A novel FADS1 isoform potentiates FADS2-mediated production of eicosanoid precursor fatty acids.

The fatty acid desaturase (FADS) genes code for the rate-limiting enzymes required for the biosynthesis of long-chain polyunsaturated fatty acids (LCPUFA). Here we report discovery and function of a novel FADS1 splice variant. FADS1 alternative transcript 1 (FADS1AT1) enhances desaturation of FADS2, leading to increased production of eicosanoid precursors, the first case of an isoform modulating the enzymatic activity encoded by another gene. Multiple protein isoforms were detected in primate liver, thymus, and brain. In human neuronal cells, their expression patterns are modulated by differentiation and result in alteration of cellular fatty acids. FADS1, but not FADS1AT1, localizes to endoplasmic reticulum and mitochondria. Ribosomal footprinting demonstrates that all three FADS genes are translated at similar levels. The noncatalytic regulation of FADS2 desaturation by FADS1AT1 is a novel, plausible mechanism by which several phylogenetically conserved FADS isoforms may regulate LCPUFA biosynthesis in a manner specific to tissue, organelle, and developmental stage.

studies reveal that a FADS1 isoform enhances FADS2 activity, the fi rst known function for a FADS isoform and the fi rst example of an AT from one gene regulating the activity coded by an adjacent gene. These data are consistent with ⌬ 6-desaturase enhancement of desaturase activity by a lipoprotein-like protein reported in the 1980s ( 28 ). Additionally, we present the detection of protein isoforms in neonate tissues and mammalian cells, protein isoform expression and fatty acid changes during cell differentiation, and isoform-specifi c subcellular localization, and fi ndings that all three FADS genes are translated at similar levels in mouse embryonic fi broblasts.

MATERIALS AND METHODS
Studies on baboons were approved by the Cornell University and Texas Biomedical Research Institute (formerly the Southwest Foundation for Biomedical Research) Institutional Animal Care and Use Committees.

RNA isolation and preparation of RACE-ready cDNA
Neonate baboon liver tissue from a 12-week-old baboon treated with RNAlater and stored at Ϫ 80°C since necropsy was used to isolate total RNA, and the RNA quality was assessed as described previously ( 25 ). First-strand 5 ′ RACE-ready cDNA and 3 ′ RACEready cDNA was prepared per the manufacturer's recommended protocol provided with SMARTer TM RACE cDNA amplifi cation kit (Clontech Laboratories, Mountain View, CA).
Human FADS1 ( 23 ) spans 17.2 kb of genomic DNA, shares 61% and 52% identity with FADS2 and FADS3 , respectively, and encodes a protein of 444 amino acids with a molecular mass of 52.0 kDa ( 24 ). The classical FADS1 transcript has been shown to be highly expressed in the liver, brain, and heart ( 23 ). Even though FADS1 operates on both n-6 and n-3 PUFA, only a single transcript has been identifi ed to date. However, currently in the National Center for Bioinformatics Information (NCBI) database, human FADS1 (GenBank accession NM_013402) is represented with an open reading frame (ORF) that encodes a 501 amino acid peptide. This new larger peptide contains 65 amino acids more than the classical ⌬ 5-desaturase protein (444 amino acids), yet the function of it is not known.
We recently showed the fi rst alternative transcripts for FADS2 and FADS3 , which were generated by alternative splicing events and expressed in a tissue-specifi c manner in a primate (neonate baboon) ( 25,26 ). They are conserved in several mammals and the chicken, and they exhibit reciprocal changes in gene expression in response to human neuronal cell differentiation (25)(26)(27). To investigate whether FADS1 is also subject to alternative splicing, we performed both 5 ′ and 3 ′ rapid amplifi cation of cDNA ends (RACE) using gene-specifi c primers from baboon FADS1 (GenBank accession EF531577). Here, we show unambiguous evidence of the existence of several FADS1 mRNA isoforms generated by alternative transcription initiation, alternative selection of poly(A) sites, and internal exon deletions resulting from alternative splicing. Our Currently accepted pathway for LCPUFA synthesis from precursors. Alternating desaturation and elongation occurs on the ER. ⌬ 8-desaturation of 20:2n-6 and 20:3n-3 yield 20:3n-6 and 20:4n-3, intermediates in the conventional pathway to 20:4n-6 and 20:5n-3 ( 18 ), which can be further elongated and desaturated.
(Bio-Rad, Hercules, CA). The samples were transferred to nitrocellulose membranes and probed with the FADS1 and ␤ -actin antibodies. FADS1 antibody (cat. #AV42384) was purchased from Sigma-Aldrich (St. Louis, MO). ␤ -actin and the secondary antibodies were from LiCor Biosciences (Lincoln, NE). Immunoblots were imaged and immunofl uorescence signal was detected by using LiCor Odyssey infrared imaging system as directed by the manufacturer (LiCor Biosciences).

