Secreted Factors from Adipose Tissue Reprogram Tumor Lipid Metabolism and Induce Motility by Modulating PPARα/ANGPTL4 and FAK

Breast cancer is the most common cancer among women and one of the leading causes of cancer-related deaths worldwide (1). Besides well-described risk factors such as a woman's age, genetic predisposition, reproductive factors, estrogen, and alcohol consumption (2), there is significant epidemiologic evidence indicating that obesity is promoting breast cancer development and progression (3–6). One of the leading hypotheses is that in obesity, fat tissue becomes hypertrophic and supports tumor progression by secreting higher levels of protumorigenic adipokines, pro-inflammatory cytokines, and free fatty acids (FFA; ref. 7). Besides these systemic effects, there is cumulating evidence that also the direct interaction of tumor cells and local adipocytes (termed cancer-associated adipocytes) supports cancer cell proliferation and progression. For instance, ovarian cancer cells and breast cancer cells take up and utilize FFAs derived from neighboring adipocytes as fuel for tumor growth by elevated β-oxidation (8–10). Moreover, treatment of breast cancer cells with exogenous fatty acids, for example, oleic acid increases fatty acid oxidation (FAO) as well as cell growth, migration, and invasion (11–14). Although the molecular link between cancer cell motility and exogenous FFA is not well established, several reports describe PPARs as potential major players in the cross-talk between metabolism and cancer (15). PPARs are ligand-activated transcription factors of the nuclear hormone receptor superfamily, regulating the expression of target genes known to be involved in lipid and glucose metabolism and in regulating expression levels of inflammatory cytokines and adipokines (15). One of the key mediators in this setting could be the PPAR-target gene angiopoietin-like 4 (ANGPTL4; refs. 15–17). Angptl4 is a secreted protein playing important roles in both lipid metabolism and cancer (15). Importantly, ANGPTL4 is expressed in several tumor types, including breast cancer, with increasing expression levels from benign to the invasive state (18). Furthermore, Angptl4 was shown to increase breast cancer cell invasion and metastasis to the lung in vitro and in vivo (19, 20). However, the role of PPAR signaling and Angptl4 in cancer cell proliferation, motility, and metastasis, especially in the context of obesity-related breast cancer, has not been established.

Another important mediator in breast cancer progression is the focal adhesion kinase (FAK), a protein tyrosine kinase, which was described to enhance proliferation, migration, and invasion capabilities in breast cancer cells (21, 22). FAK was first identified at extracellular matrix and integrin receptor cell adhesion sites, but was also found at adherens junctions, in endosomes and the nucleus where it modulates integrin-recycling activation, vascular permeability, cell survival, and transcriptional regulation (23). Interestingly, recent studies demonstrated that Angptl4 activates integrin and downstream FAK signaling in keratinocytes (24) and in head and neck squamous cell carcinoma cells (16). However, the interrelation of Angplt4 and FAK signaling in breast cancer cells, especially in the interaction with adipose tissue, has not been investigated so far.

Here, we investigated the impact of obesity on triple-negative breast cancer (TNBC) on a molecular and cellular level. Specifically, we were interested in analyzing changes in molecular pathways, lipid homeostasis, and cancer cell behavior in response to factors secreted from fat tissue of obese individuals. We show that culture with adipose tissue conditioned media (ACM) increases PPAR-signaling and Angptl4 expression in TNBC cells, likely triggered by a surplus of exogenous FFA secreted from adipose tissue. In addition, exposure to ACM induced metabolic reprogramming in TNBC cells characterized by lower de novo lipogenesis and elevated FAO, along with increased FAK activation and cell motility. Here, cell invasion and migration was dependent on Angptl4 and FAK. Together our data indicate that the interaction of breast cancer cells with adipose tissue of obese individuals leads to a drastic increase of tumor promoting pathways, with an elevated supply of FFA playing an important part in these processes.

Extended material and methods are available in the supplement, including Supplementary Tables S1 to S6.

Human breast cancer cell lines MDA-MB-231 (#HTB-26) and HCC38 (#CRL-2341) were obtained from the ATCC; the murine breast cancer cell line E0771 was obtained from Tebu-bio (#940001-A) with respective certificates provided by the companies. No further cell line authentification was performed. After thawing, cells were used for up to 8 to 10 passages. Cell lines were grown in DMEM (Thermo Fisher Scientific, #31966-021; MDA-MB-231, E0771) or RPMI1640 (Biochrom, #FG1415; HCC38) supplemented with 10% FBS (Biochrom, #S0615) and 1x Antibiotic-Antimycotic (Thermo Fisher Scientific, #15240-062) and incubated at 37°C in 5% CO₂. Cells were periodically tested for mycoplasma contamination using MycoAlert Mycoplasma Detection (Lonza, #LT07-318). Cells were counted manually using a Neubauer chamber.

C57BL/6J mice were fed a high-fat diet (HFD, 58%kcal from fat; Research Diets D12331) or normal chow diet (NC; 11%kcal from fat) ad libitum for 16 weeks. Subcutaneous and visceral abdominal adipose tissue was removed and collected from euthanized mice. Tissue explants were washed once with DPBS (Thermo Fisher Scientific, #14040-091) and cut to small pieces (100 mg). Adipose tissue was then incubated with DMEM-F12 (Thermo Fisher Scientific, #31330-038) supplemented with 10% FBS and 1x Antibiotic-Antimycotic for 24 hours at 37°C in 5% CO₂ and washed with DPBS before co-culture with human breast cancer cells. Experiments were performed in accordance with the rules for animal care of the local government authorities and were approved by the animal care and use committee of Leipzig University and the Bezirksregierung Leipzig, Germany (TVV46/13, T10/16).

