ACSL5 Mediates Adaptation to the Palmitic Acid-Enriched Pulmonary Microenvironment to Enhance Metastatic Breast Cancer Cell Survival and Lung Metastasis.

Introduction
Metastasis accounts for 90% of cancer-related deaths and remains a major challenge in tumor therapy (1). Distant metastasis follows a nonrandom pattern, termed “organotropism”; for instance, breast cancer preferentially metastasizes to the lung, bone, brain, and liver (2). Patients with metastases in different organs respond differently to treatments, raising an intriguing but unresolved scientific question (3, 4). The lung is a common site for breast cancer metastasis, yet the underlying mechanisms and treatment strategies for lung metastasis remain unclear.
Metabolic reprogramming is crucial for tumor metastasis. Primary cancer cells use diverse metabolic strategies, including alterations in lipid, glucose, amino acid, and nucleotide metabolism, to facilitate their dissemination to distant organs (5). However, the metabolic microenvironment of distant organs is different from that of primary tumors, implying that the disseminated tumor cells (DTC) are facing new energy stress. Metabolic adaptation is an emerging concept in organotropic metastasis. Each metastatic site exhibits unique metabolic characteristics, including variations in nutrient and oxygen availability as well as oxidative stress (6, 7). Notably, although metastatic dissemination generally occurs early in tumor development, the formation of clinically detectable metastases often takes years (8), suggesting that metastatic sites exhibit selective pressure and that DTCs must undergo corresponding adaptation to enable their survival and growth. Metabolic plasticity and flexibility of DTCs across different organs are vital for metabolic adaptation and metastasis formation in distant organs (9). Breast cancer cells show organotropic metastases by remodeling distinct metabolic pathways: breast cancer cells preferentially metastasize to the brain and lung by remodeling lipid metabolism pathways and oxidative phosphorylation; breast cancer cells have a tendency to metastasize to the liver by adjusting the amino acid metabolism pathways and glucose metabolism pathways and tend to metastasize to bone by altering mitochondrial metabolic pathways (10). Although clinical observations and experimental investigations have unveiled the metabolic landscape of DTCs from distinct organ metastases in some tumors (11, 12), understanding of the metabolic landscape of the lung metastatic niche in patients with breast cancer remains inadequate.
Our study delineated a significant lipid metabolism feature in lung metastases shared by multiple cancer types, including breast cancer, renal cell carcinoma, colorectal cancer, and osteosarcoma. Furthermore, we demonstrated that palmitic acid (PA) was enriched in the lung metastatic niche and was essential for the lung metastasis. Elevated PA levels in the lung metastatic niche were identified as originating from lung-resident alveolar type II (AT2) cells, consistent with the results of a recent study conducted by Altea-Manzano and colleagues (13). Mechanistically, we found that AT2 cells were regulated by exosomal USP47 derived from lung metastatic breast cancer cells in a YAP-dependent manner, thereby activating fatty acid synthesis in AT2 cells. Moreover, the remodeling of prostaglandin metabolism mediated by ACSL5 contributed to PA adaptation, which subsequently facilitated the survival and proliferation of metastatic breast cancer cells in the lung. Mechanistically, ACSL5high LM3 cells mediated the COX2-induced PGE2 accumulation in a PA-dependent manner, subsequently activating PI3K/AKT and ERK signaling via the PGE2 receptor EP4. In addition, we identified that the combination of paclitaxel (PAC) with the genetic inhibition of ACSL5, celecoxib (a selective COX2 inhibitor), or ONO-AE3-208 (a selective EP4 inhibitor) significantly reduced lung metastasis in mice. Dietary PA decreased PAC sensitivity in mice with lung metastases. Taken together, these findings highlight the importance of PA and ACSL5-dependent PA adaptation in regulating breast cancer lung metastasis and provide a promising therapeutic strategy for patients with breast cancer with lung metastasis.

Materials and Methods

Cell culture
Triple-negative breast cancer cell lines 4T1, MDA-MB-231, and HCC1806 and mouse alveolar epithelial cell line MLE-12 were purchased from the ATCC. Utilized for less than 6 months and kept within 20 passages, all cell lines were authenticated by means of short tandem repeat profiling. Furthermore, these cell lines were devoid of Mycoplasma contamination and underwent monthly PCR analysis for the purpose of authentication. MDA-MB-231 cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM; Gibco), supplemented with 10% fetal bovine serum (FBS; HyClone). 4T1 and HCC1806 cells were cultured in RPMI 1640 medium (Gibco), supplemented with 10% FBS. Breast cancer cells transfected with lentivirus expressing GFP and firefly luciferase (Luc) were used for organotropic metastasis cells screening in mice. Organotropic metastatic cells, namely LM3 (breast cancer cells with lung-preferential metastasis), HM3 (breast cancer cells with liver-preferential metastasis), and BM3 (breast cancer cells with brain-preferential metastasis), were obtained after at least three rounds of in vivo selection, as described previously (14). GFP was used to isolate the cancer cells, whereas Luc was used for bioluminescence imaging (BLI). MLE-12 cells were cultured in MLE-12’s cell culture medium (ServiceBio). For the primary culture of AT2 cells, sorted CD326+MHCII+CD49flow AT2 cells were cultured in the complete DMEM medium (supplemented with 10% dialyzed FBS, 2 mmol/L l-glutamine, 1 × MEM NEAA, and 25 mmol/L HEPES). All cells were supplemented with 100 U/mL penicillin and streptomycin (Beyotime) and cultured at 37°C in a 5% CO2 incubator.

