The lipidomic profile of the tumoral periprostatic adipose tissue reveals alterations in tumor cell’s metabolic crosstalk

All patients signed an informed consent. Clinical characteristics are summarized in Additional file 1. Patients were stratified based on the International Society of Urological Pathology (ISUP) consensus conference on Gleason grading of prostatic carcinoma [11] as high-risk (III, IV, and V) (n = 20) and low-risk (I and II) (n = 20). The study was performed according to the provisions of the Declaration of Helsinki and was approved by the local ethics committee and adhered to current legal regulations (Biomedical Research Law 14/2007, Royal Decree of Biobanks 1716/2011, Organic Law15/1999 of September 13 Protection of Personal Data). All methods were approved and performed in accordance with guidelines and regulations of the Ethical Committee for Clinical Research (CEIM) from the Pere Virgili Research Institute (CEIM205/2020).

To characterize adipose tissue metabolic profiles, FA methyl ester (FAME), acylcarnitine, and lipidomic analysis was carried out using gas chromatography-mass spectrometry (GC–MS/MS), liquid chromatography-mass spectrometry (LC–MS/MS), and liquid chromatography quadrupole time-of-flight mass spectrometry (LC-QTOF-MS), respectively. For lipidomic analysis, two different LC‐QTOF‐MS‐based platforms were used to analyze methanolic tissue extracts (LIP-I) and chloroform/methanol tissue extracts (LIP-II). The methanol extract platform included free FA, bile acids, steroids, oxylipins, and lysophospholipids (LP). The chloroform/methanol extract platform provided coverage over sphingolipids (SM), monoacylglycerols (MG), diacylglycerols (DG), triacylglycerols (TG), phosphatidylcholines (PC), and ether-phosphatidylcholines (PC-O).

Total FAs were derived from the quantitative analysis of FAME. The acylcarnitine and lipidomic characterizations were performed using internal standards to correct the response of each detected compound based on their family similarity, providing a semi-quantitation as an internal standard response ratio. Specific extraction protocols and mass spectrometry-based analyses are detailed in Additional file 2: Additional Material and Methods.

The number of metabolites obtained by the three different techniques was as follows: LIP-I, n = 120 lipids; LIP-II, n = 122 lipids; and FAME, n = 22.

We designed Transwell co-culture assays of PPAT explants (7 low-risk and 3 high-risk) with 2 PCa cell lines. The prostate cancer cell lines PC-3 (androgen-insensitive) and LNCaP (androgen-sensitive) were purchased from Sigma-Aldrich (Barcelona, Spain). PC-3 cells were cultured in Ham’s F-12 K (Kaighn’s) Medium (1:1 mixture) with L-glutamate (Invitrogen/Gibco, Fisher Scientific SL, Madrid, Spain). LNCaP cells were cultured in RPMI 1640 medium (Merck KGaA, Darmstadt, Germany) supplemented with 1 mM sodium pyruvate (Gibco). PC-3 and LNCaP cultures were supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, and 5 μg/mL Plasmocin® (Invivogen, IBIAN Technologies, Zaragoza, Spain). Cells were seeded into Transwell 0.4-μm pore size cell culture inserts (Fisher scientific, Barcelona, Spain) at 50,000 cells/0.9 cm² in the same medium at 37 °C and 5% CO₂ during 24 h. The next day, the medium was exchanged for serum-free medium, for 24 h. Subsequently, 50 mg of fresh PPAT was washed with PBS twice and added to the lower Transwell chamber in 1 mL of M199 medium with 10% FBS and 1% penicillin/streptomycin in 25 mM HEPES. Each sample was tested in duplicate. The co-culture was maintained at 37 °C and 5% CO₂ for 48 h in the same medium. Subsequently, cells and tissue explants were removed, and RNA was extracted.

