The polypyrimidine tract binding protein regulates desaturase alternative splicing and PUFA composition.

The Δ6 desaturase, encoded by FADS2, plays a crucial role in omega-3 and omega-6 fatty acid synthesis. These fatty acids are essential components of the central nervous system, and they act as precursors for eicosanoid signaling molecules and as direct modulators of gene expression. The polypyrimidine tract binding protein (PTB or hnRNP I) is a splicing factor that regulates alternative pre-mRNA splicing. Here, PTB is shown to bind an exonic splicing silencer element and repress alternative splicing of FADS2 into FADS2 AT1. PTB and FADS2AT1 were inversely correlated in neonatal baboon tissues, implicating PTB as a major regulator of tissue-specific FADS2 splicing. In HepG2 cells, PTB knockdown modulated alternative splicing of FADS2, as well as FADS3, a putative desaturase of unknown function. Omega-3 fatty acids decreased by nearly one half relative to omega-6 fatty acids in PTB knockdown cells compared with controls, with a particularly strong decrease in eicosapentaenoic acid (EPA) concentration and its ratio to arachidonic acid (ARA). This is a rare demonstration of a mechanism specifically altering the cellular omega-3 to omega-6 fatty acid ratio without any change in diet/media. These findings reveal a novel role for PTB, regulating availability of membrane components and eicosanoid precursors for cell signaling.

pathways, but to our knowledge, PTB has never been directly linked with lipid metabolism.
Preliminary bioinformatics predictions suggested that PTB might function in regulation of FADS2 alternative splicing. To investigate this, we examined binding of PTB in vitro to a putative exonic silencer site, effects of PTB knockdown with small interfering RNA (siRNA), and expression of PTB relative to FADS alternative transcripts in baboon tissues. Finally, we demonstrated changes in the ratio of omega-3 to omega-6 fatty acids with PTB knockdown in liver-derived cells.

Cell culture
For all experiments, human HepG2 hepatocellular carcinoma and SK-N-SH neuroblastoma cells were maintained within 10 passages of the original passage received from the ATCC. HepG2 cells were grown in MEM with 10% FBS, and SK-N-SH cells were grown in DMEM/F-12 with 10% FBS (media and serum obtained from HyClone) in a humidifi ed environment at 37°C with 5% CO 2 .

