Chemopreventive Effects of Dietary Eicosapentaenoic Acid Supplementation in Experimental Myeloid Leukemia

Leukemia remains one of the leading causes of cancer death, despite established well-defined standards of care. Each year, an estimated 43,000 new cases of leukemia will be diagnosed and 22,000 people will die from the disease (1). This observation underscores the need for further research into alternative therapies. Previously, we have demonstrated that certain cyclopentenone prostaglandin (CyPG) metabolites of polyunsaturated fatty acids (PUFA) are effective in treating murine models of leukemia (2). While PUFAs composed of omega-6 and omega-3 fatty acids are essential for healthy growth and development, the ratio of these lipids is of particular interest in disease (3). This is because prostanoids produced from eicosapentaenoic acid (EPA, C20:5; n-3) show increased anti-inflammatory or “pro-resolving” functions when compared with some of their omega-6–derived counterparts (4–6). Furthermore, alternative pathways produce other bioactive anti-inflammatory lipid mediators, including resolvins and protectins, from EPA and docosahexaenoic acid (DHA, C22:6; n-3), respectively (7–9).

Membrane incorporation of dietary EPA provides the substrate for the production of CyPG, Δ¹²-PGJ₃ that is of particular interest in our laboratory, given its antileukemic properties (2). Release of EPA from the membrane by phospholipase A₂ commits the PUFA to CyPG synthesis through COX activity; EPA is converted to PGH₃ (10). PGH₃ is acted upon by prostaglandin synthases, such as PGD synthases, hematopoietic (H-PGDS) and lipocalin type (L-PGDS), to produce PGD₃ that undergoes dehydration and isomerization to Δ¹²-PGJ₃. However, in cells, arachidonic acid (ARA, C20:4; n-6) is the preferred substrate for COX2, whereas EPA is used at approximately 30% efficiency of ARA by COX2 (11). Thus, to increase production of 3-series prostaglandins, EPA must be saturated in the membrane by increasing intake of omega-3 fatty acids. It is therefore conceivable that inhibition of COX2 by nonsteroidal anti-inflammatory drugs (NSAID) or H-PGDS by HQL-79 can impact leukemia progression. However, the results of epidemiologic studies are inconclusive regarding the relationship between non-aspirin NSAID use and incidence of leukemia (12–14), necessitating further research before embarking on new studies.

Myeloid leukemias are myeloproliferative disorders that arise from mutations within the hematopoietic stem cell (HSC) population in the bone marrow (15). These mutations lead to defects in the generation of healthy myeloid lineage cells. In chronic myelogenous leukemia (CML), the Philadelphia chromosome is the result of a 9:22 chromosomal translocation, which leads to a constitutively active tyrosine kinase fusion protein BCR-ABL that is found in more than 90% of cases (15, 16). Current therapy for BCR-ABL⁺ CML includes treatment with tyrosine kinase inhibitors (TKI). However, this maintenance therapy has various negative side effects and targets bulk tumor cells along with healthy cells while leaving leukemia stem cells (LSC) unharmed. Disease relapse occurs due to the ability of LSC to self-renew and differentiate into bulk tumor cells (16, 17). In addition, many patients become resistant to the therapy (15, 18, 19). The T315I mutation in BCR-ABL confers resistance to nearly all TKI therapies, except ponatinib (20). Finding a new mechanism that is independent of ABL kinase inhibition is essential to eliminating the cancer cells. Acute myeloid leukemia (AML) is characterized by immature blood cells that grow quickly and progress to leukemia in a few months. However, AML is a more heterogeneous disease with multiple etiologies including translocations. One common MLL translocation generates the MLL-AF9 fusion protein that can confer self-renewal ability on hematopoietic progenitor cells leading to aggressive leukemogenesis (21).

In the present study, we show that EPA supplementation of mice at pharmacologic doses alleviates symptoms and signs of leukemia in models of BCR-ABL CML (hereafter referred to as CML), T315I CML, and MLL-AF9 AML. Our data further support the key role of COX/H-PGDS pathway–derived endogenous metabolite of EPA to specifically target the LSC population to correct leukocytosis, splenomegaly, and overall disease index in these models of experimental leukemia.

Three-week-old C57BL/6 mice (Taconic Biosciences) were randomly assigned to 2 dietary groups: animals were housed 3 to 4 mice per cage in a temperature- and humidity-controlled room with a 12-hour light–dark cycle. All mice were 11 to 16 weeks of age at sacrifice, unless noted otherwise. For most experiments, total animal number was 8 to 12; this was the summation of 3 to 4 independent experiments run in at least biological triplicate. The Institutional Animal Use and Care Committee (IACUC) at The Pennsylvania State University (University Park, PA) preapproved all procedures.

