Stromal CAVIN1 Controls Prostate Cancer Microenvironment and Metastasis by Modulating Lipid Distribution and Inflammatory Signaling

Caveolae, ultrastructural microdomains at the plasma membrane, are involved in protein and lipid trafficking and function as scaffolds for signaling proteins (1–3). Caveolae also provide a physical buffering capacity for cells under mechanical stress (4). Caveolae are formed by the assembly of caveolins (CAV) and caveolae-associated protein 1 (CAVIN1; also known as polymerase-1 and transcript release factor, PTRF), and are rich in lipids, in particular cholesterol (5–10). Intriguingly, CAVIN1 was first identified as RNA polymerase I transcription termination factor (11), and subsequently, as a critical component of caveolae (7, 8). CAVIN1 is involved in adipocyte lipid storage and hence contributes to energy metabolism. Adipocytes deficient of either CAVIN1 or caveolins are defective in lipid uptake and storage, have reduced cholesterol transport, and contribute to increased levels of circulating triglycerides and free fatty acids (2, 8, 9, 12). Although caveolin-1 (CAV1) is also involved in lipid trafficking (9), less is known how CAVIN1 and CAV1 regulate lipid droplet formation.

Mice and humans with targeted disruption or mutations of CAVIN1 are affected by numerous abnormalities including lipodystrophy, muscular dystrophy, cardiovascular disease, and diabetes (8, 13–16). Cavin1-knockout mice have decreased insulin-dependent glucose uptake, reduced lipid storage and impaired lipid tolerance, adipose tissue fibrosis, and increased macrophage infiltration (12, 13). Furthermore, loss of Cavin1 impairs insulin-mediated focal adhesion formation and remodeling required for a mechanical stress response, concomitant with activation of ERK and p38 stress signaling (17). In adipocytes, CAVIN1 also adjusts ribosomal activity to the nutrient state suggesting that CAVIN1 functions both in the regulation of lipid metabolism and RNA polymerase I transcription (18). Also, CAVIN1 may behave as an adipokine and partially contribute to the well-known detrimental effects of visceral fat accumulation (19). Under starvation or exercise, fatty acids from lipid droplet triacylglycerol stores are released as a source of energy (20). Given that certain lipids are cytotoxic, the uptake of the lipids by caveolae also has cytoprotective functions. CAVIN1 deficiency may, hence, drastically reprogram both cellular energy metabolism and the microenvironment (21).

Both CAVIN1 and CAV1 are largely absent in the normal prostate epithelium but present in the stroma (3, 22). During prostate cancer progression, the expression of CAV1 in the tumor cells increases, but the expression of both CAV1 and CAVIN1 is lost in the tumor stroma (23–25). The decrease in stromal CAV1 and CAVIN1 correlates with reduced relapse-free survival, higher Gleason score, and poor outcome (23). Ectopic expression of CAVIN1 in prostate cancer cells reduces their aggressive phenotypes (proliferation, anchorage-independent growth, migration, and invasion), lymphangiogenesis, and angiogenesis in vitro and in vivo (24, 26, 27). In contrast to CAVIN1, expression of CAV1 in prostate cancer cells increases their anchorage-independent growth, invasive and angiogenic potential, and castration resistance (3, 22, 28). The anchorage-independent growth is reversed by coexpression of CAVIN1, suggesting that their functional association can mitigate the oncogenic activity of CAV1 (24). Furthermore, CAVIN1 was shown to modulate the dynamics of cholesterol and actin cytoskeleton and impair prostasome secretion in prostate cancer (29).

The function of stromal CAVIN1 in prostate cancer has not been studied before. Given the abundant changes of CAVIN1 in prostate cancers, we implemented in vitro and in vivo orthotopic tumor models to systematically analyze how stromal CAVIN1 affects tumor growth, phenotype, and lipid regulation. We show in coculture models that prostate stromal cells lacking CAVIN1 increase prostate cancer cell lipid content, cause an inflammatory tumor microenvironment, and promote invasion and metastasis. We propose that stromal fibroblasts contribute to prostate cancer aggressive phenotypes through control of lipid and cytokine balance.