For isolation of murine adipocytes, adipose tissue from mice was collected, washed with DPBS, and minced into very small pieces. Four grams of fat tissue were incubated for 30 minutes at 37°C in 5 mL collagenase buffer, containing 1 mg/mL collagenase (Sigma-Aldrich, #C1764). The digested tissue suspension was filtered twice using a sterile 260 μm and a 190 μm filter. After each filtration step, the filter was rinsed with 10 mL DMEM-F12/10% FBS. Adipocytes were washed by centrifuging the suspension for 1 minute at 400× g. The upper layer containing mature adipocytes was transferred to a fresh tube with 40 mL DMEM-F12/10% FBS and 1x Antibiotic-Antimycotic. Both steps were repeated three times. Next, isolated adipocytes were incubated for 24 hours at 37°C in DMEM/10% FBS and 1x Antibiotic-Antimycotic. After 24 hours, the suspension was transferred to a fresh tube and centrifuged for 1 minute at 400× g. The upper layer containing mature adipocytes was used for co-culture experiments with human breast cancer cells.

MDA-MB-231 cells were co-cultured with murine adipose tissue or isolated adipocytes using ThinCert cell culture inserts (0.4 μm pore size; Greiner #657641) in six-well tissue culture plates. MDA-MB-231 cells were seeded in the bottom of a six-well plate and cultured for 24 hours prior to co-culture. Fat tissue (100 mg/insert) or mature adipocytes (35 μL/insert) were transferred into cell culture inserts. MDA-MB-231 cells and murine adipose tissue or adipocytes were co-cultured in DMEM-F12/10% FBS and 1x Antibiotic-Antimycotic for 72 hours at 37°C in 5% CO₂.

Adipose tissues were obtained from overweight [body mass index (BMI) ≥25 to <30; n = 6] and morbidly obese (class III obesity, BMI ≥40; n = 8) female patients undergoing elective surgery or gastric bypass surgery. Informed consent was obtained from each patient and the study was approved by the ethics committee of the University of Leipzig (159-12-21052012 and 017-12-23012012). All experiments were performed according to the declaration of Helsinki. ACM was prepared as previously described following a standardized protocol (25). In brief, adipose tissue was cut into small pieces and incubated with DMEM 1% BSA (Sigma-Aldrich, #A7030), 1% antibiotic-antimycotic (basal medium) for 24 hours at 37°C, 5% CO₂. For all experiments, 1 g of adipose tissue was cultured in 10 mL basal medium. The same protocol was used for both ACM of overweight (ACM_<30) and ACM of morbidly obese (ACM_>40) individuals. Following incubation, the medium was collected and centrifuged at 4,000× g for 15 minutes. To avoid contamination with cellular debris from the bottom or adipocytes from the upper layer, the intermediate phase was used and passed through a 40 μm cell strainer. ACM was stored at −80°C until usage.

MDA-MB-231, HCC38, and E0771 cells were grown to 70 % confluence. Medium was changed to ACM or basal medium for 24 hours. Incubation of breast cancer cells with BSA-conjugated fatty acids was performed for 24 hours using 0.1 mmol/L BSA-oleic acid (BSA-OA) or 0.1 mmol/L BSA-linoleic acid (BSA-LA) in DMEM unless indicated otherwise. Oleic acid (Sigma-Aldrich, #01008) and linoleic acid (Sigma-Aldrich, #L1376) were conjugated with fatty acid-free BSA (Sigma-Aldrich, #A7030) in a molar ratio of 5:1 (FFA:BSA) as described previously (26).

MDA-MB-231, HCC38, and E0771 cells were grown to 70 % confluence. Medium was changed to ACM or basal medium for 24 hours with or without 10 μmol/L PPARγ antagonist GW9662 (Sigma-Aldrich, #M6191), 0.5 μmol/L PPARβ/δ inverse agonist ST247 (Sigma-Aldrich, #SML0424), or 10 μmol/L PPARα antagonist GW6471 (Sigma-Aldrich, #G5045). Cell treatment with PPAR agonists was performed with 0.1, 1, or 10 μmol/L PPARα agonist GW7647 (Sigma-Aldrich, #G6793), PPARβ agonist GW0742 (Sigma-Aldrich, #G3295) or PPARγ agonist GW1929 (Sigma-Aldrich, #G5668) for 24 hours in ACM or basal medium. Phosphorylation of FAK was inhibited using 10 μmol/L FAK inhibitor PF573228 (Absource Diagnostics, #S2013) if not indicated otherwise.

