Phytogalactolipid dLGG Inhibits Mouse Melanoma Brain Metastasis through Regulating Oxylipin Activity and Re-Programming Macrophage Polarity in the Tumor Microenvironment

Simple Summary Metastatic brain melanoma is a common metastatic cancer with a high mortality rate. Current clinical regimens use the anti-angiogenesis drug bevacizumab (Avastin) and/or Lipo-DOX, a drug capable penetrating the blood–brain barrier; however, both commonly result in adverse side effects and limited treatment results. This study provides evidence to support the function of a phyto-glyceroglycolipid, 1,2-di-O-α-linolenoyl-3-O-β-galactopyranosyl-sn-glycerol (dLGG) in inhibiting melanoma brain metastasis (MBM) in mice through reprogramming the tumor microenvironment and interacting with melanoma cells and macrophages. The novel function of oxylipin 9,10-EpOMEs + 12,13-EpOMEs in preventing melanoma cell invasion and microglia/macrophage distribution and polarization in the tumor microenvironment is presented. The novel anti-melanoma function and underlying molecular mechanism of dLGG proposed herein can be considered as a novel therapeutic strategy to combat MBM. Abstract Current conventional cancer therapies for melanoma brain metastasis (MBM) remain ineffective. In this study, we demonstrated the bioefficacy of a phyto-glyceroglycolipid, 1,2-di-O-α-linolenoyl-3-O-β-galactopyranosyl-sn-glycerol (dLGG) alone, or in combination with liposomal doxorubicin (Lip-DOX) or Avastin against MBM in a syngeneic B16BM4COX−2/Luc brain-seeking melanoma mouse model. Treatment with dLGG–10, dLGG–25, dLGG–10 + Avastin–5, Lipo-DOX–2, dLGG–10 + Lipo-DOX–2 or Lipo-DOX–2 + Avastin–5 suppressed, respectively, 17.9%, 59.1%, 55.7%, 16.2%, 44.5% and 72.4% of MBM in mice relative to the untreated tumor control. Metastatic PD-L1+ melanoma cells, infiltration of M2-like macrophages and CD31+ endothelial cells, and high expression levels of 15-LOX/CYP450 4A enzymes in the brain tumor microenvironment of the tumor control mice were significantly attenuated in dLGG-treated mice; conversely, M1-like resident microglia and cytotoxic T cells were increased. A lipidomics study showed that dLGG promoted B16BM4 cells to secrete oxylipins 9,10-/12,13-EpOMEs into the culture medium. Furthermore, the conditioned medium of B16BM4 cells pretreated with dLGG or 9,10-EpOMEs + 12,13-EpOMEs drove M2-like macrophages to polarize into M1-like macrophages in vitro. An ex vivo 3D-culture assay further demonstrated that dLGG, 9,10-EpOME or 9,10-EpOME + 12,13-EpOME pretreatment attenuated B16BM4 cells invading brain tissue, and prevented microglia/macrophages infiltrating into the interface of melanoma plug and brain organ/tissue. In summary, this report provides a novel therapeutic strategy and mechanistic insights into phytogalactolipid dLGG for combating MBM.


Introduction
Epidemiological evidence shows that approximately 37% of cancer patients with metastatic melanoma develop melanoma brain metastasis (MBM), which causes high morbidity and mortality, with only about 4 to 5 months life expectancy after brain metastasis is diagnosed [1]. Different drug delivery strategies have been developed to cross the blood-brain barrier (BBB) and to increase drug availability and efficiency in the brain. For example, monotherapy using Lipo-DOX, stable vesicle-like liposome-encapsulated doxorubicin, an anti-cancer drug, was effective in treating metastatic melanoma in a phase II clinical trial [2,3]. Bevacizumab (Avastin), a monoclonal antibody directly targeting vascular endothelial growth factor-A (VEGF-A), has been widely used in clinical trials for patients with melanoma, HER2-negative breast cancer, and so on [4,5]. High levels of VEGFs, VEGF-A, VEGF-C and VEGF receptors (VEGFR-1, -2, and -3) are commonly identified in benign melanocytic tumors, malignant melanoma, and stromal cells surrounding tumors, e.g., macrophages, endothelial cells, and fibroblasts [6]. A combination of bevacizumab with an alkylating chemotherapeutic drug temozolomide as a first-line treatment had a suppressive effect on melanoma with uveal metastasis in a phase II study [7]. Although a combination of targeted therapy (e.g., BRAFi/MEKi) increased the intracranial response rate (58%) compared with the BRAF inhibitor only (25-40%) in MBM patients, the durability of the response rate was shorter than in patients with extracranial disease. Targeting the immune check point, PD-1 antibody therapy alone or combined with CTLA-4 could induce response rates of 20% and 55%, respectively, in MBM clinical treatment [8]. However, the effectiveness of target therapies remains unsatisfactory due to various issues, including melanoma stages, metastatic recurrence, drug resistance, and accompanying side effects [8,9].
In the central nervous system, microglia cells (F4/80+ or Iba1+) are important immune cells that serve as tissue-resident macrophages (TMEM119+) with M1-like (iNOS+) polarity that have a functional effect on brain development and neuron environment homeostasis [10][11][12]. On the other hand, the bone marrow-derived macrophages that infiltrate the brain tissue showing M2-like (CD163+) polarity might promote tumorigenesis by creating an immunosuppressive tumor microenvironment (TME) [10][11][12]. Tumor-associated macrophages (TAMs) are the most abundant leukocytes in tumors, and are attributed to cancer dissemination and metastasis, including breast cancer, gliomas, and melanoma [13]. Accumulating evidence has revealed that TAMs predominantly resemble M2-like polarized macrophages and interaction with melanoma cells and other stromal cells, such as neutrophils, Treg cells, and dendritic cells, dominate the immunosuppressive milieu in the TME [14]. The development of therapies that target TAMs to redirect the polarization from M2 to M1 phenotype in brain tumors, including glioma and melanoma, has been proposed [11,15]. Microglia/macrophages displaying M1 polarity have been shown to have a tumoricidal effect that provides a strategy for clinical cancer therapy [11,[14][15][16].
A glyceroglycolipid 1,2-di-O-α-linolenoyl-3-O-β-galactopyranosyl-sn-glycerol (dLGG) isolated from the medicinal plant Crassocephalum rabens (Asteraceae) has been demonstrated to have anti-inflammatory activity in sepsis mouse models and anti-cancer activity in an orthotopic melanoma mouse model [21,22]. Of note, dLGG suppressed B16 melanoma lung metastasis by maintaining pulmonary vasculature tight junction permeability that resulted in preventing melanoma cell extravasation into the lung in syngeneic mice [23]. In the present study, we demonstrate that dLGG effectively attenuates melanoma brain-seeking cell aggressive properties in vitro and in vivo by inhibiting M2-marophage assembly and melanoma cell infiltration and growth into brain parenchyma as seen in a 3D-organoid culture model. This study also provides evidence for the therapeutic effect of dLGG alone or in combination with current anti-cancer drugs Lip-DOX or Avastin against brain metastatic melanoma.