In vivo mouse experiments
Female BALB/c mice and BALB/c nude mice (4–6 weeks old) were purchased from Hua Fukang Co. All animal experiments were performed in accordance with the Guidelines for the Care and Use of Laboratory Animals, and the experimental protocols were reviewed and approved by the Ethics Committee of Chongqing Medical University (reference number 2021084).
For lung metastasis assays, breast cancer cells or their derivatives (1 × 106/mouse) suspended in 100 μL PBS were orthotopically injected into the mouse mammary fat pad or injected into mice via the tail vein.
For fatty acid treatment, tumor sizes were observed every 3 days; when the primary tumor could be observed, mice were administered with PA (10 mg/kg, once daily, orally), stearic acid (SA; 10 mg/kg, once daily, orally), linoleic acid (LA; 10 mg/kg, once daily, orally), or oleic acid (OA; 10 mg/kg, once daily, orally) purchased from MCE.
For drug treatment, mice (20–25 g, ages 6–8 weeks) were intravenously injected with indicated breast cancer cells. Seven days after injection, the mice were administered with PGE2 (40 μg/day, i.p.; MCE), PAC (5 mg/kg, i.p.; MCE), celecoxib (5 mg/kg, i.p.; MCE), ONO-AE3-208 (10 mg/kg, orally; MCE), LY294002 (25 mg/kg, i.p.; MCE), or SCH772984 (5 mg/kg, i.p.; MCE).
For BLI, all mice were euthanized after administering intraperitoneal injection of D-luciferin (150 mg/kg, i.p.; Sigma) for 10 minutes. Thereafter, tissues and organs were rapidly dissected and used for ex vivo BLI using a small animal living imaging system (Berthold LB983). The BLI densities were analyzed using IndiGo (version 2.0.5.0).
For Bouin fixation, lung tissues were collected and immediately stained with Bouin fixative solution (Phygene Scientific) for 30 minutes. The lung nodules were counted. For paraffin embedding, the lungs were gently washed with Dulbecco’s Phosphate-Buffered Saline (DPBS; Biosharp) and transferred to 75% ethanol to remove the fixative solution.
Mice used in in vivo mouse experiments were euthanized when weakness or loss of body weight (70% of control littermates) was observed.

RNA preparation, qRT-PCR, and RNA sequencing
Total RNA was extracted using the RNAiso Plus reagent. Reverse transcription was performed using the PrimeScript RT Reagent Kit. mRNA levels were assessed by qRT-PCR using SYBR Green qPCR Master Mix (MCE) and analyzed on a CFX96 Real-time PCR Detection System (Bio-Rad). β-Actin was used as an internal control for normalizing mRNA levels. Details of the primers used are provided in Supplementary Table S1. RNA sequencing (RNA-seq) was conducted by Shengyin Biotech and Obio Technology.

Single-cell suspension preparation and single-cell RNA-seq
Fresh lung tissues and primary tumors were harvested and washed twice with DPBS. The tissues were chopped into pieces and dissociated using a Tumor Dissociation Kit (Miltenyi Biotec). Cell suspensions were filtered through 70- and 40-μm cell strainers (Biosharp), lysed with Red Blood Cell Lysis Solution (Biosharp), and washed with DPBS containing 2% FBS. GFP+ tumor cells were sorted using fluorescence-activated cell sorting (FACS) for single-cell RNA-seq (scRNA-seq). The library was constructed using the Chromium Single Cell 3′ Reagent Version 2 Kit and the Chromium Controller (10× Genomics). Using the GemCode technology, the single cells, reagents, and gel beads (containing barcoded oligonucleotides) were encapsulated into nanoliter-sized gel beads in emulsion (GEM). Lysis and barcoded reverse transcription of polyadenylated mRNA from individual cells was conducted inside each GEM. After RT, GEMs were cleaned, and cDNA was subsequently amplified. cDNA was fragmented, end-repaired, and A-tailed. The adapters were ligated to the fragments, followed by double-sided SPRI selection. A second round of double-sided SPRI selection was performed after the sample index PCR. Fragment size distribution was analyzed using an Agilent 2100 Bioanalyzer, and the library was quantified by qRT-PCR using a TaqMan probe. Finally, scRNA-seq was conducted using the MGISEQ 2000 platform (BGI).

Lipid analysis–mass spectrometry
All harvested lung tissues from mice were collected from three stages: normal, premetastatic, and metastatic stages. The mouse models at premetastatic and metastatic stages were established as previously described (15). The lung tissues were immediately washed with DPBS, excess water was removed using Whatman filter paper, weighed on dry ice, minced into little pieces, briefly immersed in liquid nitrogen for 10 seconds, and then stored at −80°C. Subsequently, all lung tissues were subjected to quantitative targeted lipidomic analysis using BIOTREE Biotechnology. Lipid concentrations were quantified using Biobud-v2.1.4.1 software, normalized to tissue wet weight, and calculated based on internal standard peaks and concentrations.

Preparation of conditioned medium
Conditioned medium (CM) was obtained by culturing MDA-MB-231 and HCC1806 cells in DMEM or RPMI 1640 medium (with 10% FBS), respectively. All cells were washed with PBS (three times) when cell confluency reached approximately 80% and were subsequently cultured in fresh FBS-free medium for 48 hours. The supernatant was centrifuged (2,000 g, 4°C) for 10 minutes and filtered using a 0.22-μm cell filter (Biosharp) to remove cell debris. The CM in each group was normalized to the ratio of CM volume to cell numbers (16) and was used for further experiments.

Flow cytometry
To identify cells contributing to PA enrichment in the metastatic niche, normal lungs (NL), premetastatic lungs (PL), and macrometastatic lungs (ML) were collected from mice. Single-cell suspensions were prepared as described above. Flow cytometry was used to isolate CD45+ immunocytes, CD31+ endothelial cells, CD140a/b+ fibroblasts, CD326+ epithelial cells, CD326+MHCⅡ+CD49flow AT2 cells, CD326+MHCⅡ‒ non-AT2 cells, and GFP+ tumor cells. First, cells (1 × 106 cells/100 μL) were resuspended in DPBS with 2% FBS (FACS buffer). Subsequently, the cells were incubated with an anti-mouse CD16/32 (BioLegend) antibody on ice for 10 minutes. Next, the respective antibodies were used and incubated on ice for 30 minutes. Then, cells were gently washed three times with FACS buffer and resuspended in FACS buffer containing 4′, 6-diamidino-2-phenylindole (DAPI; BioLegend). Flow cytometry was performed using the FACSCanto plus machines (BD) and analyzed using FlowJo v10. The following antibodies were used: CD45 (BioLegend), CD31 (BioLegend), CD326 (BioLegend), CD140a (BioLegend), CD140b (BioLegend), CD49f (BD), and MHCⅡ (BioLegend). Antibodies and reagents used in this study are listed in Supplementary Table S2.
For Annexin V–APC/propidium iodide (PI) staining, breast cancer cells (1 × 106) were analyzed using an Annexin V–APC/PI apoptosis detection kit (YEASEN).

