and S. senegalensis enabled us to propose potential evolutionary scenarios that explain the distribution of such subfunctionalized Fads2s among fi sh lineages.