An energetically equivalent mouse diet, in terms of omega-3 fatty acid, equivalent to the FDA-approved omega-3 supplement Lovaza was developed (22). Diets remained isocaloric to the base diet (AIN-76A; Research Diets, Inc.) with 5g of total fat (12% of total calories). A portion of the corn oil in the base diet was replaced with the ethyl esters of oleic acid (OA; C18:1 n-9) (100%) or EPA (>90%; Nu-Chek Prep) at a level of 0.82% (w/w), or 1.8% of total energy (kcal; Supplementary Table S1). The OA diet was used as an experimental control diet. The diets were stored in individual portions in airtight Whirl-Pak bags at −80°C until the day of feeding (to prevent oxidation). The mice were fed ad libitum with fresh diet daily. In all experiments, mice were sacrificed after overnight fast (12–16 hours).

Fatty acid methyl esters of the tissue phospholipids were prepared and analyzed as described previously (22), with the following modifications. Briefly, tissues were homogenized in ice-cold saline. Three milliliters of chloroform–methanol (1:2, v/v) was added to each homogenate, and lipids were extracted with chloroform (1 mL) plus saline (1 mL), followed by chloroform (1 mL; ×2). The internal standard 1,2-diheptadecanoyl-sn-glycero-3-phosphocholine (17:0; Avanti Polar Lipids) was added to each sample prior to lipid extraction. The pooled chloroform extracts were evaporated, resuspended in a small volume of chloroform, and subjected to thin layer chromatography (TLC). The phospholipids were separated by TLC on silica gel 60 HPTLC plates (Merck) with chloroform/methanol (8:1 v/v) as the solvent system (phospholipids remained at the origin). Bands corresponding to the phospholipids were scraped, transferred to screw-topped test tubes, and saponified in the presence of 14% BF₃ in methanol (Thermo Scientific) at 86°C for 10 minutes, followed by the addition of 0.9% saline (2 mL). Fatty acid methyl esters were isolated with equal volumes of hexane (×2), pooled, and evaporated under nitrogen. The fatty acid methyl esters were resuspended in hexane and analyzed using a Hewlett Packard model 5890 series II gas chromatograph equipped with flame ionization detector and a DB23 fused silica capillary column (0.25 mm × 30 m; Agilent Technologies) with hydrogen as the carrier gas, and temperature programming from 160°C to 250°C at 3.5°C/min. The fatty acid methyl esters were identified by comparing the retention times with those of known standards (NuChek Prep). The fatty acids are presented as mole percent.

Murine macrophage-like cell line RAW 264.7 (ATCC) was cultured as previously described (23). In brief, cells were cultured in DMEM containing 5% FBS (Atlanta Biologicals), 2 mmol/L l-glutamine, 100 units/mL penicillin G (CellGro), and 100 μg/mL streptomycin (Invitrogen). For routine culture, the cells were passaged every 3 days at a ratio of 1:10 as per the recommendations by ATCC. For experimental assays, 100-mm plates were seeded with 2 × 10⁶ cells. Exogenous fatty acid treatment began when cells were approximately 40% confluent. EPA or OA as ethyl ester (Cayman Chemicals) was incubated with cell culture grade and endotoxin-free bovine serum albumin (BSA; Sigma-Aldrich) in serum-free DMEM for 1 hour at 37°C. Final concentration of EPA or OA in culture was 50 μmol/L conjugated to 1% (v/v) BSA. Cells were treated with fatty acid for at least 3 days, with media changed daily, prior to 50 ng/mL lipopolysaccharide (LPS) stimulation for 12 hours. In cultures treated with 10 μmol/L indomethacin (Cayman Chemicals), the NSAID was added at least 2 hours prior to addition of fatty acid (as EPA or OA conjugated to BSA). Indomethacin was added to culture daily and was present during LPS stimulation.

Protocol for harvesting and culturing bone marrow–derived macrophages (BMDM) was modified from previously described work from our laboratory (23, 24). In short, BMDMs were prepared by extracting femoral bone marrow plugs from C57BL/6 mice maintained on either EPA- or OA-supplemented diet (as described above) for 8 weeks. These cells were cultured in the same medium as described earlier with the addition of 20% (v/v) L929 fibroblast–conditioned medium (as a source of macrophage colony-stimulating factor, M-CSF) for at least 3 days. No additional fatty acid was added when culturing BMDMs. On day 3, LPS (50 ng/mL) was added for 12 hours. Clarified media were collected for lipid extraction as described below.