Exosome isolation and characterization
CM harvested after 48 hours culture in exosome-free FBS was sequentially centrifuged (2,000 g 20 minutes, 10,000 g 30 minutes, and 12,000 g 30 minutes, 4°C) to remove cellular debris. Furthermore, exosomes were harvested by ultracentrifugation (120,000 g, 70 minutes, 4°C) and washed with PBS. The pellets were resuspended in an appropriate volume of PBS and used for further study. Western blotting was conducted to examine exosome-positive markers, such as CD9 (Abcam), CD63 (Bimake), TSG101 (Millipore), and HSP70 (Abcam), and an exosome-negative marker, Calnexin (Abcam). Exosome size and concentration were measured by nanoparticle tracking analysis (NTA) using ZetaView PMX 110 (Particle Metrix). Exosome particles were detected using Brownian motion and the diffusion coefficient. Data were analyzed using ZetaView 8.02.28. All experiments were performed in triplicates. Exosomes were imaged using a JEM-1011 transmission electron microscope (TEM; Hitachi).

Western blotting
Cells were lysed in RIPA buffer (Beyotime) containing phosphatase and protease inhibitors. Proteins were quantified using a bicinchoninic acid kit (Beyotime), separated by 10% SDS-PAGE, transferred to polyvinylidene difluoride (PVDF) membranes (Bio-Rad), blocked with defatted milk, and then probed with primary antibodies. After 1.5 hours of room temperature incubation with horseradish peroxidase (HRP)-conjugated secondary antibodies, the PVDF membranes were visualized with enhanced chemiluminescence substrate detection reagents (Bio-Rad). The following primary antibodies were used: USP47 (SAB), FASN (Abcam), ACLY (Abcam), YAP (SAB), ACSL5 (SAB), COX2 (HUABIO), AKT (Abmart), pAKT (Ser473; Abmart), PI3K (Abmart), pPI3K p85-alpha/gamma (Tyr467/199; Abmart), ERK1/2 (Abmart), pERK1/2 (Thr202/Tyr204; Abmart), and β-catenin (Abmart).

Immunofluorescence staining
To evaluate the internalization of exosomes by AT2, breast cancer cell–derived exosomes were labeled with 10 mmol/L PKH26 (Sigma-Aldrich) prior to incubation with AT2 cells. The labeled exosomes were washed, collected using ultracentrifugation, and resuspended in PBS. Incubated with labeled exosomes at a concentration of 10 mg/mL for approximately 12 hours, the AT2 cells were fixed (4% paraformaldehyde, 15 minutes), permeabilized (0.1% Triton X-100, 15 minutes), blocked (goat serum, 30 minutes), and incubated with primary antibody overnight at 4°C. Then the cells were incubated with fluorescent secondary antibody for approximately 1 hour and stained using DAPI (Beyotime). The internalized exosomes in AT2 cells were visualized using a confocal microscope (Leica).

Colony formation assay
Breast cancer cells (1 × 103/well) were incubated into six-well plates at 5% CO2 incubator for 14 days. Thereafter, formed colonies were fixed with methanol (10 minutes), washed with PBS, and then stained with crystal violet for 10 minutes (Beyotime). The colonies were scanned using a digital camera and counted.

Cell viability assay
Cell viability was evaluated using Cell Counting Kit-8 (CCK8; YEASEN). Breast cancer cells were incubated in 96-well plates (2 × 103/well) in a 5% CO2 incubator and treated with the indicated reagents. CCK8 solution was added and cultured for approximately 2 hours, and the optical density (OD) at 450 nm was measured using a microplate reader (BioTek). All experiments were performed in triplicate.

Immunohistochemistry
Paraffin-embedded sections (4 μm) were deparaffinized, rehydrated, and boiled in citrate buffer (100°C, 15 minutes), blocked for endogenous peroxidase, and then incubated overnight with anti-ACSL5 at 4°C. After washing with PBS, the slides were incubated with the HRP secondary antibody (30 minutes) at room temperature, stained using a diaminobenzidine kit (Beyotime), and photographed. The immunohistochemistry (IHC) results were evaluated by two pathologists. IHC staining intensity was scored as 0 (none), 1 (weak), 2 (moderate), and 3 (strong). The percentage of stained cells was scored as 0 (none), 1 (<10%), 2 (10%–50%), and 3 (>50%). The IHC score was derived by multiplying these two scores, and samples with a score more than 4 were defined as having high ACSL5 expression.

Lentivirus-mediated RNA interference
The short hairpin RNAs (shRNA) specifically against ACSL5, USP47, COX2, EP4, Fasn, Acly, Yap1, and control hairpins were cloned into a pLV-Puro vector. Cells were infected with lentivirus for 24 hours. The shRNA sequences used in this study are listed in Supplementary Table 3. Lentivirus-mediated ACSL5, COX2, USP47, Fasn, Acly, and Yap1 overexpression constructs and the control vector were transfected into the indicated cells. The infected cells were selected using puromycin for 2 weeks. The lentivirus was acquired from GenePharma.

Transwell migration assay
The culture medium (500 μL) was added to the lower wells, and breast cancer cells (2 × 104) in 250 μL serum-free medium were incubated into the upper wells of 8 µm pore chambers (Millipore). Migrated cells were fixed (4% formaldehyde, 15 minutes), stained (0.5% crystal violet, 10 minutes), photographed, and counted.

Mammosphere formation assay
Breast cancer cells were seeded in six-well plates coated with 2% poly-2-hydroxyethyl methacrylate (poly-HEMA; Sigma-Aldrich) at a concentration of 1 × 104 cells/mL (5 × 103 cells/mL in the following passages) with regular passages every 5 days. Cells were grown in serum-free DMEM/F12 (Gibco) medium with 2% B27 (Gibco), 20 ng/mL epidermal growth factor (Invitrogen), 20 ng/mL basic fibroblast growth factor (Invitrogen), 2 μg/mL heparin (Sigma-Aldrich), insulin–transferrin–selenium (Invitrogen), and 0.4% bovine serum albumin (Sigma-Aldrich). The mammospheres were photographed under an OLYMPUS microscope.

