Omega-3 Eicosapentaenoic Acid Reduces Prostate Tumor Vascularity

Prostate cancer accounts for a large proportion of diagnosed cancers in men from western countries. The 5-year prevalence is among the highest for all cancers and the factors driving progression of low-risk prostate cancer into aggressive prostate cancer are unclear (1). The geographic distribution of prostate cancer incidence combined with the increase of prostate cancer risk following “westernized” lifestyle, both suggest that environmental factors, including diet may be linked to prostate cancer development and progression. In support to this, genetically predisposed mouse models of prostate cancer develop significantly more aggressive cancers when fed a high-fat diet (HFD; refs. 2–4). Autologous prostate cancer cells injected ectopically or orthotopically (intraprostatic injection) are also sensitive to HFD-driven tumor growth (5). In humans, the studies describing a link between dietary fat intake and prostate cancer remain conflicting (6–8). In contrast, targeting beneficial dietary fatty acids intake was shown to decrease prostate cancer progression in preclinical models (9–13). Altogether, clinical and preclinical evidence suggest that HFD is an environmental factor–driving prostate cancer.

Epidemiological studies of prostate cancer risk in relation to long-chain omega-3 fatty acids (LCω3) molecules, mainly eicosapentaenoic acid (EPA) and docosahexanenoic acid (DHA), are controversial. The only two studies of the prostate tissue were inversely associated with prostate cancer progression risk during active surveillance (8, 14). In contrast, two blood biomarker studies reported positive associations with prostate cancer risk (15, 16), supporting that more work is needed to decipher how fatty acid molecules contribute to prostate cancer. Both EPA and DHA can be incorporated into cell membranes, modulating fluidity, hampering cell signaling pathways and thus inhibiting prostate cancer cell proliferation (8, 17, 18). Experimental studies have shown that LCω3-rich diets have beneficial effects against tumor growth in prostate cancer mouse models (9–11, 13) and patients (19–21). However, animal studies used very large fatty acid dosages, not possible in humans, thus hampering the clinical translatability of findings (22). What is more, the vast majority of studies investigating the effects of LCω3 on prostate cancer generally use fish oil (mix of EPA, DHA and other fatty acids), leaving the effects of individual LCω3 subtypes on prostate cancer largely unknown both clinically and in vivo.

We sought to examine the effects of a clinically relevant dose of purified monoacylglyceride (MAG)-conjugated LCω3 EPA, DHA (MAG-EPA and MAG-DHA) and control ω9-high oleic acid sunflower oil (HOSO), to decipher the anti-cancer function of these molecules in prostate cancer. The putative anti-prostate cancer mechanisms found in our syngeneic immunocompetent prostate cancer mouse model were investigated using unbiased gene expression profiling. In parallel, the anti-angiogenic effect of MAG–EPA was validated in patients with prostate cancer taking a well-tolerated dose of MAG–EPA supplement or placebo (i.e., HOSO) as part of a daily routine prior radical prostatectomy. In the mouse experiments, we also measured the effects of LCω6 arachidonic acid (MAG–AA) supplementation on prostate cancer growth.

PC3 and DU145 human prostate cancer cell lines were directly purchased from the ATCC and cultured in EMEM with 20% FBS (Wisent). TRAMP-C2 cells were obtained from the ATCC and cultured in DMEM media (Wisent); 5% FBS added with 5% NUserum growth medium supplement (Corning), 0.03% insulin (Sigma-Aldrich) and 0.01 nmol/L dihydrotestosterone (DHT). All cell lines used in experiments were Mycoplasmas-free and regularly tested using PCR. Cells were cultured for 8–12 passages after been thawed. For cell proliferation assay 12,500 cells/well were seeded in 12-well plates. The day after, cells were treated with 6.12 and 12.5 μmol/L MAG–EPA and MAG–AA along with HOSO and vehicle control (Dimethyl sulfoxide-DMSO, Sigma-Aldrich). MAG–DHA was used at 3 μmol/L because cell death was observed at higher concentration. Proliferation was measured by cell count using Cytation-5 scanner plate (BioTek). For spheroids assay, 1,500 cells were plated on agarose pre-coated 96-well plates to promote cell–cell adhesion, plates were swirled and the next day each well contained a single spheroid mass. Tumor spheroid size was assessed via bright field imaging using Zeiss Axio Vert or Motic microscope. Spheroid diameter was measured every 48 hours and media changed at the same frequency. For the endothelial cell tube formation assay, immortalized human umbilical vein endothelial vascular cells (HUVEC/TERT2) cells, kindly provided by Dr. Olivier Barbier (CHU de Québec-Université Laval, Canada), were seeded on Matrigel (Geltrex, Thermo fisher). HUVECs were incubated with serum-free conditioned media from prostate cancer cells pretreated with MAG–EPA or control HOSO.