The estimation of Δ¹²-PGJ₃ was carried out as reported previously (25). Briefly, Δ¹²-PGJ₃ was quantified using an HPLC system consisting of LC-20AD UFLC pumps with a SIL-20AC autosampler (Shimadzu Corporation), a Luna phenyl-hexyl analytical column (2 × 150 mm², 3 μm; Phenomenex) developed with a 30-minute isocratic elution with methanol/water (70:30 v/v) containing 0.1% acetic acid at a flow rate of 150 μL/min and injection volume of 50 μL. The negative ion electrospray tandem mass spectrometric analysis was conducted using API 2000 triple quadruple mass spectrometer (AB Sciex) at unit resolution with multiple reaction monitoring mode (MRM). The source temperature was maintained at 450°C, electrospray voltage was −4,500 V, and the declustering potential was set at −16 V. Nitrogen was used as collision gas at −20 eV, and the dwell time was 150 ms/ion. During MRM, Δ¹²-PGJ₃ was measured by recording the signal for the transition of the deprotonated molecule of m/z 331 to the most abundant fragment ion with m/z 269. Data were acquired and analyzed using Analyst software program version 1.5 (AB Sciex).

To induce experimental CML and T315I CML, retroviral stocks were generated by transfecting HEK293T cells with either pMIG-BCR-ABL (a gift from Dr. Warren Pear, University of Pennsylvania, Philadelphia, PA) or MIGR-BCR-ABL-315T/I (a gift from Dr. Ravi Bhatia, University of Alabama, Birmingham, AL), respectively, with pEco for packaging using TransIT 293 reagent (Mirus Bio). For experimental AML, pMIG-MLL-AF9 (a gift from Dr. Hong-Gang Wang, Penn State College of Medicine, Hershey, PA) was used. Isolation and transduction of HSCs was performed as described earlier (2, 26). HSCs transduced with BCR-ABL virus are hereafter referred to as BCR-ABL LSCs (BCR-ABL⁺GFP⁺Kit⁺Sca1⁺Lin⁻). T315I CML LSCs are similarly defined. AML LSCs have not been defined with regard to cell surface markers. For all AML experiments, total bone marrow was transduced and transplanted (1 × 10⁶ total transduced cells) to establish the model. Studies using GFP (C57BL/6-Tg(UBC-GFP)30Scha/J; ref. 27) bone marrow transplanted directly into OA- and EPA-supplemented mice clearly indicated that EPA does not negatively impact engraftment (Supplementary Fig. S1).

BCR-ABL LSCs were isolated from spleen and bone marrow by FACS using the Influx Cell Sorter (BD Biosciences) or Astrios Cell Sorter (Beckman Coulter) as described earlier (2, 28). LSCs (5 × 10⁵ cells) in 100 μL of sterile PBS were transplanted by retro-orbital injection into irradiated (450–950 rads) recipient mice maintained on EPA or OA diet groups for at least 8 weeks. Disease progression was monitored by complete blood count (CBC) analysis, evaluating peripheral blood for leukocytosis on a Hemavet 950FS equipped with a veterinary software program. All mice were sacrificed by day 14 posttransplant. Spleen and bone marrow were isolated and used for characterization as described below. Any deviation from this transplantation system will be described in figure legends. In all experimental models, transduced bone marrow cells were passaged through mice at least 2 times prior to transplantation into OA- or EPA-supplemented mice.

Inhibition of COX1/2 was accomplished by use of indomethacin (Cayman Chemicals). Indomethacin (0.00325%, w/v) was dissolved in ethanol (0.5%, v/v) and administered by means of oral drinking water for 1 week prior to bone marrow transplantation with BCR-ABL LSCs, as described previously (23, 28, 29). In a subset of Δ¹²-PGJ₃ rescue experiments, administration of exogenous Δ¹²-PGJ₃ was performed as described previously (2). In brief, mice were given daily intraperitoneal (i.p.) injection of Δ¹²-PGJ₃ (0.025 mg/kg) in 500 μL of sterile PBS for 7 days, starting at day 7 posttransplant until sacrifice at day 14. HQL-79 (Cayman Chemicals) was used to inhibit the activity of H-PGDS. HQL-79 (30 mg/kg body weight) was administered by intraperitoneal injection every other day (3–4 times a week) from day zero (day of transplant) until day 14 (total of 7 injections). HQL-79 was resuspended in a solution of 0.05 mmol/L citric acid and 11.1 mmol/L hydroxypropyl β-cyclodextrin (HPBCD; Sigma) and incubated for >10 minutes at 37 °C, until dissolved. Mice were injected with this formation within 10 minutes of preparation. All mice were euthanized on day 14 posttransplant.

Whole spleen and bone marrow were collected and red blood cells (RBC) were lysed with ACK lysis buffer (1.5 mmol/L NH₄Cl, 100 μmol/L KHCO₃, 10 μmol/L EDTA-2Na). Lineage-negative cells were isolated following manufacturer's protocol (Stem Cell Technologies). Cells were stained for Ly-6A/E (sca-1) and CD117 (c-kit; BD Biosciences). Lineage-negative WBCs and the presence of LSCs were analyzed on Accuri C6 or Fortessa LSR flow cytometers (BD Biosciences). KSL (c-kit⁺sca-1⁺Lin⁻) LSCs were defined as Lin⁻ selected cells that stained positive for c-kit, sca-1, and GFP (present in BCR-ABL plasmid).