Cell Lines
The B16 murine melanoma cell line (NRAS mutation) and A375 human melanoma cell line (BRAF V600E mutation) obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) were grown in the product sheet suggested medium with 10% heatinactivated fetal bovine serum, 100 units/mL penicillin and 100 mg/mL streptomycin at 37 • C in a humidified 5% CO 2 incubator. THP-1, a human monocytic cell line from ATCC, was grown in RPMI1640 (Thermo Fisher Scientific, Waltham, MA, USA) medium supplemented with 10% fetal bovine serum, 4.5 g/L glucose and 0.05 mM β-mercaptoethanol. Primary human umbilical vein endothelial cells (HUVECs) were a kind gift from Professor Li-Wha Wu from the Institute of Molecular Medicine, National Cheng Kung University, Taiwan [23]. HUVECs were cultured in EndoGRO-LS medium (EMD Millipore, Burlington, MA, USA). A stable B16 COX−2/Luc cell clone established previously was used in the mouse melanoma brain metastasis experiments [21]. A stable A375 cell clone (designated A375 eIF4g/Luc ) carrying a hEF1alpha-eIF4g promoter-driven luciferase (pIF4g.As2.luc.bla) reporter gene was created by lentivirus transfection. After transfection of the reporter gene construct into A375 cells and a few rounds of antibiotic-based (Blasticidin S, 10 µg/mL; Invivogen, San Diego, CA, USA) selection, the A375 eIF4g/Luc cell clone containing stably expressed luciferase gene/protein was obtained (approximately 1 × 10 6 RLU/s/µg protein/ 1 × 10 5 cells).

Isolation of dLGG
Whole C. rabens plants (voucher specimen CB001, Agricultural Biotechnology Research Center, Academia Sinica, Taipei, Taiwan) at the flower blooming stage were collected and the compound dLGG was isolated with >98% purity following previously described protocols [21].

Cell Viability Assay
Melanoma cells (5 × 10 3 cells/well) were seeded in a 96-well plate for 12 h and incubated with the indicated concentration of compounds for 24 h. The 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT)-based colorimetric assay was used to measure cell viability based on the quantification data of the MTT salt absorbance at 570 nm. Cell viability was calculated using the following formula: viable cell number (%) = OD 570 (treated cell culture)/OD 570 (vehicle control) × 100.

Cell Clonogenic Assay
Colony-forming cell growth was attained by growing the parental cells and brainseeking cells on 24-well plates with the indicated treatments for 6 days [24]. The treated cell colonies were further stained with 0.1% (w/v) crystal violet and then dissolved in 20% (v/v) acetic acid solution for quantification by measuring absorbance at 595 nm. The percentage inhibition of compound treatment was compared with the vehicle control-treated cells.

Tube Formation Assay
Low-passage HUVECs (1 × 10 4 cells/well) were grown in endothelial growth basal medium (EndoGRO-LS medium; Millipore, MA, USA) for 12 h in growth factor-reduced Matrigel (BD Biosciences, NJ, USA)-coated 96-well plates. The tube formation of HUVECs was stimulated by adding 100 ng/mL human recombinant VEGF (Millipore, Burlington, MA, USA) in the presence of vehicle or compound for 12 h. Phenotypic tube formation was monitored every hour from 0 h to 12 h by microscope (Zeiss Imager Z1).

Cell Cycle and Cell Apoptosis Assay
Brain-seeking melanoma cells (B16BM4 COX−2/Luc and A375BM4 eIF4g/Luc ) were synchronized by incubation with culture medium with 1% FBS for 24 h, and further treated with 140 µM dLGG for 12 h and 24 h. The treated-cells were fixed with cold 70% ethanol and stained with propidium iodide (PI; Sigma-Aldrich, St. Louis, MO, USA) for 30 min at room temperature. The results were analyzed by flow cytometry (BD Accuri C6 flow cytometer). For apoptosis detection, the melanoma cells were seeded in a 10 cm culture dish and treated with 140 µM dLGG for 24 h and 48 h. The apoptotic fraction was stained with an FITC/Annexin V Apoptosis Detection Kit (BD Pharmingen, Franklin Lakes, NJ, USA) according to the manufacturer's instructions and analyzed by flow cytometry.