Normal prostate stromal line WPMY-1, HEK293T cells, and prostate cancer cell lines PC3, DU145, and CW22Rv1 were purchased from ATCC. Normal primary prostate stromal line, PrSc was purchased from Lonza. MR49F cell line was a kind gift from Dr. Martin Gleave (Vancouver Coastal Health Institute, Vancouver, British Columbia, Canada; ref. 30). All cells were maintained in the culture media as per the manufacturer's instruction, cultured at 37°C in a humidified atmosphere containing 5% CO₂, and were authenticated by short tandem repeat analyses at the Johns Hopkins Genetic Resources Core Facility (Baltimore, MD). Cell lines were tested for Mycoplasma with Venor GeM Mycoplasma Detection Kit (Sigma-Aldrich) for negativity. Upon thawing from liquid nitrogen, cell lines were passaged once before being used for experiments. Before reaching 20 passages, cells were discarded and a new vial was thawed.

CAVIN1 (catalog no. HPA049838) antibody was purchased from Millipore Sigma. CAV1 (catalog no. ab2910), SV40 T antigen (catalog no. ab16879), GAPDH (catalog no. ab8245), and alpha tubulin (catalog no. ab7291) antibodies were from Abcam. SREBP1 (catalog no. NB600-582) antibody was from Novus Biologicals and FASN (catalog no. 3180) antibody was from Cell Signaling Technology.

Lentiviral short hairpin RNA (shRNA) plasmids were purchased from Johns Hopkins High Throughput Biology Core Facility (shCAVIN1_1 5′-CCGCAACTTTAAAGTCATGAT-3′ and shCAVIN1_2 5′-GTGGAGGTTGAGGAGGTTATT-3′). Following transfection in HEK293T cells for production of lentiviral particles, the viruses were transduced into WPMY-1 cells, and selected for puromycin resistance to generate stable CAVIN1-knockdown clones. CAVIN1 targeting and scrambled control siRNA were purchased from Ambion (Assay ID: s49507 and s49508; Life Technologies). Cells were transfected with 10 nmol/L of siRNA before harvesting at 72 hours for subsequent experiments.

Oleic acid (catalog no. O1383) and water-soluble cholesterol (catalog no. C4951) were purchased from Sigma-Aldrich and prepared according to the manufacturer's instructions.

Cell culture medium was collected 48 hours after seeding and centrifuged at 670 × g for 5 minutes to remove cell debris before being used for subsequent experiments.

Oil Red O was purchased from Sigma-Aldrich and was used as per the manufacturer's protocol. Cells were cultured in culture media supplemented with 20% FBS. Following, culture medium was aspirated and cells were washed with PBS, followed by fixing with 10% formalin for 45 minutes. Formalin was then removed and cells were washed with deionized water followed by incubation with 60% isopropanol for 5 minutes. Isopropanol was then removed and cells were stained with Oil Red O solution for 5 minutes before rinsing under running tap water. Cells were then visualized with EVOS FL Auto Microscope (Life Technologies) and images were analyzed with ImageJ (NIH). Oil Red O staining was quantified from five fields per sample and normalized for total cellular area.

PC3 cells were seeded into a 6-well plate and incubated overnight. Culture medium was removed and cells were scratched with a 1 mL pipette tip and conditioned medium from stromal cells was added. Images were taken at 8, 16, and 24 hours with EVOS Microscope (Thermo Fisher Scientific) and images were analyzed with ImageJ (NIH).

Stromal cells were seeded into a 24-well plate. Following day, Transwell inserts were inserted into the 24-well plate and prostate cancer cells were seeded into the inserts. After 14 hours, the inserts were washed, fixed with 3.5% PFA, and stained with 0.5% crystal violet. Images were taken with EVOS FL Auto microscope and images were analyzed with ImageJ (NIH).

RNA was extracted with TRizol (Life Technologies) as per the manufacturer's instruction. Three biological samples were prepared for each WPMY-1 clone. RNA sequencing (RNA-seq) library for Illumina platform sequencing was prepared using Illumina TruSeq Stranded Total RNA Sample Kit following the manufacturer's recommended procedure. Briefly, 250 ng of total RNA was first depleted of rRNA using Ribo-Zero Gold (Epicentre) and fragmented. Fragmented RNA was converted to double-stranded cDNA with the second strand marked. The resulting cDNA was polyA tailed and ligated to barcoded sequencing adaptors and amplified by PCR. The amplified libraries were quality controlled and pooled. The pooled library was further quantitated using KAPA Library Quantification Kit (Kapa Biosystems) and sequenced on NextSeq 500 (Illumina) for 2 × 75 bp paired-end reads. Sequencing data were analyzed using Tophat 2 and Cuffdiff 2.0, respectively, for alignment to reference genome and differential expression detection. Pathway analyses were conducted using Gene Set Enrichment Analysis (GSEA) as in ref. 31. The data are deposited to Gene Expression Omnibus as GSE146229.