Quantitative determination of PGE2
Lung interstitial fluid and culture medium were collected and analyzed using the Prostaglandin E2 Assay Kit (R&D Systems) according to the manufacturer’s instructions. Briefly, samples were centrifuged at 1,000 g for 15 minutes and added to a microplate. The plate was treated with primary antibody solution and incubated at room temperature for 1 hour with shaking, followed by the addition of PGE2 conjugate and incubation for 2 hours. After four washes with buffer, substrate solution was added and incubated for 30 minutes in the dark. OD was measured at 450 nm within 30 minutes after adding stop solution.

Clinical breast cancer samples
Human breast tumors and metastatic lung tissues were collected from patients at the First Affiliated Hospital of Chongqing Medical University. Written informed consent was obtained from all patients. The study was approved by the Ethics Committee of Chongqing Medical University (reference number 2021084) and followed the principles of the Declaration of Helsinki.

Statistical analysis
All in vitro experiments were repeated at least three times independently. GraphPad Prism (v8.2.1) was used for statistical analyses. Student t test, Pearson correlation test, Mann–Whitney test, one-way and two-way ANOVA, Kruskal–Wallis test, and log-rank test were conducted. Data are presented as the means ± SD. P < 0.05 indicates significance.

Results

Lung metastatic breast cancer cells display increased lipid metabolism
Metabolic plasticity allows DTCs to dynamically adapt to various organ-specific environments (17). To elucidate the metabolic characteristics of breast cancer lung metastasis, we analyzed RNA-seq data from different metastatic sites in patients with breast cancer using the Gene Expression Omnibus (GEO) dataset. Lung metastases demonstrated increased lipid metabolism, whereas nonpulmonary metastases exhibited distinct metabolic signatures: bone metastases were enriched in thiamine and nitrogen metabolism, brain metastases in glycosaminoglycan pathways, and liver metastases in amino acid metabolism (Supplementary Fig. S1A). Compared with paired primary breast tumors, lung metastases also exhibited increased lipid metabolism (Supplementary Fig. S1B and S1C). Notably, activated lipid metabolism was observed in pulmonary metastases of renal cell carcinoma (Supplementary Fig. S1D), colorectal cancer (Supplementary Fig. S1E), and osteosarcoma (Supplementary Fig. S1F), indicating that this metabolic signature may be a common feature in pulmonary metastasis across various tumor types.
To investigate the role of lipid metabolism in breast cancer lung metastasis, we derived organ-preferential metastatic cell lines through in vivo selection (14) and analyzed public gene expression datasets and our mRNA-seq data. Lung metastatic E0771, MDA-MB-231, and HCC1806 cells exhibited enhanced lipid metabolism (Supplementary Fig. S2A and S2B). Given the intratumoral heterogeneity in breast cancer (18), we performed scRNA-seq on freshly isolated primary tumor cells (PRI) and LM3 (breast cancer cells with lung-preferential metastasis) from BALB/c nude mice to evaluate metabolic heterogeneity. We identified three distinct clusters (Supplementary Fig. S2C), with cluster T3 nearly absent in lung metastases (Supplementary Fig. S2C and S2D). Subsequent bulk mRNA-seq and Kyoto Encyclopedia of Genes and Genomes analysis revealed that lung metastatic T1T2 (designated LT1T2) were enriched in metabolic pathways and survival advantages (Supplementary Fig. S2E). Consistent with these findings, LT1T2 cells showed proliferative and survival advantages (Supplementary Fig. S2F–S2I). Notably, lipid metabolism activation was negatively correlated with apoptosis but positively correlated with the cell cycle (Supplementary Fig. S2J). These findings unveil that lung metastatic breast cancer cells possess increased lipid metabolism and metastatic outgrowth-promoting advantages.

Enrichment of PA in the metastatic niche promotes breast cancer lung metastasis
Nutrients, such as glucose, pyruvate, fatty acids, and glutamine, support the seeding and colonization of DTCs in distant organs (10). To explore the lipid dynamics during pulmonary metastasis, we injected MDA-MB-231/LM3 cells into BALB/c nude mice and used LC-MS/MS to analyze altered lipid metabolites in NL, PL, and ML. We found that PA, OA, LA, and SA (both free and as acyl chains) were enriched in tissue (Fig. 1A) and the interstitial fluid of PL and ML (Fig. 1B) compared with NL. Notably, PA levels were significantly higher in the interstitial fluid from lung metastases in patients with breast cancer (Fig. 1C). Cholesterol esters were enriched in ML, whereas lysophosphatidylcholines and lysophosphatidylethanolamines were abundant in NL (Fig. 1D). We then investigated whether increased availability of fatty acids correlates with enhanced metastasis. Feeding mice with PA, OA, LA, and SA, we discovered that PA, but not OA, LA, or SA, increased lung metastatic burden in spontaneous BALB/c nude mice (Fig. 1E and F). Thus, a PA-rich pulmonary environment promotes DTC lung metastasis.

