The FABP12/PPARγ pathway promotes metastatic transformation by inducing epithelial‐to‐mesenchymal transition and lipid‐derived energy production in prostate cancer cells

FABP12 promotes fatty acid (FA) uptake and facilitates FA trafficking to lipid droplets (storage), mitochondria (beta‐oxidation), and the nucleus (PPARγ activation). In turn, FABP12‐induced PPARγ activation regulates expression of mitochondrial FA metabolic enzymes (for energy production) and Slug (for EMT), thereby inducing metastatic transformation.


Introduction
Prostate cancer (PCa) is the most prevalent cancer in men. While localized PCa can usually be cured, metastatic disease is often resistant to treatment, with eventual relapse and death [1]. It is therefore imperative that we understand the underlying causes of PCa metastasis and identify the key molecules that drive PCa metastasis. Three features limit the clinical management of PCa: cancer heterogeneity, lack of molecular signatures to stratify tumor subtypes [2], and reliance of PCa cells on fatty acid oxidation for their energy supply rather than glucose consumption [3].
Lipid metabolic reprogramming is regarded as a hallmark of cancer progression, and it is broadly accepted that cancer cells have a markedly increased need for lipids [4,5]. This is particularly true for PCa [6][7][8][9]. Epidemiological studies reveal a positive relationship between the consumption of dietary fats, lipid metabolism, and development of PCa [7,10,11]. It is believed that aberrant metabolic adaptations, such as enhanced aerobic glycolysis and increased lipid utilization, are crucial for cancer cells to separate from the primary tumor mass, invade the surrounding stroma, overcome nutrient and energy deficits, and metastasize to secondary sites [12]. Recent studies suggest emerging roles for dysregulated lipid accumulation and metabolism in metastatic cancers [13,14], especially in PCa [15,16]. Lipids or fatty acids not only provide energy for cancer cell growth and dissemination [3,17,18], but also serve as cell membrane components that exert profound effects on signal transduction and cell growth properties [19]. Furthermore, fatty acids function as activating ligands for nuclear receptors such as peroxisome proliferator-activated receptors (PPARs), which regulate gene pathways implicated in PCa lipid homeostasis, tumorigenesis, and cancer progression [19][20][21].
Because of their hydrophobic nature, fatty acids require chaperones such as fatty acid-binding proteins (FABPs) for their transport within the cell [22]. FABPs are important determinants of intracellular accumulation, distribution, utilization, and fate of fatty acids [23,24]. There are ten FABP genes in mammals, with each FABP displaying distinct tissue distribution patterns and ligand preference [22,25]. FABPs are receiving increasing attention in the field of oncology because of their demonstrated roles in cancer progression, and proposed roles in the prevention and treatment of cancer [26][27][28][29], particularly as related to PPAR function [30,31]. While a number of studies point to FABP involvement in PCa metastasis [32][33][34], the precise mechanism underlying FABP action remains elusive. Particularly, the roles of FABPs in epithelial-to-mesenchymal transition (EMT), a priming process leading to metastasis of epithelium-derived carcinomas [35], and dysregulation of energy production, a critical event promoting metastasis in various cancer types [13,14], have yet to be determined in PCa.

PCa xenograft mouse model and immunohistochemistry
Three million cells in 100 lL PBS were injected into the right flank of NSG mice. Mice were euthanized 8 weeks postinjection. Ipsilateral and contralateral axillary lymph nodes, and sites of metastasis previously reported for PC3 cells [42] were examined. Tumors were removed and fixed in 4% paraformaldehyde containing 10% sucrose for 48 h at 4°C. Tissue sections were processed as previously described [43] and immunostained with anti-FABP12 antibody (Customized Ab by BIOMATIK, Kitchener, Ontario, Canada; 1 : 300 at 0.79 lgÁmL À1 ), followed by EnVision + anti-rabbit HRP-labeled polymer (DakoCytomation, Carpinteria, CA, USA). Signal specificity was determined using FABP12 peptide (antigen for FABP12 antibody production, 13 lgÁmL À1 ) as a competitor. Tissues were counterstained with hematoxylin.
Animal ethics approval was obtained from the Cross Cancer Institute Animal Care Committee-protocol AC17233. All methods used for monitoring tumor growth, euthanasia, and collection of tumor tissue were carried out in accordance with the Canadian Council on Animal Care (CCAC) Guidelines and Policies. Four to five mice were used for each experiment, with experiments carried out four times.