Subcellular localization of FADS1 isoforms
To determine the localization of FADS1 and FADS1AT1 , the coding regions were cloned into pEGFP-N1 vector driven by the CMV promoter (pE FADS1 and pE FADS1AT1 , respectively). NB cells were transfected with the constructed vectors using Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA), and organelle-specifi c stains MitoTracker Red CMXRos or ER-tracker Blue-White DPX (Molecular Probes, Invitrogen) were used to stain organelles. Live cells were visualized with an inverted Meta Confocal Microscope (LSM 510 META, Zeiss). Mitochondria were isolated from the transfected NB cells using mitochondria isolation kit (Thermo Scientifi c). Western blot was performed using GFP antibody (Abcam), COX IV (inner mitochondrial membrane-specifi c) antibody (Cell Signaling), and PDI (endoplasmic reticulumspecifi c) antibody (Cell Signaling).

Translation of FADS transcripts in mouse cells
MEF cells were fi rst treated with cycloheximide (100 µg/ml) for 3 min at 37°C to free the translating ribosomes on mRNAs. Cells were then harvested by ice-cold polysome lysis buffer (10 mM HEPES, pH 7.4, 100 mM KCl, 5 mM MgCl2, 100 µg/ml cycloheximide, 5 mM DTT, 20 U/ml SUPERaseIn, and 2% Triton X-100) followed by profi ling using 10-50% sucrose gradient. Polysome fractions were collected and treated with RNase I to digest the mRNA segments not protected by ribosomes . The ribosomeprotected fragments were enriched and converted into a DNA library suitable for Illumina sequencing. Qualifi ed sequencing reads were aligned to the cDNA database using SOAP2-based mapping software allowing up to two mismatches.

RESULTS
We cloned and sequenced the entire coding region of baboon FADS1 (GenBank accession EF531577). The baboon FADS1 coding sequence shares 97% identity with the human FADS1 sequence as well as exon architecture. The 5 ′ and 3 ′ RACE was performed using FADS1 gene-specifi c primers.

′ RACE
5 ′ RACE PCR experiments generated four products. The long base pair (bp) product was prominent on the ethidium bromide-stained agarose gel, whereas the remaining three products were weakly visualized. All four gel-purifi ed products were sequenced, and the Basic Local Alignment Search Tool (BLAST) confi rmed the products (750 bp, 537 bp, 430 bp, and 403 bp) as FADS1 amplicons. Promoter and TSS prediction was performed using Neural Network Promoter Prediction software ( 30 ). For the 750 bp sequenced product, the promoter region ranging from Ϫ 95 bp to Ϫ 45 bp from the fi rst translation start site was identifi ed with a minimum promoter score of 0.54; the TSS (TSS1) is located at Ϫ 55 bp from the fi rst translation start site ("TSS1" for the classical transcript FADS1CS ). The pGEM-T Easy vector (Promega) and sequenced at Cornell University Life Sciences Core Laboratories Center.

Cell culture
Three human cell lines, SK-N-SH neuroblastoma (NB), MCF7 breast cancer, and HepG2 hepatocellular carcinoma cells, were grown in DMEM/F-12, MEM-␣ , and MEM/EBSS media with 10% FBS (media and serum obtained from HyClone), respectively, at 37°C in a humidifi ed environment with 5% CO 2 . At confl uence, the cells were harvested for RNA and protein extraction. RNA was isolated using the QIAshredder and RNeasy Mini kit (Qiagen, Valencia, CA). Protein extract was carried out using RIPA lysis buffer (Thermo Scientifi c, IL). Protein concentrations were determined by a bicinchoninic acid (BCA) assay (Thermo Scientifi c). NB cell differentiation assays were carried out in serum free DMEM/F-12 with 1× N-2 supplement (Invitrogen) as described earlier ( 25 ).