Immunocompetent C57BL/6 mice (purchased from Charles River Laboratories) were injected subcutaneously with TRAMP-C2 prostate cancer cells to evaluate the effect of specific fatty acids on tumor growth. The animal protocol was approved by our institutional review board of CHU de Québec—Université Laval, Canada (#2015112) and was performed in compliance with institutional guidelines and regulations. Briefly, twelve 6–8 weeks male C57BL-6 mice per group were single housed and fed with a low-fat diet (LFD, #D12450H, Research Diet.). Two weeks after LFD, mice were daily supplemented by oral gavage with 618 mg/kg of body weight of purified MAG–EPA (SCF Pharma), MAG–DHA (SCF Pharma), MAG–AA (Nu-Check-Prep Inc.) or HOSO (SCF Pharma). Two weeks after gavage, mice were injected subcutaneously with 2.0 × 10⁶ TRAMP-C2 cells on both flanks and body weight of the animals were taken every week. After occurrence of the initial mass, tumor size was measured every other day and tumor volume was calculated using the formula: (4/3) × 3.14159 × (longest diameter/2) × (shortest diameter/2)⁁2. Mice were sacrificed when combined tumor volume reached 2 cm³. Some mice were sacrificed earlier due to tumor ulceration. Plasma, red blood cells (RBC) and tumors were collected from each mouse at sacrifice and stored at −80°C.

Fatty acid profiles of RBC and TRAMP-C2 cells and tumors were determined by gas-chromatography mass-spectrometry (GC-MS) after total lipid extraction, as described previously (11). Relative abundance of fatty acids is expressed as percentages of total fatty acids for RBC membranes and tumor tissues.

RNA extractions were performed at the Genomics Center of CHU de Québec-Université Laval Research Center, Québec, QC, Canada. Snap-frozen mouse tumors were homogenized using polytron tissue homogenizer (Thomas Scientific). RNA was extracted with the miRNeasy Micro Kit (Qiagen, #217084), treated with DNase I (Qiagen, #79254), and resuspended in RNAse-free water. RNA quality was checked using a TapeStation 2100 (Agilent Technologies). RNA integrity number (RIN) for all samples was >9.0. TruSeq Stranded mRNA Kit (Illumina) was used to prepare mRNA sequencing libraries, according to the manufacturer’s instructions. 1 μg of total RNA was used for poly(A) mRNA selection using oligo-dT–conjugated magnetic beads. Reverse transcription of the fragmented mRNA was done using random primers. This cDNA was converted into double-stranded DNA, end-repaired to incorporate the specific index adaptor for multiplexing, purified with Agencourt AMPure XP beads (Beckman Coulter) and amplified for 15 cycles. The final libraries’ quality was examined using a D1000 DNA ScreenTape assay on a 2200 TapeStation System (Agilent Technologies) and quantified using a QuBit 3.0 fluorometer (Thermo Fisher Scientific) and by qPCR using KAPA library quantification (KAPA Biosystems). RNAseq libraries with unique index were pooled in equimolar ratio (20 samples/pool) and sequenced on an HiSeq 2500 Sequencing System (Illumina) for a paired-ends 125 pb sequencing at the next-generation sequencing platform, CHU de Québec-Université Laval. Reads were quality-trimmed using Trimmomatic v0.32 and aligned to the mm10 build of the murine genome using STAR v2.5.2b. Read counts for the genes described in the 2012-02-07 UCSC annotation of the murine genome were then obtained using htseq-count. Unwanted variation was removed from RNAseq raw counts using the RUVSeq v.1.10.0 package. Differential expression analyses were performed using the DESeq2 v1.16.1 package. Removal of unwanted variation and differential expression analysis were performed in R-3.4.0.

RNA extracted from the samples not analyzed by RNAseq was used to validate the expression of VEGFR2, Lrg1, Mmm2, Cd248, Plvap, DUSP1, Sfp1, Aplnr, c-Jun, and Sntb1 genes using the housekeeping gene b-actin as a reference gene. For HUVECs, the expression of VEGFR2, ANGPT2, and VWF was compared with reference genes; GAPDH, b-actin, and 18S rRNA. Primers were ordered from Integrated DNA Technologies (IDT Company), qPCR reactions were prepared using Advanced qPCR mastermix (Wisent) and cDNA produced using the high-capacity cDNA reverse transcription Kit (Thermo Fisher scientific). The cycling conditions consisted of 1 cycle at 95°C during 5 minutes, followed by 40 three-segment cycles for amplification (94°C 30 sec; 55–60°C 45 sec; 72°C 30 sec). Melting curve was performed at the end of PCR amplification to verify specificity. Experiments were run on CFX96 real-time system C1000 thermal cycler (Bio-Rad). For western blotting, total proteins were extracted using RIPA buffer and samples separated by SDS-PAGE electrophoresis before PDVF transfer and immunoblotting. Protein-bound membranes were incubated overnight with primary antibodies at 4°C, washed with TBS-T and secondary horseradish peroxidase–conjugated goat anti-rabbit or anti-mouse antibodies (The Jackson Laboratory) were incubated to perform ECL-based revelation of specific peptides. The primary antibodies were: Ang-2 (Santa Cruz Biotechnology, sc-74402), GAPDH (Cloud-clone, CAB932Hu01), VEGFR-2 (Novus, NB100-627), b-Actin (Sigma, A2228), and Flag (Fisher, F3165).