In vitro RNA binding assay
HepG2 nuclear extracts were prepared from cell lysates using the Qproteome Nuclear Protein Kit (Qiagen). Protein concentration was determined using a BCA Protein Assay Kit (Pierce). Biotin-labeled RNA oligos (Integrated DNA Technologies) were suspended in nuclease-free PBS. The sequences used were: FADS2wt GAUUAUGGCCACCUGUCUGUCUACAGAAAA, and FADS2mut GAUUAUGGACAGAGAUAGACGGACAGAAAA. PTB binding to wild-type (wt) but not to the mutant version was predicted using the online tool Splicing Rainbow (EMBL-EBI Alternative Splicing Workbench, <http://www.ebi.ac.uk/asd-srv/wb.cgi>) ( 24 ). For each binding reaction, 1 mg of magnetic streptavidinlabeled M-280 Dynabeads (Invitrogen) were fi rst prewashed to remove nucleases, following the manufacturer's recommended protocol. Beads were mixed with 27 g of biotin-RNA and 15 g of nuclear extract in PBS for 30 min with gentle rotation at 4°C. Beads were washed four times with PBS, then bound proteins were eluted with the addition of Laemmli sample buffer and boiling for 5 min. Eluted proteins were loaded onto a Mini-Protean precast Any KD SDS-PAGE gel (Bio-Rad) along with 15 g nuclear extract as a positive control. After separation, proteins were transferred to a nitrocellulose membrane (Li-Cor). The membrane was probed with antibodies at the following dilutions: mouse SH54 anti-PTB (EMD), 1:100; rabbit anti-␤ -actin (Li-Cor) as a negative control, 1:1000; goat anti-mouse 680LT (Li-Cor), 1:10,000; goat anti-rabbit 800CW, 1:15,000. The blot was visualized on a Li-Cor Odyssey two-color fl uorescence imager. siRNA PTB siRNA (100 nM) and control nontargeting siRNA (Dharmacon siGENOME SMARTpool) were used in triplicate treatments with Dharmafect 4 (HepG2) or Dharmafect 1 (SK-N-SH), following the manufacturer's protocol. SMARTpool reagents contained four different siRNA, allowing lower concentrations of individual siRNA to be used to prevent off-target effects. In addition, a combination of sense and antisense strand chemical modifi cations that reduce off-target effects were incorporated when needed ( 25 ). Cells were treated for 72 h before RNA or lipid extraction. For fatty acid treatment, omega-3 docosapentaenoic acid (22:5n-3) was fi rst noncovalently bound to fatty-acid free BSA (BSA). Fatty acid sodium salts were suspended in PBS, mixed in mouse uterus at the site of embryo implantation (15)(16)(17). Although the genetic evidence suggests that FADS3 is important for lipid metabolism, mechanisms by which genetic variants in FADS3 affect blood lipids and disease have not been identifi ed.
Here, we investigated regulation of fatty acid desaturase alternative pre-mRNA splicing. Two major classes of splicing factors are known to function in opposing ways to regulate splicing: the serine-arginine (SR) proteins, which typically enhance splicing, and the heteronuclear riboprotein (hnRNP) family, known for splicing repression ( 18 ). Polypyrimidine tract binding protein (PTB, also known as PTBP1 or hnRNP I) is a member of the hnRNP family that usually acts as a repressor but also sometimes enhances splicing, depending on the location it binds pre-mRNA and proximity to binding sites for other splicing factors ( 19 ). The majority of PTB protein is normally found in the nucleus, but under some circumstances, it can relocalize to the cytoplasm, where it affects mRNA stability and translational effi ciency ( 20 ). Although sometimes described as ubiquitous because it is expressed by most or all cell types at some point during development, PTB protein levels vary in different tissues, and control of function by phosphorylation allows for dynamic temporal and spatial adjustment of PTB activity.
PTB is known to be a primary regulator of alternative splicing for several genes ( 20 ), and it presumably affects numerous others. However, relatively little is known about high-level functional changes resulting from PTB activity. PTB appears to promote growth in some cancer cell types ( 21 ), and it binds pyrimidine tracts in several growth-promoting genes, encouraging transcription ( 22 ). Cytoplasmic PTB has been shown to enhance production of both insulin and insulin secretory granules by stabilizing mRNA, leading to increased translation ( 23 ). These functions are consistent with switching from lipid utilization to glucose as an energy source and enhancing fatty acid biosynthetic Fig. 1. FADS2 codes for a desaturase that adds cis double bonds to the growing fatty acid chains in omega-3 (n-3) and omega-6 (n-6) long chain polyunsaturated fatty acid synthesis. Nomenclature refers to the number of carbons and double bonds (e.g., 20:5n-3 has 20 carbons and 5 double bonds, with n-3 describing distance of the double bonds from the terminal methyl carbon of the fatty acid). The end products of the biosynthetic pathway give rise to eicosanoids and docosanoids with opposing effects. sites identifi ed from our previous work were examined for predicted binding sites. Numerous splicing enhancer binding sites were predicted, but there was also a striking pattern of predicted PTB binding directly on splice sites identified from baboon cDNA for FADS2 AT1 (NCBI accession FJ901343) and FADS3 AT7 (NCBI accession FJ641203). The alternative splice site in FADS2 exon 4 occurred at base 7 of the 14-mer predicted PTB binding site. FADS3 exon 8 included a 35 nt stretch predicted to bind PTB, with the alternative splice site located at nucleotide 19 of the 35-mer. Replacing the sequences with human cDNA sequences yielded identical binding predictions. In addition, FADS3 AT1 (NCBI accession EU780004) exhibited a predicted PTB binding site near the splice site but not overlapping with it.