Total splenocytes or bone marrow from leukemic mice were plated in M3231 Methocult (Stem Cell Technologies). Cell density is described in figure legends. Media were supplemented with IL3 (2.5 ng/mL; R&D Systems), stem cell factor (SCF; 50 ng/mL; Gold Biotechnology), growth differentiation factor 15 (GDF-15; 30 ng/mL; Biomatik), and sonic hedgehog (Shh; 25 ng/mL; Gold Biotechnology). On days 7 to 10, colonies were counted. LSC-CFU (colony forming unit) were defined on the basis of their size (large, >100 cells at high magnification), shape (perfectly circular), density (dark brown/black color), and edge (clearly defined edge with limited differentiation). Any variation from LSC-CFU formation was also identified: in some cases, granulocyte-macrophage progenitor colony-forming units (CFU-GM) were observed (30). Each biological sample was plated in triplicate. The average number of LSC colonies from each biologic sample was tabulated for analysis.

Whole spleen tissue was fixed in 4% paraformaldehyde at time of sacrifice. Sections were prepared (5-μm thickness) and stained with hematoxylin and eosin (H&E) at the Animal Diagnostic Laboratory, Penn State University. Experimental groups were blinded before being examined and scored by a board-certified laboratory animal veterinarian with training in pathology. Six characteristics were evaluated: disruption of splenic architecture (DSA), tumor infiltrates (TI), mitotic figures (MF; scored as 0, rare to none; 1, 1–3 per high power field (hpf); 2, 3–6 per hpf; 3, 6–9 per hpf; 4, >7 per hpf), megakaryocytes, extramedulary hematopoiesis (EMH), and necrosis. With the exception of MF, tissue characteristics were scored on a 0 to 4 scale as follows: 0, within normal limits; 1, minimally increased/elevated/affected; 2, mildly increased/elevated/affected; 3, moderately increased/elevated/affected; and 4 = extensively or severely increased or affected.

TUNEL assay was subsequently performed on representative tissues following histopathological examination using the In Situ Cell Death Detection Kit (Roche Diagnostics Corp.). Tissue sections were deparaffinized in Histo-Clear (National Diagnostics), washed in 100% ethanol, 95% ethanol, 70% ethanol, and distilled water. Slides were then boiled in 10 mmol/L sodium citrate, pH 6.0, with 0.05% Tween-20 and allowed to cool. TUNEL staining was performed according to the manufacturer's instructions. Images were collected using an Olympus BX51 fluorescence microscope (Olympus America). Total fluorescent area (TUNEL⁺ cells) of stained slides was determined using ImageJ software program (NIH, Bethesda, MD).

Statistical analysis was performed using GraphPad Prism version 6 (GraphPad Software). Unless noted, nonparametric tests, including one-way ANOVA and unpaired one-tailed Student t test, were used where appropriate. The OA group was used as the control, unless otherwise noted, for t tests (comparing EPA only with OA only). For ANOVA, all groups were compared. Variation from these analyses is described in figure legends. Bars represent mean ± SEM: *, P < 0.05; **, P < 0.01; ***, P < 0.001; #, P < 0.0001. The Biostatistics, Epidemiology, and Research Design (BERD) Core of Penn State Clinical and Translational Sciences Institute (CTSI) provided statistical consulting.

Two experimental diets were designed such that the only difference between dietary groups was the replacement of a subset of the base diet (AIN76-A) corn oil with either EPA or OA as ethyl esters at 0.82% (w/w) or 1.8% of total kcal (Supplementary Table S1). Mice on either diet consumed a similar amount of diet and gained weight at the same rate (Fig. 1A). A widely accepted method for assessing omega-3 status is the omega-3 index, which is the relative abundance of EPA + DHA in RBCs (31). To examine whether EPA supplementation led to changes in the membrane lipid composition in the spleen, liver, and RBCs, the omega-3 index was determined by gas chromatography-mass spectrometric (GC-MS) analysis (22, 31). The amount and type of omega-3 fatty acid esters increased with EPA supplementation, as shown by the membrane lipid composition of the spleen, liver, and RBCs (Fig. 1B). Compared with OA control diet, EPA supplementation led to a significant increase in the omega-3 index in the membrane of all examined tissues (Fig. 1C; and Supplementary Table S2). The content of omega-3 in the liver amounted to 19.4% of the membrane with EPA supplementation versus 7.8% in OA-supplemented mice, indicating a 2.5-fold increase. In the spleen, there was a 3.5-fold increase in the omega-3 in EPA mice; the omega-3 index was 11.1% in EPA mice of compared with 3.2% in OA mice. The largest increase was in the RBCs where EPA supplementation increased the omega-3 index to 5.5% versus 1.0%, a 5.5-fold increase. These data demonstrate that an EPA-supplemented diet increases the omega-3 content of cells, which will allow us to determine whether these changes lead to changes in Δ¹²-PGJ₃ production, which in turn will affect CML and AML progression.