Total proteins from TRAMP-C2 tumors, cells and prostate tissue from patients were first extracted using Bio-Plex Cell Lysis Kit (Bio-Rad, #171304011) and total protein concentration measured by DC Protein Assay Kit (Bio-Rad, #5000111). Tumor and normal prostate samples were extracted from whole optimal cutting temperature (OCT)–embedded prostate specimen and punched in the respective compartment previously characterized by a pathologist, then confirmed to be benign or tumoral. Measurements were performed on the whole-tissue punch, without microdissection. To quantify VEGF, the Bio-Plex Pro Mouse VEGF set (#171G6008M) with Bio-Plex Pro Mouse Cytokine Standards Group II (9 cytokines, including VEGF, #171I60001) from Bio-Rad were used according to the manufacturer’s instructions. Bio-Plex assays were run on Bio-Plex 200 Systems and data analyzed using Bio-Plex Manager software 6.1 and GraphPad Prism 7.0c.

Human prostate tumor specimens were obtained from the consecutive first 20 men participating into a phase IIb double-blind randomized controlled trial testing 3 g/d of MAG-EPA versus placebo (HOSO, composed of 80% ω9-monounsaturated oleic acid, 10% saturated fat, 10% ω6-polyunsaturated linoleic acid, no ω3) for 4–10 weeks before radical prostatectomy (n = 10 per group; ref. 23). This clinical study was conducted in accordance with recognized ethical guidelines, approved by our institutional review board of CHU de Québec–Université Laval (2012–1012) and registered to clinicaltrials.gov (NCT02333435). Participants also signed a written informed consent to provide their prostate tissue for additional analyses (2012–671). For the preparation of tissue microarrays (TMA), one to three paraffin blocks containing tumor and/or normal tissue were selected for each patient. Sections of these blocks were stained with hematoxylin/eosin and examined by a pathologist. Tumor areas were selected to represent the dominant and secondary Gleason patterns. To build the TMA, four representative 1.0-mm tumor cores and two normal zones (tumor-proximal and tumor-distal) were taken and placed on a recipient paraffin block with appropriate alignment using a Tissue Arrayer (Beecher Instruments). Sections of 5 μmol/L were cut from the TMA blocks to perform IHC using Dako platform. Sections were deparaffinized, heat-induced antigen retrieval treated with PT Link (PT-11 #PT10027 Dako/Agilent) in Envision Flex Target Retrieval Solution High pH (#K8004, Dako/Agilent). Ready-to-use monoclonal mouse antibody clone JC70A was used for human CD-31 antigen (#IR610, Dako/Agilent) with Autostainer Link48 instrument (#AS480, Dako/Agilent) for 10 minutes. Slides were Harris hematoxylin counter-stained. Digital images of stained TMA slides were obtained at ×20 magnification using a slide scanner (NanoZoomer 2.0-HT, Hamamatsu). TMA map is described in Supplementary Fig. S3D, original images are available at https://doi.org/10.6084/m9.figshare. 13067423.v1

A portion of TRAMP-C2 tumors were OCT compound embedded and stored at −80°C. For IHC, frozen sections (5 μm thick) were fixed with 4% paraformaldehyde and submitted to heat-induced epitope retrieval in presence of citrate buffer (10 mmol/L, pH 6.0). Endogenous peroxidase activity was inhibited using Dako Dual Endogenous Enzyme Block (Dako S2003, Agilent Technologies). Sections were permeabilized with 0.5% triton X-100 in PBS and blocked with 5% serum (#927503, BioLegend), then stained with an anti-CD31 polyclonal antibody (#Ab28364, Abcam). Detection was performed using DAB substrate (Dako #k3468, Agilent Technologies) as per the manufacturer’s instructions. Images were captured using Zeiss Axio-Observer Z1 microscope. For both patients and mice, CD31⁺ blood vessel number and circumference/diameter were measured by two independent and blind reviewers using NPD view and Zeiss ZEN suite, respectively.

Difference of tumor growth between groups was performed using a linear mixed model with repeated measures (LMMRM) analysis. The repeated measured were the tumor growth data obtained at days 14, 18, 20, 22, 24, 26, 28, 30, 32, 36, 38, 40 and 42, respectively. Three intervention groups were noted as MAG-AA, MAG-EPA, MAG-DHA and control group was noted as HOSO. Because of small sample size and missing data observed at baseline tumor data at time 0, to ensure the model to converge, baseline tumor measurements were removed from the model. The model included fixed categorical effects for treatment, time of tumor growth, treatment by time interaction. Statistical modeled for repeated measures data is noted as logY_ijk = n + a_i +b_z+b_ij+g_k+(ag)_ik+e_ijk, where, logY_ijk denotes to the tumor growth measurement at day k for subject j (j = 1…N) assigned to treatment i (i = HOSO, MAG-AA, MAG-EPA or MAG-DHA, respectively). n + a_i +b_z +g_k+(ag)_ik is the mean for treatment i at day k, containing effects for treatment, time and treatment by time interaction. b_ij is a random effect for subject j assigned to treatment i. e_ij refers to the within-subjects error term that assumed a normal distribution. SAS PROC MIXED with restricted maximum likelihood estimation (REML) and an unstructured (UN) within-subject covariance structure were used to account for within-subject correlations. From this model, least squares means, standard errors, treatment differences in least squares means, and 95% confidence intervals were estimated for each time point. Primary inference was based on the pairwise comparison of least squares means from days 14 to 42 from this model between each intervention group and HOSO group, and a P value was presented accordingly for each time point. The null hypothesis is that the mean difference between the two treatment groups is zero, versus the alternative hypothesis that this difference is not zero. The hypotheses can be expressed as follows: H₀: μ_Ti – μ_HOSO = 0 versus H₁: μ_Ti – μ_HOSO ≠ 0, where, μ_Ti refers to the mean tumor growth at each timepoints from days 14 to 42, respectively, for MAG–AA, MAG–EPA or MAG–DHA-treated group and μ_HOSO refers to the mean tumor growth at HOSO-treated groups from days 14 to 42, respectively. The test was performed using the final, LMMRM model-based t test with a two-sided significance level of 5%. Change in mice survival (time to sacrifice) was measured by log-rank test. Differences between fatty acid profiles, cell proliferation, in vitro, mRNA fold changes and blood vessels’ number and size were assessed using two-tailed unpaired student t test. VEGF difference was calculated by the Mann–Whitney test. P values below 0.05 were considered significant. NS or no asterisk = not significant; P ≥ 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001.

