Mangiferin (MGF) was purified and characterized from the aqueous extract of the stem bark and leaves of Mangifera indica L to 95,5% purity at the Departments of Pharmacognosy, Institute of Food and Pharmacy, Havana University and Department of Analytical Chemistry of Center of Drugs Research and Development, CIDEM (Núñez Sellés et al., 2002; García-Rivera et al., 2011; Gil et al., 2014). Cisplatin (cis-diamminedichloroplatinum(II), CDDP) was purchased from Shandong Boyuan Pharmaceutical Co., Ltd, China. The culture medium RPMI-1640, dimethylsulfoxide (DMSO), fetal bovine serum, penicillin-streptomycin (100x), L-Glutamine, trypsin, ethylenediaminetetraacetic acid (EDTA), sodium dodecyl sulphate (SDS), reduced Matrigel (196 USP unit/ mg) and set of Drabkin reagents were purchased from Sigma-Aldrich Inc. (Saint Louis, MO, United States). 3-(4,5-Dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT), 2-propanol, ethanol, hydrochloric acid (HCl), formaldehyde and acetic acid were obtained from Merck KGaA (Germany). Solutions of test compounds were routinely prepared in DMSO or MilliQ water and sterilized using syringe-driven 0.22 µm filters. The final concentration of DMSO in culture medium did never exceed 0.5%.

CT26. WT is an N-nitroso-N-methylurethane-(NNMU) induced undifferentiated murine adenocarcinoma of the colon, syngeneic with the BALB/c mouse, which was provided by the Molecular Immunology Center, Havana. CT26. WT cells were grown in cell culture as monolayers in T-75 cm² culture flasks (BD Falcon) in RPMI-1640 with 2 mM of L-Glutamine supplemented with 10% fetal bovine serum, 100 UI/mL of penicillin, and 100 μg/ ml of streptomycin. Only cells of the first two serial passages after cryostorage were used in the syngeneic allograft models (Evans et al., 2016). At the day of implantation, the cells were harvested from subconfluent cultures (70–85%) by trypsinization (0.05% trypsin and 0.02% EDTA), and washed twice in phosphate-buffered saline solution before they were resuspended in non-supplemented RPMI culture medium. Human colon adenocarcinoma cell line HT29 (ATCC HTB-38) was cultured in McCoy’s 5 A medium (Gibco, ThermoFisher, United States) supplemented with 1.5 mM L-glutamine, 2.2 g/L sodium bicarbonate, 100 ml/ L of fetal bovine serum and 100 U/ mL penicillin and 100 g/ L streptomycin. All cell lines were cultured in a humidified atmosphere (37°C, 5% CO₂).

MTT assays were performed to assess cell viability and/or growth inhibition of CT26 and HT29 colon cancer cell lines after exposure to MGF essentially, as previously described (Hernandez-Balmaseda et al., 2021). Briefly, cells were seeded in 96-well tissue culture plates at the concentration of 5,000 cells per well. The next day cells were treated with MGF in a concentration range of 10–400 µM for 24, 48, and 72 h. Control cells were exposed to 0.5% DMSO solvent control. Following treatment, the medium supernatant was aspirated from the wells and replaced with fresh medium. Next, 50 µL of MTT solution was added to all the wells and cells were incubated for another 4 h at 37°C. Subsequently, medium was aspirated and formazan crystals were dissolved in 50 µL of DMSO. The measurement of absorption was performed at 595 nm using (Envision 2103 multilabel reader, Perkin Elmer, United States) plate reader. The experiments were repeated with four technical replicates per treatment.

HT29 cells were seeded at a concentration of 35,000 cells per well and left to grow at 37°C under 5% CO₂ until they reached 95–100% confluence. Then, cells were treated with 10 μg/ ml of MMC for 4 h to block cell proliferation and distinguish it from cell migration. When the incubation time was finished, a scratch was made in each well with the Incucyte Woundmaker tool. Next, cells were gently washed with PBS to remove any cell debris formed after making a scratch. Afterwards PBS was removed and exchanged with serum-free McCoy’s 5 A medium to prevent cell division. The cells were treated either with MGF alone (at 1, 50 and 400 µM concentrations) or in combination with a proangiogenic cell migration growth factor VEGF (at 10 ng/ ml) for 24 h and incubated at 37°C under 5% CO₂ in the Incucyte Live-Cell Analysis system (Sartorius, Germany). Control cells were treated with an appropriate solvent, which in the case of MGF was 0.5% DMSO and in the case of VEGF–MilliQ water. Imaging was performed every 2 h for a total duration of 24 h with a × 10 objective. The area of wound closure was calculated and analysed using the Incucyte Scratch Wound Cell Migration software module of the ZOOM imaging system. The experiments were performed in three biological replicates with three technical replicates for each treatment.