On the basis of our previously reported studies demonstrating the ability of Δ¹²-PGJ₃ to specifically target LSCs (2), endogenous production of Δ¹²-PGJ₃ was measured. EPA-supplemented cells and mice were treated with LPS to mobilize membrane-esterified EPA and to upregulate COX2 activity (Fig. 2). As expected, EPA supplementation without LPS stimulation did not lead to detectable levels of Δ¹²-PGJ₃ in either RAW 264.7 cells or BMDMs (data not shown). Primary BMDMs from mice on either OA- or EPA-supplemented diets were cultured ex vivo and stimulated with LPS. No additional fatty acid was added to the cultures. BMDMs derived from EPA-supplemented mice produced significantly higher levels of Δ¹²-PGJ₃ than OA-derived BMDMs (Fig. 2A). We next tested whether Δ¹²-PGJ₃ production was COX-dependent. RAW 264.7 cells treated with EPA generated Δ¹²-PGJ₃, in response to LPS. In contrast, cells treated with BSA (used as a carrier) alone or OA-conjugated BSA failed to produce any Δ¹²-PGJ₃ (data not shown). As shown in Fig. 2B, indomethacin treatment blocked the production of Δ¹²-PGJ₃ by EPA-treated RAW 264.7 cells. Although EPA-treated macrophages produced Δ¹²-PGJ₃ in vitro in response to LPS treatment, we needed to determine whether mice fed on EPA-supplemented diet produced significantly higher Δ¹²-PGJ₃, in vivo, following LPS injection. Lipid extracts were isolated from plasma and subjected to LC-MS/MS analysis (Fig. 2C). While there was a nonsignificant increase in Δ¹²-PGJ₃ in EPA mice versus OA mice without LPS treatment, stimulation with LPS for 12 hours led to a 3.5-fold increase in Δ¹²-PGJ₃ in the plasma of EPA-supplemented mice (Fig. 2C). Taken together, these results correlate an increase in the omega-3 index in EPA-fed mice with increased endogenous Δ¹²-PGJ₃ production.

To relate increases in the omega-3 index and the ability of in vivo EPA supplementation to form endogenous Δ¹²-PGJ₃ during CML, mice on the specific diets were given a bone marrow transplant of BCR-ABL LSCs. As reported previously, transplantation with BCR-ABL LSCs leads to leukocytosis and splenomegaly (2, 28, 31). Interestingly, mice on an EPA-supplemented diet showed less severe symptoms of BCR-ABL–induced CML. Splenomegaly was decreased in EPA mice when compared with OA mice (Fig. 3A and Supplementary Fig. S2A), whereas healthy mice on diet showed no difference in spleen weight (Supplementary Fig. S1A). Furthermore, H&E-stained sections of whole spleen were scored on the basis of six parameters of tissue health as mentioned in Materials and Methods (Fig. 3B). OA-supplemented LSC-transplanted mice had deterioration of normal spleen architecture with no clear demarcation between the red and white pulp areas (Fig. 3B and Supplementary Fig. S2B). In contrast, H&E staining of splenic sections prepared from EPA-supplemented LSC-transplanted mice showed normal structure in the spleen with well-demarcated red and white pulp areas (Fig. 3B and Supplementary Fig. S1B and S2B). To further demonstrate the effect of EPA in BCR-ABL–induced CML, we examined the BCR-ABL LSC population in the spleen and bone marrow by flow cytometric analysis on day 14 posttransplantation. In both the spleen and bone marrow, OA-supplemented mice showed an approximate 2-fold increase in the percentage of LSCs compared with EPA-supplemented mice (Fig. 3C). This increase in BCR-ABL LSCs was further corroborated with ex vivo investigation of BCR-ABL LSC-CFU from the spleen and bone marrow. EPA mice failed to produce LSC-CFU colonies, whereas OA mice formed numerous LSC-CFUs (Fig. 3D). EPA mice exhibit dense colonies consistent with CFU-GM progenitor colonies (30). These assays showed that there were not only a limited number of LSCs in the EPA-supplemented group (as shown from flow cytometric results) but also these cells were not functional nor able to expand ex vivo.