Fatty acids, including LCω3, have anti-proliferative activities on prostate cancer cells. Notably, purified DHA and EPA fatty acids can reversibly impair human and mouse prostate cancer cell proliferation (Supplementary Fig. S1A–S1C). However, the maximal recommended daily intake of LCω3 for humans is below most growth inhibitory concentrations tested in vitro (24, 25). Because prostate cancer cells are capable of directly bioaccumulating pure DHA and EPA fatty acids (Supplementary Fig. S1D), we decided to test the effect of a clinically relevant dose of purified LCω3 molecules (618 mg/kg, corresponding to 3 g/d for men) on prostate cancer growth in vivo (experimental design Supplementary Fig. S1E). Syngeneic C57Bl/6 male mice were injected subcutaneously with TRAMP-C2 prostate cancer cells after being fed a defined LFD (Research Diet #D12450H) and receiving daily oral gavage, started two weeks before tumor implantation, of either MAG-EPA, MAG-DHA, MAG-AA or HOSO fatty acids at matched concentrations and continued throughout the study (n = 12 mice/group). Mice had their body weight taken every week (Supplementary Fig. S1F). We first measured the effects of fatty acid supplementation on tumor growth. There was no change for time to minimal mass (palpable tumor) appearance between the fatty acid–supplemented groups, and all mice developed tumors. However, MAG–DHA-supplemented animals had significantly reduced early tumor growth compared with HOSO (Fig. 1A). At later time points, MAG–DHA-supplemented tumors resumed growth and only tumors from MAG–EPA-supplemented mice displayed significantly reduced tumor growth compared with HOSO control (Fig. 1A). In addition, time to sacrifice was significantly longer for MAG–EPA- and MAG–DHA- compared with HOSO-gavaged animals (P = 0.042 and P = 0.040, respectively).

End point fatty acid profiles were generated from RBC membranes (Fig. 1B; Supplementary Fig. S1G) and prostate tumor tissues samples (Fig. 1C; Supplementary Fig. S1H). Both RBC and tumors were significantly enriched for DHA and EPA following targeted supplementation. MAG–AA-supplemented mice showed no significant change in AA levels compared with HOSO but displayed marked increase of LCω6:LCω3 ratio. Conversely, MAG–DHA and MAG–EPA supplementation significantly decreased LCω6:LCω3 ratios in RBC membranes and tumor samples (Fig. 1B and C, black bar graphs). Overall, the results confirmed that a clinically relevant dose of a pure fatty acid can be accumulated in tumors and impacts prostate cancer growth.

To investigate molecular pathways modulated in prostate tumors by clinically relevant LCω3-supplementation dosage, we measured global gene expression using high-throughput RNAseq profiling. Five representative tumor samples from each group were tested. Principal component analysis (PCA) showed that tumors from MAG–EPA-supplemented mice had a distinct mRNA profile from the others (Fig. 2A). Differential gene expression analyses showed that out of all comparisons, MAG–EPA-supplemented mice had significant changes in gene expression (Fig. 2B; Supplementary Fig. S2). Interestingly, the 67 genes with significant lower expression in tumors of MAG–EPA-supplemented mice were mostly enriched for angiogenesis- and vasculature-related genes (Fig. 2C). We then compared the various sets of differentially expressed genes that distinguished MAG-EPA from the other groups (HOSO, MAG–AA, and MAG–DHA). We identified two vasculature-related genes, VEGFR-2 and Lrg1, that were common to all these comparisons (Fig. 2D), suggesting that these two genes are important drivers of the MAG–EPA effect.