Transfection assay
The open reading frame of FADS transcripts ( FADS1, FADS2, FADS1AT1 ) were cloned into a pcDNA3.1 expression vector containing cytomegalovirus (CMV) promoter (Invitrogen). FADS1AT1 and empty vector stable MCF7 cells were dosed with 100 M of albumin-bound 18:2n-6 or 20:3n-6 fatty acid and were incubated for 24 h. For cotransfection studies, MCF7 cells stably expressing FADS1AT1 or empty vector were transfected with equal amounts of FADS1 or FADS2 DNA using Lipofectamine LTX (Invitrogen). Twenty-four hours after cotransfection, the cells were incubated with 100 M of albumin-bound 18:2n-6 or 20:3n-6 fatty acid for additional 24 h. After incubation, the cells were washed twice with 1× PBS and removed by trypsinization. Then fatty acids were analyzed.

Fatty acid analysis
Fatty acid methyl ester (FAME) preparation and structural identifi cation of FAME was carried out as described earlier ( 18 ). Briefl y, cells were isolated and total fatty acids hydrolyzed and converted to FAME by a one-step reaction mixture. Structures were identifi ed by gas chromatography covalent adduct chemical ionization tandem mass spectrometry (GC-CACI-MS/MS), which provides positive structural assignments of double-bond positions for all monoene and homoallylic FAME, for low abundance FAME ( 29 ). Quantitative analysis was performed with GC coupled to a fl ame ionization detector using an equal weight mixture for response factor calibration.

FADS1 transcript expression in baboon tissues and human cells
Expression levels of FADS1 transcripts was measured using cDNA from nine normal tissues from a 12-week-old baboon neonate and three human cell lines by RT-PCR. cDNA from baboon tissues is from a previous study ( 25 ). RT-PCR analysis was performed using primers designed from unique regions specifi c for each transcript. (Primer sequences are provided upon request.) RT-PCR reactions were carried out using 1 M of each primer, 0.25 mM each of dNTPs, 1.5 mM MgCl2, and AmpliTaq II (ABI, Foster City, CA) in a fi nal volume of 30 l. Thermal cycling conditions were as follows: initial denaturation at 95°C for 5 min followed by 35 cycles of denaturation at 95°C for 30 s, annealing at 65°C for 45 s, and extension at 72°C for 1 min, with a fi nal extension at 72°C for 5 min. PCR products were separated on 2% agarose gels.

Immunoblot
Total protein from baboon neonate liver, brain, and thymus tissue lysates and cultured cells was resolved by SDS-PAGE motifs, losing the third due to splicing ( Fig. 2 ). Our screening did not identify the novel NCBI-annotated transcript that codes for a predicted 501 amino acid protein.

Expression of FADS1 transcripts in neonate baboon tissues and human cells
To determine tissue-specifi c expression levels of the transcripts initiated from TSS1 ( FADS1CS ) and TSS2 ( FADS1AT1 ), we performed RT-PCR analysis using transcript-specifi c primers. The transcripts initiated from TSS3 have very short 5 ′ UTRs embedded within the exon 2; it was not possible to check the expression levels of these transcripts as the sequence overlaps with the TSS1 transcripts. FADS1CS expression was higher than FADS1AT1 in all the tissues and human cells examined ( Fig. 5A , B ). No amplifi cation product was detected in skeletal muscle, and expression in spleen was very low for FADS1AT1 . A trace FADS1AT1 band was visible in MCF7 and SK-N-SH (NB) cells.

FADS1 protein detection by Immunoblot
To determine whether the newly identifi ed FADS1 mRNA transcripts encode a protein expressed in tissue, we performed immunoblot analysis using protein lysates extracted from baboon neonate liver, brain, and thymus tissues 750 bp amplicon closely corresponds to previously reported human sequences (GenBank accessions AK027522 and AF084558). The 537 bp amplicon arose because of internal splicing within the distal end of exon 1, followed by a 81 bp N-terminal or 5 ′ end sequence extension, resulting in an alternative exon ("E1A"). The promoter region (minimum promoter score of 0.27) for this amplicon ranged from Ϫ 106 bp to Ϫ 56 bp, and the TSS (TSS2) is located Ϫ 66 bp from the fi rst translation start site ("TSS2" for the alternative transcript FADS1AT1 ). In this amplicon, the initial 81 bases ( Ϫ 136 bp to Ϫ 55 bp) from the fi rst translational start site are strikingly similar to a newly identifi ed sequence from the NEDO splicing variants project (GenBank accession AK299762), which is also highly similar to FADS2 . Internal splicing within exon 2 gives rise to 430 bp and 403 bp amplicons. They are located at +224 and +251 of the baboon FADS1 gene (GenBank accession EF531577) and have very short 5 ′ UTR. We refer to this cluster as "TSS3."