Animals
Male C57BL/6JNarl mice were bred and obtained from the National Laboratory Animal Center, Taiwan. Female NOD/SCID (NOD.CB17-Prkdcscid/IcrCrlBltw) mice were bred and obtained from the Laboratory Animal Core Facility, Agricultural Biotechnology Research Center (ABRC-LACF), Academia Sinica, Taiwan. All of the mice were given a standard laboratory diet and distilled H 2 O ad libitum and kept on a 12 h light/dark cycle at 22 ± 2 • C in ABRC-LACF [23]. All experimental protocols (Protocol # 13-05-552) are approved by the Institutional Animal Care and Utilization Committee (IACUC), Academia Sinica, Taiwan [23].

Establishment of the Melanoma Brain Metastatic Mouse Model and Brain-Seeking Melanoma Cell Lines
Experimental mice (5 weeks old) were anesthetized by intraperitoneal (i.p.) injection of zoletil (0.16 mg/kg) and restrained on the platform of a dissecting microscope. The mice received intracarotid (ica.) injection surgery through ligating the left carotid artery at the distal part near the heart by surgical sutures and the second ligature was tied into a slipknot proximal to the injection site. A PE-10 (Intramedictm, No. 427401) tube was inserted into the lumen of the blood vessel to connect with the artery to inject melanoma cells (1 × 10 5 cells/ 100 µL PBS). Luciferase expression was measured every two days, starting from day 7 after tumor cell injection [23]. The male C57BL/6JNarl mice, in which positive signals were detected by IVIS system in the brain region after ica. injection of B16 COX−2/Luc cells, were sacrificed at day 17. The brain tissues were taken out of the skull and cut into small pieces in a culture plate containing 5 mL Hank's balanced salt solution (HBSS; Sigma-Aldrich, St. Louis, MO, USA) containing 5 mg/mL papaya enzyme (papain; Sigma-Aldrich, MO, USA), then incubated in a CO 2 incubator for 15 min to release B16 COX−2/Luc melanoma cells into the HBSS solution. Five milliliters of RPMI1640 medium were added to stop the enzyme reaction and the tissues/cells containing media were centrifuged at 100× g for 10 min. The tissue/cell pellets were re-suspended in fresh medium and then filtered through a 0.40 µm filter. The collected cell solution was cultured in a T150 culture flask with RPMI 1640 medium and replaced by fresh culture medium after 24 h. Acclimated brainseeking melanoma cell lines were obtained from 4 cycles of ex vivo primary cell culture processes ( Figure S1a), and the final brain metastatic cells (BM) for further study were designated as B16BM4 COX−2/Luc . We further established human brain-seeking melanoma cells A375BM4 eIF4g/Luc using NOD/SCID mice and the same procedure.
The in vivo bioefficacy of dLGG against the human A375BM4 eIF4g/Luc brain-seeking melanoma mouse model was studied by dividing NOD/SCID mice (5 weeks old) into three groups and n = 10 per group: sham, tumor control, and dLGG-25. At day 0, the sham group mice received 0.1 mL PBS and the other groups had A375BM4 eIF4g/Luc cells (1 × 10 5 cells/100 µL PBS) implanted by ica. injection. Starting from day 7, the tumor control group mice were given 0.2 mL PBS p.o. every day, and the dLGG-25 group mice received 25 mg/kg p.o. every day. The tumor control group mice were given vehicle (0.2 mL PBS) p.o. every day.
In vivo bioluminescence imaging of each tested mouse group was acquired using the IVIS spectrum system (Xenogen) every other day, starting from day 7 after tumor cell inoculation until the animals were sacrificed. Bioluminescent signals were quantified using Living Image 2.5 (Xenogen) as photons/sec/region of interest (ROI) [22].

Tight Junction Permeability Analysis
At 14 days after the tumor inoculation in the MBM model, experimental mice received FITC-conjugate dextran (fluorexcin isothiocyanate-dextran, Sigma) by tail injection and 30 min later, the mice were sacrificed by systemic perfusion. The brain tissues were collected and underwent 10% formalin-fixation and OCT-embedded preparation. After immunofluorescence staining, the CD31 protein and FITC-dextran were visualized by microscope (Zeiss Imager Z1) and quantified by AxioVision software (Carl Zeiss MicroImaging) [23].

Ex Vivo 3D Tumor-Brain Organoid Co-Culture Model
Brains were taken from 5-day-old mice and the brain coronal sections (1 mm) were prepared by rodent brain matrix (RBM-2000C; ASI instrument, Eugene, OR, USA). The brain slices were placed onto a 6-well plate cell culture insert (CAT.35306; SPL life Sciences, Korea) and incubated with 100 µL RPMI1640/DMEM culture medium overnight at 37 • C in a 5% CO 2 incubator. Tumor plugs were prepared using B16BM4 COX−2/Luc or A375BM4 eIF4g/Luc cells (5 × 10 5 ) with (pretreatment group) or without (post-treatment group) compound treatment for 24 h, mixed with ECM (Basement Membrane Extract; CAT.354234, BD Falcon); and filled into a sterilized metallic spacer (5 mm diameter) and incubated for 2 h at 37 • C in a 5% CO 2 incubator. The tumor plug was placed in contact with the brain slice for 6 days and the compound-containing RPMI1640/DMEM culture medium was changed every 2 days [26]. Immunofluorescence staining with specific antibodies to targeting melanoma cells (Mel-A) and macrophages (F4/80+) was conducted and visualized by light and fluorescent microscopes (Zeiss Imager Z1) and quantified by AxioVision software (Carl Zeiss MicroImaging) [23].