RNA was extracted using TRizol as per the manufacturer's instruction and converted into cDNA using Super Script II cDNA Synthesis Kit (Life Technologies). qPCR was performed using iTaq Universal SYBR Green Supermix purchased from Bio-Rad on a CFX6100 qPCR Instrument (Bio-Rad).

Protein lysate was collected from cells using RIPA buffer supplied with Protease Inhibitor Cocktail and quantified using BCA Assay (Life Technologies). Twenty micrograms of protein were electrophoresed on precast gel purchased from Life Technologies for 90 minutes. Electrophoresed proteins were transferred onto Nitrocellulose Membrane (Bio-Rad) and blocked with 10% milk (Bio-Rad) for 1 hour at room temperature. Primary antibody was then added and incubated overnight. Following day, nitrocellulose membrane was washed thrice, 5 minutes each with TBST buffer before being incubated with horseradish peroxidase (HRP)-conjugated secondary antibody (Agilent Dako) for 1 hour at room temperature. Membrane was then washed thrice, 5 minutes each, followed by addition of Lighting ECL Reagent (GE Amersham) and the membrane was visualized with Bio-Rad GelDoc+ System (Bio-Rad).

Multiplex cytokine array was performed at the Johns Hopkins Immune Monitoring Core Facility on a Bioplex 200 Platform (Bio-Rad) using Luminex Bead-based Immunoassays (Millipore) according to the manufacturer's protocols. The HAGE1MAG-20K panel was used to detect IL6 and IL18, HCMBMAG-22K panel for DKK1, HCYP3MAG-63K panel for CSF-1 (MCSF), HCYP4MAG-64K for IL32, and HMMP1MAG-55K for MMP3. IL6, IL18, and IL32 were below detection limit and are not reported.

QuickChange II Site-directed Mutagenesis Kit (Agilent Technologies) was used to generate CAVIN1 lentiviral plasmid resistant to shRNA degradation and sequence verified. The resulting plasmid was used to generate lentiviral particles in HEK293T cells and used to transduce stromal CAVIN1-knockdown (KD) cells.

The orthotopic mouse experiment was conducted under an approved protocol by the Johns Hopkins University Animal Care and Use Committee. We generated PC3 cells stably expressing firefly luciferase (PC3-Luc) and mixed them (1.25 × 10⁵ cells, 1:1 ratio) with WPMY-1 shCtrl, shCAVIN1, or CAVIN1-R stromal cells and implanted orthotopically to the right anterior prostate of NSG (NOD-SCID) mice. Bioluminescence of the tumors was determined using IVIS Spectrum In Vivo Imaging System (Perkin Elmer) weekly or biweekly. At the end of the study, urogenital block, tumor, lungs, and liver were harvested. Invasion and metastasis were determined by counting visible metastases at necropsy. Frozen sections were prepared. Tissues were then fixed in 10% formalin, and 4-μm sections were stained with hematoxylin and eosin. The number of micrometastases were assessed in a blinded fashion.

CD163 (catalog no. ab182422) antibody was purchased from Abcam and used to stain for the presence of M2 macrophages in tumor tissues. Briefly, paraffin-embedded slides were dewaxed, rehydrated, and antigen retrieval was performed with Antigen Unmasking Buffer (Vector Laboratories). Slides were then treated Dual Endogenous Enzyme Blocker (Agilent Dako). Primary antibody was then added to the slides and incubated at room temperature for 1 hour. Slides were washed, stained with HRP-labeled secondary antibody (catalog no. PV6119, Leica Microsystems) followed by detection using 3, 3′-Diaminobenzidine (Sigma-Aldrich), counterstained with hematoxylin, dehydrated, and mounted with cover slip. Slides were then imaged using EVOS FL Auto microscope on three independent fields. Signal intensity or presence of stained cells were quantified using ImageJ software.

Statistics were computed with GraphPad Prism 6 using Student two-tailed t test, Mann–Whitney nonparametric t test, and one-way ANOVA with Tukey or Kruskal–Wallis post hoc test as indicated. N represents biological replicates, unless otherwise stated. Data are presented as mean ± SD.

