Trans-vaccenate is Δ13-desaturated by FADS3 in rodents.

Fatty acid desaturases play critical roles in regulating the biosynthesis of unsaturated fatty acids in all biological kingdoms. As opposed to plants, mammals are so far characterized by the absence of desaturases introducing additional double bonds at the methyl-end site of fatty acids. However, the function of the mammalian fatty acid desaturase 3 (FADS3) gene remains unknown. This gene is located within the FADS cluster and presents a high nucleotide sequence homology with FADS1 (Δ5-desaturase) and FADS2 (Δ6-desaturase). Here, we show that rat FADS3 displays no common Δ5-, Δ6- or Δ9-desaturase activity but is able to catalyze the unexpected Δ13-desaturation of trans-vaccenate. Although there is no standard for complete conclusive identification, structural characterization strongly suggests that the Δ11,13-conjugated linoleic acid (CLA) produced by FADS3 from trans-vaccenate is the trans11,cis13-CLA isomer. In rat hepatocytes, knockdown of FADS3 expression specifically reduces trans-vaccenate Δ13-desaturation. Evidence is presented that FADS3 is the first “methyl-end” fatty acid desaturase functionally characterized in mammals.


Extinction of native FADS3 and expression of recombinant FADS3 in cultured rat hepatocytes
The experimental procedure was approved by the French Animal Care Committee in compliance with recommendations of the 2013-118 directive for animal experimentation. The agreement reference of the laboratory to realize rat liver perfusion is R-2012-FP-01. Hepatocytes were obtained from male rats after an in situ collagenase perfusion of the liver, as described ( 26 ). Small interfering RNA (siRNA) was obtained from Sigma. Negative siRNA control targets unrelated mRNA sequences, and siRNA against FADS3 targets 5 ′ -gcgacaaguggcuggucau-3 ′ sequence. After the rat liver perfusion, hepatocytes were seeded at 2.0 × 10 6 cells in 6 cm diameter culture dishes. They were transfected at the same time with 25 nM of either siRNA control or siRNA targeting FADS3 by using the N-TER Nanoparticle siRNA transfection system (Sigma). For the expression of recombinant FADS3, hepatocytes were transfected after 4 h of culture with 10 µg of plasmid/1.6 × 10 6 cells by using the jetPEI-Hepatocyte kit (Polyplus-transfection).

Fatty acid analysis
Cellular lipids were extracted with hexane-isopropanol (3:2, v/v) after acidifi cation with 1 ml HCl 3 M, as described ( 26 ). After saponifi cation for 15 min at 70°C with 1 ml of 0.5 M NaOH in methanol, FAs were methylated with BF 3 (14% in methanol) at 70°C for 10 min. The fatty acid methyl esters (FAME) were extracted with pentane and analyzed by GC-MS using an Agilent Technologies 7890N (Bios Analytic) with a bonded silica capillary column (60 m × 0.25 mm; BPX 70; SGE). Helium was used as carrier gas (average velocity 24 cm/s). The column temperature program started at 150°C, ramped at 2°C/min to 220°C, and held at 220°C for 10 min. Mass spectra were recorded with an Agilent Technologies 5975C inert MSD with triple-axis detector. The mass spectrometer was operated under electron impact ionization conditions (electron energy 70 eV, source temperature 230°C). Data were obtained in scan mode with a mass range of m/z 50-550 atomic mass unit (amu). Identifi cation of FAMEs was based upon retention times (R t ) obtained for methyl ester of authentic standards, when available. The National Institute of Standards and Technology (NIST) database (version 2.01) was sometimes used to identify unknown FAs.
For positional analysis of carbon double bonds, FAMEs were converted to 4,4-dimethyloxazoline (DMOX) derivatives ( 29,30 ) and analyzed by GC-MS in electron impact ionization mode. The column, gas vector, and detector were similar to those used for the analysis of FAMEs (see above). To preconcentrate the minor FA be a mammalian membrane-bound front-end desaturase. However, no function has been attributed to the FADS3 protein yet ( 15 ).
