Saturated fatty-acids regulate retinoic acid signaling and suppress tumorigenesis by targeting fatty-acid-binding protein 5

Long chain fatty acids (LCFA) serve as energy sources, components of cell membranes, and precursors for signalling molecules. Here we show that these biological compounds also regulate gene expression and that they do so by controlling the transcriptional activities of the retinoic acid (RA)-activated nuclear receptors RAR and PPARβ/δ. The data indicate that these activities of LCFA are mediated by FABP5 which delivers ligands from the cytosol to nuclear PPARβ/δ. Both saturated and unsaturated LCFA (SLCFA, ULCFA) bind to FABP5, thereby displacing RA and diverting it to RAR. However, while SLCFA inhibit, ULCFA activate the FABP5/PPARβ/δ pathway. We show further that, by concomitantly promoting activation of RAR and inhibiting the activation of PPARβ/δ, SLCFA suppress the oncogenic properties of FABP5-expressing carcinoma cells in cultured cells and in vivo. The observations suggest that compounds that inhibit FABP5 may constitute a new class of drugs for therapy of certain types of cancer.


Introduction
Saturated and unsaturated LCFA (SLCFA, ULCFA) share common roles as energy sources and key membrane components, but also display distinct biological activities. Hence, while high concentrations of SLCFA trigger acute endoplasmic reticulum (ER) stress, induce Users may view, print, copy, and download text and data-mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use:http://www.nature.com/authors/editorial_policies/license.html#terms The data thus demonstrate that RA activates both RAR and PPARβ/δ in vivo. Reporter mice were then crossed with FABP5 −/− mice. Ablation of FABP5 enhanced lacZ expression in RARE reporter mice (Fig 1c), and markedly decreased luciferase expression in PPRE-luc reporters (Fig 1d, Supplementary Fig. 1d, 1e). As FABP5 does not deliver ligands to other PPAR isotypes 36 , the effect of its ablation in PPRE-luc mice must have specifically stemmed from alterations in the activation of PPARβ/δ. The data thus show that FABP5 suppresses the activation of RAR and promotes activation of PPARβ/δ in vivo. Importantly, RA failed to upregulate luciferase expression in PPRE-luc mice lacking FABP5 (Fig. 1d, Supplementary Fig. 1f), demonstrating that this protein is critical for RA-induced activation of PPARβ/δ. FABP5 can bind multiple ligands, including RA and LCFAs. The equilibrium dissociation constants (Kd) for the association of FABP5 with the SLCFA palmitate (16:0) and stearate (18:0), and the ULCFA linoleate (18:2) and oleate (18:1) were measured by fluorescence competition titrations 37 using bacterially-expressed recombinant FABP5 ( Supplementary  Fig. 1g). Binding of the fluorescent lipid 1-anilinonaphthalene-8-sulfonic acid (ANS) to the protein was examined by fluorescence titrations (Supplementary Fig. 1h), which yielded a Kd of 70±6.4 nM. The affinities of LCFAs for FABP5 were then assessed by monitoring their ability to displace ANS from the protein (Fig. 1e). Kds for binding of 16:0, 18:0, 18:2, and 18:1 to FABP5 were found to be 20.4±4.2, 15.3±2.4, 19.3±3.3, and 18.5±4.1 nM (data are mean±SD, n=3), respectively, a somewhat stronger affinity than that of RA (42.3±6.4 nM 28 ). Human keratinocyte HaCat cells, which express high levels of FABP5 15 , were used to examine whether FABP5 links cellular responses to its different ligands. Cells were cultured in charcoal-treated medium to deplete them of retinoids and transactivation assays were carried out. Cells were co-transfected either with a vector encoding an RARE-driven luciferase and an expression vector for RARα, or with a PPRE-driven luciferase and an expression vector for PPARβ/δ, treated with LCFA, and luciferase activity was measured. In the absence of RA, neither saturated nor unsaturated LCFA affected the activity of RAR (Fig. 1g, 1i). SLCFA also did not activate PPARβ/δ ( Fig. 1h) but, as previously reported 32,38 , ULCFAs functioned as agonists for this receptor (Fig. 1j, Supplementary Fig.  1i). Strikingly, in the presence of RA, treatment with <10 μM concentrations of all LCFAs modulated the transcriptional activities of both receptors. Both SLCFA and ULCFA activated RAR (Fig. 1g, 1i). PPARβ/δ was inhibited by SLCFA ( Fig. 1h) but activated by ULCFA (Fig. 1j).
A HaCaT cell line in which the expression of FABP5 is stably decreased was then generated (Fig. 1f). Lowering the level of FABP5 abrogated the ability of both 16:0 and 18:2 to activate RAR in the presence of RA (Fig. 1g, 1i). Reducing FABP5 expression decreased the activity of PPARβ/δ even the absence of RA, indicating that cells contain other endogenous PPARβ/δ ligands that rely on FABP5 for their nuclear delivery (Fig. 1h, 1j). Decreasing the expression of FABP5 also diminished the ability of both SLCFA and ULCFA to regulate RA-dependent PPARβ/δ activity (Fig. 1h, 1j).
Modulation of the transcriptional activities of RAR and PPARβ/δ by LCFA was further examined by monitoring their effects on expression of endogenous target genes for these receptors in NaF mammary carcinoma cells. NaF cells are derived from tumors that arise in the MMTV-neu mouse model of breast cancer 30 and they express a high level of FABP5 16 . Similarly to their effects in transcriptional activation assays, both SLCFA and ULCFA induced the expression of the RAR target genes Rarb and Cyp26a (Fig. 2a, 2b, and Supplementary Fig. 2a-2c). Also in accordance with transactivation assays, SLCFA decreased (Fig. 2c, and Supplementary Fig. 2a, 2b), and ULCFA increased (Fig. 2d, Supplementary Fig. 2c) expression of the PPARβ/δ target genes Pdpk1 and Pln2. Cells were then treated with Triacsin C (TriC), an inhibitor of fatty acyl CoA ligase, the enzyme that catalyzes the first step in fatty acid metabolism 39 . TriC elevated the level of total free fatty acids in the cells by about 2 fold (Supplementary Fig. 2d) and augmented the respective activities of both SLCFA and ULCFA ( Fig. 2e-2h, and Supplementary Fig. 2e, 2f). The data thus demonstrate that these effects are exerted by the LCFAs themselves and not by their metabolic products. Notably, TriC upregulated the expression of RAR target genes even in the absence of ectopic administration of LCFA (Fig. 2e, 2f), indicating that RA signalling is controlled by alterations in endogenous LCFA levels. TriC treatment per se did not significantly affect expression of PPARβ/δ targets (Fig. 2g, 2h), likely reflecting that TriC elevates the levels of both SLCFA, which inhibit, and ULCFA which activate PPARβ/δ, resulting in an overall neutral effect.