Ultra-Performance Liquid Chromatography-Electrospray Ionization Tandem Mass Spectrometry Analysis of Serum Oxylipin Metabolome
Serum (70 µL) was collected immediately from sacrificed mice and added to extraction solvent containing CHCl3:MeOH (2:1), 0.1 M 2,6-di-tert-ubutyl-4-methylphenol (ACROS ORGANIC, MA), and 0.01 M triphenylphosphine (SIGMA). Internal oxylipin standards (0.1 ppm of DHA-d 5 , 9-HODE-d 4 , PGE2-d 4 , and 1 ppm of 14,15-EET-d 4 , EPA-d 5 , 20-HETEd 6 , 5-HETE-d 8 ) purchased from Cayman Chemical (Ann Arbor, MI, USA) were added into samples before extraction [23]. The organic layer was collected by centrifugation. Five biological serum samples/repeats and five technical repeats for each biological sample were analyzed using ultra-performance liquid chromatography-electrospray ionization tandem mass spectrometry (UPLC-ESI-MS/MS). The culture medium of B16BM4 cells was collected immediately after the indicated treatment times, and stored at −80 • C before sample preparation for mass spectrometry analysis. The collected cultural media were supplemented with 10 µL of internal standards (0.1 ppm of DHA-d 5 , 9-HODE-d 4 , and PGE2-d 4 ; 1 ppm of 14,15-EET-d 11 , and 5-HETE-d 8 ), and then extracted using Oasis HLB cartridge (solid-phase extraction). The column was conditioned first with 6 mL of 100% MeOH and then with 6 mL of ddH 2 O. After loading the sample, the column was washed with 10% MeOH (20 mL), and the oxylipins were then eluted with 6 mL of 100% MeOH. The eluent was dried out under vacuum and redissolved in 100% MeOH for UPLC-ESI-MS/MS analysis. Samples were separated with ACQUITY UPLC BEH C18 column (particle size 1.7 µm, 2.1 × 100 mm 2 , Waters; Milford, MA, USA) at a 400 µL/mL flow rate using a 25 min gradient for analysis; the mass spectrometry setting and operation were according to our previous study [23]. Accuracy and precision were analyzed with quality control (QC) samples for each experimental group. Chromatogram acquisition, detection of mass spectral peaks, and their waveform processing were performed using Thermo Xcalibur 2.1 SP1 software (Thermo Fisher Scientific, MA, USA). The peak area of each quantified ion was calculated and normalized to the peak area of the corresponding internal standards.

Statistical Analysis
Experimental data were expressed as mean ± SD by using PASW Statistics 18. The statistical significance of differences between treatments was determined by analysis of variance (ANOVA) with Fisher's post hoc test and Kruskal-Wallis test. A p value of greater than 0.05 was considered to be statistically significant. Different letters represent significant differences (one-way ANOVA, p ≤ 0.05). Kaplan-Meier survival plots were compared using a log-rank test.

Establishment of the Melanoma Brain Metastatic Mouse Model and Brain-Seeking Melanoma Cell
In order to assess the mechanism of melanoma cell brain metastasis and the deregulating activities of phytogalactolipid dLGG and Lipo-DOX, we established a B16 and B16 COX−2/Luc -carrying luciferase reporter gene melanoma brain metastatic mouse model by ica. injection. Then, brain-seeking melanoma cells (designated BM) with specific and high brain metastasis frequency compared to the parental cells in the mouse brain were acclimated and obtained from four cycles of repeating primary melanoma cells culture (B16BM1-4 and B16BM1-4 COX−2/Luc ) ( Figure S1a). In order to select and decide the most appropriate brain metastatic clone for in vivo animal study, we used Western blotting to examine the expression of proliferation and metastatic markers including p-Src, tyrosinase, TGF-β, GABA A R-α3, and Ki-67 in the B16, B16BM3 and B16BM4, and B16 COX−2/Luc , B16BM3 COX−2/Luc and B16BM4 COX−2/Luc cell lines. Compared to parental cells, the BM cell lines (either B16BM3 and B16BM4 or B16BM3 COX−2/Luc and B16BM4 COX−2/Luc ) had higher protein expression levels in most of the tested proliferation and metastasis-associated pro-teins ( Figure S1b). We then established melanoma brain metastases in mice using B16BM4 and B16BM4 COX−2/Luc . The bioimaging data in Figure S1c show that B16BM4 COX−2/Luc cells were specifically detected in the mouse brain. The final survival time of mice carrying parental brain B16 melanoma was 17 days, and those of mice carrying B16BM4 and B16BM4 COX−2/Luc were 13 days and 15 days, respectively ( Figure S1d). We examined the coronal section of mouse brains and observed that most of the B16BM4/B16BM4 COX−2/Luc tumors grew in mouse brain parenchyma, yet B16/B16 COX−2/Luc tumors grew in the third ventricle and hippocampus area ( Figure S1e). These data indicate that primary melanoma brain-seeking metastatic cell clones, B16BM4 and B16BM4 COX−2/Luc , were successfully established. These clones were then used for subsequent studies in vitro and in vivo.