In the absence of any described functional role, investigations have been focused on the tissue distribution of the FADS3 mRNA transcripts in different species. FADS3 mRNA was shown to be ubiquitously expressed in a large variety of tissues in humans ( 7 ), baboons ( 12 ), and rats ( 11 ), with a specifi c distribution compared with FADS1 and FADS2 mRNAs. In neonate baboon tissues, seven alternative transcripts of FADS3 were described ( 12 ), but when they were cloned and expressed into yeast and mammalian cells, no activity toward a variety of PUFA substrates was detected ( 16 ). Interestingly, the expression of the FADS3 transcripts was upregulated by dietary long-chain PUFAs, whereas FADS1 and FADS2 mRNA levels were decreased ( 17 ). In addition, gene polymorphism studies carried out in humans showed that several single nucleotide polymorphisms (SNP) located within the FADS3 gene were associated with variations in triglycerides, HDL-and LDL-cholesterol plasma levels ( 18,19 ), and with variations in plasma PUFA levels (20)(21)(22)(23). These correlations underlined the possible implication of FADS3 in lipid and FA homeostasis.
More studies were required to elucidate the function of the FADS3 protein. Here we have identifi ed FADS3 as a trans -vaccenate ( trans 11-18:1) ⌬ 13-desaturase. FADS3 is therefore the fi rst methyl-end fatty acid desaturase functionally characterized in mammals. This discovery suggests that FADS3 may be an exception to the consensus dogma that rodents and higher animals are devoid of desaturases inserting double bonds in the methyl-end of fatty acids.

Plasmid constructs for expressing rat FADS1, FADS2, and FADS3 in COS-7 cells
The plasmids constructed for expressing rat FADS2 (pCMV/ FADS2) and FADS3 (pCMV/FADS3) in mammalian cells have been described ( 11,24 ). A plasmid coding for rat FADS1 was also constructed. Briefl y, from the published ( 25 ) rat FADS1 cDNA sequence (GenBank accession number AF320509), oligonucleotide primers were designed to PCR amplify the entire coding sequence from the rat liver marathon-ready cDNA library (BD Biosciences). The forward primer (5 ′ -gact cc atg g ctcccgacccggtgcaga-3 ′ ) included the translation start codon (in bold) and Nco I restriction site (underlined). The reverse primer (5 ′ -ctag gcggccg c ta ttggtgaaggtaagcatccagccaga-3 ′ ) contained the translation stop codon (in bold) and Not I restriction site (underlined). The resulting PCR product was cloned into the pCMV expression vector using Nco I and Not I sites. The integrity of the plasmid, now referred to as pCMV/FADS1, was confi rmed by DNA sequencing.

Rat FADS3 converts trans -vaccenate into an unknown 18:2
By using COS-7 cells expressing each rat FADS protein, many FA were unsuccessfully tested as putative desaturation substrates of FADS3. When TVA ( trans 11-18:1) was used, interesting FAME profi les were obtained ( Fig. 2 ). First, whatever the experimental condition (control or transfected cells), TVA was converted to RA ( cis 9, trans 11-18:2) by the native ⌬ 9-desaturase (SCD1) present in COS-7 cells ( Fig. 1A ). RA (2.4 ± 0.6% of FA, n = 5) was identifi ed with its retention time (R t = 27.5 min) and MS profi le after conversion to DMOX derivatives (see below). Second, in the FAME profi le obtained with cells expressing FADS3, an additional specifi c FAME (with R t = 28.2 min) was shown ( Fig. 2 ), which was absent in control cells and in cells expressing FADS1 or FADS2. This FA (0.6 ± 0.2% of FA, n = 5) was identifi ed as an octadecadienoic acid (18:2) with a FAME molecular ion at m/z 294 amu. It was in the range of R t in which conjugated linoleic acids (CLA) are detected. Third, in the FAME profi les obtained with cells expressing either rat FADS1 or FADS2 ( Fig. 2 ), additional octadecadienoic acids (18:2) with molecular ions at m/z 294 amu were detected with R t = 21.9 min and R t = 22.1 min, respectively.
Thereafter, similar experiments were carried out with [1-13 C]TVA. The obtained FAME profi les were identical to those described in Fig. 2 , and they confi rmed the presence of the three unknown 13 C-18:2, with FAME molecular ions at m/z 295 amu, when recombinant FADS3, FADS2, and FADS1 were expressed.