Cell motility, migration, and invasion assays
The scratch (wound healing) assay was used to measure cell motility [44]. Cells transiently (DU145) or stably (PC3) transfected with FABP12 expression constructs, or their corresponding empty vectors, were seeded in triplicate in 12-well plates. A scratch was made across the monolayer of cells with a P20 pipette tip when cells reached confluency. Cells were cultured for an additional 36 h (DU145) or 30 h (PC3). Dynamic images (15-minute intervals) at two separate positions in each well were captured using digital phase-contrast microscopy (Axiovert 200M, Zeiss). The percentage open area of the scratch at different time points was analyzed with TScratch software [45]. The open area at each time point is represented relative to the 0-h time point which was set at 100%. Cell migration and invasion properties were analyzed using Transwell inserts (8 lM pore size) and BioCoat Matrigel Invasion Chambers, respectively (BD Biosciences, San Jose, CA, USA). Fifty thousand cells in serumfree medium were seeded in the upper chambers of the inserts. Cells were allowed to migrate through the Transwell or Matrigel membranes toward the lower chamber containing DMEM supplemented with 10% FBS for 48-72 h (Transwell) or 72 h (Matrigel). Cells were then fixed, stained with crystal violet, and counted.

Gel shift assays
Gel shifts were carried out using a peroxisome proliferator-activated receptor response element (PPRE) probe (Table S1) as previously described [43] Briefly, annealed oligonucleotides were radiolabeled with [a-32 P]dCTP. Nuclear protein extracts (1 lg) were preincubated with 1 lg poly(dIdC) in binding buffer, followed by addition of labeled probe. For supershift assays, 1 lL of anti-PPARa, anti-PPARb/d, or anti-PPARc antibody was added to the binding reaction 10 min after addition of the labeled probe and incubated at room temperature for 30 min. DNA-protein complexes were resolved in a 5% native polyacrylamide gels in 0.59 Tris/borate/EDTA buffer for 1.5 h. The gels were vacuum-dried and exposed to Xray film.

Luciferase reporter assay
Cells were seeded in 12-well plates (30 000 cells/well) and transfected with the luciferase reporter construct (PPRE 3-TK-Luc; Addgene, Waterton, MA, USA) using polyethyleneimine (PEI). Transfected cells were then harvested and protein lysates prepared using the luciferase cell culture lysis reagent (CCLR; Promega). Luciferase activity was measured using the Luciferase Assay System (Promega, Madison, WI, USA) and quantified with the FLUOstar OPTIMA microplate reader (BMG Labtech, Ortenberg, Germany) following the manufacturer's instructions.

Lipid droplet analysis
Cells were cultured on coverslips for 2 days, fixed with 4% paraformaldehyde, and stained with Nile Red (0.5 lgÁmL À1 ) for 15 min. To measure fatty acid-induced lipid droplet formation, we used the Lipid Droplet Fluorescence Assay Kit (Cayman Chemical, Ann Harbor, MI, USA). Briefly, cells were seeded in 96well plate at 15 000 cells/well and treated with oleic acid at increasing concentrations (0, 125 lgÁmL À1 , 500 lgÁmL À1 ) for 24 h. Cells were then stained with Nile Red, and fluorescence intensity was measured using a FLUOstar Omega microplate reader (BMG Labtech).

High-resolution respirometry analysis of mitochondrial function
Cell respiration was measured at 37°C using an Oxygraph-2k (Oroboros Instruments, Innsbruck, Austria). Cells (in 2 mL culture medium) were added to each chamber of the Oxygraph and the O 2 flow (pmolÁmin À1 Ámillion À1 cells) measured under routine condition (ROUTINE state), resting state (LEAK, after inhibition of ATP synthase with 2 lgÁmL À1 of oligomycin), electron transfer capacity (ET state; maximal O 2 consumption after stepwise titration to an optimal concentration of the protonophore carbonyl cyanide p-[trifluoromethoxy] phenylhydrazone, FCCP), and residual oxygen consumption (ROX state, after inhibition of complex III with 2.5 lM antimycin A). ATP-linked respiration is defined as ROUTINE with subtraction of LEAK. For detection of fatty acid-related respiration, cells were treated with 100 lM of etomoxir (Sigma) [46], a CPT1 inhibitor, for 24 h prior to mitochondrial respirometry analysis.

Measurement of cellular ATP levels
Cells were grown in 12-well plates (30 000 cells/well) for three days. Cells were then trypsinized, resuspended in DMEM, and counted. ATP levels of intact cells (in triplicate) were measured using the ATP Determination Kit (Invitrogen, Thermo Fisher) following the manufacturer's instructions. ATP levels for each sample were normalized to cell numbers.