Exosome-derived USP47 mediates PA synthesis in AT2
To explore the source of PA enrichment in the PL niche, we isolated CD45+ immunocytes, CD31+ endothelial cells, CD140a/b+ fibroblasts, and CD326+ epithelial cells from PL and ML of BALB/c nude mice injected with MDA-MB-231/LM3, using the NL cells as controls (Supplementary Fig. S3A). Analysis revealed that CD326+ epithelial cells, which mainly include AT1 and AT2 cells in NL and PL but also tumor cells in ML, exhibited elevated PA levels at different metastatic stages (Fig. 2A). Notably, PA levels progressively increased from PL to the ML (Fig. 2B). Given that AT2 cells are the predominant epithelial cells in the lung and secrete lipids containing palmitic acyl chains, which constitute 90% of the pulmonary surfactant (19), we hypothesized that AT2 contributed to PA enrichment. Further comparison of freshly isolated CD326+MHCⅡ+CD49flow-AT2 cells, CD326+MHCⅡ− non-AT2 cells, and GFP+ tumor cells confirmed that AT2 cells exhibited significantly higher PA levels than non-AT2 cells and tumor cells at various metastatic stages (Fig. 2C). PA levels also progressively increased from PL to ML (Fig. 2D). In summary, lung-resident AT2 cells drive PA enrichment in the metastatic lung niche.
To determine whether LM3 cell–derived factors create a PA-enriched niche, we measured PA levels in tumor-derived CM (TCM) from PRI, HM3, BM3, and LM3 cells. No significant differences in PA levels were found among these groups (Supplementary Fig. S3B), suggesting that metastatic tumor cells do not directly secrete PA. We then investigated whether AT2 cells could respond to TCM. Treatment of primary isolated AT2 cells with LM3-CM enhanced PA production (Fig. 2E) and upregulated key genes involved in de novo fatty acid synthesis (Fig. 2F), including Fasn and Acly. Similar findings were observed in MLE-12 cells, a well-established model for studying AT2 cells (Supplementary Fig. S3C and S3D). We subsequently isolated exosomes from TCM, characterized them by NTA, TEM, and Western blotting (Supplementary Fig. S3E–S3G), and identified 120 upregulated proteins in LM3-derived exosomes (Supplementary Fig. S3H) involved in metabolic pathways (Fig. 2G). In parallel, analysis of the global secretome identified 27 differentially expressed secretory proteins in MDA-MB-231/LM3 and HCC1806/LM3 cells compared with PRI; however, these candidates showed no significant association with lipid metabolism (Supplementary Fig. S3I). Of these 120 upregulated proteins, USP47, highly expressing in LM3-derived exosomes (Fig. 2H) and potentially regulating SREBPs (key fatty acid synthesis regulators; ref. 20), was of particular interest. Efficient exosome internalization was observed in AT2 cells incubated with PKH26-labeled exosomes from LM3 (LM3-exo), HM3 (HM3-exo), BM3 (BM3-exo), and PRI (PRI-exo) cells. Notably, USP47 expression was elevated only in AT2 cells treated with LM3-exo (Fig. 2I). FASN and ACLY levels were significantly increased in AT2 cells isolated from metastatic lungs and MLE-12 cells cocultured with LM3-exo (Fig. 2J). These findings unveil that LM3-exo promotes PA synthesis in AT2 cells.
To further explore whether LM3-exo–derived USP47 mediates PA synthesis, we knocked down USP47 in LM3 cells, reducing its levels in LM3-exo, whereas ectopic overexpression of USP47 increased its levels in PRI-exo (Supplementary Fig. S3J and S3K). A previous study reported that USP47 activated YAP signaling, which regulates fatty acid synthesis. As expected, knockdown of USP47 in LM3-exo decreased FASN, ACLY, YAP, and PA levels in AT2 cells, whereas USP47 overexpression in PRI-exo increased these levels (Fig. 3A and B). YAP knockdown in MLE-12 cells significantly reduced endogenous and USP47-elicited FASN, ACLY, and PA levels (Fig. 3C and D). Ectopic YAP partially rescued the decrease in fatty acid synthesis–related gene expression and PA levels in AT2 cells caused by USP47-deficient exosome (Fig. 3E). To test whether FASN and ACLY can rescue the effects of USP47 on YAP signaling, ACLY or FASN was knocked down or overexpressed in the MLE-12 cells. Knockdown of ACLY or FASN decreased YAP-elicited PA levels in MLE-12 cells treated with shUSP47 exosomes (Fig. 3F), whereas overexpression of ACLY or FASN increased the PA levels in YAP-knockdown MLE-12 cells treated with oeUSP47 exosomes (Fig. 3G). Furthermore, mice injected with shUSP47-LM3 cells had reduced PA levels and metastatic burden in the lungs compared with those injected with shNC-LM3 cells (Fig. 3H–J). Collectively, exosomal USP47 from lung metastatic breast cancer cells promotes PA synthesis in AT2 cells via YAP signaling, contributing to breast cancer lung metastasis.

ACSL5-dependent adaptation to PA promotes cell survival and lung metastasis
To investigate the role and underlying mechanisms of PA in lung metastasis, we analyzed 1,660 metabolism-related genes using RNA-seq. Eight metabolism-related genes were commonly upregulated in lung metastases compared with their corresponding MDA-MB-231, HCC1806, and T3 cells isolated from PRI (Fig. 4A). Of these, ACSL5, a member of the acyl-conenzyme A (CoA) synthetase family, attracted our attention. ACSL5 catalyzes the activation of fatty acids like PA, OA, and LA (21), and ACSL5-knockout mice exhibit a 60% reduction in palmitoyl-CoA levels (22). Notably, ACSL5 expression was high in LM3 cells (Fig. 4B; Supplementary Fig. S4A–S4C, left) and clinical lung metastases (Fig. 4C–E), and its expression progressively increased in lung-preferential metastatic breast cancer cells (Fig. 4B; Supplementary Fig. S4A, right). Furthermore, we observed a trend toward poorer overall survival in patients with breast cancer with higher ACSL5 expression in lung metastases (Fig. 4F). To strengthen these observations, ACSL5 was silenced in LM3 cells and stably overexpressed in PRI cells (Supplementary Fig. S4D). ACSL5 overexpression in PRI facilitated lung metastasis, whereas ACSL5 knockdown in LM3 significantly reduced metastases (Fig. 4G; Supplementary Fig. S4E). However, ACSL5 did not affect metastasis to other organs, including the brain, liver, spleen, and kidney (Supplementary Fig. S4F and S4G). Ectopic ACSL5 expression in HM3 and BM3 also increased lung metastatic burden (Supplementary Fig. S4H and S4I). Taken together, these results indicate that ACSL5 mediates lung-preferential metastasis of breast cancer cells.
Surprisingly, ACSL5 knockdown did not significantly affect migration, mammosphere formation, colony formation, proliferation, or survival of breast cancer cells in vitro (Supplementary Fig. S5A–S5E), suggesting that ACSL5-mediated lung metastasis is influenced by specific lung environmental factors. Given that PA is highly enriched in the lung metastatic niche and serves as a ACSL5 substrate, we hypothesize that ACSL5 relies on PA to facilitate lung metastasis. To test this hypothesis, ACSL5-knockdown LM3 cells and their control counterparts were inoculated into the fat pads of mice subjected to either conventional feeding or PA feeding. PA feeding remarkably increased the lung metastatic burden in mice bearing ACSL5high LM3 cells (Fig. 5A and B; Supplementary Fig. S5F) but not in other organs (Supplementary Fig. S5G). Subsequently, we demonstrated that the exposure of LM3 cells to PA resulted in decreased cell survival; however, ACSL5high LM3 cells rapidly adapted to the high-PA environment than ACSL5-deficient LM3 cells (Supplementary Fig. S5H and S5I), enhancing their survival and proliferation in a PA-enriched environment (Fig. 5C–F). All subsequent in vitro studies were therefore performed under PA-adapted conditions. Comparable results were observed in ectopic ACSL5-expressing PRI cells (Fig. 5G–J). These findings indicate that ACSL5-dependent adaptation to PA enhances metastatic breast cancer cells survival and proliferation, fueling lung metastasis.