Rat FADS3 is a trans -vaccenate desaturase, not a rumenate isomerase
The results described above showed that COS-7 cells expressing FADS3 and incubated with TVA gained the ability to produce an unknown 18:2. However, at this stage, we were not able to discriminate between a desaturase activity of the recombinant FADS3 toward TVA and an isomerase activity toward RA, which was concomitantly produced from TVA by the endogenous ⌬ 9-desaturase ( Fig. 2 ). Therefore, RA was subsequently incubated with COS-7 cells expressing rat FADS1, FADS2, or FADS3. In the FAME profi le obtained after RA incubation with cells expressing FADS3 ( Fig. 3 ), no additional peak with R t = 28.2 min was detected compared with control cells. This result indicated that, in our from some experiments before the DMOX derivatization, FAMEs were separated according to the number and confi guration of their double bonds on preparative silver ion thin-layer chromatography using silica gel H impregnated with 10% (w/w) AgNO 3 and a mixture of 90/10 (v/v) hexane/diethylether for development ( 31 ).

Immunoblot analysis
Immunoblotting was performed as described ( 11 ) using the antibodies described here. Rat FADS1 protein was detected with a polyclonal antibody to FADS1 targeting the antigenic peptide 100 QPSFEPTKNKALTDE. Rat FADS2 was detected with an antibody to FADS2 directed against the peptide 3 KGGNQGEGSTEL. Rat FADS3 was detected with an antibody to FADS3 ( 11 ) directed against the peptide 352 PKEIGHEKHRDWAS. An antibody to stearoyl-CoA desaturase 1 (SCD1) was obtained from Santa Cruz Biotechnology. A quantitative estimation of protein loading was performed with an antibody to actin (Sigma).

RNA isolation and quantitative RT-PCR
Total RNA was extracted from hepatocytes with Trizol (Invitrogen). RNA was converted into cDNA with the IScript cDNA synthesis kit (Bio-Rad). RNA was quantifi ed by real-time PCR with the Taqman Universal PCR Master Mix (Applied Biosystems) containing 40 ng of retrotranscripted RNA, 0.5 µM of primers, and 0.25 µM of Taqman probe. Primers and 5 ′ -FAM/TAMRA-3 ′ Taqman probes specifi c to each rat FADS gene were described previously ( 11 ). Primers (forward primer 5 ′ -caaagagaagggcggaaagc-3 ′ ; reverse primer 5 ′ -aaagtttcgccccagcagta-3 ′ ) and a Taqman probe (5 ′ FAMctggcctcctgctcatgtgcttca-TAMRA3 ′ ) were also designed to quantify SCD1 mRNA. Amplifi cation was performed by an ABI PRISM 7000 (Applied Biosystems) over 40 cycles of 95°C for 15 s and 60°C for 1 min. The 18S gene expression was quantifi ed as an endogenous control using the 18S RNA Control kit Yakima-Yellow Eclipse Dark Quencher (Eurogentec), and relative gene expression was determined from the cycle thresholds (Ct) using the ⌬ ⌬ Ct method.

Expression of results and statistical analysis
All data values shown represent the mean ± SD for the number of replicates (n) indicated in each fi gure legend. Student t -test was used to determine the statistical signifi cance between groups, where indicated.

Rat FADS3 is not functionally redundant with FADS1 or FADS2
To investigate the putative desaturase activity of FADS3, COS-7 cells expressing rat FADS1, FADS2, or FADS3 were grown for 24 h in culture media supplemented with various potential FA substrates. Immunoblot analyses ( Fig. 1A ) revealed that transient transfection resulted in the specifi c expression of each FADS protein. Rat FADS1 and FADS2 were detected as single immunoreactive bands with respective apparent molecular weights of 48 and 45 kDa. As previously described ( 11 ), recombinant rat FADS3 was specifi cally detected at about 48 kDa. Endogenous FADS1, FADS2, and FADS3 were not detected in control cells with our anti-rat antibodies, but COS-7 cells endogenously expressed the ⌬ 9-desaturase (SCD1).
After 24 h of incubation with a FA, the FAME profi le of cells expressing each FADS protein was analyzed by GC-MS to detect the presence of newly desaturated FA. Cells expressing rat FADS2 were able to synthesize 18:4 n-3 from FADS3, cellular FA obtained after incubation with [1-13 C] TVA were converted to DMOX derivatives and analyzed by GC-MS. The mass spectrum of the unknown 13 C-18:2 derivative ( Fig. 4A ) contained DMOX-characteristic intense ions at m/z 114 and 127 amu and a molecular ion at m/z 334 amu, which confi rmed that it possessed a di-unsaturated 13 C18 structure. A gap of 12 amu was detected between ions containing 10 and 11 carbons ( m/z 225 and 237 amu ) and between ions containing 12 and 13 carbons ( m/z experimental conditions, the recombinant FADS3 was not able to catalyze the isomerization of RA and suggested that the 18:2 shown in FADS3-expressing cells ( Fig. 2 ) was likely to result from TVA desaturation.