dLGG Inhibits B16BM4 Brain-Seeking Cell Activities, HUVEC Tube Formation, and Selected Marker Protein Expression
We first investigated the in vitro bioactivity of dLGG against B16BM4 cells. The anti-B16BM4 cell (5000 cells/well) proliferation activity (IC 50 ) of dLGG was determined to be 140.5 µM after treatment for 24 h (Figure 1a). The clonogenic ability of the B16BM4 cells was inhibited 35% to 62% when the lower melanoma cell population (500 cells/well) were treated with dLGG for 6 days at concentrations between 5 and 35 µM (Figure 1b). Further, when B16BM4 cells (8.6 × 10 5 cells/10 cm culture dish) were treated with 140 µM dLGG for 24 h, the sub-G1-phase cell population was slightly induced (up to 11.8%) compared with the vehicle control (5.9%) as measured by a flow cytometer using a PI staining probe (Figure 1c). The G1 (49.2%) and G2/M (26.5%) phases in the dLGG treated-B16BM4 cells were similar to the vehicle-treated control cells (56.8% and 25.7%, respectively), and there was no difference in the S phase in any of the tested cells ( Figure 1c). Furthermore, flow cytometry analysis using the annexin V and PI double staining method showed that dLGG treatment time-dependently induced apoptosis in B16BM4 cells. At 24 h and 48 h treatment, apoptotic cells were detected to be 24.8% and 48.3%, respectively, compared to vehicle control cells for which apoptotic cells made up 11.5% and 0.85%, respectively (Figure 1d).
Brain tumors are commonly observed to have high levels of angiogenesis and thus, anti-angiogenesis drugs are used as first-line therapy for patients with brain tumors [4,27,28]. We examined whether dLGG plays a role in inhibiting the VEGF-stimulated tube formation phenotype in primary HUVECs, thus representing the potential for anti-angiogenesis activity. When HUVECs were grown in Matrigel containing 100 ng/mL VEGF for 6 h, tube formation was significantly elevated (199%) compared to vehicle control cells (100%) (Figure 1e). When VEGF-stimulated cells were treated with 20 and 40 µM dLGG, the tube formation phenomena were inhibited 56.6-70% (Figure 1e).
We further examined the time-dependent effect of dLGG on the expression of some protein markers related to cell apoptosis, angiogenesis, and PD-L1 expression in B16BM4 cells. dLGG treatment significantly down-regulated anti-apoptotic Bcl-2 protein and increased the cleaved form of PARP, an apoptotic marker. dLGG also inhibited B16BM4 cells expressing VEGF (Figure 1f). Furthermore, dLGG significantly inhibited protein levels of PD-L1 and original forms or phosphorylated forms of MEK, ERK, c-Jun, Src and STAT3 in B16BM4 cells (Figure 1e). Together, our data demonstrated the in vitro effect of dLGG on anti-angiogenesis, and anti-cell proliferation and induction of apoptosis in melanoma brain-seeking cells.
1 Figure 1. dLGG inhibits brain-seeking-cell proliferation and colony formation, and is anti-angiogenic. (a) The brain-seeking melanoma cells (B16BM4) were treated with vehicle (0.5% DMSO) or the indicated concentrations of dLGG at 37 • C for 24 h, and cell viability (%) was determined by MTT assay. (b) B16BM4 cells were treated with dLGG for 6 days. The colonies of melanoma cells were analyzed by crystal violet staining. Data are mean ± SD, n = 3; Kruskal-Wallis test; * p ≤ 0.05 compared with vehicle control; # p ≤ 0.05 compared with dLGG treatment for 12 h. (c) B16BM4 cells were treated with 140 µM dLGG for 12 h and 24 h before being stained with propidium iodide for cell cycle analysis using flow cytometry. Data are mean ± SD; n = 3; Kruskal-Wallis test; * p ≤ 0.05 compared with 12 h of vehicle control. # p ≤ 0.05 compared with 24 h of vehicle control. (d) B16BM4 cells were treated with 140 µM dLGG for 12, 24 and 48 h. Apoptotic cells were detected by annexin V/propidium iodide double staining and flow cytometry. Data are mean ± SD, n = 3; Kruskal-Wallis test; * p ≤ 0.05 compared with vehicle control. (e) The VEGF-stimulated (100 ng/mL) tube formation of primary HUVECs cells was observed after incubation for 6 h. The mean area of the vascular tube was analyzed by Image J software. Data are mean ± SD, n = 3; Kruskal-Wallis test; * p ≤ 0.05 compared with vehicle control. # p ≤ 0.05 compared with VEGF-treated group. (f) B16BM4 cells were treated with 140 µM dLGG for 6, 12, 24, and 48 h. The total proteins were prepared and subjected to Western blotting using the specific antibodies indicated. Actin protein intensity was used as the loading control. The target protein band intensity was analyzed by Image J software. Specific target protein expression level was normalized to respective actin band intensity first, followed by comparison with 6 h vehicle-treated protein band intensity.