Given that loss of CAVIN1 in prostate cancer stroma associates with aggressive prostate cancer and poor survival, we hypothesized that CAVIN1 could influence these events by modifying the lipid content in the stroma. To study this, we knocked down CAVIN1 in an established, well-studied WPMY-1 prostate stromal fibroblast line (32). For this purpose, we used CAVIN1 targeting shRNAs to silence the expression of CAVIN1 and generated two stable lines, shRNA-CAVIN1_1 and shRNA-CAVIN1_2, and a nontargeting control, shCtrl (Fig. 1A). Unless otherwise noted, we chose to conduct subsequent experiments with the clone shRNA-CAVIN1_2 cells, called hereafter shCAVIN1. To generate an alternative model, we used siRNAs to silence CAVIN1 in PrSc stromal fibroblasts (Fig. 1B). Although PrSc cells are not fully characterized for all mesenchymal markers, they provide another representation of prostate stromal fibroblasts. CAVIN1 silencing did not affect the growth of these cell lines (Supplementary Fig. S1A and S1B). To test the robustness of our model, we reintroduced CAVIN1 expression in the shCAVIN1 stromal cells. Using site-directed mutagenesis, we generated a lentiviral CAVIN1 expression plasmid that is resistant to the shRNA used to KD CAVIN1. After viral transduction, a mixed pool of cells was collected. We observed rescue of CAVIN1 expression in the shCAVIN1 cells following transduction and selection of a stable clone, hereafter called CAVIN-R (Fig. 1C). We then tested the presence of lipid droplets in these cells as determined by Oil Red O staining. To facilitate the detection of lipid droplets, we cultured the cells in 20% FCS. We found that lipid droplets were dramatically reduced in the CAVIN1-silenced WPMY-1 and PrSc stromal cells compared with controls (Fig. 1D). Lipid droplets were restored in the CAVIN1-R cells affirming that the phenotype is CAVIN1 dependent (Fig. 1E).

To further test the model, we fed the WPMY-1 shCAVIN cells with oleic acid, a fatty acid stored in lipid droplets. We observed that shCAVIN cells were not able to store the lipid, whereas the shCtrl cells robustly did so (Fig. 1F). To assess whether the changes in intracellular lipid droplets are due lipid uptake from the extracellular space, we used a cholesteryl ester uptake inhibitor, ML278, and found that this reduced lipid droplets in the WPMY-1 shCtrl cells (Supplementary Fig. S1C). ShCAVIN1 cells remained devoid of lipid droplets. To ask whether CAVIN1 KD leads to major perturbation in lipid synthetic pathways of the cells, we analyzed the levels of fatty acid synthase, FASN, and the lipid metabolism controller, SREBP1, by Western blotting, but did not detect any obvious changes (Supplementary Fig. S1D). These findings suggest that CAVIN1 depletion in stromal cells leads to defective lipid uptake. Given this, we postulated that CAVIN1-depleted stromal cells could have augmented amounts of lipids in their microenvironment.

To probe the impact of stromal cell lipid cycling on prostate cancer cells, we used conditioned medium from WPMY-1 shCtrl and shCAVIN1 cells and applied the media to cultures of four prostate cancer cell lines, PC3, DU145, MR49F, and CWR22RV1. We found that prostate cancer cell lines cultured in shCAVIN1 medium had significantly more lipid droplets compared with those cultured in shCtrl medium (Fig. 2A; Supplementary Fig. S2A–S2C). We also assessed for perturbations in prostate cancer cell growth by the conditioned medium but did not observe any changes in growth or expression of FASN or SREBP1 lipid synthesis proteins (Supplementary Fig. S2D–S2F). These experiments suggest that CAVIN1 KD in the stromal cells increases lipid uptake by the prostate cancer cells.

To address how the stromal cells impact prostate cancer cell phenotypes, we first cultured PC3 cells in shCtrl and shCAVIN1 conditioned medium and performed a scratch wounding assay. We observed that PC3 cells applied with shCAVIN1 conditioned medium had increased migration as measured by closure of the wound scratch (Fig. 2B). We then used a Transwell coculture model, where we seeded WPMY-1 shCtrl or shCAVIN1 stromal cells in the bottom well and prostate cancer cells in the top inserts. Two prostate cancer cell lines, PC3 and DU145, were tested. After 14 hours of culture, the inserts with the prostate cancer cells were stained and analyzed for their ability to penetrate the insert membrane (migration) or invade the Matrigel (invasion). We observed both increased migration and invasion of PC3 and DU145 cells cocultured with shCAVIN1 cells compared with shCtrl cells (Fig. 2C). Using siRNA to silence CAVIN1 in PrSc stromal cells, we repeated the Transwell invasion assay and observed increased invasion of PC3 cells cocultured in CAVIN1-silenced PrSc cells compared with siRNA control cells (Supplementary Fig. S2G). Furthermore, the invasion was decreased when PC3 cells were cocultured with CAVIN1-R cells compared with coculture with shCAVIN1 stromal cells (Fig. 2D), suggesting a successful rescue of the phenotype. These findings suggest that CAVIN1 downregulation in the stromal cells augments prostate cancer cell invasive properties.