PA-ACSL5–induced COX2 expression and PGE2 accumulation activate EP4 to foster DTC survival and proliferation through PI3K/AKT and ERK signaling
To investigate the role of ACSL5-dependent adaptation to PA, we performed bioinformatics analysis on differential expressed genes in ACSL5 wild-type (WT) and ACSL5-deficient LM3 cells treated with or without PA. We identified several PA-dependent metabolic pathways in ACSL5high LM3 cells, with arachidonic acid (AA) metabolism being the most prominent (Supplementary Fig. S6A). This is consistent with previous observation of robust AA metabolism in lung metastases of breast cancer and other malignancies (Supplementary Figs. S1A–S1F and S2B). In patients with breast cancer, genes associated with AA metabolism were enriched in lung metastases with higher ACSL5 expression (Supplementary Fig. S6B). Of these AA genes, PTGS2 (encoding COX2) was identified as a highly expressed, PA-ACSL5–dependent gene (Fig. 6A). Elevated COX2 levels in lung metastases are associated with reduced lung metastasis–free survival in patients with breast cancer (Fig. 6B). COX2 catalyzes the conversion of AA into prostanoids (23, 24). Of these prostanoids, PGE2 plays a predominant role in various tumors, including lung, colorectal, breast, and head and neck cancers (24–27). Nonetheless, the involvement and mechanism of COX2-driven PGE2 production in breast cancer lung metastasis remain largely unknown. Interestingly, elevated PGE2 levels were detected in lung metastases of patients with breast cancer (Fig. 6C). To determine whether PA-ACSL5 promotes lung metastasis via COX2-induced PGE2 accumulation, we measured COX2 in ACSL5-WT and -deficient LM3 cells treated with or without PA, and PGE2 levels in indicated culture medium were detected as well. ACSL5high LM3 cells showed significant PA-dependent increase in COX2 (Fig. 6D) and PGE2 levels (Fig. 6E). Similar results were observed in lung metastatic breast cancer cells isolated from mice orthotopically injected with shNC LM3 cells compared with shACSL5 LM3 cells (Supplementary Fig. S6C and S6D). In vitro (Supplementary Fig. S6E and S6F) and in vivo (Supplementary Fig. S6G and S6H) studies with ectopic ACSL5 expression in PRI cells also confirmed these findings. Using COX2-knockdown LM3 cells and ectopic COX2-overexpressing PRI cells (Supplementary Fig. S7A and S7B), we confirmed that the loss of COX2 resulted in decreased PGE2 in vitro (Supplementary Fig. S7C) and in vivo (Supplementary Fig. S7D), along with decreased cell survival and proliferation in LM3 cells (Supplementary Fig. S7E–S7J). Conversely, COX2 overexpression results in the opposite phenotypes (Supplementary Fig. S7C–S7J). Restoring COX2 expression or supplemented exogenous PGE2 in ACSL5-deficient cells increased cell survival and proliferation (Fig. 6F–J). These data confirmed that ACSL5 induces COX2 expression and PGE2 accumulation to enhance cell survival and proliferation in a PA-enriched environment.
PGE2 work by binding to G protein–coupled receptors (EP1, EP2, EP3, and EP4; ref. 24). RNA-seq data revealed that EP4 is the predominantly expressed receptor in metastatic breast cancer cells (Supplementary Fig. S8A, left), breast cancer tissues from the TCGA dataset (Supplementary Fig. S8A, right), and clinical breast cancer lung metastases (Fig. 7A; and Supplementary Fig. S8B). Treating LM3 cells with PGE2 and EP1/2 inhibitor AH6809 (EP1/2i), EP3 inhibitor ONO-AE3-240 (EP3i), or EP4 inhibitor ONO-AE3-208 (EP4i) revealed that ONO-AE3-208 significantly inhibited PGE2-mediated survival and proliferation (Fig. 7B–C, and Supplementary Fig. S8C and S8D). Collectively, these data demonstrate the pivotal role of EP4 in breast cancer lung metastasis.
PGE2–EP4 has been implicated in the regulation of multiple signaling pathways related to cellular proliferation and survival, including the PI3K/AKT, ERK, and β-catenin pathways (26, 28). To elucidate the downstream signaling pathways that contribute to the oncogenic role of the PA-ACSL5/COX2/PGE2–EP4 axis in breast cancer, we assessed the activation of these pathways under the PA-enriched condition. Whereas β-catenin signaling was not affected by ACSL5 or COX2 knockdown (Fig. 7D and E), loss of ACSL5 or COX2 markedly reduced the pPI3K, pAKT, and pERK levels, whereas overexpression increased these levels (Fig. 7D and E). EP4 knockdown or ONO-AE3-208 treatment inhibited the PI3K/AKT and ERK pathways (Fig. 7F; Supplementary Fig. S8E). In breast cancer cells overexpressing ACSL5, PI3K/AKT and ERK signaling pathways were inhibited by COX2 knockdown, whereas COX2 overexpression or PGE2 supplementation attenuated the inactivation of these signaling pathways caused by ACSL5 silencing (Fig. 7G; Supplementary Fig. S8F). Furthermore, the ACSL5/COX2/PGE2–EP4 axis, along with PI3K/AKT and ERK signaling, was shown to be critical for breast cancer lung metastasis in mice (Fig. 7H; Supplementary Fig. S8G–S8I).
PA cannot be directly converted into AA to boost PGE2. Instead, PA may aid COX2-mediated PGE2 increase in LM3 cells via other processes. PA can enter mitochondria for β-oxidation, providing energy and acetyl-CoA for acetylation processes (29), and can also modify cysteine through palmitoylation (30). Both acetylation and palmitoylation are crucial for regulating gene expression. Our RNA-seq analysis revealed no significant changes in genes related to fatty acid β-oxidation but significant upregulation of several palmitoyl acyltransferases (ZDHHCs) in LM3 cells (Supplementary Fig. S9A), particularly in a PA-ACSL5–dependent manner (Supplementary Fig. S9B). This suggests that palmitoylation may regulate COX2 expression. ACSL5 has been shown to mediate the palmitoylation of proteins (e.g., WNT2B and WNT3A) and transcription factors potentially modified by palmitoylation (e.g., P23 and STAT5A), as well as transcription coactivators (e.g., P38, P300, and SP1), which may be involved in COX2 regulation (30). Therefore, we investigated the role of palmitoylation in COX2 expression and found that 2-BP, a palmitoylation inhibitor, effectively reduced COX2 levels in LM3 cells treated with PA (Supplementary Fig. S9C). Furthermore, ACSL5 knockdown or treatment with 2-BP significantly decreased COX2 levels in LM3 cells treated with PA (Supplementary Fig. S9D).
Our findings revealed an activated ACSL5–COX2–EP4 axis in lung metastatic breast cancer cells, which was also confirmed in several lung metastatic tumors. Elevated COX2 levels were observed in lung metastases of breast cancer, colon cancer, osteosarcoma, renal carcinoma, and pan-cancer from MET500 datasets compared with primary tumors or other metastases (Supplementary Fig. S10A). Increased EP4 levels were also observed in lung metastases of breast cancer and pancreatic cancers (Supplementary Fig. S10A). A positive correlation exists between ACSL5 and COX2 levels in patients with breast cancer, sarcoma, and osteosarcoma (Supplementary Fig. S10B), and between ACSL5 and EP4 levels in patients with breast cancer, colon cancer, and head and neck squamous cell carcinoma (Supplementary Fig. S10C). These clinical data suggest that ACSL5–COX2–EP4 axis activation in lung metastases may be a common event across various cancer types, warranting further validation.