Mitochondrial respiration and glycolysis were analyzed using the Seahorse XF Cell Energy Phenotype Test Kit (Agilent, #103325-100) according to manufacturer's protocol. MDA-MB-231 cells (15,000/well) were seeded in Seahorse XF24 Cell Culture Microplates (Agilent, #100777-004) in DMEM/10% FBS. After 48 hours, medium was replaced by ACM or basal medium for 24 hours. Forty-five minutes prior to the assay, cells were washed and incubated in assay medium [Seahorse XF DMEM pH 7.4 (Agilent, #103575-100), supplemented with 1 mmol/L pyruvate, 2 mmol/L glutamine, and 10 mmol/L glucose] in a non-CO₂ incubator at 37°C. Measurements of baseline and stressed phenotypes of oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were conducted in a Seahorse XF24 Bioanalyzer. The stressed phenotype was induced by injection of the stressor mix including 1 μmol/L oligomycin and 1 μmol/L carbonyl cyanide-4 (trifluoromethoxy) phenylhydrazone (FCCP) at indicated time points. OCR was normalized to protein concentration of MDA-MB-231 cells.

β-Oxidation was determined using the Seahorse XF Palmitate FAO Substrate Kit (Agilent) according to manufacturer's instructions. In brief, MDA-MB-231 cells (20,000/well) were seeded in DMEM/10% FBS in a XF-96 cell culture microplate (Agilent, #101085-004). After 24 hours, cells were cultured with basal medium or ACM for 24 hours and then switched to substrate-limited medium (Seahorse XF DMEM pH 7.4, supplemented with 0.5 mmol/L glucose, 1 mmol/L glutamine, 1% FBS, and 0.5 mmol/L carnitine) for 2 hours before assaying. Forty-five minutes prior to the assay, medium was replaced with FAO assay medium (Seahorse XF DMEM pH 7.4 supplemented with 0.5 mmol/L glucose and 0.5 mmol/L carnitine) and cells were put in a non-CO₂ incubator at 37°C. Etomoxir (4 μmol/L) and 200 μmol/L palmitate (XF Palmitate-BSA conjugated FAO substrate; Agilent, #102720-100) were added directly before analysis in a Seahorse Bioanalyzer XF96. OCR was determined in the basal state and after consecutive treatment with: (i) 1.5 μM oligomycin, (ii) 1.5 μmol/L FCCP, and (iii) 0.5 μmol/L rotenone/antimycin A. To evaluate coupled respiration, OCR in the presence of oligomycin was subtracted from basal OCR. Maximal respiration was determined after injection of FCCP. Etomoxir inhibition of OCR was calculated by subtracting basal OCR with etomoxir from basal OCR without etomoxir.

FAO during ACM treatment was determined using the FAO Kit from the Biomedical Research Service Center at State University of New York, Buffalo, NY (Catalog No. #E-141; ref. 27). This assay measures FAO activity levels based on the oxidation of octanoyl-CoA, which is coupled to NADH-dependent reduction of INT to INT-formazan. The signal is proportional to the activity of FAO in the samples. Briefly, MDA-MB-231 cells were incubated in basal medium or ACM. After 24 hours, cell lysates were harvested using the provided 1 × Cell Lysis Solution (50 μL/well). Sample protein concentrations were determined and normalized. 50 μL of FAO Assay Solution or control solution was added to 10 μL of the protein sample. Following incubation for 60 minutes at 37°C, the reaction was terminated by adding 50 μL of 3% acetic acid and the plate was read at 492 nm. Blank reading was subtracted from the sample reading.

The data were analyzed by either ANOVA or unpaired two-tailed Student t test using GraphPad PRISM. P < 0.05 was applied as threshold for significance. In figures, data are shown as mean ± SD and significance is relative to basal medium if not indicated otherwise. Levels of statistical significance are denoted by asterisks: *P < 0.05, **P < 0.01, ***P < 0.001.

To investigate the effect of factors secreted from adipose tissue on TNBC cells, we used a two-dimensional co-culture system, which enables both cell types to communicate via secreted factors through the pores of a membrane. Using this system, we addressed the influence of obesity by co-culturing MDA-MB-231 cells with visceral adipose tissue of diet-induced obesity mice fed a high-fat diet for 16 weeks (HFD, C57BL/6 mice) and with visceral adipose tissue of chow-fed normal weight mice. After three days of co-culture, the RNA of MDA-MB-231 cells was subjected to genome-wide mRNA expression analysis. Coherent with data from our previous co-culture study with murine 3T3-L1 adipocytes (28), pro-inflammatory ILs, such as IL1B and IL1A, were among the most upregulated genes in MDA-MB-231 cells co-cultured with adipose tissue of obese mice (Fig. 1A), thus confirming their involvement during the interaction of breast cancer cells and adipocytes. Concomitantly, the cytokine-cytokine receptor interaction pathway was enriched in a pathway analysis of all differentially expressed genes (≥1.5-fold change; Fig. 1B). However, the 50 top up- and downregulated gene list also contained several key regulators of lipid metabolism. We observed that genes involved in the biosynthesis of fatty acids or cholesterol, such as stearoyl-CoA desaturase (SCD1), insulin induced gene 1 (INSIG1), and sterol regulatory element binding transcription factor 1 (SREBF1) were downregulated in co-cultured MDA-MB-231 cells, whereas genes involved in β-oxidation such as carnitine palmitoyltransferase 1a (CPT1A) and the carnitine-acylcarnitine translocase encoding gene SLC25A20, were upregulated. The expression of these genes was also altered in MDA-MB-231 cells co-cultured with adipose tissue from normal chow-fed mice, albeit to a lesser extent (Fig. 1A). Further, the enrichment analysis of differentially expressed genes for KEGG pathways revealed a downregulation of biosynthesis of cholesterol and unsaturated fatty acids pathways, whereas the fat digestion and absorption, fatty acid degradation, and PPAR signaling pathways were upregulated (Fig. 1B; Supplementary Fig. S1). These data were corroborated by Ingenuity Pathway Analysis, where an increased activity for PPAR transcription factors was identified as potential upstream regulator of genes with increased expression in co-cultured MDA-MB-231 cells (Supplementary Table S2). Top upregulated PPAR target genes included angiopoietin-like 4 (ANGPTL4), cAMP responsive element binding protein 3 like 3 (CREB3L3), pyruvate dehydrogenase kinase 4 (PDK4), and perilipin 2 (PLIN2) and we next aimed to validate these using qPCR. To investigate whether induction of PPAR target genes was dependent on adipose tissue of obese mice or if this effect could also be induced by adipose tissue of lean mice, we studied their expression after co-culture with adipose tissue from lean and obese mice. The results confirmed an elevated expression of PPAR target genes which was even more pronounced following co-culture with adipose tissue of obese mice, as shown by higher fold changes in gene expression of ANGPTL4, CREB3L3, PDK4, and PLIN2 (Fig. 1C). Interestingly, co-cultivation with subcutaneous murine adipose tissue led to similar gene expression changes in MDA-MB-231 cells, but to a lesser extend (Supplementary Fig. S1B). The activation of PPAR target genes was also observed upon co-culture with isolated adipocytes of lean and obese mice, indicating that factors secreted from adipocytes are sufficient to induce these expression changes (Supplementary Fig. S1C). Collectively, these data suggest that co-culture with adipose tissue induces distinct expression changes in genes and pathways related to lipid metabolism in MDA-MB-231 cells. Of note, these changes were more drastic when MDA-MB-231 cells were cultured with adipose tissue of obese mice.