A defining characteristic of LSCs is their ability to cause relapse of disease upon serial transplantation (32, 33). To further test the role of EPA in limiting experimental BCR-ABL–induced CML, whole bone marrow from BCR-ABL LSC primary transplanted OA- or EPA-supplemented mice was used for a secondary transplant in mice fed on normal chow (Supplementary Fig. S2C). The disease was allowed to progress for 1 month prior to analysis. This time point was chosen on the basis of humane endpoints. At the time of sacrifice, recipients of leukemic OA bone marrow had lost 20% of body weight (Fig. 3E). Bone marrow from EPA-supplemented BCR-ABL leukemic mice caused less severe disease as compared with OA-supplemented mice as shown by inhibition of leukocytosis (Fig. 3F and Supplementary Fig. S1C) and reduction in splenomegaly (Fig. 3G). Mice receiving OA leukemic bone marrow showed increased proliferation of splenocytes into LSC-CFUs (Fig. 3H). The number of colonies was too numerous to count (TNC) in all but one OA group, whereas the EPA group consistently showed limited cell growth (Fig. 3H). Taken together, these results supported our findings that EPA supplementation limited the LSC population in the spleen.

The T315I mutation in BCR-ABL is resistant to most current TKI therapeutics (20). These therapies are not only ineffective but also extremely damaging to healthy cells. EPA-supplemented mice transplanted with T315I bone marrow showed a decrease in disease, as compared with OA controls (Fig. 3I and J and Supplementary Fig. S2D–S2F). Mice were monitored for overall survival and were euthanized on the basis of humane endpoints, including drastic changes in behavior (data not shown) and in body weight (Supplementary Fig. S2D). The experiment was ended at day 18 posttransplant. EPA mice showed a significant decrease in splenomegaly at time of sacrifice (Fig. 3I). Of particular interest, the EPA diet significantly limited the LSC population as seen from LSC-CFU assay (Fig. 3J) and from flow cytometric analysis (Supplementary Fig. S2F). No T315I LSC-CFUs formed ex vivo from EPA-supplemented bone marrow or spleen (Fig. 3J). These results suggest a potential breakthrough in limiting TKI-resistant CML with dietary change, where a COX-derived metabolite of EPA, such as Δ¹²-PGJ₃, was likely mediating this process.

We hypothesized that COX activity was critical in mediating the antileukemic effects of EPA supplementation. To examine such a relationship, we used the NSAID indomethacin, a nonspecific inhibitor of COX1 and COX2, to limit the endogenous production of prostanoids, including Δ¹²-PGJ₃. Mice were given indomethacin in drinking water ad libitum 1 week prior to transplantation of BCR-ABL LSCs. Interestingly, the protective property of EPA supplementation in BCR-ABL LSC–transplanted mice was blocked by indomethacin (Fig. 4A–E). Indomethacin-treated mice on both OA and EPA diet showed the same severity of disease as measured by the development of splenomegaly, leukocytosis, and disruption of normal splenic architecture, represented by total histology scoring (Fig. 4A–C and Supplementary Fig. S3A and S3B). We previously reported Δ¹²-PGJ₃ as mediating apoptosis in LSCs (2). To determine the level of apoptosis in EPA-supplemented LSC-transplanted mice, TUNEL staining was performed on splenic sections. EPA-supplemented BCR-ABL mice displayed increased apoptosis in the spleen compared with OA-supplemented mice (Fig. 4D). In contract, indomethacin treatment eliminated the antileukemic property of EPA as measured by the decreased TUNEL⁺ cells following NSAID treatment. There was no difference in apoptosis between dietary groups in healthy (nonleukemic) splenic tissue (Supplementary Fig. S1D). These data suggest that a metabolite of the COX pathway was responsible for the increased apoptosis, in EPA-supplemented CML mice. Furthermore, flow cytometric analysis of LSCs showed that indomethacin blocked the protective effects of EPA. EPA-supplemented CML mice treated with indomethacin showed no difference in the percentage of LSCs in the spleen and bone marrow when compared with OA alone and OA + indomethacin groups (Fig. 4E). However, treatment of EPA-fed mice on indomethacin with exogenous Δ¹²-PGJ₃ starting 1-week posttransplant of LSCs rescued the antileukemic effect in these mice as seen by significantly lowered BCR-ABL LSCs in the spleen and bone marrow, which were similar to that observed in EPA-treated mice (Fig. 4E). Surprisingly, the general phenotype of these mice was only partly restored to that of EPA-supplemented mice, as seen by decreased splenomegaly (Supplementary Fig. S3C and S3D).

To further examine whether PGD₃-derived CyPGs mediated the EPA effect, BCR-ABL mice were given intraperitoneal injections of HQL-79, a specific H-PGDS enzyme inhibitor, from time of transplantation to experiment end (Supplementary Fig. S3E). At day 14 posttransplant, HQL-79 treatment of EPA-fed mice had significantly higher total WBCs, as compared with EPA-supplemented mice (Fig. 4F and Supplementary Fig. S3F) and displayed similar disease index to OA-supplemented and OA + HQL-79 BCR-ABL mice. The BCR-ABL LSC population was analyzed by flow cytometry and LSC-CFU analysis. HQL-79 treatment of EPA mice doubled the LSC population in the spleen (Supplementary Fig. S3G). BCR-ABL LSC-CFU formation was significantly increased in the EPA + HQL-79 group compared with EPA only. In the bone marrow, there was a 6.5-fold increase in LSC-CFU formation, and splenocytes formed 5.5-fold more LSC-CFUs (Fig. 4G and H). There was no difference in LSC-CFU formation between OA diet only and OA diet + HQL-79 in BCR-ABL transplanted mice (Fig. 4G and H). HQL-79 treatment eliminated the protective effect of EPA supplementation and returned BCR-ABL LSCs to levels as in OA controls. These studies reveal that COX and H-PGDS play an important role in the antileukemic function of EPA in BCR-ABL–transplanted mice.