We delineated the changes in transcriptome after knocking down CAVIN1 in WPMY-1 cells using RNA-seq. We detected significant alterations in over 6,000 transcripts as compared with the shCtrl cells. GSEA pathway analyses indicated significant enrichment of inflammatory and IFN response pathway transcripts (Fig. 3A and B). This was highly interesting, as it suggested that CAVIN1 KD led, not only to a change in cellular lipid uptake, but also activation of an inflammatory response including increased expression of several secreted cytokines and proteases (Fig. 3C). We validated the RNA-seq results using qPCR and by biochemical analyses. We detected significant upregulation of gene transcripts such as CCL2, CSF1, DKK1, IL18, IL32, MX1, MMP3, and TLR3 in the shCAVIN1 fibroblasts (Fig. 3D). We used ELISA and zymography to detect MMP3 and observed its robust upregulation (Supplementary Fig. S3A and S3B). To validate these findings in the PrSc fibroblasts, we silenced CAVIN1 using siRNAs and observed similar increases in gene transcripts (Supplementary Fig. S3C). Furthermore, as determined by qPCR, changes in the inflammatory gene signatures in shCAVIN1 cells were reversed in CAVIN1-R stromal cells (Fig. 3E). In addition, we performed multiplex cytokine assays to determine the cytokines at protein level. Of those measurable, we found that MMP3, DKK1, and CSF-1 were upregulated in shCAVIN1 stromal cells. Conversely, the cytokine levels showed a trend of reduction in the CAVIN1-R stromal cells (Fig. 3F). We also explored whether the change in inflammatory pathways was accompanied with markers for cancer-associated fibroblasts (CAF). On the basis of RNA-seq, we observed robust increases in several CAF markers such as ACTA2 (α-SMA), S100A4 (FSP1), and decorin (Supplementary Fig. S3D). Although other CAF markers such as SDF1, COL1A2, and tenascin were downregulated and absolute markers for CAFs remain to be defined, these findings suggest that stromal cells with CAVIN1 KD had assumed CAF-like features. On the other hand, there was no systematic activation of lipid synthetic pathways (Supplementary Fig. S3E).

Obesity and high fat diet are linked to prostate cancer progression and mortality (33–35). To test whether extracellular lipids promote prostate cancer cell migration or invasion, we added cholesterol or oleic acid to Transwell bottom wells and plated PC3 cells on the top inserts. We observed increased migration and invasion by the lipid addition in a dose-dependent manner (Fig. 4A and B). This simple experiment suggested that the lipid content in the surrounding microenvironment influences prostate cancer cell malignant properties. We next explored potential intervention strategies and used methyl-β-cyclodextrin (MBCD), a lipid scavenger, or cholesteryl ester uptake inhibitor, ML278. In a Transwell assay, both agents reduced invasion of PC3 cells stimulated by shCAVIN1 stromal cells, suggesting that targeting accessibility of lipids by the prostate cancer cells restrains their metastatic susceptibility (Fig. 4C and D). To assess whether the excess of lipids would be a driving factor that triggers inflammatory response, we treated WPMY-1 shCtrl cells with cholesterol and performed qPCR on the gene panel. We detected upregulation of gene expression of most transcripts observed in the shCAVIN1 cells (Fig. 4E). Conversely, we used MBCD to scavenge cholesterol in shCAVIN1 cells and observed downregulation of several of the inflammatory genes (Fig. 4F). Taken together, this suggests that loss of CAVIN1 in stromal cells triggers a lipid-induced inflammatory response in the tumor microenvironment.