Limiting PA intake and targeting the ACSL5–COX2–EP4 axis increase sensitivity of lung metastatic breast cancer to chemotherapy
Given the critical roles of the ACSL5–COX2–EP4 axis and PA in breast cancer lung metastasis, we assessed the clinical potential of targeting this axis or restricting PA intake. Using the ROC Plotter website, we demonstrated that patients with breast cancer with lower ACSL5 levels were sensitive to PAC (Fig. 8A). Given PAC’s efficacy in treating metastatic breast cancer (31), we evaluated its therapeutic effects in conjunction with celecoxib, ONO-AE3-208, ACSL5 knockdown, or limited PA intake in a mouse model. The results showed that ACSL5 knockdown, celecoxib, ONO-AE3-208, or PAC dramatically mitigated lung metastasis, whereas targeting ACSL5, COX2, or EP4 in combination with PAC led to a more pronounced reduction in lung metastasis (Fig. 8B and C) and increased overall survival compared with PAC monotherapy (Fig. 8D). In addition, PA feeding worsened lung metastasis, reduced PAC sensitivity (Fig. 8E), and decreased overall survival (Fig. 8F). In summary, targeting the ACSL5/COX2/EP4 axis or limiting PA intake enhances PAC sensitivity and mitigates lung metastasis in breast cancer.