Next, we investigated the impact of secreted factors of human adipose tissue on TNBC. Therefore, we generated human ACM for the following cell culture experiments. Human adipose tissue was cultured for 24 hours in basal medium (DMEM/1% BSA) to obtain ACM. To address the effect of obesity, we used ACM prepared from adipose tissue of overweight (BMI ≥25 <30) and morbidly obese patients (BMI ≥40). Treatment of MDA-MB-231 and HCC38 cells with ACM from overweight patients (ACM_<30) significantly increased gene expression levels of PPAR target genes (up to 7.2-fold), whereas the incubation with ACM from morbidly obese patients (ACM_>40) resulted in an even more profound induction of PPAR target genes (up to 50-fold; Fig. 2A). This effect was also detectable for the murine TNBC cell line E0771 treated with ACM of lean and obese mice (Supplementary Fig. S2). Thus, these data are coherent with the results obtained from our co-culture experiments with murine adipose tissue and point in the direction that especially factors secreted by fat tissue of obese individuals promote PPAR signaling in TNBC cells.

As mentioned before, one of the most upregulated PPAR-target genes in ACM-treated cells was ANGPTL4. This was of specific interest, because Angptl4 is not only described for its role in regulating lipid metabolism in adipose tissue, but was also found in several tumor types, with higher expression levels in aggressive cancers (18). For instance, ANGPTL4 expression was associated with metastatic spread of breast cancer cells to the lung (19, 20). Analyzing data from the TCGA breast cancer collection, we found that mRNA expression of ANGPTL4 was higher in TNBC as compared with luminal and HER+ subtypes (Supplementary Fig. S3A). Further, Kaplan–Meier curve analysis indicated that high ANGPTL4 expression was associated with reduced progression free survival, particularly in patients with TNBC (Supplementary Figs. S3B and S3C), suggesting that ANGPTL4 expression may also be of clinical relevance.

To investigate if the higher expression of ANGPTL4 observed in our cell models would also translate into increased protein level, we quantified secreted Angptl4 in the supernatant of ACM-treated MDA-MB-231 and HCC38 cells (Fig. 2B). We found that treatment of MDA-MB-231 cells with human ACM_<30 resulted in a significant increase of Angptl4 secretion as compared with basal medium. However, Angptl4 secretion was even more drastically increased following treatment with ACM_>40, which is in line with our results from ANGPTL4 gene expression. In HCC38 cells, elevated secretion of Angplt4 protein was also detected after treatment with ACM, but only with ACM_>40 (Fig. 2B), as these cells generally secret markedly lower levels of Angptl4 compared with MDA-MB-231 cells_.

Because PPARs comprise of different subtypes, namely PPARα, PPARβ/δ, and PPARγ, we sought to determine if a specific PPAR subtype could be identified as main regulator in ACM-treated breast cancer cells. First, we examined the inducibility of all three PPAR subtypes in MDA-MB-231 cells. MDA-MB-231 cells were treated with 0.1, 1, or 10 μmol/L of PPAR agonists GW7647, GW0742, and GW1929, to activate PPARα, PPARβ/δ, and PPARγ, respectively (Supplementary Figs. S4A–S4C). All three PPAR subtypes were inducible by PPAR agonists, as indicated by an upregulation of the PPAR target genes ANGPTL4, CREB3L3, PDK4, and PLIN2. Next, we cultured MDA-MB-231 and HCC38 cells with human ACM with or without PPARα antagonist GW6471. The treatment with GW6471 decreased expression of the PPAR target genes and reduced Angptl4 protein secretion (Fig. 2C and D; Supplementary Fig. S4D). The reduced ANGPTL4 gene expression following GW6471 treatment was also observed in murine E0771 cells (Supplementary Fig. S4E). Of note, the PPARβ/δ antagonist ST247 and PPARγ antagonist GW9662 did not reduce the expression of the selected PPAR target genes in MDA-MB-231 cells (Supplementary Fig. S4F). Together, our findings show that human ACM induces PPAR target gene expression and Angptl4 protein secretion in TNBC cells in a PPARα-dependent fashion.