To assess how stromal CAVIN1 affects the metastatic ability of prostate cancer in vivo, we chose to use orthotopic coimplantation of the stromal and cancer cells to the mouse prostate. We used PC3 prostate cancer cells for this purpose, given their robust metastatic capacity. We first generated a stable PC3 cell line that expresses firefly luciferase (hereafter PC3-Luc cells) to facilitate bioluminescence monitoring of the cells in vivo. PC3-Luc cells were then mixed 1:1 with either shCtrl or shCAVIN1 WPMY-1 stromal cells, and implanted orthotopically to the mouse anterior prostates. The growth of the tumors was determined using bioluminescence over the course of 8 weeks (Fig. 5A and B) and by determining the urogenital block and tumor sizes at the end of the experiment. We observed that coinjection of PC3-Luc cells with stromal fibroblasts with or without CAVIN1 did not affect the growth rate or size of the tumors (Fig. 5B and C; Supplementary Fig. S4A). However, based on the macroscopic analysis during necropsy, we observed an increase in the number of lung metastasis when PC3-Luc cells were coinjected with the shCAVIN1 stromal cells (Supplementary Fig. S4B and S4C). The lungs and livers were harvested and subjected to histology (Fig. 5D). We observed a significant increase in micrometastases in both distant sites when the orthotopic tumors were generated by coinjection with shCAVIN1 stromal cells (Fig. 5E and F). Strikingly, assessment of the primary tumors showed a significant increase in the lipid content in mice with coimplantation of shCAVIN1 stromal cells as compared with shCtrl cells (Fig. 5G). To analyze this further, we used qPCR to detect changes in the inflammatory gene signatures. We observed up to 10-fold increases of CSF1 and MMP-3 in tumors coimplanted with shCAVIN1 stromal cells albeit with an intracohort variation (Fig. 5H). These findings strongly suggest that stromal cells lacking CAVIN1 condition the tumor microenvironment and promote metastasis of the primary tumor.

Prompted by these findings, we established another cohort of orthotopic PC3-Luc–bearing animals using coimplantation with stromal cells from shCtrl, shCAVIN1, and CAVIN1-R clones (Fig. 6A). Replication of the model was robust. We neither observed differences between the cohorts in their rate of primary tumor growth during the study, nor tumor or urogenital block weights at the end of the study (Fig. 6B and C; Supplementary Fig. S5A). However, we observed increased numbers of macrometastases in liver and lungs in the shCAVIN1 cohort, and this phenotype was rescued in the CAVIN1-R cohort (Fig. 6D and E). We analyzed the liver and lung samples for histology and counted the number of micrometastases. As previously, we observed an increase in micrometastases in the lungs and liver of shCAVIN1 cohort compared with shCtrl cohort. This phenotype was robustly rescued in the CAVIN1-R cohort (Fig. 6F and G; Supplementary Fig. S5B).

We analyzed the primary tumors for lipid content using Oil Red O. Again, we found a significant increase in the lipid content when shCAVIN1 stromal cells had been used for coimplantation as compared with shCtrl stromal cells. This phenotype was fully reverted when CAVIN-R stromal cells were used (Fig. 6H). We then assessed the primary tumors for M2 inflammatory macrophages using CD163 as a marker. We detected a significant increase in M2 macrophage infiltration in primary tumors from shCAVIN1 cohort as compared with shCtrl cohort and abrogation of macrophage infiltration in the CAVIN1-R cohort primary tumors (Fig. 6I). To assess whether the human WPMY-1 stromal fibroblasts continued to reside in the primary tumors, we performed SV40 T-antigen immunostaining. We observed occasional staining of SV40 T-antigen–positive stromal cells in and around the primary tumors that varied from less than a few hundred per tumor section to no detectable signal. Their presence in the primary tumors was independent of CAVIN1 expression (Supplementary Fig. S5C). This suggested that it is unlikely that the persistent changes observed in the primary tumors were due to the continuous influence of the resident human stromal cells. These findings suggest that stromal cells cause permanent changes in the primary tumors and their metastatic capacity.

Lipids fuel cancer cell metabolism, and serve as macromolecules for membranes and building blocks for hormone synthesis (36). Epidemiologic studies have linked obesity and high fat diet with cancer incidence, and mechanistic studies have shown that cancer cells utilize exogenous lipids for many of their pathologic pathways and functions. Our study provides a conceptual advance by showing that loss of CAVIN1, an essential factor for caveolae formation in the prostate stromal cells promotes an inflammatory microenvironment that fuels prostate cancer aggressive phenotypes. We demonstrated using in vitro coculture models that loss of stromal CAVIN1 expression negates the ability of stromal cells to sequester lipids leading to upregulation of inflammatory signatures such as expression of cytokines, cytokine receptors, matrix metalloproteinases, and markers for CAFs. We showed that CAVIN1-deficient stromal cells promoted prostate cancer cell migration and invasion, and in vivo, metastasis and M2 macrophage infiltration. This study reaffirms the critical function of the stromal component in tumorigenesis and reveals a metastasis-suppressing role of CAVIN1.