RT-PCR and western blot analysis
Total RNA from PCa cell lines was prepared using the TRIzol reagent (Invitrogen). First-strand cDNA was synthesized using SuperScript reverse transcriptase II (Invitrogen). PCR amplification was performed with gene-specific primers (Table S1) as previously described [25].
Cell protein lysates were prepared as previously described [47]. Whole-cell (40 lg per lane), cytoplasmic (40 lg per lane), or nuclear protein extracts (20 lg per lane) were separated by SDS/PAGE and transferred to nitrocellulose membranes. Membranes were immunoblotted with primary and secondary antibodies (listed in Table S2).

Databases and statistical analysis
Gene copy numbers and mRNA levels in human PCa tissues were measured using different PCa cohorts. Clinical data from the MSKCC dataset [48] were used for patient survival analysis. All datasets were accessed through cBioportal (www.cbioportal.org/datasets). Statistical analyses were performed using MEDCALC version 14.12.0 (MedCalc Software, Ostend, Belgium). Logrank test was used to compare Kaplan-Meier survival probabilities between patient populations stratified based on mRNA levels (FABP12, PPARc) or amplification status (FABP12). One-way ANOVA (when comparing more than two groups) or two-sided Student's t-test (when comparing two groups) was employed to compute the significance of the difference between patient groups or experimental treatments. Two-way ANOVA was used to examine the significance of effects from two independent experimental factors (i.e., scratch assay, Nile Red assay).

FABP12 is preferentially amplified and overexpressed in metastatic PCa
We examined the gene amplification status of FABP12 in nine human PCa patient cohorts from cBioportal. Three cohorts that consisted mainly of patients with metastatic disease (> 80% of the whole population) showed frequencies of FABP12 amplification ranging from 17% to 30%. The other six cohorts which consisted mainly of localized tumors (> 80%) had much lower frequencies of FABP12 amplification (5-8%) (Fig. 1A). We then compared the frequencies of tumors with FABP12 diploid (CN = 0), copy gain (CN = 1), and amplification (CN ≥ 2) between a primary (MSKCC) and a metastatic (SU2C) cohort. Metastatic PCa had significantly higher frequencies of FABP12 copy gain (54% vs 15%) and amplification (17% vs 5%) compared to primary PCa, but lower frequencies of FABP12 diploid tumors (29% vs 80%) (Fig. S1A,B). Kaplan-Meier survival analysis showed that patients with FABP12 copy gain, amplification, and high mRNA levels have significantly lower recurrence-free survival probability compared to those with FABP12 diploid (P < 0.0001) and low mRNA levels (P = 0.015) (Fig. 1B,C). The dramatic drop in survival probability for patients with FABP12 amplification may be attributable to the additive effects of other oncogenes residing within the same amplicon [37,[39][40][41]. Furthermore, elevated FABP12 levels were significantly associated with high Gleason scores (Fig. S1C) and metastatic status (Fig. S1D). These results indicate an association between FABP12 and poor prognosis in PCa.

Increased FABP12 levels in metastatic xenograft PCa tumors
We examined FABP12 RNA levels in four human PCa cell lines (VCaP, PC3, DU145, LNCaP) derived from metastatic tumors using semiquantitative PCR. These cell lines showed weak (VCaP, PC3, and LNCaP) to barely detectable levels of FABP12 mRNA (DU145) (Fig. S2A). C4.2, an LNCaP subline derived from a mouse xenograft, had slightly higher levels of FABP12 RNA compared to the parent cell line.
To investigate a possible role for FABP12 in metastasis, we immunostained primary and metastatic mouse PC3 xenograft tumor tissues with anti-FABP12 antibody. The FABP12 signal was considerably weaker in primary tumors (Fig. 1D, upper panels) compared to metastatic tumors ( FABP12's subcellular distribution indicates roles in both the cytoplasm and nucleus. We also observed a striking increase in FABP12 transcript levels in PC3derived metastatic tumor tissue compared to PC3 parental cells (Fig. 1G,H). Thus, elevation of FABP12 in our xenograft metastatic tumors mimics the preferential expression of FABP12 in metastatic tumors observed in PCa patient cohorts (Fig. S1D). These results support an association between FABP12 and PCa metastasis.

FABP12 promotes migration and invasion in PCa cells
To investigate the effect of FABP12 expression on cellular properties, we transiently transfected DU1 45 with HA-pcDNA3.1 and HA-pcDNA3.1-FABP12 expression constructs (Fig. S3A) and examined cell motility using the scratch assay. The relative open area (percentage of the scratched area at time 0) was 67%, 46%, and 36%, respectively, for control cells, and 52%, 27%, and 11%, respectively, for the FABP12overexpressing cells after 12, 24, and 36 h (Fig. S3B,  C), indicating a role of FABP12 in enhancing cell motility in DU145.