FABP5 and RA are critical for LCFA function
NaF cells express FABP3 and FABP5 but the latter displays a markedly higher level ( Supplementary Fig. 2g). Decreasing FABP5 expression in NaF cells ( Supplementary Fig.  2h) upregulated the RAR target gene Rarb ( Supplementary Fig. 2i), and suppressed the PPARβ/δ target gene Pdpk1 ( Supplementary Fig. 2j). The pan-RAR antagonist LE540 abolished the ability of 16:0 to induce RAR targets (Supplementary Fig. 3a) but had no effect on the responsiveness of PPARβ/δ target genes ( Supplementary Fig. 3b). These data demonstrate that induction of RAR target genes by LCFA does not stem from an RARindependent function of these compounds. These observations also show that RAR is not involved in modulation of PPARβ/δ activity by 16:0. To examine whether RA is necessary for these effects, cells were depleted of retinoids by culturing in charcoal-treated medium. The depletion decreased the expression of both RAR and PPARβ/δ target genes (Fig. 2i, 2j). 16:0 did not induce the expression of RAR target genes in the absence of retinoids, and the response was restored following replenishment with RA ( Fig. 2i). Unlike the absolute RAdependence of the responsiveness of RAR targets, 16:0 downregulated the expression of PPARβ/δ targets even in the absence of retinoids (Fig. 2j). These observations likely reflect that, in contrast with CRABP2 and RAR which are specifically activated by RA, FABP5 and PPARβ/δ can be activated by other endogenous ligands. Hence, 16:0 displaces all PPARβ/δ ligands from FABP5. RARE-lacZ and PPRE-luc reporter mice were separated into three groups which were fed a regular chow, a diet enriched in 16:0, or a diet enriched with safflower oil in which the predominant fatty acid is 18:2 (SF/18:2). Feeding RARE-lacZ mice with diets enriched with either 16:0 or SF/18:2 markedly enhanced x-gal staining, demonstrating activation of RAR (Fig. 2k, Supplementary Fig. 4a). In PPRE-luc reporter mice, 16:0-enriched diet decreased, and SF/18:2-enriched diet increased PPAR activation (Fig. 2l, Supplementary Fig. 4b).