′ RACE
We performed 3 ′ RACE to detect alternative selection of poly(A) sites and splice variants. Seven amplicons with varying lengths were detected on agarose gels. The sizes of the sequenced amplicons were 1114 bp, 1156 bp, 1460 bp, 1606 bp, 1620 bp, 2603 bp and 3750 bp. BLAST (http:// www.ncbi.nlm.nih.gov/BLAST) searches confi rmed the products as FADS1 amplicons. Six out of the seven amplicons differ only in the 3 ′ UTR length because of alternative use of poly(A) sites. The 2603 bp amplicon has several exon deletions resulting from internal exon splicing and closely (99% identity) resembles a Rhesus Macaque sequence (GenBank acc# AC194778).

In silico analysis
In silico analysis of the 5 ′ /3 ′ RACE sequences for FADS1 show that they have at least four 5 ′ UTRs and several 3 ′ UTRs of varying lengths (GenBank accessions JF518968, JF518969, JF518970, JF518971, JF518972, JF518973, JF518974, and JF518975). Putative coding regions of these transcripts were predicted using ORF fi nder software. The four conserved functional features of most PUFA desaturases are an N-terminal cytochrome b5 domain and three histidine motifs (HXXXH, HXXHH, and QXXHH) conserved from humans to microalga ( 31 ). The amplicon that starts at TSS1 can uniquely be translated from the exon 1 to produce a protein product with 444 amino acids, retaining all four conserved desaturase domains: a cytochrome b5 domain and three histidine motifs. However, the inframe translation start site for TSS2 and TSS3 amplicons, located within exon 2, all have identical protein-coding sequences. They produce 360 amino acid proteins lacking the N-terminal cytochrome b5 domain ( Fig. 2 ).
We also found a 2,603 bp transcript generated by internal exon deletions using 3 ′ RACE. This transcript had a truncated exon 6, skipped exons 7-11, and truncated exon 12. For this splice variant, ORF fi nder predicts a 270 amino acid protein using TSS1. The 270 amino acid protein retains the cytochrome b5 domain and the fi rst two histidine b5 domain; this immunogen was predicted to recognize both the classical and the putative isoforms. We detected two prominent (42 and 48 kDa) and two faint (54 and 65 kDa) products in liver, two prominent (48 and 65 kDa) and one faint (60 kDa) product in brain, and one prominent (48 kDa) product in thymus ( Fig. 6A ). We also found 42 kDa and 48 kDa products in HepG2 and MCF7 human cells. A signifi cant observation was increased expression of the 65 kDa product in the cells compared with the tissues ( Fig. 6B ).
Previously, we showed reciprocal changes in FADS3 alternative transcript expression changes in undifferentiated versus differentiated NB cells ( 25 ). We carried out a similar cell differentiation experiment and performed Western blot to check for isoform expression differences. We detected 54 kDa FADS1 protein isoform to be highly expressed in undifferentiated cells, whereas no product was seen in differentiated cells ( Fig. 7A ). To establish a functional signifi cance of 54 kDa isoform expression limited only to undifferentiated cells, the fatty acid composition of NB cells was analyzed. The fatty acid concentrations of undifferentiated versus differentiated NB cells showed and three human transformed cells (HepG2, MCF7, and NB). An immunogen directed against the C-terminal third histidine motif was selected because the new putative protein isoform does not contain the N-terminal cytochrome  band was visible using PDI antibody, demonstrating that there is no ER contamination in the mitochondrial isolate and showing conclusively that FADS1CS localizes to mitochondria.