FABP12 drives epithelial-to-mesenchymal transition
Epithelial-to-mesenchymal transition is associated with invasion and early stages of metastasis [35,49]. We found that the morphology of PC3 cells dramatically changed upon stable expression of FABP12. The PC3-pREP4 cells had a roundish shape with short extensions, whereas the PC3-pREP4-FABP12 cells were larger and fibroblast-like with elongated processes (Fig. 3A). This morphology is typical of cells undergoing EMT. Similar morphological changes were also observed in DU145 cells with ectopic expression of FABP12 (Fig. S5A). We further found that ectopic expression of FABP12 resulted in loss of CDH1 (Ecadherin) expression, at both the RNA (Fig. 3B) and protein (Fig. 3C) levels. E-cadherin is a factor critical for maintenance of epithelial cell structure [50]. Notably, there was a much higher negative correlation between FABP12 and CDH1 mRNA levels in metastatic tumors (r = À0.66, P = 0.002) than in primary tumors (r = À0.25, P = 0.004). FABP12 overexpression also resulted in upregulation of vimentin, a survival probability (log-rank test) for patients with FABP12 amplification/copy gain (B) and high mRNA levels (C). The cutoff point for stratifying FABP12 mRNA levels (z-score = À0.0835) was determined by receiver operating characteristic (ROC) analysis using recurrence status as a classification factor. All datasets were obtained from cBioportal (http://www.cbioportal.org/datasets), and the dataset used for survival analysis was MSKCC, Cancer Cell 2010 [48]. n, sample size; HR, hazard ratio; P, statistical significance level. mesenchymal marker [51], and induction of Slug, a key transcription factor that triggers EMT and directly suppresses E-cadherin [52,53] (Fig. 3C). Induction of Slug in FABP12-expressing cells was also observed in two independent transient transfection experiments (Fig. 3D). FABP12-induced loss of E-cadherin and upregulation of Slug were verified by immunofluorescence staining of stable PC3 control and FABP12overexpressing cell lines (Fig. 3E). Finally, we immunostained our xenograft PCa tumor tissue sections with EMT markers. In contrast to primary tumors, and in agreement with our cell line data, we observed loss of E-cadherin, and elevation of Slug and fibronectin in metastatic tumors (Fig. S6).
Next, we analyzed the subcellular localization of bcatenin, an E-cadherin-interacting partner critical for maintenance of the epithelial phenotype. Loss of Ecadherin facilitates b-catenin release and translocation into the nucleus, which in turn promotes EMT through nuclear signaling [54]. We observed an increase in b-catenin nuclear accumulation, concomitant with E-cadherin loss in PC3-pREP4-FABP12 cells compared to control cells. There were no obvious changes in the cytoplasmic levels of b-catenin (Fig. 3F). These combined findings suggest a role for FABP12 as a driver of EMT in PCa cells.
To examine cell motility, we carried out the scratch assay using stable PC3-pREP4 control and PC3-pREP4-FABP12 cells treated with the FABP inhibitor, CAS300657-03-8 (T4B). While T4B has previously been used to block FABP4 activity [55], our protein-ligand binding modeling analysis using COACH [56] identified T4B as a top-ranked ligand for FABP12 with a TM-score (indicative of structure similarity) of 0.94. We observed significant FABP inhibitor-dependent reduction in cell motility in FABP12-expressing PC3 cells but not in control cells which express abundant FABP4 and FABP5 (Fig. 4A-D and Fig. S2B), suggesting that FABP12, rather than FABP4 and FABP5, is the predominant driver for cell motility for these cell models. Semiquantitative PCR analysis showed that ectopic expression of FABP12 did not affect the levels FABP4, FABP5, and FABP9, paralogs known to be expressed in PCa cells (Fig. S2B). These results suggest that the effect of T4B on PCa cell motility is due to specific inhibition of FABP12.
Furthermore, FABP12-induced Slug expression was reduced in a dose-dependent manner upon FABP inhibition (T4B) in PC3-pREP4-FABP12 cells, but not control cells (Fig. 4E), suggesting that T4B-mediated inhibition of cell motility may be through inhibition of Slug activity. Depletion of Slug in FABP12-expressing PC3 cells using siRNAs targeting SNAI2 (Slug) caused reversal of the increase in cell motility associated with FABP12 expression (Fig. S7). These results suggest that the EMT-inducing factor Slug is a main downstream effector of FABP12 in triggering cell motility and EMT.