LCFA differentially modulate cancer cell growth
Considering the pro-proliferative activities of the FABP5/PPARβ/δ path [27][28][29] , the opposing effects of SLCFA and ULCFAs on PPARβ/δ activation suggest that they may differentially modulate cell growth. Indeed, SLCFA suppressed while ULCFA facilitated NaF cell proliferation (Fig. 3a, 3b). Normal human mammary epithelium cells (HMEC) and MCF-7 mammary carcinoma cells, which express low levels of FABP5, NaF and MBA-MD-231 (231) mammary carcinoma cells and PC3M prostate cancer cells, which highly express the binding protein (Fig. 3c) were used to examine the involvement of FABP5 in these opposing effects of the two types of LCFA. 16:0 downregulated the expression of the RAR target gene Rarb and upregulated the PPARβ/δ target Pdpk1 in PC3M and 231 cells but not in MCF-7 cells (Fig. 3d). RA facilitated proliferation of 231 and PC3M cells, but not of HMEC and MCF-7 cells (Fig. 3e). Strikingly, SLCFA potently inhibited the growth of 231 and PC3M cells, but had no effect on proliferation of either HMEC or MCF-7 ( Fig. 3f, Supplementary Fig. 5a). This activity was abolished upon depletion of retinoids and restored upon replenishing depleted media with RA ( Fig. 3g). 18:2 promoted the growth of the FABP5-expressing 231 and PC3M cells but, similarly to SLCFA, had no effect on growth of MCF-7 and HMEC cells (Fig. 3h). Both RA and 18:2 promoted cell proliferation when administered alone, and the activity was additive when these two PPARβ/δ ligands were added together (Fig. 3i). Treatment of cells with the PPARβ/δ antagonist, PT-S58 40 , inhibited cell proliferation with an additive effect observed upon co-treatment with LCFA ( Fig. 3j, 3k). Moreover, despite inhibition of PPARβ/δ in the presence of PT-S58, 18:2 suppressed cell proliferation (Fig. 3k). Considering that PT-S58 inhibits PPARβ/δ 40 but does not associate with FABP5, the activities of LCFA in the presence of the inhibitor likely stemmed from activation of RAR. Taken together, the data indicate that, by binding to FABP5, both SLCFA and ULCFA activate RAR by shifting RA towards this nuclear receptor.
It is well established that, at high concentrations, SLCFA induce apoptosis by triggering ER stress 1,2,6 . Indeed, an 8 h. treatment with >50 μM 16:0 increased the expression of the ER stress markers Chop and Grp78 (Fig. 4k). However, while 16:0 induced apoptosis at <10 μM ( Fig. 4f, 4g), it did not trigger ER stress at these concentrations. In fact, treatment with low concentrations of 16:0 reduced the expression levels of both Chop and Grp78 (Fig. 4l). Induction of apoptosis in NaF cells and in MCF-7 cells ectopically expressing FABP5, in response to ≤ 10 μM concentrations of 16:0 was abolished in the presence of antagonists of RARα, β, or γ (Fig. 4m, Supplementary Fig. 5d). Hence, at low levels, 16:0 induces apoptosis through controlling RA signaling and not via induction of ER stress.