Structural characterization strongly suggests that the 18:2 produced by rat FADS3 is the trans 11, cis 13-CLA isomer
To locate the double bonds in the carbon chain of the unknown 18:2 produced from TVA in cells expressing as FAME standards (the trans 11, cis 13-CLA isomer was not commercially available). GC-MS analysis of these standards ( Fig. 4B ) showed that the cis 11, cis 13 and trans 11, trans 13-CLA isomers had signifi cantly higher retention times (R t = 28.8 and 29.1 min, respectively) than the ⌬ 11,13-CLA produced by FADS3 from TVA (R t = 28.2 min), while the cis 11, trans 13-CLA isomer had a slightly lower retention 251 and 263 amu ), showing that the unknown FA was a ⌬ 11,13-CLA ( Fig. 4A ). Analysis of this derivative strongly suggested that, after its conversion from [1-13 C]TVA, the 13 C label was still positioned at the 1st carbon.
To get more information on the geometrical structure of this ⌬ 11,13-CLA, we used a commercial mixture of cis 11, trans 13-CLA, cis 11, cis 13-CLA, and trans 11, trans 13-CLA  The identity of each important FA is indicated above its respective peak. It shows that only the cells expressing FADS2 were able to use RA as a substrate. Indeed, detectable amounts of an octadecatrienoic acid (18:3) with FAME molecular ion at m/z 292 amu and R t = 29.1 min were produced when compared with control cells. Although not fully characterized in the present study, this FA was identifi ed as the cis 6, cis 9, trans 11-18:3, the known ⌬ 6-desaturated product of RA (56)(57)(58), by using the NIST database. Representative example of n = 3 distinct experiments.
suggested that the unknown ⌬ 11,13-CLA was not the cis 11, cis 13 or the trans 11, trans 13 isomer. However, because the difference in R t between the cis 11, trans 13 standard isomer and the unknown ⌬ 11,13-CLA was small (only 0.2 min), time (R t = 28.0 min). Knowing that the four geometrical isomers of a CLA elute with a specifi c order ( cis -trans , transcis , cis -cis , and trans -trans ) when separated on a capillary column with a polar stationary phase ( 32 ), this result fi rst complete conclusive identifi cation, these observations strongly suggested that this FA was likely the trans 11, cis 13-CLA corresponding to a cis -⌬ 13-desaturated product from TVA in cells expressing FADS3. This result was in agreement with the ⌬ 11 double bond preexisting in TVA keeping its trans confi guration when the second cis -⌬ 13 double bond was inserted by the recombinant FADS3.
GC-MS analyses of DMOX derivatives also allowed the identifi cation of 13 C-cis 9, trans 11-18:2 (RA) produced by endogenous ⌬ 9-desaturation of [1-13 C]TVA ( Fig. 5A ), 13 Ccis 5, trans 11-18:2 produced by recombinant FADS1 ( Fig. 5B ) identifi cation remained uncertain between the trans 11, cis 13 and the cis 11, trans 13 geometric isomers. Therefore, we compared the DMOX mass spectrum pattern of the unknown ⌬ 11,13-CLA ( Fig. 4A ) with that of the cis 11, trans 13 standard isomer with the nearest R t ( Fig. 4C ). We noticed that, in the case of a trans -⌬ 13 double bond, the ion at m/z 250 amu was slightly more abundant than the ion at m/z 262 amu, whereas the ion at m/z 251 amu of the unknown 13 C-labeled ⌬ 11,13-CLA was signifi cantly less abundant than the ion at m/z 263 amu ( Fig. 4A ). Although there was no commercially available standard for comparison and of TVA by recombinant FADS2 in COS-7 cells ( Fig. 2 ), was also shown ( Fig. 7 ).