FABP12 induces PPARc activation
PPARc, a nuclear receptor implicated in fatty acid signaling and lipid metabolism, has been reported to be a driver of metastasis in PCa [57]. We therefore analyzed the effect of FABP12 on PPARc activation (measured by binding to the peroxisome proliferator response element, PPRE) using the gel shift assay. We first transiently transfected PC3 cells with pcDNA3.1-HA, pcDNA3.1-HA-FABP12, or pcDNA3.1-HA-FABP7. FABP7 is not normally expressed in PC3 cells but has been implicated in the regulation of PPARc activity [58]. Compared to control cells, we observed an increase in the intensity of the PPRE DNA-nuclear protein complex (arrow) in PC3 cells transfected with FABP12 but not with FABP7 expression constructs (Fig. S8, left panel). The supershift assay revealed a slower-migrating band (arrowhead) with anti-PPARc, but not anti-PPARa or anti-PPARb/d antibodies, in FABP12-expressing PC3 cells (Fig. S8, right panel). These results indicate that PPARc is part of the PPRE-nuclear protein complexes observed in the gel shifts.
We then repeated the gel shift assay using nuclear extracts from PC3-pREP4 and -pREP4-FABP12 stable cell lines. We observed PPRE DNA-protein complexes (arrow) in all four PC3 clonal populations tested. However, the signal intensity was markedly increased in PC3-pREP4-FABP12 (FABP12_#1 and #2) cells compared to control cells (C1 and C2). Addition of 509 and 1009 excess unlabeled probe to the binding reaction effectively competed with the labeled probe, with little to no signal detected, indicating probe-specific interactions (Fig. 5A).
To further address FABP12-mediated PPARc-specific binding to PPRE, we carried out gel shift experiments with nuclear lysates from PC3-pREP4 and PC3-pREP4-FABP12 cells with or without PPARc depletion (Fig. 5B,C). Again, increased intensity of the binding complex was observed in PC3-pREP4-FABP12 cells (lane 3) compared to PC3-pREP4 cells (lane 2). PPARc depletion using two different siRNAs resulted in reduced protein-DNA binding, with signal intensities similar to those observed in control PC3 cells (lanes 4 and 5). Supershift experiments with anti-PPARc antibody resulted in a slower-migrating band (arrowhead), in support of the presence of PPARc in the binding complex (Fig. 5D). Next, we used a PPRE-driven luciferase reporter construct to further investigate the effect of FABP12 on PPARc activation. Ectopic expression of FABP12 in PC3 cells increased PPRE-driven luciferase expression/activity by 2.25-fold (pREP4 vs FABP12, P < 0.01) (Fig. 5E). The effect of FABP12 on luciferase activity was substantially reduced (~5-fold) upon depletion of PPARc (Fig. 5E). Similar to PC3 cells, ectopic expression of FABP12 in DU145 cells increased PPRE-driven luciferase activity (Fig. S5B), despite reduction in cell growth (Fig. S5C).  Together, these results indicate a role for FABP12 in the activation of PPARc.

PPARc promotes progression-related properties in FABP12-expressing PCa cells
We analyzed the prognostic significance of PPARc using a PCa dataset from cBioportal [48]. We found that patients with high levels of PPARc had a significantly worse prognosis compared to those with low PPARc levels in the whole population (HR = 3.15, P = 0.0016) (Fig. 5F). Notably, this effect was further magnified in the subpopulation of patients with high levels of FABP12 (HR = 5.28, P < 0.0006) (Fig. 5G), but became insignificant in the population with low levels of FABP12 (HR = 1.41, P = 0.55) (Fig. 5H).
Stable PC3-pREP4-FABP12 cells treated with vehicle control showed a 21-fold increase in cell migration compared to control cells using the Transwell assay. Importantly, when cells were treated with PPARc inhibitor GW9662, we observed a significant decrease (~2-fold) in cell migration in PC3-pREP4-FABP12 cells, but not in PC3-pREP4 cells (Fig. 6C,D). Similar results were observed upon depletion of PPARc with siRNA (Fig. S9). Finally, inhibition of PPARc with GW9662 ( Fig. 6E) or depletion of PPARc with siR-NAs (Fig. 6F) in stable PC3-pREP4-FABP12 cells reversed FABP12-dependent upregulation of Slug. However, PPARc depletion in PC3-pREP4 control cells resulted in only a slight reduction in Slug levels (Fig. 6F). These results support an important role for PPARc in mediating the function of FABP12 in PCa progression.