16:0 induces genome-wide regulation of cancer-related genes
NaF cells, cultured in delipidated medium, were treated with the PPARβ/δ-selective agonist GW1516 (GW), the pan-RAR agonist TTNPB, or RA (4 h., 1 μM each). Transcriptome analyses (Affymetrix® Mouse Gene 2.1 ST Arrays) revealed that 1047 and 1474 genes were commonly regulated by RA and GW1516 and RA and TTNPB, respectively (Supplementary Fig. 6a). Notably, more genes were regulated by RA (3960 genes) than by either GW (1979 genes) or TTNPB (2598 genes). Transcriptome analyses were also carried out in cells treated with 16:0 (10 μM, 4h.) in the presence of retinoids. The analysis identified 258 and 446 genes commonly regulated by 16:0 and GW1516 and 16:0 and TTNPB, respectively (Fig. 5a, 5b). Hierarchical clustering identified 69 genes whose expression was up-regulated by GW1516 and downregulated by 16 Validating the transcriptome analysis, Q-PCR showed that the PPARβ/δ target genes Cd47 and Tgfb2 were downregulated, and expression of the RAR target genes Skt3 and Cereblon (Crbn) increased upon treatment with 16:0 ( Fig. 4f). Functional analyses of the 120 genes regulated by 16:0 in common with GW1516 and TTNPB (Ingenuity Pathway Analysis) indicated that the most significantly enriched network is "cellular development, cancer and embryonic development" (p=0.0002). Notably, most genes in this network were downregulated by 16:0 treatment (Fig. 5g, green), implying an anti-oncogenic activity.

16:0 suppresses tumor growth in vivo
Female NCr athymic mice were separated into three groups and fed a regular chow diet, a diet enriched with 16:0, or a SF/18:2-enriched diet. Following a week of feeding, NaF cells were injected into the mammary fat pad and tumor growth was monitored. Consumption of either LCFA-enriched diet elevated serum levels of free fatty acids (Fig 5h). Mice in the three groups consumed similar amounts of food ( Supplementary Fig. 6b), and displayed similar weight gain (Fig. 5i). Total FFA levels in tumors that arose in mice fed either of the LCFA-enriched diets were similarly elevated ( Supplementary Fig. 6c) but level of 16:0 was higher only in tumors of mice fed 16:0-enriched diet (Fig. 6a). Strikingly, tumor development was markedly suppressed in mice fed 16:0-enriched diet vs. in mice fed either regular chow or SF/18:2-enriched diet (Fig 6b). While not statistically significant, tumor development appeared to be facilitated in mice fed SF/18:2-enriched diet. In line with their opposing effects on the activity of PPARβ/δ in cultured cells, tumors that arose in mice fed 16:0-enriched and SF/18:2-enriched diets respectively expressed higher and lower levels of the pro-proliferative PPARβ/δ target genes Pdpk1, Vegfa, Tgfb2 and Cd47 as compared with tumors in chow-fed animals (Fig. 6c, 6e). Tumors in mice fed either 16:0 or SF/18:2 expressed higher levels of the RAR target genes Rarb, Caspase 9, Skt3, and Crbn although the effect of 16:0 was more pronounced than that of SF/18:2 (Fig. 6d, 6e, Supplementary  Fig. 6d). In addition, expression of cyclin D1, which is suppressed by RAR 41 , was markedly lower in tumors of 16:0-fed mice (Fig. 6d). Notably, expression of the proliferation marker Ki67 was lower and apoptosis was more pronounced in tumors of 16:0-fed mice (Fig. 6f). Expression of the ER stress markers Chop and Grp78 were similar in tumors of mice fed the three diets ( Supplementary Fig. 6e), indicating that the growth inhibitory activity of 16:0 did not originate from initiation of ER stress.