Total RNA was extracted from 50 mg of either frozen PPAT or fresh PPAT explants (after co-culture) using the RNeasy Lipid Tissue Mini Kit (Qiagen, Germantown, MD). PC-3 and LNCaP RNA was extracted using the RNeasy Mini Kit (Qiagen). Total RNA was quantified by absorbance measurement, and its purity was assessed by the OD260/OD280 ratio. RNA was retrotranscribed to cDNA using a High Capacity cDNA-to-RNA Kit (Applied Biosystems, Foster City, CA) and the following genes were tested using Taqman assays (Applied Biosystems): SREBP1 (hs01088691_m1), PPARG (hs00234592_m1), FASN (hs01005622_m1), ACACA (hs01046047_m1), LIPE (hs00193510_m1), IL-6 (hs00985639_m1), IL-1B (hs01555410_m1), TNFα (hs99999043_m1), CD36 (hs00169627_m1), FABP4 (hs00609791_m1), CPT1A (hs00912671_m1), PNPLA2 (hs00386101_m1), FATP5 (hs00202073_m1), ADIPOQ (hs00605917_m1), ESRRA (hs01067166_g1), TWIST1 (hs00361186_m1), and MMP-9 (hs00957662_m1). The value for each sample was normalized to the expression of GAPDH. SDS software 2.3 and RQ Manager 1.2 (Applied Biosystems) were used to analyze the results with the comparative Ct method (2^−∆∆Ct). To compare low- versus high-risk, PPAT data were expressed as an n-fold difference relative to a calibrator (a mix of 1 sample of PPAT and 1 sample of abdominal adipose tissue from our Biobanc collection). Co-culture cell lines and co-culture explant PPAT data were expressed as an n-fold difference relative to control.

PCa cells were seeded in 96-well clear flat-bottom black plates (Thermo Fisher Scientific, Barcelona, Spain) and co-cultured with 10 mg of PPAT for 48 h. To quantify fatty acid uptake, the growth medium was exchanged with TF2-C12 fatty acid (Sigma-Aldrich), and cells were incubated at 37 °C for 30 and 60 min. Cellular uptake was measured on Varioskan Lux Reader (Thermo Fisher Scientific) by measurement of fluorescence intensity (λ exposition = 485/λ emission = 515 nm). Lipid content was measured with Nile Red (Sigma-Aldrich). Briefly, media were removed, cells were washed twice with PBS, and lipids were stained with 1.1 mg/mL Nile Red for 15 min at 37 °C. After incubation, cells were washed with PBS followed by measurement of fluorescence intensity (λ exposition = 488/λ emission = 590 nm).

For the PPAT study, the sample size was calculated using G*Power 3.1.9.7 (https://​www.​psychologie.​hhu.​de/​arbeitsgruppen/​allgemeine-psychologie-und-arbeitspsycholog​ie [12]. Assuming a change of twofold between groups and similar group variances, with an average power > 80% and a false discovery rate of 5%, a minimum of 17 patients were needed in each group. Variables not normally distributed are presented as medians and interquartile ranges and were compared with the Mann–Whitney U test. Multivariate stepwise backward regression analysis was employed in order to evaluate the independent predictors associated with PCa aggressiveness. Statistical analyses were performed using the Statistical Package for the Social Sciences, version 22 (SPSS, Chicago, IL). The R software (https://​cran.​r-project.​org/​web/​packages/​Information/​index.​html) was downloaded. To achieve the best discriminatory model between studied groups, we performed variable importance in projection (VIP) score analysis and partial least squares-discriminant analysis (PLS-DA) using MetaboAnalystR package (https://​github.​com/​xia-lab/​MetaboAnalystR). Pathway enrichment was performed by using MetaboAnalyst 5.0 software (https://​www.​metaboanalyst.​ca), and we selected the metabolite set library from SMPDB (The Small Molecule Pathway Database). GraphPad Prism 7.0 was used for the box plot analysis of the different metabolites. For ex vivo experiments, at least 6 co-culture experiments were performed with different human PPAT explant tissues and 2 PCa cell lines. Statistical significance was evaluated with Mann–Whitney tests. Results with p < 0.05 were considered statistically significant.

We performed untargeted lipidomic profiling of 40 PPAT samples using LC–MS/MS and GC–MS/MS platforms. Patients were stratified according to ISUP GG into two categories: low-risk (ISUP groups I and II, n = 20 tissues) and high-risk (ISUP groups III, IV, and V, n = 20 tissues). Patients’ characteristics are summarized in Additional file 1. Of 264 individual lipid metabolites detected, 67 were significantly differentially expressed between low- and high-risk PPAT samples (Additional file 3).

Analysis of total FA profile obtained after FAME analysis (reveals all FA cellular composition, including acyl chains and free fatty acids) showed that the percentage of total saturated FA (SFA) was not different between the low- and high-risk groups whereas a higher percentage of monounsaturated FA (MUFA) was observed in high-risk PPAT tissue when compared with low-risk counterpart’s (27.59% vs 28.53%, respectively; p = 0.001). Analysis of polyunsaturated FA (PUFA) revealed a significant difference for ω-6 PUFA, which were lower in abundance in high-risk PPAT samples compared with low-risk (17.05% vs 18.98% respectively, p = 0.001) (Fig. 1A, Additional file 4).