The interplay of FABP12 and PPARc modulates intracellular lipid accumulation and energy metabolism
To address lipid accumulation, we stained stably transfected PC3-pREP4 and PC3-pREP4-FABP12 cells with Nile Red, a lipophilic dye that measures intracellular lipid droplets. Lipid droplets are site of lipid storage and have been associated with desaturation of fatty acids and cancer growth and survival [59,60]. We observed a 1.79 increase of Nile Red staining intensity in PC3-pREP4-FABP12 cells compared to control cells (P < 0.01) (Fig. 7A,B). We also observed marked increase in ADRP (a lipid droplet marker) immunoreactivity in PC3-pREP4-FABP12 cells compared to control cells (Fig. S10).
Next, we asked whether FABP12 affects uptake of exogenous fatty acids. We found that supplementation of oleic acid (a monounsaturated omega-9 fatty acid) did not significantly change lipid droplet accumulation in PC3 control cells, but significantly increased lipid droplets in PC3-pREP4-FABP12 cells in a dose-dependent manner (Fig. 7C). These results suggest an active role for FABP12 in the uptake of fatty acids leading to increased lipid droplet accumulation.
We then examined a possible role for FABP12 in energy production using the Oxygraph-2k system. Cellular energy, in the form of ATP, can be generated via either anaerobic glycolysis in the cytoplasm or oxidative phosphorylation in the mitochondria. The rate of oxygen consumption by cells is a measure of mitochondrial oxidative phosphorylation. In addition to pyruvate (derived from glucose), fatty acids can also serve as electron donors to produce ATP through oxidative phosphorylation. Here, we measure the respiration of ROUTINE (total cellular oxygen consumption in the physiological coupling state), ATP-linked (ROUTINE strictly coupled to phosphorylation of ADP after subtracting the nonphosphorylating LEAK respiration), and ET (maximum oxygen flux as the determinant of electron transfer capacity) states. Compared to control cells, FABP12-expressing PC3 cells showed significant increases in ROUTINE (84 vs 191, P = 0.02), ATP-linked (Routine-LEAK) (56 vs 116, P = 0.03), and ET (maximum) state (188 vs 408, P = 0.02) respiration (oxygen flow in pmolÁmin À1 Á10 À6 cells). Of note, this FABP12-dependent effect was no longer observed upon PPARc depletion (79 vs 109, P = 0.24 for ROUTINE; 52 vs 70, P = 0.12 for ATPlinked; 213 vs 257, P = 0.73 for ET states). These results demonstrate the importance of FABP12 for increased energy production, which is in turn dependent on PPARc (Fig. 7D).
To test whether FABP12 facilitates mitochondrial energy production from fatty acids, we analyzed mitochondrial respiration in stable PC3-pREP4 and PC3-pREP4-FABP12 cells cultured in the presence or absence of a potent inhibitor of CPT1 (etomoxir), a rate-limiting enzyme mediating fatty acid b-oxidation in mitochondria [61]. CPT1 inhibition in PC3-pREP4-  (Fig. 7E). The difference in fatty acid oxidation observed in PC3-pREP4-FABP12 compared to control cells is significant (P = 0.015) for ROUTINE, and very significant for both ATP-linked (P = 0.004) and ET state (P = 0.004) mitochondrial O 2 consumption (Fig. 7E). Together these results indicate a role for FABP12 in promoting lipid accumulation and lipid utilization for energy production in PCa cells. We then analyzed the activity of citrate synthase (CS), the first and rate-limiting enzyme in the Kreb's cycle. Ectopic expression of FABP12 in PC3 cells resulted in a 2.2fold induction in CS activity (P < 0.01) compared to control cells (Fig. 7F), supporting a critical role for FABP12 in mitochondrial energy production.
We also measured ATP levels in our cell models. ATP levels in two independent pREP4-FABP12 clones were increased by 2.7-fold (Fig. 7G, P < 0.0001) and 1.7-fold (Fig. 7H, P = 0.004), respectively, compared to their corresponding control cells (pREP4_1 and pREP4_2). Unexpectedly, we did not observe significant changes in ATP levels upon PPARc depletion in either pREP4 control or pREP4-FABP12 cells (data not shown). We postulate that this is the result of an overall balanced energy supply and demand in FABP12-expressing PCa cells, with the FABP12-PPARc pathway stimulating mitochondrial oxidation for ATP production, while at the same time triggering processes requiring ATP production such as EMT, cell motility, and invasion.
Next, we examined the effect of PPARc depletion on several key enzymes in cellular fatty acid accumulation and utilization. We observed upregulation of ADRP (a lipid droplet marker), FASN (fatty acid synthase), and CPT1A (rate-limiting enzyme for conversion of fatty acids from acyl-coA chains to acylcarnitine for utilization as energy source in mitochondria; inhibited by etomoxir) in cells with ectopic expression of FABP12 (Fig. 7I,J). PPARc depletion reduced CPT1A expression, but had no effect on ADPR and FASN in FABP12 expressing cells. Interestingly, the expression of ACLY, an ATP citrate lyase responsible for the conversion of citrate to acetyl-CoA, a key step in fatty acid biosynthesis, was not affected by FABP12 expression, but upregulated upon PPARc knockdown (Fig. 7I,J). These results suggest that the FABP12-PPARc pathway primarily drives fatty acid b-oxidation in mitochondria, with FABP12 playing a PPARc-independent role in the induction of de novo fatty acid synthesis and lipid storage.
Finally, we examined whether the increase in PCa cell migration observed upon FABP12 expression was dependent on fatty acid utilization in mitochondria. When the transfer of fatty acids into mitochondria was blocked by etomoxir, a potent CPT1 inhibitor, there was a significant reduction (P < 0.0001) in the migration of PC3-FABP12-expressing cells (Fig. 7K,  L). There was no effect on PC3 control cell migration. These results directly implicate FABP12 in fatty acid beta-oxidation-driven PCa cell migration.