Next, we sought to determine adipose tissue secreted factors, which may activate PPARα signaling in TNBC cells. Because lipids are potential activators of PPAR signaling, we characterized major lipid species present in ACM using Fourier transform mass spectrometry (FIA-FTMS). We found that FFA were most abundant in ACMs, with significantly higher concentrations in ACM_>40 as compared with ACM_<30 (Fig. 3A). Both, ACMs_<30 and ACMs_>40, also contained significant amounts of mono-, di-, and triacylglycerols, but these did not differ between ACMs. Furthermore, we analyzed the composition of fatty acids in the ACMs by gas chromatography-mass spectrometry (GC-MS). We observed that oleic (FA18:1 n-9), linoleic (FA18:2 n-6), and palmitic acid (FA16:0) were most abundant in the ACMs (Supplementary Table S4; Supplementary Fig. S5A). ACMs_>40 contained significantly higher concentrations of these fatty acids than ACMs_<30 (Fig. 3B). To explore a potential link between fatty acids and PPARα activation in TNBC cells, we correlated fatty acid concentrations of the ACMs and the ANGPTL4 gene expression levels in MDA-MB-231 cells treated with the corresponding ACMs. Indeed, ANGPTL4 and total FFA concentrations were highly correlated (r = 0.80; Fig. 3C) and comparable correlations were detected when each of the three most abundant fatty acids (oleic, palmitic, and linoleic acid) were analyzed individually (Fig. 3C). Thus, to find out more about the role of exogenous fatty acids in PPAR-signaling in breast cancer cells, we incubated MDA-MB-231 and HCC38 cells with basal medium complemented with FFAs. Coherent with the data obtained from ACM treatment, incubation with BSA-conjugated oleic acid (BSA-OA) and linoleic acid (BSA-LA) significantly increased expression of ANGTPL4 (Fig. 3D and E) and other PPAR-target genes (Supplementary Figs. S5B–S5D) in a dose-dependent manner and resulted in drastically elevated Angptl4 protein secretion (Fig. 3G). In both cell lines, the PPARα inhibitor GW6471 diminished PPAR target gene expression and Angptl4 secretion (Fig. 3F and G). The PPARα-dependent upregulation of Angptl4 and other PPAR-target genes by oleic and linoleic acid could also observed in murine E0771 cells (Supplementary Figs. S5D and S5E). Together, these results suggest that FFAs released by adipose tissue activate PPARα in TNBC cells. This effect was dose-dependent as culturing breast cancer cells with ACM of obese individuals enhanced PPARα-signaling, highly likely by providing higher amounts of FFA. Because previous studies demonstrated that ANGPTL4 expression can also be regulated by TGFβ1 in cancer cells (20), we also quantified the concentration of TGFβ1 in ACMs and explored their association with ANGPTL4 expression. Interestingly, we only detected a weak correlation (r = 0.37) between ANPGTL4 and TGFβ1 levels of the corresponding ACMs (Supplementary Fig. S5F). These data corroborate that fatty acid induced PPARα-signaling plays a major role in ANGPTL4 expression in ACM-treated breast cancer cells.

A hallmark of many tumor types is the upregulation of de novo fatty acid synthesis to produce lipids needed for tumor growth, progression, and signaling (29). However, it was recently demonstrated that cancer cells are able to take up, store, and metabolize exogenous FFA supplied by surrounding adipocytes (8–10). Given these observations, we explored the effects of ACM-treatment on cellular lipid homeostasis in TNBC cells. We recently reported that treatment with ACM leads to accumulation of intracellular lipids in MDA-MB-231 cells (22). Likewise, we now observed significantly increased formation of intracellular lipid droplet (LD) in MDA-MB-231 and HCC38 cells treated with oleic acid (Supplementary Fig. S8) in a dose-dependent manner in response to an increased supply of exogenous FFAs.

Coherent with the microarray data (Fig. 1A), mRNA levels of FASN and SCD1, key genes of de novo fatty acid synthesis were consistently diminished upon ACM-treatment in MDA-MB-231, HCC-38, and E0771 cells (Fig. 4A; Supplementary Figs. S6A and S6B). This effect was also observed when TNBC cells were incubated with BSA-OA (Supplementary Fig. S6C). Further, overrepresentation analyses of the microarray data for gene ontology terms indicated that “fatty acid synthesis” was enriched in downregulated genes in MDA-MB-231 cells co-cultured with adipose tissue (Supplementary Tables S5A and S5B). In line with these findings, a decreased activity for sterol regulatory element-binding protein (SREBP) transcription factors, which are pivotal in inducing the transcription of de novo fatty acid synthesis genes, was predicted by Ingenuity Pathway analysis (Supplementary Table S3). Therefore, we determined Srebp1 protein levels and detected significantly lower amounts of the active fragment (nSrebp) in ACM-treated TNBC cells (Fig. 4B; Supplementary Fig. S6D). To confirm a lower rate of de novo fatty acid synthesis upon treatment with ACM, we labeled MDA-MB-231, HCC38, and E0771 cells with ¹³C-acetate and analyzed its enrichment in newly synthesized fatty acids via GC-MS. We found that co-culture with ACM reduced fatty acid synthesis in all three cell lines, with ACM_>40 showing an even more pronounced reduction than ACM_<30 (Fig. 4C; Supplementary Figs. S6E and S6F). Likewise, fatty acid synthesis was reduced upon treatment of TNBC cells with BSA-OA (Supplementary Fig. S6G). Collectively, these data indicate that ACM-treatment of TNBC cells leads to the inhibition of SREBP mediated de novo lipid synthesis.