Silencing of FADS3 expression in rat hepatocytes
The synthesis of ⌬ 11,13-18:2 from TVA described above suggested that native FADS3 was present and biologically active to catalyze this unique ⌬ 13-desaturation in rat hepatocytes. To correlate the level of ⌬ 11,13-18:2 synthesis with the level of FADS3 protein expression, hepatocytes were transfected with siRNA control or siRNA targeting FADS3 for 24 h and incubated for a further 24 h with TVA. Similarly, hepatocytes were transfected with the pCMV empty or pCMV/FADS3 for 24 h before incubation with TVA. Compared with siRNA control, FADS3 siRNA signifi cantly suppressed the FADS3 mRNA expression by more than 50% ( Fig. 9A ). FADS1, FADS2, and SCD1 mRNA levels were not affected by the treatment. Conversely , pCMV/FADS3 transfection drastically increased the FADS3 mRNA expression ( Fig. 9B ). To a much lesser extent, FADS2 mRNA was increased in hepatocytes transfected with the pCMV/FADS3. In these experiments, because of the massive presence of FADS3 mRNA and high homology with FADS2 mRNA sequence, FADS2 mRNA may have been artifactually increased during RT-PCR analyses.
Although integral membrane-bound desaturases are notoriously diffi cult to study in native conditions ( 33 ), we also analyzed the expression of the FADS3 protein in each experimental condition ( Fig. 9C ). An immunoreactive band was hardly detected but reproducibly shown at 48 kDa with and 13 C -cis 6, trans 11-18:2 produced by recombinant FADS2 ( Fig. 5C ).

Biosynthesis of the suspected trans 11, cis 13-CLA in rat hepatocytes incubated with TVA
Because the liver is known to possess high FA desaturase activities ( 3 ), we chose cultured rat hepatocytes to investigate the ability of the native FADS3 to desaturate TVA, like its recombinant counterpart. Rat hepatocytes were incubated for 24 h with media supplemented with TVA or [1-13 C]TVA and with a control medium without FA. RA (R t = 27.5 min) produced by ⌬ 9-desaturation of TVA was easily detected (7.6 ± 3.2% of FA, n = 5) in the cells incubated with TVA or 13 C-TVA ( Fig. 7 ). A small peak (0.4 ± 0.1% of FA, n = 5) with R t = 28.2 min and FAME molecular ion at m/z 294 amu (incubation with 12 C-TVA) or m/z 295 amu (incubation with 13 C-TVA), similar to the suspected trans 11, cis 13-18:2 previously characterized in COS-7 cells expressing FADS3 ( Fig. 2 ), was specifi cally detected ( Fig. 7 ). The presence of another small 18:2 (0.4 ± 0.1% of FA, n = 5), with R t = 22.1 min and FAME molecular ion at m/z 294 amu or m/z 295 amu (incubation with 12 C-TVA or 13 C-TVA), similar to the ⌬ 6-desaturated product ( cis6 , trans 11-18:2) The identity of each important FA is indicated above its respective peak. It shows that only the cells expressing FADS1, but not FADS2 or FADS3, were able to desaturate CVA into an octadecadienoic acid (R t = 22.4 min), which was identifi ed as the cis 5, cis 11-18:2 by the NIST database. Representative example of n = 3 distinct experiments. n-6, 18:1 n-9, and 16:1 n-7 for FADS2), FADS1 and FADS2 were specifi cally able to produce ⌬ 5-and ⌬ 6-desaturated FAs in detectable amounts ( Fig. 1 and data not shown). These results are in accordance with previous studies that have used this recombinant model to demonstrate the surprising diversity of ⌬ 6-desaturase substrates ( 34 ), including the conversion of 24:5 n-3 into 24:6 n-3, 24:4 n-6 into 24:5 n-6 ( 24 ), and 16:0 into 16:1 n-10 ( 29 ).
As opposed to these positive results, COS-7 cells expressing FADS3 did not show any desaturase activity toward these common FAs ( Fig. 1 ). These results demonstrated that recombinant FADS3 displayed no ⌬ 5-and ⌬ 6-desaturase activities, consistent with previous reports describing FADS2null and FADS1-null mice. In these studies, FADS1 or FADS2 defi ciency ( 6,35,36 ) was not rescued by the presence of FADS3 in terms of PUFA composition of tissues. These data and the present study therefore support the idea that FADS3 is not a ⌬ 5-or ⌬ 6-desaturase isoform capable of compensating the absence of FADS1 or FADS2.