Prostate cancer is a lipid-rich tumor and periprostatic white adipose tissue inflammation associates with prostate cancer aggressiveness (37–39). Dietary fat promotes prostate cancer development and progression through SREBP1-regulated lipogenic and MYC programs (34, 40, 41). Chemical inhibition of fatty acid synthase reprograms castration-resistant prostate cancer and reduces the expression androgen receptor and its treatment-resistant variant V7 (42). Circulating prostate tumor cells have high lipid uptake and increased intracellular lipids (43), further suggesting that lipids and lipogenic programs contribute to prostate cancer progression and metastasis. Genetic models have established that CAVIN1 is essential for caveolae formation and alterations in lipid content in adipocytes (8, 9, 18). CAVIN1 is highly expressed in normal prostate stroma, and its loss confers poor clinical outcomes (24, 25). However, the function of CAVIN1 in the regulation of lipid metabolism has not been previously studied in the prostate stroma. Given our finding that loss of stromal CAVIN1 exposes prostate cancer microenvironment to an excess of lipids, and consequently, proinflammatory environment, our study identifies a molecular event that contributes to lipid-driven pathogenesis of prostate cancer.

Caveolar integrity is essential for cellular signal transduction and cholesterol transport (3). Caveolae formation requires both CAVIN1 and caveolins. Both are abundantly expressed in adipocytes, endothelial, and stromal cells. They are typically not expressed by epithelial cells, but curiously, CAV1 expression varies broadly among cancer types such that its high expression correlates with poor outcomes in breast, prostate, and lung cancer and good outcomes in head and neck and biliary cancer (3). In prostate cancer, CAV1 is bestowed with protumorigenic properties that can be reset by ectopic expression of CAVIN1 (7, 25, 44, 45). Reintroduction of CAVIN1 into prostate cancer cells reduced the aggressive phenotypes (migration, invasion, lymphangiogenesis, and angiogenesis) in vitro and in vivo (7, 24, 26, 27). CAVIN1 is downregulated in prostate cancer stroma (24, 25). Loss of CAV1 in tumor stroma also confers a poor outcome in prostate, breast, esophageal, gastric, and pancreatic cancers (21, 25). Cav1-null mice have stromal abnormalities, increased epithelial hyperplasia, and support increased growth of ectopically implanted breast cancer cells or tumors (21). Genetic ablation of Cav1 increases proliferation of primary and transformed fibroblasts via an increase in MAPK signaling pathway (21). Knocking down CAV1 increases lipid uptake, the amount of cholesterol and testosterone in fibroblasts, and promotes cancer cell proliferation, primary tumor growth, and metastasis (21, 23). Here, we show that CAVIN1 KD abrogates stromal lipid uptake and has no effect on proliferation of either stromal cells or prostate cancer cells or primary tumor growth. This suggests that even if functions of CAVIN1 and CAV1 are coupled through caveolae formation, their loss exerts different outcomes in the stromal cells, both of which may exacerbate the aggressiveness of the primary tumor.

Here, we found that loss of stromal CAVIN1 activated inflammatory and CAF-like gene signatures and upregulated secretion of MMP3, DKK1, and CSF1. Given that the addition of lipids caused an inflammatory response, and that it was reverted by depleting cholesterol, we suggest that these effects are lipid-enacted and reveal remodeling of the tumor microenvironment via CAVIN1. CAVIN1 was recently reported to interact with suppressor of cytokine signaling 3 (SOCS3), and is required for SOCS3 localization and function, suggesting at least one mechanism how CAVIN1 limits JAK–STAT inflammatory signaling (46). Cavin1^−/− mice have increased inflammatory gene signatures including TNFα, IL6, F4/80, and CD11c in the epididymal fat tissues, and these signatures are further upregulated in the presence of high fat diet (17). The mechanisms that direct increases in inflammatory pathways require further exploration.