Next, we investigated the effect of ACM on mitochondrial oxidation and glycolysis in MDA-MB-231 cells by performing Seahorse XF Cell Energy Phenotype Tests. Treatment of cells with ACM significantly increased baseline OCR and ECAR, suggesting higher basal levels of both mitochondrial respiration and glycolysis after ACM_>40 treatment (Fig. 4D). The stressed cell phenotype was induced by simultaneous inhibition of ATP synthase and mitochondrial uncoupling after adding the stressor compounds oligomycin and FCCP. This resulted in a compensatory increase in the rate of glycolysis and mitochondrial oxidation, indicated by enhanced ECAR and OCR in both groups. Interestingly, under stressed conditions, OCR and ECAR were even higher in MDA-MB-231 cells pretreated with ACM_>40 than in cells pretreated with basal medium, implying a higher mitochondrial and glycolytic capacity of the cells.

We reasoned that increased amounts of fatty acids provided by ACM might result in a higher utilization of fatty acids for β-oxidation and thereby contribute to the increased OCR observed in ACM-treated TNBC cells. This hypothesis was supported by our microarray data, which indicated that “FAO” and related gene ontologies were enriched in upregulated genes in MDA-MB-231 cells co-cultured with adipose tissue (Supplementary Tables S6A and S6B). Further, expression levels of CPT1A, SLC25A20 and ACAA2, known PPARα target genes and key regulators of mitochondrial β-oxidation, were consistently elevated in TNBC cells following ACM treatment. Although ACM_<30 modestly increased CPT1A, SLC25A20, and ACAA2 expression, cultivation with ACM_>40 resulted in a highly significant increase of these genes (Fig. 4E; Supplementary Fig. S7A). Similarly, the expression of FAO genes was also induced in TNBC cells upon cultivation with oleic acid (Supplementary Fig. S7B). To address whether these changes in gene expression also translate to increased β-oxidation, we applied the Seahorse XF Palmitate Oxidation Assay, in which the measurement of OCR reflects the oxidation of fatty acids. Indeed, treatment of MDA-MB-231 cells with ACM resulted in higher basal and maximal respiration, as well as respiration coupled to ATP production (Fig. 4F). Importantly, the addition of the FAO inhibitor etomoxir profoundly reduced the OCR in ACM-treated cells, whereas cells in basal medium only showed a modest reduction in OCR (Fig. 4G). These results were further corroborated by testing the FAO enzymatic activity with a FAO Assay Kit. We found that treatment of MDA-MB-231 cells with ACM significantly induced FAO enzymatic activity (oxidation of octanoyl-CoA) by 33% (Fig. 4H). Collectively, these findings indicate that ACM-derived lipids can be utilized in FAO, decreasing the dependency of TNBC cells on de novo lipid synthesis.

Given that ACM_>40 treatment induced a metabolically more active cell phenotype, we next investigated, if cultivation with ACM_>40 affects energy-demanding cellular processes such as proliferation, migration, and invasion in MDA-MB-231 cells. Indeed, treatment of MDA-MB-231 and HCC38 cells with ACM_>40 significantly elevated cell proliferation compared with cells treated with basal medium as detected by increased BrdUrd incorporation (Fig. 5A). Moreover, in scratch assays, we detected a 2.7- to 7.2-fold higher wound healing capability of MDA-MB-231 and HCC38 cells treated with ACM_>40 (Fig. 5B). Finally, cell invasion of MDA-MB-231 and HCC38 cells was drastically stimulated by treatment with ACM_>40 (Fig. 5C). Increased cell proliferation and cell motility following ACM treatment were also observed in E0771 cells (Supplementary Figs. S9A and S9B). Together, these results indicate that adipose tissue released factors promote a more aggressive phenotype in TNBC cells with elevated proliferation, wound healing, and invasion.