Active translation of FADS transcripts in mammalian embryonic cells
The range of alternative transcripts now identifi ed for all three FADS genes prompted us to look for the translation of FADS mRNA in mouse embryonic fi broblast (MEF) cells using ribosome foot-printing technology. On the basis of a previously published protocol ( 34 ), we investigated the translation (mRNA → protein) of all three FADS mRNA in MEF cells. Briefl y, MEF cell mRNA was isolated, retaining between 1 and 20 ribosomes attached in the process of translation, the mRNA resembling a string and the ribosomes resembling pearls. Bare mRNA was digested, leaving ribosomes protecting mRNA of length averaging 28 mers. Ribosomes were then proteolyzed, releasing mRNA for sequencing and assignment to genes. Fig. 9 presents a histogram of the results, showing the exact copy numbers of various 28 mers positioned along the coding regions of FADS1 , FADS2 , and FADS3 . mRNA fragments with high copy numbers may refl ect translational pausing. The total fragments appear at approximately the same abundance from each FADS gene, providing the fi rst positivesequence-specifi c proof that FADS1 , FADS2 , and FADS3 are translated at similar levels in mammalian embryonic stage cells.

DISCUSSION
Here we report that a novel human FADS1 alternative transcript, FADS1AT1 , with no desaturase activity potentiates the function of the FADS2 classical transcript on its native substrates, 18:2n-6 and apparently 20:2n-6. The net process promotes the accumulation of the downstream eicosanoid precursor 20:3n-6 synthesized via rapid (not ratelimiting) elongation either subsequent or prior to ⌬ 6-or ⌬ 8-desaturation, respectively. Alternative transcripts are known to modulate the function of the classical transcript of their respective genes (e.g., the estrogen receptor); however, this is the fi rst instance of the alternative transcript modifying the function of another gene.
Biochemical studies of ⌬ 6-desaturation demonstrated that the microsomal activity required a low density protein ("lipoprotein-like" saline density 1.26-1.27 g/ml) from cytosol for normal activity ( 28 ). The reported protein bound fatty acids and fatty acid CoA, especially 18:2n-6 CoA, suggesting it is involved in transport and is noncatalytic; no other properties were reported. Speculation on mechanisms of action included binding and transport of product fatty reciprocal changes in the composition of polyunsaturated fatty acids (PUFA) compared with monounsaturated fatty acids (MUFA). The physiologically important LCPUFA [AA, EPA, and 22:6n-3 (docosahexaenoic acid )] was greater in undifferentiated NB cells compared with MUFA [18:1n-7 (vaccenic acid) and 18:1n-9 (oleic acid)]; an opposite pattern was found in differentiated cells ( Fig. 7B ). The most signifi cantly increased fatty acid was AA (>4-fold change) in undifferentiated cells; the protein encoded by FADS1 , the ⌬ 5-desaturase, catalyzed the synthesis of AA.