No changes were observed between the two studied groups for the following metabolites obtained by LIP-II analysis: DG, p = 0.565; SM, p = 0.565; lysophosphatidylethanolamine (LPE), p > 0.9; LPC, p = 0.873; PC, p = 0.155; and PC-O, p = 0.337. By contrast, a significant difference (p < 0.05) was detected in the total levels of lysophosphatidylinositol (LPI), which were higher in high-risk PPAT samples. Also, a trend for lower levels of TG (p = 0.063) was found in high-risk PPAT samples (Fig. 1B).

To identify the most relevant lipid features that would allow us to correctly stratify low-risk versus high-risk patients based on the PPAT lipid profile, we performed PLS-DA on a data set of 70 variables (including the 67 significantly deregulated lipid metabolites shown in Additional file 3 and we also included the variables total SFA, total MUFA, and total PUFA). The score plot of the PLS-DA model showed a separation between patients regarding PCa aggressiveness (Additional file 5: Fig. S1A). We then performed VIP analysis to examine the contribution of the 70 variables in determining the degree of PCa aggressiveness, finding that 33/70 variables had a VIP score ≥ 1 and were therefore considered as important in the model for determining PCa aggressiveness (Additional file 5: Fig. S1B).

The selected 33 variables were then back evaluated with PLS-DA to test the strength of the model, and again the patients were segregated into two differentiated groups (Fig. 2A). The PLS-DA model over fitting, measured as the Q2/R2 ratio (R2—how well the model predicts the calibration of variables, and Q2—how well the model predicts PCa aggressivity) was 0.59, indicating that the model fitted well (Fig. 2B). A model is considered predictive when the Q2/R2 ratio is greater than 0.5 [13]. To obtain the minimum number of significant lipidomic PPAT signatures that could separate the low- and high-risk PCa groups, we performed a second VIP analysis and the model showed that only 16 of the 33 variables had a VIP score ≥ 1 (Fig. 2C). The heat map in Fig. 3 shows that based on these 16 features the PPAT samples, categorized by ISUP grade group, naturally clustered into 2 separate groups corresponding to low- and high-risk PCa.

Multivariate stepwise backward regression analysis with the 16 signatures from VIP panel allowed us to evaluate the independent lipid metabolite predictors associated with PCa aggressiveness. Results showed that 12,13-EpOME (B = 0.008, p = 0.039, 95% CI = 0–0.779) and MG (18:0) (B = 0.942, p = 0.005, 95% CI = 0.9–0.982) were independently associated with PCa aggressiveness.

The PPAT lipid signatures revealed an apparent disorder in lipid metabolism according to PCa pathogenesis. To obtain a global overview of the altered metabolic pathways, we performed metabolite set enrichment analysis using MetaboAnalyst 5.0 and SMPDB metabolite set library with the all-lipid metabolites outlined in Fig. 3. These functional approaches revealed that alterations in linoleic acid metabolism, biosynthesis of FA, and β-oxidation of very long-chain fatty acid had the highest impact in the PPAT lipidome (Fig. 4A) (p < 0.05). Then, several SFA metabolites’ profiles obtained by FAME (Additional file 4) were mapped onto de novo lipid synthesis pathways, and we observed that the amount of palmitic acid and the total amount of its intermediate products, which may be further elongated to form other FA, showed a gradually decreasing trend when patients were stratified by ISUP group. When samples were grouped into low- and high-risk PPAT, significant differences were observed in the amounts of palmitic acid, stearic acid, arachidic acid, and behenic acid (Fig. 4B).

When mapping metabolites related to the linoleic acid pathway, we observed that in general, metabolites showed a gradual decrease when PPAT samples were stratified by ISUP group (Fig. 4C). Significant lower amounts of linoleic acid (LA) only appeared in high-risk PPAT samples when grouped by lower and high-risk PPAT (Fig. 4C). Linoleic acid can be oxygenated by 15-lipoxygenase 1 (LOX-1) in humans, primarily to 13-oxoODE or 9-oxoODE, which were also found to be significantly lower in high-risk than in low-risk PPAT.

Linoleic acid can also be converted by cytochrome P450 to epoxy-octadecenoic acids (EpOMEs) in the form of either 9(10)-EpOME (leukotoxin, coronaric acid) or its regioisomer 12(13)-EpOME (isoleukotoxin, vernolic acid); both metabolites were also lower in abundance in high-risk PPAT than in low-risk PPAT (Fig. 4C).