Gene expression analysis suggested that MAG–EPA affected angiogenesis and/or vasculature-related pathways in TRAMP-C2 tumors. Using qPCR, we validated the differential expression of several genes modulated by MAG–EPA supplementation (VEGFR-2, Lrg1, Mmm2, Cd24B, Plvap, DUSP1, Aplnr, c-Jun, Sfrp1 and Sntb1) in randomly selected tumor samples from the seven MAG–EPA- and HOSO-supplemented mice not profiled by RNAseq. With the exception of Sntb1 and Aplnr, all other genes were expressed at lower levels in tumors of EPA- compared with HOSO-supplemented mice (Fig. 2E). We also confirmed lower protein levels of angiogenesis-related marker, VEGFR2. Unfortunately, Lrg1 levels were below western blot detection. Because Lrg1 is a proangiogenic molecule implicated in vascular remodeling pathways, we tested the level of another marker of vascular remodeling, angiopoietin-2 (Ang2), and confirmed its depletion in MAG-EPA-treated tumors (Fig. 2F). We also measured VEGF protein levels in tumor tissues, a key cytokine involved in the growth and survival of vascular endothelial cells and development of new blood vessels. As shown in Fig. 2G, VEGF level was significantly lower in tumors of MAG–EPA- compared with HOSO-supplemented mice. To further confirm that the depleted angiogenic program in tumors of MAG–EPA-supplemented mice translated into an altered blood vessel phenotype, tumors were histologically analyzed for blood vessel morphology using a CD31-specific antibody, a marker of endothelial cells (Fig. 2H). This IHC analysis showed that tumors from EPA-supplemented mice presented fewer total blood vessels and significantly smaller blood vessels as well (Fig. 2I, left and right, respectively). Therefore, these results suggest that MAG–EPA supplementation is sufficient to alter tumor-related vascularization.

To test whether the above findings from the prostate cancer mouse model could be translated to patients with prostate cancer, we analyzed tumor vasculature of archived primary prostate tumor tissues of men randomized to MAG–EPA or a placebo before radical prostatectomy (n = 20). These men were recruited in a randomized controlled trial to test the effects of MAG–EPA supplementation on prostate cancer and received 8 weeks on average of either 3 g/d MAG-EPA or placebo (HOSO) per day. As shown in Fig. 3A, IHC staining of prostate tumour tissues displayed smaller blood vessels in MAG–EPA- compared with placebo-supplemented patients (Fig. 3B, left), but the total number of blood vessels was similar between both patient groups (Fig. 3B, middle). In addition, we found that MAG–EPA-supplemented patients had a greater proportion of smaller blood vessels and significantly less medium- and large-sized blood vessels in their tumors compared with those in the placebo group (P = 0.03; Fig. 3B, right). VEGF levels in tumor tissues were not significantly different in MAG–EPA-treated patients (Supplementary Fig. S3A). Interestingly, the effect of MAG–EPA supplementation was tumor-specific because the size of blood vessels in adjacent normal prostate was not significantly different between MAG–EPA- and placebo-treated patients (Fig. 3C). The differences observed were not due to tumor characteristics as the distribution of tumor grades in MAG–EPA and placebo-supplemented patient groups was similar (Supplementary Fig. S3B). In addition, we confirmed that EPA was specifically enriched in RBC membranes and prostate tumor samples of MAG–EPA-supplemented patients compared with placebo (Supplementary Fig. S3C). These results support that MAG–EPA alters the size of blood vessels and thus specifically modulate tumor vasculature in patients with prostate cancer.

Cancer cells promote de novo angiogenesis to sustain their unlimited growth. To reconcile the tumor-specific anti-angiogenic phenotype of MAG–EPA, we tested for a growth factor–dependent effect of MAG–EPA in TRAMP-C2 cells. First, we observed that serum-deprivation induced an increase in VEGF level, a bona fide pro-angiogenic signaling molecule (Fig. 4A). More importantly, MAG–EPA treatment restrained the ability of TRAMP-C2 cells to stimulate VEGF expression in serum-deprived condition, suggesting that MAG–EPA was directly restraining the paracrine activity of cancer cells to foster tumor angiogenesis (Fig. 4A). Next, we tested whether MAG–EPA was sufficient to alter in vitro tumor cell growth independently of vasculature. Using an avascular spheroid assay, we observed that TRAMP-C2 spheroid tumor growth was blocked by MAG–EPA in a dose-dependent manner (Fig. 4B). This effect on tumor growth was confirmed in ω3 fatty acid desaturase 1 (FAT-1)–expressing TRAMP-C2 cells (Supplementary Fig. S4A–S4D), which convert ω6 in ω3 fatty acids. The resulting increase of endogenous ω3 fatty acid levels reversibly reduced prostate cancer cell growth (Supplementary Fig. S4E), blocked tumor spheroid growth (Supplementary Fig. S4D) without affecting cell viability (Supplementary Fig. S4F), altogether supporting that ω3 fatty acids can directly hinder prostate tumor cell growth. VEGF secretion by spheroids was also affected by MAG–EPA treatment (Fig. 4C).