The MLL-AF9 bone marrow transplant model was used to explore whether EPA was protective in AML. At day 13 posttransplant, analysis of WBC showed severe leukocytosis in all OA-supplemented mice, whereas 2 EPA-supplemented mice had WBC counts within the normal healthy range (Fig. 5A). Further analysis of mice showed a slight increase in spleen weight in OA- as compared with EPA-supplemented AML mice (Fig. 5B). However, of greatest interest was the effect of EPA on suppressing AML LSC-CFUs. There was a significant decrease in the formation of AML LSC-CFUs from EPA-supplemented spleen and bone marrow. OA splenocytes and bone marrow showed a nearly 2-fold increase and a 4-fold increase, respectively, compared with EPA (Fig. 5C and D). Although EPA colonies were formed, the morphology of these colonies appeared to be CFU-GM like in nature (Fig. 5D).

Secondary transplantation with either OA or EPA total bone marrow was used to determine the functionality of AML LSCs in EPA-supplemented mice (see Supplementary Fig. S2C for experimental setup). Although the peripheral WBC count was not different between diets at 14 days posttransplant, there was a general decrease in WBCs in mice transplanted with EPA-supplemented bone marrow (Fig. 5E). There was a significant decrease in the spleen size in secondary transplanted EPA mice (Fig. 5F). AML LSC-CFUs were lower in the bone marrow of mice transplanted with EPA-supplemented AML cells, but the differences did not reach significance (Fig. 5G and H). In contrast, there was a significant decrease in AML LSC-CFUs from the spleen of secondary EPA-transplanted mice (Fig. 5G). Representative examples of AML LSC-CFUs from secondary transplanted mice showed clear differences in the morphology of the colonies (Fig. 5H). OA colonies were similar to those from primary transplant, whereas EPA colonies are clearly differentiated from the typical AML LSC-CFU. These results also suggest that the antileukemic effect of EPA supplementation is not limited to a specific oncogene.

This study demonstrates a novel role for EPA at pharmacologic levels in three experimental models of myelogenous leukemia, with a focus on BCR-ABL–induced CML. To our knowledge, no previous study has investigated the in vivo role of EPA supplementation on CML or AML progression. Moreover, previous studies cite the beneficial role of omega-3 in limiting COX activity (34), whereas our data indicate that the activity of COX in EPA-supplemented conditions contributes toward the endogenous production of CyPGs, such as Δ¹²-PGJ₃. Mice on an EPA-supplemented diet displayed decreased signs and symptoms of leukemia, including a decrease in the LSC population when compared with mice on an OA-supplemented diet. Furthermore, indomethacin and HQL-79 blocked the antileukemic effect of EPA. This supports our hypothesis that EPA supplementation limits CML due to generation of endogenous CyPGs, most importantly those derived downstream of COX-2 and H-PGDS, namely Δ¹²-PGJ₃. The dose of EPA used in this study was selected specifically to translate to the human pharmaceutical dosage. The FDA-approved drug Lovaza, composed of 1.86 g EPA (55%) and 1.5 g DHA (45%), provides an equivalent dosage of EPA as in our study. We limited the supplementation only to EPA to understand the role of EPA-derived CyPGs via the COX/H-PGDS pathway.

We observed a wide range of detectable Δ¹²-PGJ₃ extracted from the cell culture media and plasma following EPA supplementation. While the trend of EPA leading to increased Δ¹²-PGJ₃ remained consistent between independent experiments, the total amount of the compound was variable. This could be due to the extractability of Δ¹²-PGJ₃ and/or the ability of this Michael acceptor electrophile to bind to proteins, thus limiting efficient extraction. In addition, metabolic transformation of Δ¹²-PGJ₃ could also contribute to the observed variability as such metabolic products were not part of LC-MS/MS–based quantitation.

Although targeting inflammation with NSAID treatment has been an attractive proposition in certain types of cancer, our studies demonstrate a potential detrimental role for NSAID use in CML, which may also be true in AML. Most epidemiologic studies, to date, are inconclusive regarding the association between nonaspirin NSAID use and leukemia (12–14). Given the controversial nature of the role of NSAIDs in CML, future studies to associate NSAIDs as a potential confounder in omega-3 PUFA trials may shed more light on the role of endogenous CyPGs.