To elucidate the function of the FADS3 gene product, many potential FA substrates were unsuccessfully incubated with the COS-7 cell model expressing FADS3. Finally, upon incubation with TVA ( Fig. 2 ), small amounts of a specifi c octadecadienoic acid were reproducibly synthesized in COS-7 cells expressing FADS3. The use of 13 C labeling confi rmed its metabolic conversion from the initial [1-13 C]TVA substrate. Additional experiments showed that the unknown FA was not synthesized by a FADS3-catalyzed isomerization of RA, which was concomitantly produced by endogenous ⌬ 9-desaturation of TVA ( Fig. 3 ). Evidence is therefore presented that COS-7 cells expressing rat FADS3 specifi cally gained the ability to introduce a double bond in the carbon chain of TVA to produce the unknown 18:2. Experiments were then carried out to further characterize the chemical structure of this unknown octadecadienoic acid. First, its DMOX derivative mass spectrum ( Fig. 4A ) demonstrated that it was a ⌬ 11,13-CLA. Then, comparison of its R t with a commercially available mixture of ⌬ 11,13-CLA standards lacking the trans 11, cis 13 isomer showed that the the anti-FADS3 antibody, matching the apparent molecular weight of the recombinant FADS3 ( Fig. 1A ). The intensity of the 48 kDa band was reduced approximately 30% in the cells treated with FADS3 siRNA compared with siRNA control ( Fig. 9D ) and increased by 1.7 in the cells transfected with pCMV/FADS3 compared with control hepatocytes ( Fig. 9E ).
In the corresponding FAME profi les ( Fig. 10 ), the peak corresponding to the suspected trans 11, cis 13-18:2 was 3-fold smaller in FADS3 siRNA transfected cells compared with siRNA control ( Fig. 10A, B ). Conversely, the ⌬ 11,13-18:2 was 2-fold greater in the cells transfected with pCMV/ FADS3 compared with the cells transfected with the pCMV empty ( Fig. 10A, C ) . The amounts of TVA, RA, and cis 6, trans 11-18:2 were not affected by the treatments ( Fig. 10B, C ). In these experiments, due to the use of transfection reagents, TVA uptake and incorporation into the cells was lower than in classical incubation experiments ( Fig. 7 ), leading therefore to a lower production of the ⌬ 11,13-CLA.
These results suggested that siRNA targeting FADS3 signifi cantly suppressed FADS3 mRNA expression, triggered the partial abolition of the native FADS3 protein expression, and specifi cally reduced ⌬ 11,13-18:2 synthesis from TVA. Conversely, the recombinant FADS3 expression increased the production of the suspected trans 11, cis 13-18:2.

DISCUSSION
No function has been attributed to the FADS3 gene yet, but it has been postulated that FADS3 may be a new animal FA desaturase. This hypothesis was tested in this study to address the functional role of FADS3. First, transfection of the 1,350 bp open reading frame of the rat FADS3 in COS-7 cells was used to investigate its putative desaturase activity ( Fig. 1 ). Rat FADS1 and FADS2 were concomitantly expressed in this model to validate the positive or negative results obtained with FADS3. Indeed, when incubated with their known FA substrates (20:3 n-6 for FADS1; 18:3 n-3, 18:2 unknown ⌬ 11,13-CLA was the trans 11, cis 13-18:2. In addition, it would be unlikely that the trans -⌬ 11 double bond preexisting in TVA would be isomerized to a cis double bond at the same time than a trans -⌬ 13 double bond would be introduced by FADS3. However, because of the lack of commercially available trans 11, cis 13-CLA standard, the identifi cation could not be considered completely conclusive. Other methodologies, such as GC linked to Fourier transform infrared spectroscopy (FTIR), have been considered for a better characterization of the cis / trans confi guration cis 11, cis 13 and trans 11, trans 13 isomers with R t signifi cantly higher (+0.6 and +0.9 min, respectively) could be discarded ( Fig. 4B ). However, uncertainty remained between the cis 11, trans 13 and trans 11, cis 13 isomers. Therefore, we analyzed the DMOX derivative mass spectrum of the commercially available cis 11, trans 13 standard ( Fig. 4C ) and showed a reproducible difference in the relative abundance of the ions at m/z 250 and 262 amu with the unknown ⌬ 11,13-CLA ( Fig. 4A ). This result strongly suggested that, based on exclusion of the cis 11, trans 13 isomer, the fourth of total CLA ( 42 ), suggesting that its concentration could be related to cows that graze mountain pasture. However, to our knowledge, this particular CLA has not yet been described in humans or in nonruminant species, even after consumption of TVA-enriched diets in studies focusing on the conversion of TVA into rumenic acid ( 47,48 ). In the present study, when rat hepatocytes were exposed to lower concentrations of TVA (50 µM instead of 200 µM, data not shown), the suspected trans 11, cis 13-18:2 was still detected but represented 0.1% of cell FA (limit of detection). This result suggested that the ⌬ 11,13-CLA may weakly appear under TVA concentrations described as physiological in the plasma of humans ( 49 ) and ruminants ( 50 ).