Discussion
Prostate cancer is invariably fatal once it metastasizes [62]. Lipid accumulation and metabolic reprogramming have previously been documented in metastatic PCa [15,63,64]. In this study, we identify an intracellular lipid transporter gene, FABP12, which is preferentially amplified and overexpressed in metastatic PCa PPRE-driven luciferase activity in PC3 control (pREP4) and PC3-FABP12 overexpression (FABP12) stable cell lines transfected with scrambled (control) or PPARc-specific siRNAs. Statistical analysis was done using Student's t-test. (F-H) High levels of PPARc mRNA are significantly correlated with a worse prognosis in a PCa patient cohort (MSKCC dataset described in [48]) (F). This prognostic significance is markedly increased in the subpopulation of PCa patients with high levels of FABP12 mRNA (G), but eliminated in the subpopulation with low FABP12 levels (H). Log-rank test was used for patient survival analysis. The cutoff point for stratifying PPARc mRNA levels was determined by receiver operating characteristic (ROC) analysis using disease-free status as a classification factor. HR, hazard ratio; n, sample size. Error bar: SD.  and correlates with poor clinical outcomes. FABP12 expression promotes properties associated with cancer progression such as EMT conversion, invasion, and metastasis. In addition, FABP12 expression in PCa cells increases fatty acid uptake, lipid droplet accumulation, and energy production from fatty acids. This is the first report demonstrating a role for a FABP in regulating energy production in PCa cells, and an association between FABP12 expression, EMT and increased lipid bioenergetics. Furthermore, our results indicate that EMT and lipid bioenergetic alterations can both be attributed to FABP12-mediated activation of PPARc. Based on our findings, we propose that increased FABP12 expression represents one of the major adaptations of PCa cells required to trigger a pro-invasive molecular pathway and to satisfy the increased energy demands associated with tumor expansion and metastasis (Graphical abstract).
FABP12-mediated increases in cell motility, cell migration, and invasion are accompanied by EMT, a process that turns epithelial cells into migratory and invasive mesenchymal cells. As such, EMT is regarded as a critical priming event for metastasis [35]. There is increasing evidence suggesting a link between EMT and metabolic reprogramming (particularly lipid metabolic reprogramming) in cancer [8,35,63]. Thus, fatty acids may not only serve as energy sources, but also be involved in protein modification, cell membrane remodeling, and cell signaling, processes that are essential for EMT [8,63]. In support of these expanded roles for fatty acids, others have demonstrated that PCa cells undergoing EMT show significant increases in lipid accumulation [65]. FABP12-induced expression of EMT-promoting Slug, elevation in mitochondria respiration, and increases in cell motility and migration were either reversed, or partially reversed, upon depletion and/or inhibition of PPARc, pointing to a role for PPARc in mediating FABP12 functions. PPARs primarily regulate the expression of genes involved in lipogenesis and energy homeostasis [66].
PPARc is generally believed to be a tumor suppressor as numerous in vitro studies show that PPARc agonists (for instance thiazolidinediones) are able to inhibit tumor cell growth and/or induce differentiation in many different cancers [21,67]. However, more recent mechanistic studies suggest that many of PPARc's previously reported anti-cancer properties may actually be PPARc-independent due to off-target effects of its synthetic agonists [67,68]. In PCa, PPARc is overexpressed and its levels are positively correlated with tumor grade and stage [57]. In particular, PPARc has recently been shown to drive PCa progression/ metastasis [57]. In light of our study, we suggest that targeting FABP12, which resides upstream of PPARc, may delay or prevent PCa progression.
A number of FABPs are expressed in PCa cells, including FABP4, FABP5, and FABP9 [29]. The biological actions of these FABPs in promoting PCa progression appear to be synergic rather than redundant and mediated through distinct pathways. The most extensively studied FABP in PCa, FABP5, promotes PCa cell and/or xenograft tumor growth, with inhibition of both PPARb/d and PPARc suppressing these effects [30,33,[69][70][71][72][73]. FABP5 has been proposed to promote PCa metastasis by triggering angiogenesis (VEGF induction) [71,74] and fatty acid synthesis [33]. FABP4 is secreted from both PCa cells and adipocytes surrounding PCa tumors, and is believed to either shuttle fatty acids between adipocytes and tumor cells [75] or act as an inducer of MMPs from cancer cells and cytokines (e.g., IL-6, IL-8) from stroma cells to promote invasion and metastasis [34]. Elevated levels of FABP4 have been reported in PCa bone metastasis in an obese mouse model [76]. Surprisingly, FABP4 expression is downregulated in primary PCa biopsies or cell lines compared to normal tissues/cell lines [77,78], and ectopic expression of FABP4 in PCa cells causes apoptosis [79]. Increased levels of FABP9 in PCa have been shown to have prognostic significance, with FABP9 depletion resulting in reduced cell invasion [80]. Unlike FABP12 shown here, none of these FABPs have been shown to act as an activator of the EMT pathway or a mediator of lipid-related energy production in PCa cells.
FABP12 is expressed at very low levels in PCa cell lines. Its ectopic expression in PCa cells has dramatic effects on cellular properties including increased migration and invasion, but significantly reduces rates of proliferation. As cells in culture are naturally selected for rapid proliferation rates, we believe that tissue culture conditions may select for low levels of FABP12. This idea is in keeping with our observation that FABP12 levels are induced in metastatic tumors derived from a mouse xenograft PC3 model and support the 'Go-or-Grow' hypothesis proposed for tumor invasion [81]. Thus, FABP12 may be at the apex of a critical oncogenic axis leading to PCa metastasis: FABP12 amplification/overexpression ? PPARc activation ? dysregulation of lipid metabolism ? EMT and increase motility, migration and invasion.
PPARc and androgen receptor are both ligand-activated nuclear receptors with important roles in normal prostate development and cancer progression. A number of studies suggest a bidirectional interaction between AR and PPAR, with each receptor influencing the expression and/or activity of the other within prostatic tissues or tumor cells [68,82]. For example, ligand-induced activation of AR is significantly reduced upon treatment with PPARc agonist pioglitazone and troglitazone in LNCaP cells [83], whereas PPARc expression and/or activation is inhibited or enhanced by increasing or suppressing AR signaling in both castration-resistant and sensitive PCa cell models [84]. We speculate that overexpression of FABP12 and the consequent activation of the PPARc pathway may contribute to castration resistance and metastasis in advanced PCa. Future work will focus on the crosstalk between the FABP12-FA-PPARc pathway and androgen signaling.