PTB is associated with the FADS2 alternative splice site in vitro
To test the validity of the predicted binding sites, we carried out an in vitro RNA pull-down assay to determine whether the putative exonic splicing silencer (ESS) in FADS2 could pull down PTB from nuclear extracts. Biotinylated RNA oligos were designed containing either the intact 14-mer human predicted binding site or a mutated binding site of equal length, with pyrimidines in the binding site replaced by purines in the mutant but retaining the same fl anking sequences (oligo sequences shown in Materials and Methods ). The mutant sequence was checked with Splicing Rainbow to ensure that no consensus binding sites for other splicing factors were introduced in the process of altering the predicted PTB binding site. Each oligo was incubated with nuclear extract, and RNA-bound proteins were eluted and detected by Western blotting. Whole nuclear extract was run in parallel as a positive control, and the blot was probed for ␤ -actin as a negative control to detect nonspecifi c protein binding. As shown in Fig. 2 , PTB bound preferentially to the predicted ESS rather than to the mutant version.

PTB knockdown upregulates FADS2 AT1 and FADS3 AT7
Because PTB has previously been shown to repress splicing by sterically hindering binding of splicing enhancers ( 20 ), we hypothesized that PTB might function as a repressor for FADS2 AT1 and FADS3 AT7 expression. To investigate this hypothesis, siRNA against PTB was used to knock down PTB in human neuronal (SK-N-SH) and liverderived (HepG2) cells. Expression of the alternative transcripts was examined by RT-PCR using primers designed with fatty-acid free BSA in a 3:1 ratio of fatty acid to albumin, and incubated for 5 h at 37°C. RNA interference (RNAi) was carried out as usual, with the addition of 25 M BSA-bound 22:5n-3 or vehicle (BSA only) in the media for the entire 72 h incubation.

RNA extraction and PCR
Banked neonate baboon tissues were obtained in a previous study from our laboratory ( 26 ) and stored from time of necropsy at Ϫ 80°C until RNA was extracted. All tissues were from a 12-week-old control animal that had been fed an infant formula containing no long-chain PUFA. Live baboon work was carried out at the Southwest Foundation for Biomedical Research (SFBR) in San Antonio, TX. Animal protocols were approved by the SFBR and Cornell University Institutional Animal Care and Use Committee (IACUC, protocol #02-105.) Baboon tissue and cell culture RNA was extracted, RNA integrity assessed, cDNA prepared, and semiquantitative RT-PCR carried out as described previously ( 14 ). PCR primers designed for FADS splice variants (Integrated DNA Technologies, sequences available upon request) were validated by cloning and sequencing PCR products. Gel bands were analyzed by densitometry using ImageJ software (National Institutes of Health). Quantitative real-time PCR was carried out using SYBR Green Master Mix (Roche) on a LightCycler 480 instrument (Roche), with ␤ -actin chosen from a panel of candidate reference genes because it was not affected by cell treatments. PCR reaction effi ciency was calculated from standard curves, and reactions were assessed by both melting curves and by running on agarose gels to verify reaction products and the absence of primer-dimers. Quantitative cycle (Cq) values were determined using LightCycler 480 SW1.5.0SP3 software, version 1.5.0.39 (Roche). Analysis was carried out with REST 2009 software (Qiagen), which employs the Pfaffl method for relative quantifi cation ( 27 ) and uses bootstrapping and randomization techniques to calculate fold changes and statistical signifi cance.

Fatty acid analysis
Cells were washed with HBSS and then trypsinized to remove from growth surfaces. After centrifugation and removal of supernatant, lipid extraction was carried out on the cell pellets. Fatty acid methyl esters (FAME) were prepared using a modifi cation of the one-step method of Garces and Mancha ( 28 ). FAME were analyzed in triplicate and quantifi ed by gas chromatographyfl ame ionization detection (GC-FID), using an equal weight FAME standard mixture to verify response factors daily ( 29 ). Peak identities were confi rmed by GC-covalent adduct chemical ionization tandem mass spectrometry (GC-CACI-MS/MS) (30)(31)(32). The standard deviation for ratios (fold change in treated/control) was calculated using a propagation-of-error approach.