The characterization of this FA as a ⌬ 11,13-CLA means that COS-7 cells expressing rat FADS3 gained the ability to catalyze its synthesis by ⌬ 13-desaturation of TVA. In rat hepatocytes incubated with TVA, we additionally demonstrated that the FADS3 mRNA silencing specifi cally reduced the of the product, but FTIR does not distinguish between cis , trans and trans , cis isomers ( 37 ) and would not have added information to this characterization. This illustrated the considerable analytical challenge of identifying CLA isomers when well-characterized standards are lacking ( 30,37 ).
Moreover, this unexpected CLA initially characterized in COS-7 cells expressing FADS3 was also found in cultured rat hepatocytes incubated with TVA and 13 C-TVA ( Fig. 7 ). Bibliographic data on the trans 11, cis 13-18:2 ( 38,39 ) revealed that this specifi c FA has already been identifi ed in the milk and digestas of ruminants (40)(41)(42)(43)(44)(45)(46), which are known to produce TVA from dietary linoleic acid (18:2 n-6) by bacteria hydrogenation in their rumen. We also found the presence of this specifi c CLA in cow milk and goat mammary gland (data not shown). The trans 11, cis 13-CLA isomer is therefore a naturally relevant FA. In the milk fat from Alpine cows compared with indoor cows, the trans 11, cis 13-18:2 was greatly increased and represented more than one synthesis of the suspected trans 11, cis 13-18:2, whereas conversely, recombinant FADS3 expression increased its production ( Figs. 9 and 10 ). Our report is the fi rst to show that recombinant and native FADS3 are able to introduce a double bond between carbons 13 and 14 of TVA. In previous studies, it has been postulated but never demonstrated that the synthesis of trans 11, cis 13-18:2 in ruminants could come from dietary 18:3 n-3. Indeed, 18:3 n-3 ( cis 9, cis 12, cis 15) can be converted to cis 9, trans 11, cis 15-conjugated triene in the rumen of ruminants ( 42,45 ) and subsequently metabolized to trans 11, cis 15-18:2 by incomplete bacteria hydrogenation. The last step (i.e., isomerization of trans 11, cis 15-18:2 to trans 11, cis 13-18:2) has been suggested but never confi rmed. We propose here an alternative new and direct pathway ( Fig. 11 ) catalyzed by FADS3 toward TVA, which is also a product of incomplete bacteria hydrogenation of dietary linoleic acid (18:2 n-6) in the rumen of ruminants. In ruminants, both pathways may coexist, whereas in nonruminant species, including humans, only the FADS3-catalyzed pathway would be possible.
The suspected cis geometrical structure of the newly introduced double bond is not surprising because all of the well-characterized mammalian fatty acyl-CoA desaturases ( ⌬ 9-, ⌬ 6-, and ⌬ 5-desaturases) are known to introduce cis double bonds ( 51 ). More surprising is the position of the inserted ⌬ 13 double bond, which shows that FADS3 is not a front-end desaturase, as previously hypothesized from its amino acid sequence ( 14 ). The so-called group of "frontend" desaturases refers to the desaturases which introduce double bonds between the preexisting double bond (usually located in ⌬ 9 position) and the carboxyl end of a FA ( 52 ). The front-end desaturases from eukaryotes are structurally characterized by the presence of an N-terminal cytochrome b5-like domain fused to the main desaturase domain and by the presence of a glutamine ( 53 ), which replaces a histidine residue (present in methyl-end desaturases of plants) in the third histidine-rich box H/QX 2-3 HH, which is the case for FADS3 ( 11 ) ( Fig. 12 ). These results show that FADS3 is an exception to the consensus pattern that higher animals, including humans, are devoid of desaturases that insert double bonds in the methyl-end of FA. This exception should be modulated , considering that the fi rst preexisting double bond of the single known FA substrate (i.e., TVA) is a trans -⌬ 11.