Subcellular localization
FADS are known to be ER proteins, but some evidence suggests they are active in other organelles, especially mitochondria ( 32,33 ). To determine the subcellular localization of FADS1CS (444 amino acids) and FADS1AT1 (360 amino acids), we performed experiments using GFP tagging system. NB cells transfected with constructed vectors were analyzed after staining with MitoTracker Red and ER-tracker Blue-White DPX to specifi cally target the organelle localization. Imaging analysis suggests that both FADS1CS and FADS1AT1 localize to the ER; FADS1CS , but not FADS1AT1 , localizes to the mitochondria ( Fig. 8A , B ). To confi rm the mitochondrial localization, we isolated mitochondria from the NB cells and performed Western blot using GFP, COX IV, and PDI antibodies to detect FADS1 and confi rm the purity of mitochondria. We detected a 75 kDa band for FADS1CS , but not for FADS1AT1 , with the GFP antibody corresponding to the intact FADS1CS -GFP band (48 kDa FADS1CS ; 27 kDa GFP). Clear bands were observed using COX IV antibody ( Fig. 8C ). No  DNA, 10% are regulated by bidirectional promoters ( 35,36 ). Moreover, bidirectional arrangements are conserved among orthologous gene pairs in the mouse genome ( 35 ). We also report a complex functional gene structure of primate FADS1 resulting from alternative promoter usage and transcription initiation, alternative use of poly(A) sites, and alternative splicing generating multiple mRNA transcripts and protein isoforms. Previously, a single FADS1 transcript was known in mammals ( 24 ); we previously reported that FADS3 encodes several splice variants that arise from alternative splicing of internal exons and that a splice variant of FADS2 arises from deletion of internal exons. The majority (74%) of splicing events are reported to occur in coding regions, followed by the 5 ′ UTR (22%) and 3 ′ UTR (4%) ( 37 ).
UTR heterogeneity is striking, with identifi cation of at least four 5 ′ UTR and seven 3 ′ UTR variants, whereas only one new transcript has internal exon deletions. 5 ′ UTR heterogeneity leads to alternative promoter usage, either encoding identical protein isoforms or different protein isoforms with distinct functional activities ( 38,39 ). Alternative promoter usage and 5 ′ UTR heterogeneity can affect tissue-specifi c expression and translational effi ciency and generate protein diversity ( 38,40 ), and it is widely acids away from the enzyme. The predicted FADS1AT1 protein is lipophilic and would likely bind fatty acids. We have described numerous AT of the FADS genes, particularly for FADS3 , but we have been unable to establish a function through transfection studies. Extension of the present FADS1AT1 results leads to the plausible hypothesis that FADS3 AT and FADS2 AT ( 25, 26 ) play a noncatalytic role in desaturase activity that may be revealed by cotransfection studies.
FADS1AT1 lacks the conserved cytochrome b5 domain characteristic of front-end desaturases. Splice variants that lack conserved domains exert dominant negative or positive effects on the active forms ( 8,9 ). A possible facilitator of this interaction may be the signifi cant overlap of the 5 ′ UTR region of the FADS1AT1 sequence with the newly identifi ed FADS2 isoform (GenBank accession AK299762), both sharing a common bidirectional promoter. For all human genes that are adjacent and transcribed from opposite strands of  AATAAA poly(A) signal 14 and 13 bp upstream of polyadenylation start site, respectively. Recently, Thomsen et al ( 46 ), proposed two hypotheses to explain the existence of distinct 3 ′ UTR during studies of the developmentally regulated Hox genes in Drosophila : i ) miRNA avoidance, in which shorter 3 ′ UTR isoforms may have evolved from longer 3 ′ UTR isoforms to avoid miRNA targeting; and ii ) miRNA "enhanced regulation," in which longer isoforms evolved from shorter isoforms to interact with miRNAs at selected spatiotemporal coordinates thereby enhancing regulatory surfaces. We queried DIANA-microT web server software ( 47 ) using Human FADS1 (GenBank accession NM_013402) to check for miRNA binding sites at the 3 ′ UTR ( 48 ). Three conserved miRNA (hsa-miR-511, hsa-miR-140-5P, and hsa-miR-150) binding sites were identifi ed at the 3 ′ UTR position of the human FADS1 . Two (2,603 bp and 3,750 bp) out of seven amplicons reported here have hsa-miR-140-5P and hsa-miR-150 binding sites. The hsa-miR-140-5P miRNA site at 3 ′ UTR position 2,401-2,429 bp is well conserved in rat, mouse, rabbit, and baboon. The hsa-miR-150 miRNA site at 3 ′ UTR position 1,145-1,173 bp is also well conserved in dog and baboon. Future studies may establish whether these miRNA regulate FADS1 mRNA degradation.
The tissue-and cell-specifi c expression of FADS1CS and FADS1AT1 transcripts can be attributed to utilization of distinct promoter elements. The ubiquitously expressed FADS1CS shows greater expression in most tissues than does FADS1AT1 , possibly due to stronger promoter activity. Protein isoforms generated by alternative use of promoters can exhibit opposing biological functions. For instance, for tumor suppressor p53 gene, the full-length isoforms promote cell-cycle arrest and apoptosis, whereas the truncated isoforms promote proliferation ( 49 ).