Because ACM treatment profoundly increased ANGPTL4 expression and promoted cellular processes involved in cancer progression in TNBC cells, we next analyzed the role of Angptl4 in tumor cell proliferation and motility in more detail. Therefore, we generated transient and stable ANGPTL4-knockdown cells using siRNAs and shRNAs specifically targeting ANGPTL4. The siRNA-mediated transient knockdown of ANGPTL4 (siANGPTL4_1 and siANGPTL4_2) reduced gene expression by 60% to 80% over at least 72 hours as compared with cells transfected with a nontargeting control siRNA (siNTC; Fig. 6A). The shRNA-mediated knockdown (shANGPTL4_1 and shANGPTL4_2) led to a reduction in ANGPTL4 gene expression levels by 40% to 80% (Fig. 6B). For further characterization of the role of Angptl4 in cancer cell motility, we used the siRNA- and shRNA-modified cells with the highest knockdown efficiency (siANGPTL4_1 and shANGPTL4_1). Depletion of ANGPTL4 resulted in a significantly reduced secretion of Angptl4 protein by the cells (55%–90% less) as determined by ELISA assay (Fig. 6C and D). To analyze the impact of Angptl4-depletion on cancer cell motility, we performed Matrigel invasion assays. We observed that invasion of ANGPTL4-depleted MDA-MB-231 cells was markedly reduced compared with the control cells, especially following ACM-treatment (Fig. 6E and F). Furthermore, ANGPTL4-knockdown resulted in a significant decrease of wound healing capabilities of MDA-MB-231 cells compared with control cells in the presence of ACM (Fig. 6G). Of note, cell proliferation was not found to be affected by the knockdown of ANGPTL4 (Supplementary Fig. S10).

Ultimately, we sought to determine which molecular pathways could potentially be involved in Angptl4-mediated migration and invasion of TNBC cells. Recently, Shen and colleagues reported that oleic acid-induced Angptl4 activates FAK signaling in head and neck squamous cell carcinoma cells, thereby increasing cell migration and invasion (16). Therefore, we investigated, if FAK signaling could also be involved in the ACM-induced and Angptl4-mediated invasion and migration of TNBC cells. We observed that MDA-MB-231 cells responded to ACM_>40 treatment with higher phosphorylation levels of FAK at Tyr576/77 and Tyr397 as compared with cells in basal medium (Fig. 6H), indicating more active FAK signaling in ACM-treated cells. A gelatin zymography revealed a significant increase of MMP2, a downstream effector of FAK, in the cell culture supernatants of MDA-MB-231 cells precultivated with different ACMs compared with basal medium-treated cells, whereas the amount of MMP9 was heterogenous among the used ACMs (Supplementary Fig. S11). Interestingly, treatment of MDA-MB-231 cells with the FAK inhibitor PF-573228 drastically decreased cancer cell motility as detected by invasion and wound healing assays, especially in ACM-treated cells (Fig. 6I and J). Consequently, we explored if Angptl4 was involved in the increased phosphorylation of FAK. Indeed, we observed that ANGPTL4-depletion significantly reduced phosphorylation of FAK at Tyr576/77 in MDA-MB-231 cells treated with basal medium or ACM (Fig. 6K). Taken together, our findings indicate that upregulation of Angptl4 by ACM enhances FAK signaling, thereby increasing cell migration and invasion of TNBC cells.

Obesity is not only a known risk factor for cardiometabolic diseases, but also for several tumor types, including breast cancer (3, 4, 6). The current knowledge about the cellular and molecular mechanisms linking obesity and breast cancer progression is still limited and an area of active investigations. Most of the research work published to date focused on the impact of obesity on hormone-receptor positive breast cancer in postmenopausal women, as this is the most common breast cancer type (30–32). However, several retrospective studies indicate that the incidence of TNBC is increased in obese patients compared with nonobese patients (33–36). Furthermore, tumor size and tumor grade were found to be larger in obese patients with TNBC than in patients with normal weight (35–37). We hypothesized that factors secreted from obese fat tissue induce genes and signaling pathways that promote cancer progression in TNBC cells.

Thus, to investigate potential molecular and cellular changes in TNBC cells, induced by the interaction with adipose tissue, we co-cultured the TNBC cell line MDA-MB-231 with murine adipose tissue of obese mice. Subsequent to co-culture, global gene expression in MDA-MB-231 cells was profiled. Besides upregulation of genes involved in inflammation (e.g., IL1A and IL1B), we detected a strong upregulation of PPAR-target genes, such as PDK4, PLIN2, and ANGPTL4 in the cells co-cultured with adipose tissue. This upregulation was much more pronounced in cells treated with fat explants of obese mice as compared with explants of lean animals. Similar BMI-dependent changes in gene expression were observed in MDA-MB-231, HCC38, and E0771 cells treated with ACM of overweight and obese individuals. The induction of pro-inflammatory genes is concordant with our previous work demonstrating that co-culture with 3T3-L1 adipocytes enhances interleukin gene expression in MDA-MB-231 cells (28). To date numerous studies showed that interleukin-signaling is important in breast cancer progression and aggressiveness (38), whereas the role of PPAR signaling remains less clear. However, several studies demonstrated that PPAR nuclear hormone receptors are involved in regulating cancer cell proliferation, survival, and tumor growth (15, 39). Of note, the most upregulated gene in MDA-MB-231 cells treated with adipose tissue or ACM of obese individuals was the PPAR-target gene ANGPTL4. This gene has been proposed as a potential link between metabolism and cancer through PPAR signaling pathways (15). ANGPTL4 regulates lipid and glucose metabolism but was also implicated in tumor growth and metastasis, including in TNBC cells (14, 16, 17, 18, 34). Using agonists and antagonists specifically targeting the different PPAR subtypes, we identified PPARα as the subtype mainly involved in ACM-induced upregulation of ANGPTL4 gene and protein expression in MDA-MB-231 and HCC38 cells. This differed from results published by Adhikary and colleagues showing that PPARβ/δ was the main isoform driving ANGPTL4 expression in MDA-MB-231 cells (20). However, expression and abundance of the different PPAR subtypes are known to be dependent on tissue type and metabolic environment (40). Thus, in MDA-MB-231 and HCC38 cells treated with ACM, rich in various lipids, PPARα might be the major isoform responsible for the strong upregulation of ANGPTL4. Noteworthy, we have previously shown considerably higher FFA levels in ACMs prepared from fat tissue of obese individuals as compared with ACMs from fat tissue of overweight patients (25). Moreover, in the present study, induction of ANGPTL4 gene expression in MDA-MB-231 and HCC38 cells strongly correlated with the concentration of FFA of the respective ACMs used in co-culture. In addition, a comprehensive GC-MS analysis of the ACMs revealed that oleic, linoleic, and palmitic acid were the most abundant fatty acids in the conditioned media with ACM_>40 containing significantly higher concentrations of these fatty acids than ACM_<30. Thus, our results validate findings of other groups demonstrating that FFA provided by ACM induce ANGPTL4 expression in cancer cells by functioning as ligands for PPAR nuclear receptors (16, 20).