dLGG Suppresses Melanoma Brain Metastases in Mice
Next, we explored the in vivo anti-brain melanoma effect of dLGG using brain-seeking B16BM4 COX−2/Luc melanoma cells and the syngeneic brain metastasis mouse model established in-house. Lipo-DOX and Avastin were used as reference drugs and combined with dLGG for treatment of MBM. The experimental schemes of the brain metastatic melanoma model are shown in Figure S2a. The experimental animals were divided into eight groups: sham, tumor control, dLGG-10, dLGG-25, Lipo-DOX-2, Lipo-DOX-2 + Avastin-5, dLGG-10 + Avastin-5 and dLGG-10 + Lipo-DOX-2 after ica. injection of melanoma cells into mice. At treatment day 15, the dLGG-25 (40.9%), dLGG-10 + Avastin-5 (44.3%), Lipo-DOX-2 + Avastin-5 (27.6%), and dLGG-10 + Lipo-DOX-2 (55.5% with p ≤ 0.07) were observed to have much less metastatic brain B16BM4 COX−2/Luc melanoma than the tumor control (100%) group (p ≤ 0.01) (Figure 2a), while the dLGG-10 (82.1%) and Lipo-DOX-2 (83.8%) treatments were less effective, as determined by quantified bioluminescence data. The tumor control and most treatment group mice showed a similar weight loss pattern over the experimental time period, but the Lipo-DOX-2 + Avastin-5 and dLGG-10 + Avastin-5 group mice had relatively less body weight loss than the tumor control group (Figure 2b). Meanwhile, the mean survival time of all compound-or drug-treated mice was prolonged with statistical significance relative to the tumor control mice (15 ± 1 days): dLGG-10 (18 ± 3 days, p ≤ 0.03), dLGG-25 (21 ± 3 days, p ≤ 0.0002), Lipo-DOX-2 (17 ± 2 days, p ≤ 0.013), dLGG-10 + Lipo-DOX-2 (18 ± 2 days, p ≤ 0.003), Lipo-DOX-2 + Avastin-5 (18 ± 2 days, p ≤ 0.004), and dLGG-10 + Avastin-5 (19 ± 2 days, p ≤ 0.001) (Figure 2c). We further examined the major organ weight index (organ weight/body weight) in all experimental mice. Sham control mice had a higher organ weight index for the spleen, kidney, and liver compared with the tumor control; and most of the treatment groups had lower indexes for the three major organs. The brain weight index showed no difference in any of the mouse groups and we did not observe peritumoral edema in the brains of any of the groups of mice (Figure 2d). We further assessed the tissue architecture of the spleen, kidney and liver for each group by H&E staining (Figure 2e). In the tumor control group mice, some destruction of the glomerular structure in the kidney, white pulp and red pulp structure in the spleen, and more infiltration of red blood cells in the kidney and liver central vein relative to sham control mice were observed. The dLGG-25 and dLGG-10 + Avastin-5 treatments in mice showed an organ protective effect compared to the tumor control with similar organ tissue structure to the sham group (Figure 2e). The Lipo-DOX-2 + Avastin-5 group of mice showed some apparent damage to the spleen and kidney tissues. Of note, the dLGG-10 + Lipo-DOX-2 treatment showed less spleen tissue damage compared with tumor control and Lipo-DOX-2 + Avastin-5 groups (Figure 2e).

dLGG Reprograms Tumor-Associated Macrophage Profiles in the Melanoma Tumor Microenvironment in Mouse Brain
Tumor-associated macrophages (TAMs) are the most abundant inflammatory cells, which orchestrate cancer development at different stages; thus, targeting TAMs has been considered to be a therapeutic strategy to combat various cancers [14]. We sought to determine whether the inhibitory mechanism of dLGG and Lipo-DOX against brain metastatic melanoma is through the re-education of TAMs and other immune cell types in the TME. First, we collected the coronal sections of brain specimens from all of the respective sham, tumor control and treated mice ( Figure S3). We observed that in a few of the MBM mice, although the luminescent intensity was seen in the brain area of the mouse, few tumors actually grew at locations near the inner brain shell, or close to the nose, ear, and dura mater areas, except the brain parenchyma. To micro-dissect the immune cell type distribution in the brain microenvironment, we selected the mouse groups, including the tumor control, dLGG-10, Lipo-DOX-2, and dLGG-10 + Lipo-DOX-2 groups, which had at least three mice with a melanoma tumor growing in the brain parenchyma area ( Figure S3) for IHC assays. We immune-stained the brain tissues of the tumor control mice for the population and expression of microglia/macrophage (Iba1), astrocyte (GFAP) and neuron (Neu N) cells along with melanoma cells (Mel-A). The Mel-A+ melanoma cells (DAB brown color) in the tumor nest and invasion front of mouse melanoma can be seen in Figure 3a. Notably, we observed that the activated microglia/macrophages (Iba1+) with round cell bodies were significantly accumulated in the tumor nest tissue and tumor invasion front site, as shown by brown DAB staining (left) or red immunofluorescent staining (right). Meanwhile, resting microglia/macrophages were also observed in the hippocampus area of brain tissue (Figure 3a). Most of the astrocytes (GFAP+) and neurons (Neu N+) were found localized in the hippocampus (H) of the brain tissue, but not found at the tumor site (T).

dLGG Inhibits Angiogenesis and Prevents Melanoma Cell-Induced Vascular Tight Junction Permeability
We observed previously that B16 tumor cells enhance the permeability of vascular tight junctions in lungs, resulting in lung metastasis [23]. In this study, we assessed the BBB tight junction permeability of brain tumors in an MBM mouse model with or without dLGG/drug treatment. The FITC-dextran intensity in the melanoma brain tumor tissues represents the BBB and tight junction permeability in the MBM mouse model. The results show that dLGG-10 and dLGG-10 + Lip-DOX-2 treatments significantly decreased the FITC-dextran intensity in tumor tissues by 28.2% and 55.8%, respectively, compared to tumor control (Figure 3c). In addition, a high level of CD31+ vascular endothelial cells was observed in the tumor control (100%) that was significantly attenuated by dLGG-10 (42.6%) and dLGG-10 + Lip-DOX-2 (38%) treatments (Figure 3c).