Intracellular lipid accumulation induced by free fatty acid exposure leads to enhanced fatty acid oxidation (FAO) and ketogenesis (beta-hydroxybutyrate production). Protein expression levels of CPT1 (carnitine palmitoyltransferase 1), a key regulatory enzyme of mitochondrial fatty acid β-oxidation were determined by Western analysis. Ketone body (beta-hydroxybutyrate) levels were assessed using a colorimetric assay kit (Cayman Chem). Briefly HT29 cells were seeded at a density of 18*10⁶ cells in a T175 flask. The next day cells were treated with 0.2 mM FFA and 40 µM or 400 µM MGF for 24 h. Stock solutions of 0.66 M oleic acid and (1.32 M) palmitic acid (Sigma-Aldrich, Germany) were prepared in isopropanol. Equal amounts of oleic and palmitic acid were mixed to prepare a FFA stock of 1 mM FFA. Free fatty acid-free bovine serum albumin (BSA) was dissolved in serum-free McCoy’s 5 A or RPMI medium without antibiotics at the final concentration of 1% and then sterilized using syringe-driven 0.22 µm filters. Afterwards, medium was supplemented with the mixture of FFA at a final concentration of 0.2 mM and sonicated for 6–8 h until FFA were completely dissolved using a Branson 3,200 sonication bath. FFA medium was protected from light and stored at 4°C. Following treatment, cell pellets were collected with a cell scraper and used immediately to extract ketone bodies according to the manufacturers protocol. The measurement of absorption was performed at 450 nm using (Envision 2103 multilabel reader, Perkin Elmer, United States) plate reader at room temperature. The experiments were performed in one biological replicate, with three technical replicates per each treatment. Similarly, RNA was extracted using RNeasy Mini Kit (Qiagen, Germany) from the collected cell pellets. RNA quantity was determined using Qubit™ RNA Broad Range Assay Kit with the aid of the Invitrogen Qubit™ 4 Fluorometer (Thermo Fisher Scientific, United States). The extracted RNA was stored at -80°C until further QPCR analysis. For protein analysis, cell pellet was washed once with cold PBS, centrifuged (1,500 rpm, 3 min, 4°C) then cells were lysed using 1x RIPA buffer (Cell Signalling, United States) with a cocktail of protease inhibitors (Sigma-Aldrich, Germany) and left on ice for 15 min. Afterwards, cell lysates were centrifuged (13,000 rpm, 15 min, 4°C) and the supernatants were transferred to the new test tubes. Protein concentration was measured using Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific, United States) and the measurement of absorption was performed at 560 nm using (Envision 2103 multilabel reader, Perkin Elmer, United States) plate reader. Prior to SDS-PAGE electrophoresis, protein lysates were mixed with Laemmli buffer (Biorad, United States) and 50 mM 1,4-dithiothreitol (DTT) and then heated at 70°C for 10 min. Then, the samples and protein ladder (BenchMark™ Protein Ladder, Thermo Fisher Scientific, United States) were loaded on SDS-PAGE polyacrylamide gels (stacking 6%, resolving 12%) at protein concentration equal to 10 µg/well. The electrophoresis was run in Mini-PROTEAN Tetra Cell System (Biorad, United States) using the high molecular weight buffer (100 mM MOPS, 100 mM Tris, 0.2% SDS, 2 mM EDTA, 5 mM sodium bisulfite) at 70 V until the samples reached the separating gel and then the voltage was increased up to 120 V. Afterwards, the proteins were transferred on pre-wet nitrocellulose membranes (Cytiva, United States) in Power Blotter 1-step transfer buffer (Invitrogen, United States) using Power Blotter XL System (Invitrogen, United States) (7 min at 1.3 A). Nitrocellulose membranes were blocked in blocking buffer (5% milk solution in TBST) for 1 h at room temperature and incubated overnight at 4°C with primary antibodies diluted in blocking buffer as follows: 1:4000 for anti-CPT1A (Proteintech, UK). Next day, membranes were washed three times with TBST and then incubated with HRP-conjugated anti-rabbit secondary antibody diluted in blocking buffer (1:2000). Anti-GAPDH (Cell signaling, United States) antibodies (diluted 1:2000) in blocking buffer were used as loading controls for CPT1A. Protein detection was performed on the Amershan imager 680 (Cytiva, United States) using Pierce™ ECL Western Blotting Substrate (Thermo Fisher Scientific, United States) and quantified using Image-J software.