To test the functional relevance of these observations for the activity of MAG–EPA on angiogenesis, we used the in vitro endothelial cell tube formation assay. TRAMP-C2, PC3, and DU145 prostate cancer cells were first treated with MAG–EPA or HOSO in serum-deprived conditions. Subsequently, the prostate cancer conditioned media were tested for its endothelial cell tube formation activity using immortalized HUVEC/TERT2 as described before (26). We measured by qPCR the mRNA expression of genes upregulated during endothelial tube formation in HUVECs, namely ANGPT2, VEGFR2 and Von Willebrand factor (VWF; Supplementary Fig. S4G and S4H). Interestingly, the MAG–EPA prostate cancer conditioned media systematically reduced mRNA levels of all endothelial cell markers (Fig. 4D–F). Media of TRAMP-C2 cells when pre-treated with MAG–EPA depleted mRNA levels of ANGPT2 and VWF (Fig. 4D and E) but caused a non-significant reduction of VEGFR2 compared with HOSO (Fig. 4F). On the other hand, media of human prostate cancer cells when pre-treated with MAG–EPA depleted mRNA levels of VEGFR2 (Fig. 4F). In HUVEC cells alone MAG–EPA treatment had no effect on mRNA levels of the selected endothelial cell markers (Supplementary Fig. S4I), suggesting that MAG–EPA modulates blood vessel remodeling capacity of prostate cancer. Finally, we tested the reversibility of prostate tumor cell growth inhibition from a temporary MAG–EPA treatment. To do so, we reverted MAG–EPA-exposed spheroids with control HOSO. We observed that the growth inhibitory effect of ω3 fatty acids was resumed after the low-dose MAG–EPA concentration was terminated. Interestingly, spheroid growth inhibition sustained after the highest concentration of MAG–EPA was terminated (Fig. 4G and H). Altogether our results suggest that fatty acids regulate cancer-specific vascularization and that a high concentration of MAG–EPA is required to irreversibly block prostate tumor cell growth.

Fish oil is the second most often consumed dietary supplement in adult Americans (27). LCω3 fatty acids contained in fish oil supplements have often been considered beneficial because ω3-rich fish oil extracts displayed anti-inflammatory, anti-angiogenic, anti-apoptotic, and anti-proliferative properties (28, 29). It was generally accepted that a mixture of ω3 molecules was more potent than individual ω3 molecules (30). However, a recent large-scale clinical trial has shown that a single purified ω3 fatty acid, EPA, given at 4 g/d decreased the risk of major adverse cardiovascular outcomes (31), whereas previous trials testing smaller dosage or a mixture of ω3 molecules did not (32). No such trial exists in oncology. We thus tested in a preclinical prostate tumor assay the effects of purified ω3 (MAG–EPA, MAG–DHA), ω6 (MAG–AA), and ω9 (HOSO) fatty acids. Our choice of this HOSO control was planned to dissect the ω3 effects from those of the ω6, typically the sole controls used in previous experiments. Indeed, HOSO contains mostly ω9, very little ω6 and no ω3. HOSO is thus considered a biologically neutral oil. What is more, is has been used in numerous randomized placebo-controlled trials (33–35), as it is odorless, tasteless, and has identical visual appearance. We demonstrated that a clinically relevant dose of purified ω3 supplementation was sufficient to beneficially affect TRAMP-C2 tumor growth. Our data further suggest that MAG–EPA decreases VEGF secretion from tumor cells and limits blood vessels’ size to restrict neovascularization in prostate tumor tissue. We found that a short-term dose of 3 g/d of purified MAG–EPA was sufficient to impact prostate tumor vascularization in clinical settings. Importantly, this anti-vascular effect of MAG–EPA was not observed in normal prostatic tissues, overall hinting at a specific activity of MAG–EPA in the tumor microenvironment. Although the sample size was small, the randomized controlled trial design strengthens the findings. Future experiments with a larger sample size would be useful to further validate the anti-angiogenic properties of MAG–EPA. Also, studying vessel size and other aspects of angiogenesis by tissue staining remains challenging. Further studies are required to determine whether changing in vessel size correlates with clinical outcomes. In prior studies using a different methodology based on CD-34 (36), a larger vessel diameter was considered a favorable finding in response to exercise.

The increase of pro-inflammatory ω6 fatty acids combined with a decrease of anti-inflammatory ω3 fatty acids was suggested to explain, at least in part, the growing prostate cancer incidence in western countries (37, 38). We were surprised to detect a somewhat limited detrimental effect of MAG–AA in our mouse experiments although MAG–AA did increase some cytokine levels in the tumor tissue (Supplementary Fig. S5A and S5B). We also noted an early TRAMP-C2 tumor growth acceleration effect. However, some MAG–AA-treated animals (n = 2) displayed the longest survival, suggesting heterogeneity of the effect. Supplementation of AA alone, may be hampered by the greater prevalence of AA in RBC membranes (18% of total fatty acids) compared with other ω6 fatty acids (linoleic acid 7% and adrenic acid 2%) in our experiments. In fact, we did not measure a significant enrichment of AA in RBC and prostate tumors, but observed a significant increase of the ω6:ω3 ratio supporting that MAG–AA supplementation was incorporated. It is possible that the harmful effect of MAG–AA was mitigated by feeding animals with a defined LFD, containing some AA (0.10 g AA/Kg diet, Research Diet; ref. 39) compared with feeding a grain-based or chow diet, poor in AA (AA undetectable in diet, Teklad), common to many other studies. Interestingly, patients from our cohort had a favorable ω6:ω3 ratio (mean of 3.7) at baseline. Many patients with prostate cancer spontaneously adopt a healthier nutritional lifestyle following a cancer diagnosis (40, 41). Complicating interpretation even more, participants in clinical trials often have a healthier lifestyle than non-participants (42). Regardless, in both mice and humans, we observed that prostate tumor vascularization was specifically affected by MAG–EPA supplementation. All things considered our results support that patients with prostate cancer may benefit from a single LCω3 supplementation independently of their ω6:ω3 ratio.