PTB is predicted to bind FADS2 and FADS3 exonic splice sites
To identify putative binding sites for splicing factors that might regulate alternative splicing of FADS genes, we investigated alternative exonic splice sites in FADS2 (NCBI accession EU780003) and FADS3 (NCBI accession EU780002) using the bioinformatics tool Splicing Rainbow (EMBL-EBI Alternative Splicing Workbench, <http:// www.ebi.ac.uk/asd-srv/wb.cgi>) ( 24 ). The alternative transcripts (AT) and in vivo expression patterns of FADS2 and FADS3 splice variants have been described in detail previously ( 12,14 ). Approximately 30 nt regions fl anking splice mRNA levels ( Fig. 4A , R 2 = 0.87), suggesting that PTB may play an important role in regulating FADS2 tissue-specifi c splicing in vivo. The 3 tissues with the lowest levels of PTB did not fi t the pattern; these results are consistent with predominant transcription-level control in these tissues, as levels of FADS2 AT1 corresponded closely with FADS2 CS levels in tissues with relatively low PTB ( Fig. 4B ). Unlike FADS2 AT1 , there was much lower correlation between PTB levels and FADS3 AT7 or FADS3 AT1 in baboon tissues ( Fig. 5 ). There was a slight inverse correlation of PTB and FADS3 AT7 ( Fig. 5A , R 2 = 0.44) in the same 8 tissues examined above, whereas FADS3 AT1 showed very little correlation with PTB ( Fig. 5B , R 2 = 0.26).

PTB knockdown specifi cally decreases omega-3 fatty acid content
To determine functional consequences of upregulation of FADS2 AT1 and FADS3 AT7 , the fatty acid composition of HepG2 cells was analyzed after PTB knockdown or control siRNA treatment ( Fig. 6 , gray bars). Both omega-3 and omega-6 fatty acids decreased signifi cantly after PTB knockdown, offset by a slight increase in monounsaturates and no change in saturated fatty acids. Interestingly, in all cases, omega-3 fatty acids declined more than omega-6 to bridge sequences unique to each splice variant. In a preliminary time course experiment, FADS2 AT1 levels rose with time, concomitant with reductions in PTB mRNA levels in SK-N-SH cells ( Fig. 3A ). Changes in expression were investigated in more detail in HepG2 cells with quantitative real-time PCR for transcripts in the FADS gene cluster ( Fig. 3B ). With > 90% knockdown of PTB , FADS2 AT1 expression increased 1.48-fold, whereas FADS3 AT7 increased 1.39-fold, consistent with splicing repression by PTB. Interestingly, FADS3 AT1 relative mRNA abundance decreased slightly, suggesting PTB might play a role in stabilizing splicing of this transcript. There were no signifi cant changes in FADS1 , FADS2 CS (classically spliced) , or any other FADS3 transcripts (data not shown). The concentration of FADS2 AT1 is about 10-fold lower than FADS2 CS in HepG2 cells (estimated from Cq values for identical cDNA dilutions), so even a 50% increase in FADS2 AT1 levels would decrease FADS2 CS by only 5%, well below the detection limit for real-time PCR; hence, no reduction in observed FADS2 CS is expected. Thus, in this liver cell model, PTB knockdown was suffi cient to regulate splicing of FADS2 AT1 and FADS3 AT7 transcripts.

PTB is inversely correlated with FADS2 AT1 in baboon tissues
To understand the role of PTB in regulating tissue-specifi c splicing of FADS genes in vivo, we examined levels of FADS2 AT1, FADS3 AT7 , FADS3 AT1 , and PTB mRNA in neonatal baboon tissues. Of 11 tissues evaluated, 8 followed an inverse correlation of FADS2 AT1 with PTB  6n-3, and the elongation product 24:6n-3 were still significantly reduced with PTB knockdown, although the differences from control cells tended to be less obvious than without supplementation. For example, 22:6n-3 concentration declined 21% in PTB knockdown compared with control ( P = 0.001) in unsupplemented cells, whereas the decrease was only 13% ( P = 0.002) in cells supplemented with 22:5n-3. Thus, the reduction in total omega-3 levels with PTB knockdown cannot be explained entirely by reduction in EPA, as 22:5n-3 supplementation failed to completely rescue levels of downstream desaturation products.