Treatment with FFA also altered lipid homeostasis in TNBC cells. For example, we found an augmented formation of LDs. These findings corroborate our previous work showing that treatment with ACM from fat tissue of obese patients leads to elevated formation of LDs in MDA-MB-231 cells (22). Interestingly, several studies report, that an elevated amount of LDs correlates with increased aggressiveness of mammary tumor cells (41, 42). Interestingly, LDs are sites of lipid storage and involved in the synthesis of eicosanoids (43, 44), known endogenous ligands of PPARs (45). In addition, here we demonstrate that MDA-MB-231 cells cultured with ACM show significantly higher levels of FAO and a downregulation of SREBP-mediated de novo fatty acid synthesis, substantiating the metabolic plasticity of aggressive breast cancer cells. Our results underpin recent reports demonstrating that cancer cells are able to take up, store and utilize exogenous fatty acids provided by conditioned medium or surrounding fat tissue (8, 9).

There is increasing evidence that adipose tissue secreted factors, including fatty acids, promote cancer progression, and metastasis (38, 46). Consistent with these observations, MDA-MB-231 and HCC38 cells treated with ACM showed increased rates of proliferation, migration and invasion. Cell invasion and migration was dependent on ANGPTL4 expression, as depletion of ANGPTL4 diminished the number of invaded cells. A recent paper by Shen and colleagues demonstrated that oleic acid-induced Angptl4 enhanced the invasive ability of head and neck squamous cell carcinoma cells by activating FAK signaling (16). Thus, the activation of FAK signaling, which is involved in cancer cell invasion and metastasis (15, 19, 20) may be another key factor in ACM-mediated breast cancer cell invasion. Indeed, we observed elevated activity of FAK signaling in MDA-MB-231 cells cultured with ACM, as indicated by increased FAK phosphorylation. Moreover, inhibiting FAK phosphorylation drastically reduced cell invasion and migration in response to ACM-treatment. Moreover, knockdown of ANGPTL4 resulted in decreased FAK phosphorylation levels in MDA-MB-231 cells, suggesting that Angptl4 is involved in activating FAK signaling in ACM-treated breast cancer cells.

In summary, our work demonstrates that adipose tissue secreted factors, especially from adipose tissue of obese individuals, promote metabolic reprogramming and a metabolically more active and more aggressive phenotype in TNBC cells (Fig. 7). Besides the activation of FAK signaling, the fatty acid-mediated induction of PPARs and their target gene ANGPTL4 play an important role in enhancing cell invasion of TNBC cells. Together our data support the hypothesis that breast cancer cells can adapt to and benefit from interaction with their environment by utilizing, for example, ILs and fatty acids provided by adipose tissue to support tumor growth and progression. Our results warrant exploring a combined inhibition of fatty acid uptake, Angptl4 and FAK signaling as a potential treatment strategy for TNBC, particularly in obese women.

A. Dietrich reports grants from study support (Humedics), study support (BMBF), and nonfinancial support from congress attention (BOWA, Ethicon) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

C. Blücher: Conceptualization, validation, investigation, visualization, writing-original draft. S. Iberl: Validation, investigation, visualization, writing-review and editing. N. Schwagarus: Investigation, visualization. S. Müller: Validation, investigation, writing-review and editing. G. Liebisch: Investigation, methodology. M. Höring: Investigation, methodology. M.S. Hidrobo: Investigation. J. Ecker: Investigation. N. Spindler: Resources. A. Dietrich: Resources. R. Burkhardt: Data curation, investigation, writing-review and editing. S.C. Stadler: Conceptualization, supervision, validation, investigation, visualization, writing-original draft, writing-review and editing.

We thank PD Dr. Knut Krohn from the Core Unit DNA Technologies of the Medical Faculty of Leipzig University, for his assistance and Prof. Antje Körner for providing access to the Seahorse XF24 Bioanalyzer. We also thank Birgit Wilhelm, Bärbel Schell, Cornelia Hasenknopf, Renate Kick, and Doreen Müller for technical assistance. This work was funded by means of the Deutsche Forschungsgemeinschaft DFG – Project number 209933838 – Collaborative Research Center 1052 “Obesity Mechanisms” (SFB-1052/B07 to R. Burkhardt) and by LIFE – Leipzig Research Center for Civilization Diseases, Leipzig University. LIFE is funded by means of the European Union, by the European Regional Development Fund (ERDF) and by means of the Free State of Saxony.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.