There has been no report in the literature on a ⌬ 13desaturase activity in mammals. The existence of such a ⌬ 13-desaturase activity has been described in insects for the multifunctional desaturase from Thaumetopoea pityocampa sex pheromone gland, which converts palmitic acid into (Z)-13-hexadecen-11-ynoic acid ( 54 ), but this enzyme presents no sequence homology with FADS3. In a recent study, the metabolic conversion of trans 11, trans 13-18:2 into cis9 , trans 11-18:2 in CaCo2 cells suggested the action of an enzyme capable of reducing the ⌬ 13 double bond, after a FADS3 were incubated with ␣ -eleostearic acid ( cis 9, trans 11, trans 13-CLnA) (data not shown).
In both COS-7 cells ( Fig. 2 ) and hepatocytes ( Fig. 7 ), the rate of the TVA conversion into the suspected trans 11, cis 13-18:2 was low, probably because of the enzyme competition of TVA for both ⌬ 9-desaturase and FADS3. fi rst ⌬ 9-desaturation step leading to the formation of cis 9, trans 11, trans 13-conjugated linolenic acid (CLnA) as an intermediate ( 55 ). We hypothesized that FADS3 could be a ⌬ 13-desaturase/ ⌬ 13-reductase bifunctional enzyme, depending on the substrate delivered; however, there was no ⌬ 13-reductase activity when COS-7 cells expressing Fig. 11. Proposed metabolic pathway of trans -vaccenate conversion into cis 9, trans 11-CLA catalyzed by the ⌬ 9-desaturase and into trans 11, cis 13-CLA by the newly discovered FADS3-catalyzed ⌬ 13-desaturase . Fig. 12. Alignment of the amino acid sequences of rat, mouse, and human FADS1, FADS2, and FADS3 deduced from their nucleotide sequences. In the N-terminal end, the cytochrome b5-like domain (present in FADS1, FADS2, and FADS3 sequences) is surrounded by the blue box and characterized by a HPGG motif (black box). In the carboxyl-terminal end, the desaturase domain is characterized by three histidine boxes (red) that are necessary for the catalytic activity, which are present in all the sequences shown . This result suggested that the affi nity of FADS3 toward TVA was lower than that of SCD1. Although this study focused on FADS3, we also showed that recombinant FADS1 and FADS2 were able to ⌬ 5-and ⌬ 6-desaturate TVA into cis 5, trans 11-18:2 and cis 6, trans 11-18:2, respectively ( Figs. 2 and 5 ). Therefore, in the recombinant model, TVA was desaturated by all known animal desaturases ( ⌬ 9-, ⌬ 5-, and ⌬ 6-desaturases, and FADS3). In rat hepatocytes, only the products of ⌬ 9-desaturation ( cis9 , trans 11-18:2), ⌬ 6-desaturation ( cis 6, trans 11-18:2), and ⌬ 13-desaturation ( trans 11, cis 13-18:2) were detected ( Figs. 7 and 8 ). Further studies will provide more information on the relative affi nity of the animal desaturases for their common TVA substrate in physiological conditions.
To conclude, 13 years after its gene identifi cation in the FADS cluster, this study demonstrates that FADS3 displays a unique and unexpected ⌬ 13-desaturase activity. Inside a large panel of FA, TVA is for now the single substrate to be identifi ed. Therefore, the high selectivity of FADS3 for TVA does not alter the current paradigm of mammalian PUFA dependence on dietary essential FA. Available evidence also indicates that the FADS3-catalyzed ⌬ 13-desaturation of TVA is probably not a major pathway, which is in agreement with its late discovery, but further studies may show that it becomes important in the presence of a high level of dietary TVA (or linoleic acid in ruminants) or under specifi c physiological conditions. Finally, at present, there is no known function for the suspected trans 11, cis 13-18:2 product or metabolic basis for thinking that it might be physiologically important. Based on the present description of its FADS3-catalyzed biosynthesis, further studies will probably focus on this particular CLA to answer this question.