DISCUSSION
Here, we have demonstrated that PTB plays a role in repressing expression of FADS splice variants and altering cellular omega-3 to omega-6 fatty acid ratio. Knockdown of PTB in liver-derived cells led to upregulation of both FADS2 AT1 and FADS3 AT7, and an RNA pull-down assay showed that PTB was associated with the putative binding site in FADS2 in vitro . Further studies are needed to determine whether PTB binds directly to the predicted binding site or whether other proteins are necessary for PTB binding and splicing repression. Our RNAi results showing roughly equivalent induction of FADS2 AT1 in liver and neuronal-derived cells, despite very different levels of PTB silencing, corroborate the idea of additional, tissue-specifi c factors contributing to splicing regulation in different cell types. However, the fact that the putative PTB binding fatty acids (23% decrease overall compared with 13%, respectively; P < 0.01 for the difference). There was a 12% reduction ( P = 0.005) in the ratio of total omega-3 to omega-6 fatty acids with PTB knockdown.
The most dramatic change with PTB knockdown was a 50% reduction in eicosapentaenoic acid (EPA, 20:5n-3) content, coinciding with upregulation of FADS2 AT1 and FADS3 AT7 . Although both EPA and arachidonic acid (ARA, 20:4n-6) levels declined with PTB knockdown, EPA levels went down much more than ARA levels; the ratio of EPA to ARA was reduced by 43% ( P = 0.0002) with PTB knockdown.
The EPA precursors 20:4n-3 and 18:3n-3, normally at very low levels in most cells, were below quantifi able limits, and thus the substrates were not accumulating as EPA product decreased. However, the ratio of 20:3n-6 to its precursor 18:3n-6 decreased 17% with PTB knockdown ( P = 0.0005), consistent with inhibition of the elongation step from 18-to 20-carbon polyunsaturated fatty acids.
To determine how specifi c the inhibitory activity was for EPA synthesis compared with longer chain omega-3 fatty acids, we asked whether adding 22:5n-3, an elongation product of EPA, would rescue levels of the desaturation product 22:6n-3 (docosahexaenoic acid, DHA). To investigate this, PTB knockdown was repeated in HepG2 cells in the presence of media supplemented with 22:5n-3 ( Fig. 6 , black bars). As before, EPA levels were sharply reduced with PTB knockdown, as were the ratio of EPA to ARA and the ratio of total omega-3 to omega-6 fatty acids. Despite  omega-3 fatty acid substrates and remove them from availability for enzymatic reactions, leading to a greater reduction in omega-3 than in omega-6 content. Our results provide only circumstantial evidence linking desaturase alternative splicing with fatty acid changes. However, this hypothesized role for FADS splice variants is appealing because dominant negative inhibition is a common role for alternative splice variants; splicing can generate variants that are able to bind substrates but do not contain domains necessary for enzymatic reactions ( 41 ). FADS2 AT1 lacks a conserved cytochrome b5 domain and the fi rst histidine motif, and FADS3 AT7 lacks the last histidine motif characteristic of front-end desaturases ( 12 ). Binding nonproductively to substrates reduces the effective concentration of substrate available for reactions and, thus, limits product formation in biosynthetic pathways. Sequestration of fatty acids would affect both elongation and desaturation reactions, while also disrupting processes such as fatty acid incorporation into phospholipids. The resulting change in omega-3 to omega-6 ratio alters cell membrane structure and is known to infl uence the function of rhodopsin, a G-protein coupled receptor ( 42 ), as well as change physiological responses such as infl ammation through the production of lipid mediators with opposing functions ( 5 ).
The desaturase splicing pattern resulting from PTB knockdown was associated with a particularly strong decrease in EPA concentration and the ratio of EPA to ARA. Although EPA is present as only a small percentage of total fatty acids, it is a precursor to several families of bioactive eicosanoids, including prostaglandins, thromboxanes, and leukotrienes. Products of EPA tend to act in direct opposition to eicosanoids derived from ARA; EPA-derived eicosanoids have anti-inflammatory, antiarrhythmia, anticoagulant, and provasodilation effects ( 5 ). Because levels of ARA in tissues tend to be tightly regulated ( 43 ), specifi c modulation of EPA may be an effective way to adjust ARA net activity without a large change in ARA concentration. Moreover, because eicosanoids are paracrine species that function with high potency but very short half-life ( 44 ), a mechanism to specifi cally modulate EPA levels in different cell types would allow for fl exible localized responses.