We find here striking enrichment of M2 macrophages in the primary prostate tumor when stromal cells lacking CAVIN1 had been used for coinjection. Using SV40 T-antigen staining as a marker, we detected little to no remaining human stromal fibroblasts in the primary tumors, suggesting that the human stromal cells that were initially grafted into the mouse prostate were priming the microenvironment and reprogrammed the prostate cancer cells. The M2 macrophage enrichment was accompanied by increased lipid deposition in the primary tumor cells, and both were reversed when stromal cells reexpressing CAVIN1 were coinjected. CAVIN1 has been shown to regulate macrophage number and phenotype in lungs of Cavin1^−/− mice and demonstrate increased macrophage infiltration into the adipose tissues (47). Several studies suggest that presence of M2 macrophages in the prostate tumor leads to a worse prognosis in men (48–50). Therefore, it is plausible that infiltration of M2 macrophages in the prostate tumors lacking stromal CAVIN1 may further fuel the aggressive disease, and is consistent with the striking increase in lung and liver metastases observed here. Strikingly, cancer cells have recently been shown to scavenge cholesterol from tumor-associated macrophages (TAM) leading to their polarization and reprogramming. Furthermore, inhibition of cholesterol efflux by deletion of the ABC transporters led to reversion of macrophage TAM phenotype and inhibition of tumor progression (51). Nevertheless, it is pertinent to further validate these observations using additional metastatic in vivo prostate cancer models, other than the androgen receptor–null PC3 cells, as they become available. Together with our findings, the emerging evidence points to intricate relationships between the cancer cells, TAMs, and stromal cells underlined by lipid metabolic needs, inflammatory signals, and metastatic proneness where loss of CAVIN1 in the stroma attracts inflammatory macrophages and promotes a tumor-supportive microenvironment.

Different intervention strategies to alter the composition or amount of lipids in cancer have been explored, and most lipid-targeting compounds have failed to move beyond preclinical models (52). Because lipids were reported to drive prostate cancer progression, strategies aimed at reducing fatty acid uptake into prostate tumors may provide a novel and viable therapeutic avenue for prostate cancer. Statins have been shown to effectively reduce the growth and metastasis of prostate cancer in preclinical models (34, 53, 54). In our study, we observed that scavenging lipids using MBCD and a lipid uptake inhibitor reduced invasion of PC3 cells. However, similar strategies will need validation in preclinical models. It is also plausible that regulation of lipid and inflammatory pathways intersect and are both culprits in driving aggressive prostate cancer. Nevertheless, this study, using a definitive model for CAVIN1 dependency emphasizes that loss of stromal CAVIN1 is a key event that contributes to a prometastatic state. CAVIN1 can be epigenetically regulated (55), but we do not know why and how stromal CAVIN1 is lost in prostate cancer stroma. This knowledge will be informative to pursue. Similarly, syngeneic genetic models that facilitate broad assessment of the immune and tumor microenvironment landscape will be needed. Further studies that monitor CAVIN1 expression in tumor stroma and its prognostic significance are warranted and may lead to early or novel intervention strategies to prevent metastatic disease. Although CAVIN1 was shown to regulate lipid metabolism and inflammatory cytokines, this study is the first to show these observations in the context of cancer. Collectively, we established that CAVIN1 expression in the prostate stromal cells is critical in control of lipid content and inflammation in the prostate microenvironment and suppresses metastatic disease.

M. Laiho reports grants from Walsh Prostate Cancer Fund and DoD CDRMP PCRP during the conduct of the study; other from Bluefield Innovations outside the submitted work; has a patent to US 8,680,107; US 10,214,491; EU 2195316, Canada 2,691,227; and Canada 2,912,456 issued, a patent to PCT/US2015/021699/15765295/2,943,022 US, EU, and Canada pending, and a patent to PCT/US2017/052863, 16/335,737, 2017330390, and 17853955.7 pending. No potential conflicts of interest were disclosed by the other authors.

J.-Y. Low: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. W.N. Brennen: Supervision, investigation, methodology, project administration. A.K. Meeker: Resources, formal analysis, validation, methodology. E. Ikonen: Conceptualization, funding acquisition, methodology, writing-review and editing. B.W. Simons: Investigation. M. Laiho: Conceptualization, resources, supervision, funding acquisition, visualization, project administration, writing-review and editing.

We thank Drs. Ken Pienta and Helen Nicholson for critical review. We would also like to thank Susan Dalrymple for her technical assistance in the orthotopic model. This work was supported by Patrick C. Walsh Prostate Cancer Research Fund and Department of Defense CDMRP award W81XWH-17-1-0458 (to M. Laiho) and Jane and Aatos Erkko Foundation (to E. Ikonen).

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.