dLGG-Primed Conditioning Medium of B16BM4 Cells Modulates Macrophage Polarity
As the bidirectional interaction between microglia/Mφ and melanoma cells is known to be involved in melanoma brain metastasis [29], we hypothesized that the culture medium from B16BM4 melanoma cell pretreatment with dLGG could affect Mφ polarity and result in interference of the communication or interaction between Mφ and melanoma cells, mimicking their relationship in the TME. We first prepared the M2-like phenotype of Mφ control cells by using TPA (162 nM)-activated monocyte THP-1 cells, which were then differentiated into M2-like Mφ (CD163+) cells by further incubating with IL-4 (25 ng/mL) for 48 h in RPMI1640 culture medium ( Figure 5a); in parallel, another portion of the activated-THP-1 cells were differentiated into M1-like Mφ (iNOS+) control by incubating with 50 ng/mL IFN-γ for 48 h in RPMI1640 culture medium (Figure 5a). Then, CM collected from B16BM4 cells, pretreated with vehicle or 140 µM dLGG for 24 h and then replenished with fresh medium for 12 h, was evaluated for its effect on Mφ polarization using TPAactivated THP-1 cells. As shown in Figure 5b, after 48 h incubation with activated THP-1 cells, vehicle-treated B16BM4 CM converted more of the cells into M2-like Mφ, and dLGGpretreated B16BM4 CM significantly increased the ratio of M1-like Mφ over M2-like Mφ cell populations (Figure 5b). We further evaluated the effect of THP-1-differentiated M2-like Mϕ (differentiated by incubating with B16BM4 CM) and THP-1-differentiated M1-like Mϕ (differentiated by incubating with CM from dLGG pretreated B16BM4) on B16BM4 cell proliferation, colony formation, and invasion ability. The two different phenotypes of THP-1-differentiated Mϕ cells were cultured in fresh RPMI1640 culture medium for 12 h and the CM were collected for subsequent assays (Figure 5c). The results in Figure 5d show that CM from M2-like Mϕ significantly promoted B16BM4 cell invasion (300%) and colony formation (253.2%) compared with RPMI1640 medium cultured cells (100%). Notably, the CM from M1-like Mϕ differentiated by the CM treatment collected from dLGG pretreated B16BM4 cells significantly inhibited B16BM4 cell invasion, proliferation, and colony formation to 55.8%, 62.5% and 36.6%, respectively, compared with the melanoma cell treated M2-like Mϕ CM (invasion: 100%, proliferation: 100%, and colony formation: 100%).  We further evaluated the effect of THP-1-differentiated M2-like Mφ (differentiated by incubating with B16BM4 CM) and THP-1-differentiated M1-like Mφ (differentiated by incubating with CM from dLGG pretreated B16BM4) on B16BM4 cell proliferation, colony formation, and invasion ability. The two different phenotypes of THP-1-differentiated Mφ cells were cultured in fresh RPMI1640 culture medium for 12 h and the CM were collected for subsequent assays (Figure 5c). The results in Figure 5d show that CM from M2-like Mφ significantly promoted B16BM4 cell invasion (300%) and colony formation (253.2%) compared with RPMI1640 medium cultured cells (100%). Notably, the CM from M1-like Mφ differentiated by the CM treatment collected from dLGG pretreated B16BM4 cells significantly inhibited B16BM4 cell invasion, proliferation, and colony formation to 55.8%, 62.5% and 36.6%, respectively, compared with the melanoma cell treated M2-like Mφ CM (invasion: 100%, proliferation: 100%, and colony formation: 100%).
3.9. dLGG and Oxylipins 9,10-EpOME and 12,13-EpOME Inhibit Melanoma Cell Activity and Invade the Brain Parenchyma We predicted that oxylipins present in the CM of melanoma cells might play a role in Mφ polarization and melanoma-Mφ interactions. We thus further analyzed the oxylipin profile/content in the CM from melanoma cells with/without dLGG pretreatment. On the basis of our UPLC/MS/MS data, dLGG (70 and 140 µM) treatment decreased the content of linoleic acid (LA) in the B16BM4 CM (Table S2). 13-HODE derived from LA catalyzed by 15-LOX, and 9,10-EpOME/12,13-EpOME derived from LA catalyzed by CYP450 (epoxygenase) in the B16BM4 CM were also significantly increased after pretreating with dLGG. In addition, the 9-HODE, 9,10,13-TriHOME, and 9,12,13-TriHOME derived from LA catalyzed by 9-LOX were also increased in the dLGG-pretreated B16BM4 CM compared to the vehicle control.