Additionally, we detected multiple protein isoforms of FADS1 in primate tissue using a C-terminal human FADS1 antibody. Considering the protein coding cDNA sequence of baboon FADS1 (GenBank accession EF531577), we predicted a protein size of 52 kDa using a protein molecular weight calculator (http://www.sciencegateway.org/tools/ proteinmw.htm). However, we found a smaller molecular size product of 48 kDa in all the three tissues, with highest expression in liver followed by thymus and brain. This shorter product may be due to protein degradation or cleavage. Similar observations were made by Pedrono et al. ( 50 ) using a FADS3 antibody. For the cDNA sequence of baboon FADS1AT1 (GenBank accession HQ440212), we predicted a protein size of 42 kDa. This 42 kDa protein is detected only in the liver tissue and closely resembles stearoyl-CoA desaturase ( SCD ), a key lipogenic enzyme that catalyses the biosynthesis of MUFA and is associated with metabolic syndrome (obesity and diabetes) ( 51 ). Both of them lack the cytochrome b5 domain but retain three well-conserved histidine motifs. The other significant products that we observed were 54 kDa, 60 kDa, and 65 kDa protein isoforms, but 5 ′ and 3 ′ RACE experiments did not successfully identify corresponding transcripts. A comprehensive analysis by using RNA-Seq technique may reveal FADS1 isoforms. However, in HepG2 and MCF7 represented in genes that are involved in transcription regulation and development ( 41 ). Two FADS1 isoforms have very short 5 ′ UTR that are embedded in the exon 2 of baboon FADS1 . Shorter 5 ′ UTR variants are translated more effi ciently than longer variants in TGF-␤ , BRCA1 , and Mdm2 ( 42 ).
We also identifi ed at least seven distinct 3 ′ UTR variants for FADS1 . In mammals, mRNA polyadenylation functions in mRNA stability, translation, transport, and processing ( 43 ). In mammals, several highly conserved hexamer sequence variants exist upstream of the polyadenylation sites ( 44 ), which mediate termination of transcription followed by polyadenylation ( 45 ). In our study, two amplicons (1,114 bp, and 1,156 bp) had a CATAAA poly(A) signal 11 bp upstream of the polyadenylation start site; one amplicon (1,460 bp) had a ACTAAA poly(A) signal 124 bp upstream of the polyadenylation start site; no proper poly(A) signal could be identifi ed for two amplicons (1,606 bp and 1,620 bp); and two amplicons (2,603 bp and 3,750 bp) had requiring FADS2 has been appeared periodically ( 33,53 ), and desaturation has been reported in nuclei ( 54 ). Here we show for the fi rst time molecular evidence that FADS1CS is localized to the mitochondria by live cell imaging and subcellular fractionation using a GFP tagging system and immunoblotting. The presence of N-terminal mitochondrial targeting sequence MAPDPVAAKTPVQGPTPRY-FTWDEVAQ and the cleavage site at amino acid position 28 with a probability score of 0.1624 predicted by MitoProt II ( 55 ) for FADS1CS is consistent with mitochondria localization. Our failure to observe FADS1AT1 in the mitochondria may be because the N-terminal region is deleted due to splicing in FADS1AT1 .
We show here that all three classical FADS transcripts are translated at similar levels in MEF cells, implying that they are essential components during early development. Recently, it has been shown that FADS1 is highly enriched in hatched blastocysts ( 56 ). Moreover, FADS2 and FADS3 were found to be upregulated greater than 2-fold at the implantation sites in mice ( 57 ). Though no biochemical function has been reported for FADS3 , these data further support protein detection studies ( 50 ) showing that FADS3 is a functional gene, not a "cryptic" noncoding gene.
The fi rst observation that FADS1 produces several mRNA isoforms shows that all three genes in the FADS gene cluster do so. The newly identifi ed transcripts are generated by alternative transcription initiation, alternative selection of poly(A) sites, and internal exon deletions resulting from alternative splicing. We also show positive detection of FADS1 protein isoforms, differential tissue-and cell type-specifi c isoform expression patterns, differential isoform expressions during cell differentiation, isoform-specifi c subcellular localization, and translation of FADS transcripts in embryonic cells. A novel sequence encoding a 360 amino acid protein having no cytochrome b5 domain, cells, the 65 kDa isoform was preferentially expressed compared with the 48 kDa and 42 kDa isoforms; this expression pattern was not observed in undifferentiated and differentiated NB cells. The isoform expression pattern appears to be tissue-and cell type-dependent, presumably based on tissue-and cell-specifi c splicing mechanisms. In undifferentiated NB cells, the 54 kDa isoform induction is directly related to an increase in AA, the metabolic product of FADS1CS . This particular isoform may be under specifi c control based on developmental stage.
It is well known that desaturation and elongation steps localize to the ER ( 52 ), but they have been suggested over the years to occur in other organelles as well ( 32 ). Data implicating mitochondrial involvement in 22:6n-3 biosynthesis   9. Translation of all FADS trapped in ribosomes of mouse MEF cells. FADS1 mRNA is translated into protein at similar to levels FADS2-3 . similar to SCD was identifi ed. We also show molecular evidence that the novel FADS1 isoform enhances the activity of FADS2 and increase the production of eicosanoid precursors, the fi rst demonstration of an isoform of one gene enhancing the activity of another. Our results indicate that FADS isoforms have critical roles as mediators of LCPUFA biosynthesis, including enhanced eicosanoid precursor biosynthesis and/or regulation, depending on tissues, organelles, and developmental stage.