Conclusions
We have identified an intracellular lipid-binding protein (FABP12) that is preferentially amplified and expressed in metastatic PCa tumors. We provide evidence that FABP12 drives aggressive properties in PCa cells by promoting EMT, lipid accumulation, and utilization of fatty acids for mitochondrial energy production. Our data indicate that FABP12 functions upstream of PPARc, a lipogenic transcription factor implicated in PCa metastasis [57]. Future work will involve targeting FABP12 to investigate its effect on PCa metastasis. 84

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Association of increased FABP12 gene copy numbers and mRNA levels with PCa metastasis and Gleason scores. Fig. S2. RT-PCR analysis of FABP transcripts in prostate cancer cell lines. Fig. S3. FABP12 promotes cell migration in DU145. Fig. S4. Effect of transient and stable expression of FABP12 on PC3 cell proliferation. Fig. S5. Effects of FABP12 overexpression on cell morphology, growth and PPAR activation. Fig. S6. Immunohistochemical staining of EMT markers in PC3-derived xenograft primary and metastatic tumor tissues. Fig. S7. FABP12-induced cell motility in PC3 cells is reduced upon SNAI2 (Slug) depletion. Fig. S8. Effect of transient expression of FABP12 on PPARc transactivation in PC3 cells. Fig. S9. PPARc depletion attenuates FABP12-induced cell migration. Fig. S10. Immunofluorescence staining of lipid droplets. Table S1. Oligonucleotide sequences for primers, probes and siRNAs. Table S2. Antibodies used for western immunoblotting.