Current chemotherapies are able to control CML for a period of time but are unable to cure the disease, as a result of the sustained LSC population (16–28, 35). We have previously demonstrated selective targeting of CML LSCs by exogenous treatment with Δ¹²-PGJ₃ (2). By making a change in dietary composition, we demonstrate that Δ¹²-PGJ₃ can be synthesized at levels capable of limiting the disease. It is intriguing to note that the effect of EPA of reducing CML was not due to general differences in the ability of these mice to engraft a foreign bone marrow following irradiation. Engraftment and general parameters of health were not different between diets (Supplementary Fig. S1E–S1I). Thus, suppression of the LSC population was primarily due to dietary manipulation, as confirmed by secondary transplantation of bone marrow from EPA-fed leukemic mice into mice fed on normal chow. These data suggest that dietary supplementation is an attractive way to attenuate the relapse of CML, particularly in TKI-resistant CML that only responds to ponatinib, a TKI dispensed with a black box warning. Surprisingly, even in the experimental model of MLL-AF9–induced AML that represents aggressive leukemogenesis, high-dose EPA proved to be effective. Thus, on the basis of our studies, it appears that the use of pharmacologic doses of EPA in combination with existing standard of care could open new avenues for the successful management of myeloid leukemias.

Our previous work showed that treatment of BCR-ABL LSCs with EPA failed to cause any appreciable increase in apoptosis, supporting the role of the microenvironment in CyPG production (2). The microenvironment, composed of immune and nonimmune cells, has been shown to be important in understanding cancer outcomes (36, 37). Macrophages have been implicated as being major producers of COX-derived metabolites, including Δ¹²-PGJ₃ (2)_. As shown in Fig. 6, Δ¹²-PGJ₃ is thought to act through paracrine signaling mechanisms on LSCs to activate pathways of apoptosis; this effect does not appear to be limited to CML alone.

Our data support the pivotal role of COX and H-PGDS in the antileukemic effect of EPA, suggesting that any variation in gene expression and/or activity of either or both of these enzymes could potentially impact the outcome. This observation is not limited to leukemia. In fact, several SNPs have been identified in the COX2 gene, PTGS2, and meta-analysis has suggested a possible association to breast cancer (38). Therefore, the translatability of these studies into human therapy would depend on the effect the SNPs have on the expression and/or activity of COX and H-PGDS. Few genome-wide association studies have explored SNPs in the COX gene and resulting leukemia outcomes (39), highlighting an area that should be more closely studied in the future.

In summary, our studies use well-established rodent models of CML, T315I CML, and MLL-AF9 AML to demonstrate that dietary EPA decreases the severity of disease. In the case of CML, the antileukemic effect of EPA was blocked by both indomethacin and HQL-79, therefore confirming the essential role of the COX pathway in the metabolism of EPA into a novel class of endogenous CyPGs, including Δ¹²-PGJ₃. Although deemed safe, the success of this therapy rests on the efficient metabolism of EPA to endogenous levels of CyPGs that target LSCs. Future clinical trials with EPA are warranted to conclusively provide a compelling case for its use as an adjuvant therapy in patients with myeloid leukemia.

R.F. Paulson has ownership interest (including patents) in OncOmega Inc. and Nemean Pharma Corporation, and K.S. Prabhu has ownership interest (including patents) in the two Start ups. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R.F. Paulson, K.S. Prabhu

Development of methodology: E.R. Finch, A.K. Kudva, R.F. Paulson, K.S. Prabhu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.R. Finch, A.K. Kudva, M.D. Quickel, L.L. Goodfield, M.J. Kennett, J. Whelan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.R. Finch, L.L. Goodfield, M.J. Kennett, J. Whelan, K.S. Prabhu

Writing, review, and/or revision of the manuscript: E.R. Finch, A.K. Kudva, M.D. Quickel, M.J. Kennett, J. Whelan, R.F. Paulson, K.S. Prabhu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.S. Prabhu

Study supervision: K.S. Prabhu

The authors thank Dr. Shailaja Hegde for irradiation of mice, Dr. Naveen Kaushal for assisting with bone marrow transplantation, Dr. Ramesh Ramachandran for help with microscopy, Flow Cytometry and Microscopy Facility (Drs. Ruth Nissly and Ningchen Xu), and all current and former members of the Prabhu laboratory for valuable suggestions and help.

This work was supported, in part, by grants from the National Cancer Institute (NIH RO1 CA 175576 to K.S. Prabhu and R.F. Paulson), USDA National Institute of Food and Agriculture Hatch Projects #4736 (RFP) and #4475 (KSP), and the National Center for Advancing Translational Sciences (NCATS) in the form of NIH UL1 TR000127 and TL1 TR000125 to E.R. Finch.

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.