The ratio of 20:3n-6 to its precursor 18:3n-6 decreased signifi cantly with PTB knockdown, suggesting that the elongation step from 18-to 20-carbon polyunsaturated fatty acids is inhibited. Sequestration of 18-carbon fatty acids by dominant negative desaturase splice variants, with a preference for omega-3 over omega-6, may be expected to decrease EPA levels by effectively suspending metabolic transformations, including acylation and recycling pathways. This, in turn, would decrease levels of all longer chain omega-3 fatty acids in the biosynthetic pathway, consistent with our observations. However, the failure of 22:5n-3 supplementation to rescue levels of longer chain fatty acids in the pathway suggests binding affi nity for both 18-carbon and >20-carbon fatty acids, as is observed in functional FADS2CS ( 45 ). Long chain omega-3 products such as 22:6n-3 may also be important in mediating PTB's high-level functions. sites overlap with the alternative splice sites in FADS2 and FADS3 strongly suggests PTB binding sterically hinders access of splicing machinery proteins to the splice sites. Moreover, the high correlation of the expression pattern of FADS2 and PTB in baboon tissues suggested that PTB alone may be an important regulator of FADS2 splicing in vivo. FADS3 splicing was not as closely correlated with PTB mRNA levels in baboon tissues. It is possible that posttranscriptional regulation of PTB is more important than mRNA level in determining splicing repression of FADS3 . Likely there are other important regulators of FADS3 splicing that vary by tissue type and interact with PTB to determine the fi nal splicing pattern. Alternatively, the paralog PTBP2 may replace PTB in repressing FADS3 splicing in some tissues, as PTBP2 has some functional redundancy with PTB ( 33 ). However, the increase in FADS3 AT7 levels in HepG2 cells in response to PTB knockdown may refl ect liver-specifi c regulation, where PTB alone may play a major role. The 40-50% changes in expression we observed in vitro were relatively subtle compared with changes often observed in disease states, but similar fold changes have routinely produced important biochemical changes in studies of nutritional effects on metabolic genes ( 34-37 ). We did not examine fatty acid composition in most of these tissues. Numerous enzymes control fatty acid incorporation and turnover in cell membranes, and many of these have tissue-specifi c expression or activity. We speculate that PTB alters fatty acid composition as a response to extracellular stimuli, which may produce relatively transient changes in eicosanoid precursor availability by modifying PTB activity (for example, by phosphorylation) and changing desaturase splicing. Testing this idea requires future studies.
Knockdown of PTB in liver-derived cells also resulted in decreased cellular omega-3 and omega-6 fatty acid content. Omega-3 fatty acids decreased by nearly one half relative to omega-6 fatty acids in PTB knockdown cells compared with controls. These results were surprising because the same enzymes operate on both omega-3 and omega-6 fatty acids ( Fig. 1 ), so it has been accepted that synthesis of both occurs largely in concert. Intervention studies to raise the proportion of omega-3 relative to omega-6 fatty acids in cell membranes have always focused on dietary modifi cations, as it was presumed that consuming altered ratios of substrates or end products was the only way to change cellular ratios of omega-3 to omega-6 fatty acids. However, puzzling biochemical evidence has occasionally emerged suggesting the possibility of separate ⌬ 6 desaturase enzymes for different fatty acids ( 38,39 ). The idea of separate enzymes lost favor after the discovery of a single FADS2 gene, which was found to code for a desaturase with dual substrate activity toward both omega-3 and omega-6 fatty acids ( 40 ). Because classical enzymes in the biosynthetic pathway accept both classes of fatty acids as substrates, it is diffi cult to explain how the omega-3 to omega-6 fatty acid ratio could change without any alteration in media components.
A reasonable alternative is that desaturase splice variants may be able to preferentially bind and sequester Fatty acid composition is known to change in many tissues and cells during development, as well as during disease states and nutrient defi ciency conditions. However, the results reported here represent a rare demonstration of a mechanism specifi cally altering the cellular omega-3 to omega-6 fatty acid ratio in well-nourished cells without any change in diet/media, and they are the fi rst associated with a splicing factor. Moreover, the sensitivity of PTB to extracellular stimuli and developmental stages provides a means to fi ne-tune cellular desaturase activity and the ratio of EPA to ARA as needed for different tissue types or physiological states. These fi ndings reveal a novel role for PTB as a modulator of the availability of key membrane components and eicosanoid precursors, with the potential to propagate extracellular signals to affect the circulatory system and infl ammatory responses.