It is still debated whether prostate cancer cells can directly uptake, store or process supplemented fatty acids. Watt and colleagues (38) recently found that fatty acid uptake and storage are increased specifically in malignant prostate cancer tissues compared with benign adjacent tissues. Other groups also showed that in vitro, prostate cancer cells capture fatty acids from the media (43). Accordingly, when directly added to the media, we were able to detect specific bioaccumulation of each supplemented fatty acid by prostate cancer cells (Supplementary Fig. S1D). As reported in the literature, we detected DHA and EPA accumulation in RBC of laboratory animals and patients (44–46). We also observed a significant accumulation of DHA and EPA in ectopic TRAMP-C2 mouse prostate tumors and increased EPA levels in human tumor tissues. Steady-state accumulation of fatty acids in tumors might not be a good proxy for compliance because of the putative changes in fatty acid metabolism. It is possible that the actual prostatic metabolism of fatty acids, which likely vary across subjects as we observed (8), could better describe bioaccumulation of fatty acids and possibly serve as a risk stratification biomarker to select patients that would benefit from targeting ω3-related pathways. However, measuring prostatic metabolism is currently difficult to perform, lacks validation, and is not clinically available yet. Regardless, we showed that a purified fatty acid can be successfully accumulated in the prostatic tissue.

Because our data showed that purified MAG–EPA inhibited the growth of TRAMP-C2 tumors compared with HOSO, we wanted to further explore reversibility of the anti-cancer effect from this naturally occurring molecule. To our surprise, cultured cells were able to resume complete growth after LCω3 removal, suggesting that growth-repressive activities of MAG–EPA are dependent on their constant bioavailability. We also showed that ω3 endogenously produced from Fat-1–expressing cells caused fully reversible growth arrest. In line with this, our spheroid experiments demonstrated that a greater dose of MAG–EPA irreversibly arrested spheroid tumor growth, suggesting that increasing MAG–EPA dosage may improve durability of the protective effect.

To identify pathways affected by MAG–EPA accumulation in tumors, we performed global gene expression analysis. Gene enrichment analysis showed that angiogenesis and vasculature development functions were those mostly affected by MAG–EPA. A large body of literature supports that ω3 molecules have anti-angiogenic properties. An in vitro study of human endothelial cells showed that ω6 fatty acids stimulate and ω3 fatty acids inhibit pro-angiogenic processes by inhibiting Ang2, matrix metalloproteinase-9 (MMP-9) and endothelial invasion (47). A study of solid mammary tumors showed that LCω3-rich fish oil or a mixture of fish oil and safflower oil significantly reduced blood vascular area, mast cells and macrophage infiltration, which ended in slower tumor growth (48). What is more, a growing body of evidence mostly in cancer sites other than prostate, support the tumor anti-angiogenesis effects of ω3 fatty acids that inhibit the production of many important angiogenesis mediators such as VEGF, platelet-derived growth factor (PDGF), platelet-derived endothelial cell growth factor (PD-ECGF), COX-2, prostaglandin-E2 (PGE2), nitric oxide, NFκB, MMPs and beta-catenin (28). Yet, we found no study discriminating the effects of individual LCω3 molecules, like we did, nor could we find one exploring the mechanisms linking LCω3 to the cancer-specific blood vessel morphogenesis in prostate cancer patients or models.

We present both preclinical and clinical data demonstrating the effects of MAG–EPA supplementation on tumor blood vessels, describing molecular interactions between MAG–EPA and the vascular system. MAG–EPA decreased VEGF secretion by prostate cancer cells and affected their paracrine activity toward endothelial cells, suggesting that MAG–EPA can restrain the capacity of prostate cancer cells to signal new blood supply. Another LCω3 molecule, docosapentaenoic acid, was described as having vascular remodeling activity in animal models of pulmonary hypertension notably by reducing VEGF level (45). In addition, DHA was also shown to reduce cell proliferation of human pulmonary arteries, altogether supporting neovascular inhibition by LCω3 (49). On the basis of the effect of these various LCω3, we could have expected a similar anti-angiogenic activity for other LCω3. We observed that MAG–DHA blocked as strongly as MAG–EPA prostate cells growth in vitro. MAG–DHA reduced tumor size at early time point; however this growth-reducing effect was mitigated at later prostate cancer growth stages, whereas MAG–EPA showed stronger effect at later time points. Taken together, our data support that MAG–EPA modulates tumor blood vessels and further suggest that LCω3 could have complementary repercussions on prostate tumors that would advocate for testing MAG–EPA and MAG–DHA in combination.