Discussion
Melanoma metastases are the second most common type of brain metastases in humans. A high influx of tumor-associated macrophages (TAMs) in tumors is associated with poor prognosis for various cancers, including breast, ovarian, cervical, melanoma, and hepatocellular cancers [14]. On the other hand, massive accumulation of TAMs was able to consistently activate and facilitate the development of melanoma brain metastasis [13]. It is thus believed that brain metastases will be hampered by the elimination of accumulated TAMs at metastatic sites [30]. In the current study, we established brain-seeking B16BM4 melanoma cells and a syngeneic mouse model to demonstrate the bioefficacy of a bioactive phytogalactolipid dLGG. dLGG revealed moderate inhibition of infiltrated Mφ (TEME119-Iba1+); however, dLGG reprogrammed the M2-like infiltrated Mφ, seen abundantly in the TME of tumor control mice, into M1-like Mφ. These data suggest that the M2-like Mφ with pro-tumorigenic properties in melanoma brain tumor were attenuated by dLGG, supporting the notion of the inhibitory effect of dLGG on melanoma brain metastases. We used a 3D co-culture model to mimic the interactions of melanoma cells and microglial/Mφ cells in vivo by observing the melanoma cells invading into the interface of the metastatic tissue and the brain parenchyma. We observed that dLGG pretreated melanoma cells inhibited Mφ infiltration into the interface of the brain tissue contacting the tumor plug. These results demonstrated that dLGG prevented brain-seeking B16BM4 cells communicating and interacting with Mφ in the brain tissue, further supporting the notion that dLGG inhibited melanoma brain metastases through reprogramming/re-educating Mφ cells in the TME. We also established a xenograft human A375BM eIF4g/Luc melanoma in the NOD/SCID mouse model. The brain metastatic melanoma tumor growth was found to be significantly suppressed by dLGG treatment with the same treatment doses as those used in the syngeneic mouse B16BM4 melanoma in mice, and the treated mice had a longer survival time (data not shown). This study therefore demonstrates that dLGG did indeed have a therapeutic effect against mouse and human melanoma cells metastasized into animal brains.
Our results also demonstrate that a combination of a low dose of dLGG (10 mg/kg) with the anti-angiogenesis drug Avastin (5 mg/kg) (dLGG-10 + Avastin-5) also profoundly inhibited brain metastasis, and was more effective than treatment with a low dose of dLGG (10 mg/kg) alone. It has been reported that addition of bevacizumab to glioblastoma patients receiving radiotherapy and temozolomide treatment could extend progression-free survival but did not improve the overall survival time in phase III trials [31]. In this study, we observed that dLGG-10 + Avastin-5-treated animals had a longer survival time than the tumor control, suggesting the beneficial effect of the combination of Avastin with dLGG. Moreover, we also observed that dLGG-10 combined with Lipo-DOX-2 (dLGG-10 + Lipo-DOX-2) also showed greater inhibition of melanoma brain metastasis in animals when compared to animals treated with dLGG-10 or Lipo-DOX-2 alone. These data suggest that the natural phytogalactolipid, dLGG, in combination with the current anti-cancer drugs Lipo-DOX or Avastin, may be a useful strategy for combating melanoma brain metastasis in humans.
In this study, we investigated the systemic oxylipin metabolome in the sera of B16BM4bearing mice and observed that dLGG significantly inhibited the levels of 12/15-LOXderived oxylipins, such as 8-HETE, 12-HETE, and 15-HETE in the mice sera. A previous study reported that 12(S)-HETE promotes melanoma cell adhesion ability and metastasis potential by activating the ERK and FAK signaling pathways [18]. Meanwhile, 20-HETE derived from AA by CYP450-4A has been proposed to play a critical role in tumor growth and angiogenesis [32] that was detected less in dLGG-10/-25-treated (1-3-fold decrease) mouse serum than the tumor control mouse serum. The expression levels of both 15-LOX and CYP450-4A proteins from either melanoma (Mel-A+) or microglia/Mφ (F4/80+) in the TME of dLGG-10-and dLGG-10 + Lipo-DOX-2-treated mice were lower than those in the tumor control mice. Our Western blotting data have also shown that dLGG treatment deceased the levels of 15-LOX and CYP450-4A in B16 melanoma cells (data not shown), supporting in part the decrease in both enzyme-derived oxylipins in dLGG-treated mice sera. Moreover, our study is the first to demonstrate that LA-derived oxylipins 9,10-EpOME and 12,13-EpOME have an inhibitory effect on melanoma cells invading the brain parenchyma. Moreover, 9,10-EpOME alone or in combination with 12,13-EpOME was able to promote M1-like Mφ phenotype in vitro or in the brain melanoma tumor microenvironment. In addition to reprogramming the macrophage phenotype or activity, the observation of promotion of cytotoxic T cell infiltration and inhibition of angiogenic endothelial cell marker CD31 and VEGF protein expression in the TME sheds light on the novel modes of action of dLGG against melanoma brain metastases.

Conclusions
In summary, this study provides a novel therapeutic strategy to combat melanoma brain metastases by using a natural phytogalactolipid dLGG alone or in combination with the anti-cancer drug liposomal doxorubicin or the anti-angiogenesis drug Avastatin through targeting M1/M2 polarization and inhibiting angiogenesis in the mouse TME. The novel function of oxylipin 9,10-EpOMEs + 12,13-EpOMEs in preventing melanoma cell invasion and increasing M1-like phenotypic microglia/macrophages in the tumor microenvironment also reveals a potential mechanism for cancer therapy development.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/cancers13164120/s1. Figure S1: Brain-seeking cells acclimated by repeating primary culture from an MBM mouse model with different batches of cell lines selected according to brain metastasisrelated protein marker expression level by using Western blot analysis. Figure S2: The experimental scheme of the MBM mouse model. Figure S3: The coronal brain tissue sections in each experimental group were matched to the IVIS image of the corresponding mice in a B16BM4 COX−2/Luc MBM mouse model. Table S1: Oxylipin metabolites identified in the sera of mouse-implanted brain metastatic melanoma after treatment with vehicle PBS, dLGG, Lipo-DOX, or combination treatments. Table S2: LA-derived oxylipins in the culture medium of B16BM4 cells with vehicle or dLGG treatments.