Discussion
These studies establish that RA activates both RAR and PPARβ/δ in vivo, and that the iLBP FABP5 is a critical regulator of both receptors. The observations further reveal that, through targeting FABP5, the important dietary components LCFA can regulate gene expression by governing the transcriptional activities of both RA-responsive nuclear receptors. We thus define a novel function for these fundamental biological building blocks. Taken together, the observations suggest the model depicted in Fig. 6g. SLCFA displace RA and other PPARβ/δ ligands from FABP5 and thereby block the delivery of such ligands to this receptor and inhibit its activation. Displacement of RA from FABP5 diverts the hormone to RAR, and as SLCFA do not induce either the nuclear import of FABP5 32 or the transcriptional activity of PPARβ/δ 15 , these LCFA concomitantly activate RAR and inhibit PPARβ/δ. On the other hand, binding of ULCFA to FABP5 similarly reroutes RA to RAR, but these compounds trigger translocation of FABP5 to the nucleus where they are delivered to PPARβ/δ and induce its activation. ULCFA thus activate both RAR and PPARβ/δ. Importantly, in vivo, LCFA shifted RA signalling when provided as dietary components, and, in cultured cells, they exerted such an effect at remarkably low concentrations. Hence, while SLCFA induce lipotoxicity and ER stress at concentrations of ~500 μM 5 , these compounds regulate the transcriptional activities of RAR and PPARβ/δ at <10 μM levels. Moreover, even in the absence of ectopic administration, short-term inhibition of fatty acid metabolism was sufficient to markedly induce expression of RAR target genes. The data thus show that the transcriptional activity of RAR is exquisitely sensitive to small alterations in cellular LCFA concentrations. Notably, while LCFA-induced activation of RAR critically depended on the presence of RA, PPARβ/δ was active and its activity was inhibited by 16:0 even in the absence of RA. These observations likely reflect that cells contain multiple endogenous ligands for PPARβ/δ and reveal that SLCFA suppress the activation of this receptor regardless of the nature of the agonist it uses under particular circumstances.
The ability of LCFA to modulate gene expression by governing RA signalling is expected to have profound and wide-ranging consequences for cell function. One example addressed here is their involvement in regulating carcinoma cell growth. We show that, in FABP5-expressing carcinoma cells, SLCFA inhibit and ULCFA induce cell proliferation. We show further that, in such carcinomas, SLCFA convert RA from a pro-proliferative to a growthinhibitory agent, induce apoptosis, suppress oncogenic properties, and inhibit tumor development in vivo. SLCFA induced the expression of multiple RAR targets and reduce the levels of multiple PPARβ/δ target genes, many of which have a known function in cancer. Taken together with the reports that expression of FABP5 is upregulated and is associated with poor survival in several types of human cancers [22][23][24]26,28 , the data indicate that compounds that inhibit FABP5 may constitute a promising new class of drugs for therapy of certain types of cancer. While SLCFA function as potent FABP5 inhibitors, their rapid metabolism and potentially detrimental activities at pharmacological concentrations preclude their use as efficacious drugs. However, the data presented here provide precise criteria for developing FABP5 inhibitors and a strong proof-of-principle for the efficacy of such compounds in suppressing carcinoma cell growth.

ER stress assay
For treatments with high 16:0 concentrations, the FA was complexed with BSA. BSA (affymetrix) was delipidated using activated charcoal. 16:0 was dissolved in ethanol and added to the BSA solution to create a stock solution of 1mM 16:0. The solution was incubated at 55°C and for 30 min, then filtered and diluted in cell media to establish the desired concentrations.