We next aimed to determine the specific contribution of selected genes in relation to the altered metabolic pathways. We evaluated the expression of sterol regulatory element binding transcription factor 1 (SREBP1), fatty acid synthase (FASN), and acetyl-CoA carboxylase (ACACA) genes, which are involved in de novo FA synthesis. Levels of FASN and ACACA were significantly lower in high-risk PPAT (Fig. 5A), whereas levels of SREBP1 were reduced in high-risk PPAT when compared with low-risk PPAT (Fig. 5B) but did not reach significance. All this data indicates a diminished de novo FA synthesis, in good agreement with the levels of these metabolite pathway products in PPAT (Fig. 4).

Linoleic acid can be converted to 13-OxoODE, which is an endogenous ligand for peroxisome proliferator-activated receptor gamma (PPARG) [14]. When we measured PPARG expression levels, we observed a downregulating trend in high-risk PPAT when compared with low-risk PPAT samples (Fig. 5B). This finding can be related to the lower levels of 13-oxoODE in high-risk PPAT (Fig. 4C). A significant reduction in ADIPOQ gene expression, a gene directly regulated by PPARG, was observed in high-risk PPAT samples when compared with their low-risk counterparts (Fig. 5C).

We also observed that the expression of hormone-sensitive lipase (LIPE) was significantly lower in high-risk PPAT samples than in low-risk samples (Fig. 5D), which has also been described in breast cancer-associated adipocytes [4].

Interestingly, we also revealed an altered inflammatory state in high-risk PPAT samples, demonstrated predominantly by the significantly higher expression of the proinflammatory cytokine interleukin 6 (IL-6) in addition to a trend for higher levels of interleukin 1 (IL-1B) and tumor necrosis factor alpha (TNFα) (Fig. 5E).

The observed reduction in metabolites from pathways associated with de novo FA synthesis and the increased inflammatory profile led us to question if these gene expression alterations were due to direct contact with tumor cells.

To address this, we co-cultured PPAT explants with 2 different PCa cell lines (PC-3 and LNCaP) and then performed gene expression analysis. Results showed that regardless of the cell line used in the co-culture experiment, the expression of membrane lipid transporter CD36 and the expression of genes implicated in de novo lipogenesis such as FASN and PPARG, a transcription factor implicated in lipid metabolism and inflammation, were all significantly downregulated in PPAT explants (Fig. 6A). Of note, the key gene regulator factor SREBP1 was slightly reduced but did not reach significance (Fig. 6A). The expression of carnitine palmitoyltransferase 1A (CPT1A), responsible for the translocation of FA from the cytosol to the mitochondrial matrix, was also reduced. By contrast, the inflammatory profile in PPAT co-cultured explants was elevated, as shown by an upregulation of cytokines such as IL-6 and IL-1B. Interestingly, fatty acid binding protein 4 (FAPB4) expression was higher after co-culture, indicating an active cytoplasmatic lipid mobilization process between PPAT and PCa cell lines (Fig. 6A). The expression of the lipase PNPLA2 was reduced in PPAT explants after co-culture, indicating that lipolysis was likely not upregulated under this culturing condition (Fig. 6A). Of note, no differences in co-cultured explants between high-risk and low-risk tumors were observed following co-culture with PCa cell lines in the lipid and inflammatory genes here analyzed (Additional file 6: Fig. S2A-S2B).

Conversely, co-cultured PCa cell lines (both PC-3 and LNCaP) showed a significant increase in the expression of lipid metabolism-related genes such as CD36, FATP5, CPT1A, and FABP4 and of the inflammation-related genes IL-6 and IL-1B (Fig. 6B, C). Noteworthy, while FASN, SREBP1, PPARG, and PNPLA2 were upregulated in PC-3 after PPAT explant co-culture (Fig. 6B), no changes were observed in these genes in LNCaP cells after co-culture (Fig. 6C).

Alteration in PCa cell aggressiveness was also observed in both cell lines after PPAT explant co-culture, demonstrated by a significant increase in expression levels of genes implicated in tumor cell proliferation (ESRRA) and in tumor invasive and metastatic potential (MMP-9, TWIST1) (Fig. 6B, C).

An increase in fatty acid uptake is shown in both cell lines (Fig. 6D) after explant co-culture. Also, both PCa cell lines exhibited enhanced lipid accumulation after explant co-culture (Fig. 6E). Of note, LNCaP co-culture cells showed a significantly higher uptake when compared with co-culture PC-3 cells. No differences in accumulation or in uptake rates were observed in the PCa cell lines studied (PC-3 and LNCaP) regardless of PPAT explant aggressiveness (Additional file 6: Fig. S2C-S2D).