Interestingly, the vast majority of genes enriched in MAG–EPA mouse tumors were coding for secreted proteins, suggesting that MAG–EPA directly affected paracrine signaling from cancer cells. In addition to the diffusion of proangiogenic molecules, tumor cells reroute existing blood vessels to allow for new nutrients to reach actively growing cancer cells, clear growth inhibitory compounds and thus sustain unrestricted tumor growth (50). Animal modeling of prostate cancer showed that a VEGFR1–VEGFR2 (kdr) angiogenic switch was associated with the transition from latent to aggressive tumor growth (51, 52) with similar observations made in humans (53, 54). In addition to reduced VEGF levels, we observed morphological reduction of blood vessels specifically in the tumor microenvironment of MAG–EPA-treated patients but not in their normal prostate epithelium. It is noteworthy that MAG–EPA could exert its effect via the adjacent normal epithelium by prohibiting rearrangement of normal blood vessels toward actively growing tumors. Recent data suggest that proangiogenic alteration of the adjacent stromal compartment could fuel tumors (50, 55–57). Altogether our results support that in addition to reducing pro-angiogenic signaling, MAG–EPA could decrease blood vessels’ remodeling in tumor microenvironment thereby hindering further tumor growth.

Some limitations of our study are worth mention. Our experimental design used pre-treatment with fatty acids before tumor implantation and therefore is not a classical treatment study design. Rather, our design suggests that fatty acids may be used as a prevention strategy to inhibit tumor neovascularization and thereby prevent prostate cancer tumor progression. Indeed, we have not formally examined if the fatty acids can be used as a treatment strategy once the tumor is established. Yet, our reversal experiments suggest that MAG–EPA given at a later growth stage also blocks tumor growth at that point, suggesting efficacy at any tumor size. Including tumor weights could have been a useful addition but was not performed. As in many similar studies, we measured tumor size, a good proxy for tumor weight given the uniform histology of this model.

In conclusion, our results show that LCω3 supplements of purified MAG–EPA reduce TRAMP-C2 tumor growth. MAG–EPA supplements directly modulate proliferation of prostate cancer cells and tumor vasculature by reducing blood vessels’ size specifically in tumor microenvironment. This key phenotype was validated in patients undergoing prostatectomy taking MAG–EPA supplements in a phase IIb randomized placebo-controlled trial. Our findings warrant further examination of MAG–EPA supplements to reduce prostate cancer growth by modulating neovascularization.

A. Marette reports grants from Danone, as well as personal fees from Plexus, grants and personal fees from Acasti and Allysta, and grants from Pfizer outside the submitted work. A. Bergeron reports other from SCF Pharma, as well as grants from CUOG-Astellas during the conduct of the study, grants and personal fees from Merck, and grants from IMV Inc. and Astellas outside the submitted work. V. Fradet reports other from SCF Pharma, as well as grants from CUOG-Astellas during the conduct of the study, grants from Sanofi, Astellas, and non-financial support from Ferring outside the submitted work. No disclosures were reported for the other authors.

N. Gevariya: Conceptualization, formal analysis, funding acquisition, validation, methodology, writing–original draft. G. Lachance: Conceptualization, analysis, validation, visualization, methodology, writing–original draft. K. Robitaille: Conceptualization, formal analysis, supervision, validation, visualization, methodology, writing–original draft, project administration. C. Joly Beauparlant: Data curation, formal analysis, writing–review and editing. L. Beaudoin: Formal analysis. E. Fournier: Formal analysis. Y. Fradet: Conceptualization, supervision, investigation, writing–review and editing. A. Droit: Data curation, supervision, methodology, writing–review and editing. P. Julien: Formal analysis, methodology, writing–review and editing. A. Marette: Resources, supervision, methodology, writing–review and editing. A. Bergeron: Conceptualization, supervision, funding acquisition, methodology, writing–review and editing. V. Fradet: Conceptualization, supervision, funding acquisition, validation, investigation, methodology, writing–original draft.

The authors thank Dr. Samuel Fortin from SCF Pharma (Ste-Luce, Canada) for kindly providing MAG-EPA, MAG-DHA, and HOSO. The authors thank Hui Wen Xu for help with statistical analyses, Myriam Tremblay for help in collecting and storing biological samples as well as Hélène Hovington for TMA preparation. The authors also thank the staff of the animal facility of Centre de recherche de l’IUCPQ—Université Laval, members of A. Marette laboratory as well as Valérie Picard for providing expert technical assistance. The authors thank Line Berthiaume for performing fatty acid profiling, Nathalie Paquet for RNA extraction and Martine Dumont for RNA sequencing, all performed at the Centre de recherche du CHU de Québec-Université Laval. They finally thank all urologist–oncologists and patients for their participation in research projects. This work was supported by the CUOG-Astellas Research Grant Program, managed by the Canadian Urological Association and the Canadian Urological Oncology Group with funding provided by Astellas Pharma Canada, Inc. through a grant to the CUA-CUOG. This grant was awarded to N. Gevariya and V. Fradet. N. Gevariya was supported by Prostate Cancer Canada PhD scholarship. V. Fradet is a recipient of Fonds de recherche du Québec—Santé (FRQS) clinician—scientist career award.

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.