Mice
PPRE-luc mice (RepTOP ™ PPRE-Luc) 35 were obtained from Charles River laboratories. RARE-lacZ mice, which globally express β-galactosidase (lacZ) gene under the control of the RA responsive element (RARE) of the RA target gene Rarb 34

Mouse experiments
RARE-lacZ and PPRE-luc males and females (4-6 weeks old, n=3 each) were treated with RA (1 mg or 4 mg). PPRE-luc mice (4-6 weeks old, n=3) were used for GW (4 mg) treatments. Ligand solutions were prepared immediately before use in a vehicle of sterile saline:polyethylene glycol:Tween 80 (80:10:10 v/v) and injected intraperitoneal. For AGN193109 treatments, total of 6 RARE-lacZ males were used (3 mice treated and 3 controls). AGN193109 solution was prepared to a final concentration of 1 mg kg −1 and administered by intraperitoneal injections. Mice were treated with AGN193109 and then coinjected with RA the following day.
Nine-week-old NCr nude females were fed ad libitum and the amount of food consumed was recorded once a week. One week after being set on the diet, 2 × 10 6 NaF cells were injected subcutaneously into the mammary fat pad of the mice. Tumor size was assessed twice per week using a digital caliper. Tumor volumes were determined by measuring the length (l) and the width (w) of the tumor and calculating the volume (V = lw 2 /2). Mice were scarified 23 days after injection. Statistical significance between the control and treated mice in both experiments was evaluated using a Student's t test. Mouse experiments were conducted after approval by the institutional animal care and use committee at Case Western Reserve University. Serum and tumors total free fatty acid concentrations were measured using the Serum/Plasma Fatty Acid Detection Kit (Zen-Bio).

X-gal staining
RARE-LacZ mice were euthanized and organs were harvested and fixed in 4% paraformaldehyde (1 h.). Organs were washed 3 times with phosphate buffered saline (PBS), transferred into X-gal staining solution (20 mM K3Fe(CN)6, 20mM K4Fe(CN)6.3H2O, 2 mM MgCl2, 1 mg/ml X-gal substrate) and incubated over night at 37°C. Organs were washed with PBS and imaged. In addition, fixed organs were embedded in paraffin, sectioned (8 μm), mounted on glass slides and counterstained with nuclear fast red prior to imaging.

In vivo Imaging
PPRE-luc mice were injected intraperitoneally with luciferin (20 mg/ml, 200 μl) and imaged 5 min. later by using IVIS 200 CCD camera (Xenogen, CA, USA). Immediately after, mice were euthanized and organs were harvested and imaged. Data acquisition and quantification were done with the software Living Image (Xenogen).

MTT proliferation assays
1000 cells were plated in quadruplicates in a 96-wells plate and treated with ligand for 4 days. Cells were incubated with MTT reagent (5 mg/ml in PBS) until the formation of formazan crystals. Crystals were dissolved in 4 mM HCl in isopropanol, and absorbance at 590 nm was measured using a microplate reader.

COS-7 or
HaCaT cells were cultured in 6-well plates and co-transfected with either a luciferase reporter driven by 3 copies of a PPRE and expression vector for PPARβ/δ, or a luciferase reporter driven by RARE and expression vector for RARα, together with a vector harboring cDNA for β-galactosidase, serving as a transfection control. 18 h. posttransfection, cells were placed in a serum-free medium and treated with ligand. 18 h. later, cells were lysed, luciferase activity was assayed (Promega, WI, USA) and corrected for transfection efficiency by the activity of β-galactosidase.

Transcriptome analyses
Two experiments were done: 1) NaF cells were cultured in media supplemented with 10% charcoal-treated FBS for 48 h. and then treated with vehicle, GW (1 μM), RA (1 μM) or TTNPB (1 μM) for 6 h. 2) NaF cells were cultured in media supplemented with 10% FBS then treated with vehicle or 16:0 (10 μM) for 6 h. Total RNA was extracted by QIAzol (Qiagene) and purified using RNeasy columns (Qiagen). Samples were amplified, labeled, and hybridized on Affymetrix® Mouse Gene 2.1 ST Arrays (Affymetrix, USA) by the Gene Expression & Genotyping Facility of the Case Comprehensive Cancer Center of Case Western Reserve University. Raw data files were analyzed using Partek-Genomics-Suite (PGS) v6.6 software. Data was normalized using Robust Multichip Average Method (RMA) which allows reduction of block effect done at the probe-set level. t-test analysis was used to select differentially expressed genes with Fold Change and P-value Cutoffs respectively fixed to at 1.2 and 0.05. Each treatment was compared to its relevant control. The network of genes with known functions in cancer was identified using IPA (Ingenuity Systems) (P=2.2×10 −4 ).