Discussion
Lung metastasis remains a major clinical challenge accounting for 60% of the tumor-related mortality in patients with breast cancer. Although substantial efforts have been made to depict metastasis-promoting roles of metabolic rewiring, how cancer cells adapt their metabolism to the dynamic metastatic microenvironment remains unclear. This study highlights the critical role of PA in breast cancer lung metastasis and how DTCs adapt to PA-enriched niche, supporting their survival and proliferation by reprogramming prostaglandin metabolism and facilitating lung-tropic metastasis.
Metabolic alterations in primary cancer cells are well documented (32, 33), but the metabolic reprogramming of DTCs in specific distant organs remains largely unexplored. In this study, we identified the preferential activation of lipid metabolism in breast cancer lung metastases. Consistent with previous findings, blocking lipid metabolism impairs lung metastasis in patients with breast cancer (34, 35). CD36 promotes macrophage infiltration to facilitate liver metastasis, whereas myeloid-specific CD36 knockout leads to reduced liver metastasis (36). Notably, we identified that enhanced lipid metabolism in lung metastases is a common characteristic shared by several tumors prone to lung metastasis, including breast cancer, renal cell carcinoma, colorectal cancer, and osteosarcoma. We propose that lung metastatic cancer cells modify their metabolic processes to adapt to the destination organ during dissemination and colonization. Recent research has shown that the gene expression profiles of metastatic tumors represent an intermediate state between primary tumors in their organs of origin and those in their destination organs, with a closer resemblance to the gene expression profiles of primary tumors within the destination organs (37). This finding supports the notion of organ-specific adaptation of metastatic breast cancer cells when colonizing in the lungs.
To explore the importance of metabolic adaptation in lung metastasis, we investigated the dynamic metabolic profiles of lung metastatic niche. Our findings unveil an increased fatty acid availability, with PA, SA, OA, and LA significantly increasing in the PL and further increasing in the ML. However, the extent to which these dietary fatty acids contribute to metastasis remains unclear. In this study, we demonstrated that PA facilitates breast cancer lung metastasis, consistent with a recent study showing that premetastatic niche formation established a PA-rich environment in the lungs, fostering breast cancer metastatic growth via prometastatic NF-κB signaling (13). These findings confirm that PA enrichment in the lung metastatic niche supports lung metastasis. Our findings revealed that lung-resident AT2 cells respond to exosomal USP47 from lung metastatic breast cancer cells, leading to PA synthesis and release in a YAP-dependent manner.
To elucidate how PA promotes lung metastasis in breast cancer, we demonstrated that ACSL5 was highly expressed exclusively in lung metastatic breast cancer cells, as opposed to primary tumor cells, brain and liver metastatic cancer cells. Elevated ACSL5 levels predicted a trend toward worse clinical outcomes in patients with lung metastasis and exacerbated lung-specific metastasis in mice. Although ACSL5 did not significantly affect the biological phenotypes of lung metastatic breast cancer cells in vitro, its lung metastasis–promoting capacity of ACSL5 was linked to PA enrichment in the lung metastatic niche. Mechanistically, ACSL5 upregulated COX2 expression in a PA-dependent manner, contributing to the accumulation of PGE2, which subsequently activated the PI3K/AKT and ERK pathways. This highlights the importance of PA adaptation and prostaglandin metabolism in breast cancer lung metastasis. Consistently, previous studies have uncovered elevated COX2 levels and increased prostaglandin metabolism in breast cancer lung metastasis, supporting the robustness of our findings (12, 38). Recent studies have identified that COX2+ lung fibroblasts reshape myeloid cells to foster an immunosuppressive microenvironment by producing PGE2. Targeting these COX2+ fibroblasts decreased lung metastasis and improved immunotherapy efficacy (39). These findings support that PGE2 plays a crucial role in breast cancer lung metastasis. However, whether PGE2 simultaneously promotes metastatic growth of lung DTCs and educates the local immune microenvironment to permit metastasis formation in breast cancer remains to be clarified. We demonstrated that elevated COX2 and EP4 levels have been detected in lung metastases across various cancer types, with a positive correlation between COX2/ACSL5 and EP4/ACSL5 levels. These findings support the ACSL5/COX2/EP4 axis as a common molecular event in lung metastasis across multiple cancer types. However, its role in other cancer types remains unexplored. Notably, the role of ACSL5 in cancer progression is highly context-dependent, promoting tumor progression in leukemia, glioma, gastric cancer, and colorectal cancer (30, 40–42). In these contexts, ACSL5 supports cancer cell proliferation, survival, and metabolic reprogramming. Conversely, ACSL5 inhibits tumor development in lung cancer (43). In non–small cell lung cancer, PA levels are significantly lower in patients than in healthy individuals, and PA reduces cell proliferation and migration in A549 cells by inducing ACSL5 expression and inhibiting ERK phosphorylation. However, the direct correlation between ACSL5 and ERK phosphorylation has not been fully elucidated, and using only A549 cells limits the generalizability of these findings. Our study and others have identified PA as a significant promoter of breast cancer development and progression (13, 44). Breast cancer cells adapt to a high PA environment and utilize elevated PA to promote tumorigenesis and distant metastasis. ACSL5 mediated this adaptation in lung-preferential metastatic breast cancer cells, possibly by regulating COX2 palmitoylation and downstream pathways associated with proliferation and apoptosis. Collectively, ACSL5’s role in cancer is highly context-dependent, reflecting the complex interplay between tumor type, microenvironment and experimental conditions, emphasizing the need for further investigation into its differential effects in cancer.
PA cannot be directly converted into AA for prostaglandin synthesis, but it may regulate prostaglandin production through posttranslational modifications of the proteins responsible for prostaglandin biosynthesis. PA can modify cysteine residues of proteins in a process known as palmitoylation. Palmitoylation of proteins like PD-L1, β-catenin, GULT1, STAT3, and IFNGR1, influences their stability, cellular localization, and functions, facilitating tumor progression (45, 46). In our study, PA-ACSL5 upregulated several ZDHHCs involved in palmitoylation in lung metastatic breast cancer cells. Treatment with 2-BP significantly reduced COX2 and PGE2 levels in ACSL5high LM3 cells, suggesting that PA-ACSL5 may regulate COX2 expression through palmitoylation. However, given that 2-BP is not entirely specific to palmitoylation and also inhibits other lipid-metabolic enzymes, including monoacylglycerol- and diacylglycerol acyltransferases, fatty acid CoA ligase, and glycerol-3-P acyltransferase (47). The precise mechanisms by which PA-ACSL5 regulate COX2 expression remains to be fully elucidated.
Limited therapeutic efficacy and drug tolerance are the biggest obstacles for clinical lung metastasis. Our study demonstrated that restricting dietary PA intake can enhance the therapeutic efficacy of PAC. Targeting the ACSL5/COX2/EP4 axis is a promising strategy for treating breast cancer lung metastasis. COX2 inhibitors, which reduce cancer risk in individuals taking nonsteroidal anti-inflammatory drugs (48). We added to this knowledge by demonstrating that COX2 inhibitors exhibited synergistic antitumor effects on breast cancer lung metastasis.
In conclusion, our study reveals the mechanism by which PA is enriched in the lung metastatic niche and metastatic breast cancer cells adapt to the elevated PA. Our findings highlight the potential of combining PAC with targeting the ACSL5/COX2/EP4 axis as a therapeutic strategy for patients with breast cancer with lung metastasis. Our findings also provide insights into the role of dietary fatty acids in promoting metastasis. However, further efforts are required to elucidate how dietary nutrient availability on DTCs in specific organs and how environmental factors are converted into regulatory signals supporting metastatic survival and growth.

Supplementary Material
Supplementary Table 1List of primer sequences utilized in the study

Supplementary Table 2Antibodies and regents utilized in the study

Supplementary Table 3Core sequences of shRNA against target gene

Figure S1Lung metastatic breast cancer cells display increased lipid metabolism

Figure S2Lung metastatic breast cancer cells display increased lipid metabolism and growth advantage

Figure S3LM3-derived exosome promotes fatty acid synthesis of lung resident-AT2

Figure S4ACSL5 mediates lung-specific metastasis

Figure S5ACSL5 mediates adaptation to palmitic acid of lung metastatic breast cancer cells

Figure S6ACSL5 induced COX2 expression in a PA-dependent manner

Figure S7COX2 promotes PGE2 production, cell survival and growth of lung metastatic breast cancer cells

Figure S8PA-ACSL5 induced COX2 expression and PGE2 accumulation activates EP4 to foster DTCs survival and proliferation through PI3K/AKT and ERK signaling

Figure S9COX2 expression is potentially regulated by palmitoylation process

Figure S10ACSL5/COX2/EP4 axis was activated in lung metastases of multiple cancer types