Colony Formation Assays
A layer of 0.8% agarose in cell media was cast in a 6-well plate and set in room temperature to solidify. Cells were suspended in 0.25% agarose in media and 1 ml from this mixture containing 5000 cells was added to each well. Cells were cultured for 21 days. Media were replenished every 3 days. Colonies were visualized by staining with 0.005% crystal violet and counted under a light microscope.

Palmitic acid analysis
Palmitic acid was quantified using stable isotope dilution liquid chromatography with online tandem mass spectrometry using a modification of a method previously described (1). For serum samples, palmitic acid-7,7,8,8-d 4 internal standard (DLM-2893, Cambridge Isotope Laboratories, Inc.) was added to serum prior to acidification with acetic acid, and the palmitic acid was extracted into hexane 44 . For tumor tissues, samples were homogenized in methanol, and to an aliquot, palmitic acid-7,7,8,8-d 4 added as internal standard. Samples were then acidified and extracted with hexane as with serum samples. For all analyses hexane extracts were dried under N 2 , resuspended in methanol, and injected onto a reverse phase C18 HPLC column (2.0 × 150 mm, 5 μm, Phenomenx, Torrance, CA) operated at a flow rate of 0.2 ml/min and resolved using a linear gradient between 0.2% formic acid in water and 0.2% formic acid in acetonitrile. HPLC column effluent was introduced into an AB Sciex API 5000 triple quadrupole mass spectrometer using electrospray ionization in negative-ion mode. Analytes were monitored using multiple reaction monitoring of parent and characteristic daughter ions: m/z 255 → 237 for palmitic acid; and m/z 259 → 241 for palmitic acid-7,7,8,8-d 4 .

Binding Assays
Assays were carried out by fluorescence titrations. FABP5 was bacterially expressed and purified and the equilibrium dissociation constants (Kd) that characterize its interactions with different FAs were measured by fluorescence competition assays. The method entails two steps 37 . In the first step, Kd for the association of the protein with the fluorescent fatty acid probe ANS was measured. Protein (1 μM) was titrated with ANS from a concentrated solution in ethanol. Ligand binding was monitored by following the increase in the fluorescence of the ligand upon binding to the protein, and Kd for the association of ANS with FABP5 was computed from titration curves 45 . Kds for binding of non-fluorescent ligands were then measured by monitoring their ability to compete with ANS for binding to the protein. FABP5 was pre-complexed with ANS at 1:1 molar ratio and titrated with the different FAs whose binding was reflected by a decrease in probe fluorescence. Kds were extracted from the EC50 of the competition curve and the measured Kd for ANS. Analyses were carried out using Origin 8 software (MicroCal Software Inc., Northampton, Mass.).

Statistics
Statistical significance was analyzed using an independent sample t-test vs. respective untreated controls.      2×10 6 ) were transplanted into the mammary fat pad of NCr athymic female mice fed denoted diets and tumor growth was monitored. Mean±SD (n=6). **p<0.01 by unpaired t-test vs. chow control. c), d) mRNAs for PPARβ/δ (c) or RAR target genes (d) in mice fed denoted diets. Mean±SD (n=6). *p<0.05, **p<0.01 calculated by unpaired t-test. e) Immunoblots demonstrating expression of PDPK1 and RARβ in tumors of mice fed denoted diets. Blots are representative data from 3 individual mice per group. f) Immunohistochemistry analyses demonstrating expression of Ki67 and cleaved caspase 3 in tumors of mice fed denoted diets. Representative images out of three tumors per group are shown. Magnification 400x, scale bars represent 50 μm. g) A model illustrating the mechanism by which SLCFA and ULCFA regulate RA signaling. Left: SLCFA compete with RA and other PPARβ/δ ligands for binding FABP5, divert RA to RAR and inhibit PPARβ/δ activation. Right: ULCFA bind to FABP5 and divert RA to RAR. ULCFA are delivered to PPARβ/δ by FABP5 and activate it.