Housing and all procedures involving animals were performed according to protocols approved by the Drug Research and Development Center Ethics Committee for Animal Research and in compliance with the Cuban National Guidelines for the Care and Use of Laboratory Animals (CECMED Resolution No. 37 2012 for Good Laboratory Practice for Quality Control). Male BALB/c mice were purchased at National Laboratory Animal Production Center, Havana, Cuba. BALB/c mice were used at 6–8 weeks of age with weight between 18 and 22 g, in all experiments. They were supplied with water and food at libitum and were kept in quarantine for 15 days for adaptation to temperature (22°C ± 2°C), humidity (77 ± 3%) and alternating cycles of 12 h light/dark environment. For allograft studies, the animals were inoculated (as specified in more detail for each specific model below) with 1x10⁵ CT26. WT cells/mouse in 100 μL of non-supplemented RPMI-1640 culture medium. The following day, they were weighted, labeled and placed in separate boxes.

Subcutaneous tissue tumors were fixed in formalin (10% w/v in phosphate-buffered saline–PBS pH 7.4) during 24 h and sections (4 µm) were processed for immunohistochemistry analysis. A biotin-peroxidase system was used with identification of the secondary antibody by polymer (ADVANCE HRP). Antigen retrieval was performed in water bath at 98°C, with a citrate buffer solution (pH 6.0) (Target Retrieval Solution) for 20 min. In order to block the endogenous peroxidase activity, slides were incubated with a solution of H₂O₂ 3% in methyl alcohol. Reagents were applied manually with 1 h incubation for the primary antibody and 30 min for the reagents, except for chromogen diaminobenzidine, incubated for 5 min. Minichromosome maintenance complex component 7 (MCM 7 or CDC47), a marker of cell proliferation, was measured using a specific monoclonal antibody CDC47 (clone 47DC141, Neomarkers, 1/300) (Zorde Khvalevsky et al., 2013). Positive and negative controls were included in each bath. The index for CDC47 staining was obtained by counting the percentage of positive cells in 500 tumor cells.

Subcutaneous and pulmonary tumor tissue samples from each animal (n = 5/group) were individually treated with RNAlater ^® (Qiagen, Benelux, Belgium) by 24 h at 4°C, and immediately stored at −80°C until the RNA isolation procedure. Up to 30 mg tissue were completely disrupted and homogenized in 700 μL QIAzol^® Lysis Reagent (Qiagen, Benelux, Belgium) by the Tissue Lyser II (Qiagen^®, Benelux, Belgium) with 5 mm stainless steel beads diameter to extract total RNA. Total RNA was isolated using RNeasy^® Microarray Tissue kit (Qiagen, Benelux, Belgium), according to manufacturer’s protocol. Total RNA was purified on RNeasy Mini spin columns according to the manufacturer’s protocol. During the purification process contaminating genomic DNA was removed by means of DNase digestion. Purity and concentration total RNA samples were evaluated by ultraviolet spectroscopy (NanoDrop, Thermo Scientific, Wilmington, United States). The integrity of the RNA samples was evaluated by the Experion Automated Electrophoresis System (BioRad, CA, United States) according to the manufacturer instructions. Good quality of the RNA material samples with A260/280 ratio between 1,9–2,1 were used for further analysis.

Five hundred ng of total RNA from 18 samples (3 replicates by group) subcutaneous and pulmonary tissue tumors were amplified using the Illumina TotalPrep RNA Amplification kit (Life Technologies, Carlsbad, CA, United States). RNA was reverse transcribed using T7 oligo(dT) primers after which biotinylated cRNA was synthesized through in vitro transcription reaction. Seven hundred 50 ng of amplified cRNA was hybridized to a corresponding array on a Mouse WG-6 2v bead chip (Illumina, San Diego, CA, United States). The beadchip array was incubated for 18 h at 58°C in a hybridization oven whilst continuous rocking. After several consecutive washing steps, bead intensities were read on Illumina iScan. Raw data intensities were read in R using the “bead array” package (v2.8.1) (Dunning et al., 2007). Intensities were quantile normalized and differential gene expression between samples was estimated using “limma” (v3.14.1) (Ritchie et al., 2015). Genes were considered to be differentially expressed when they had a p-value < 0.05 and a fold change of at least 1.25 in comparison to controls. Transcriptomic analysis was done by uploading Illumina bead array gene expression data into Ingenuity Pathway Analysis (IPA) software (Ingenuity^® Systems, www.ingenuity.com, Redwood City, CA, United States and QIAGEN Inc., https://digitalinsights.qiagen.com/products-overview/discovery-insights-portfolio/analysis-and-visualization/qiagen-ipa/?), and performing a core analysis according to the instructions provided. Each identifier was mapped to its corresponding gene object in the Ingenuity knowledge base. Fisher’s exact test was used to calculate a p-value determining the probability that each biological function and/or disease assigned to that data set is due to chance alone. Metascape systems biology freeware (https://metascape.org/) was used for correlating the transcriptomic profiles of the different in vivo data (Zhou et al., 2019). The Circos plot visualization allows to show how genes from different input gene lists overlap. Heatmaps show Metascape enrichment analysis of all statistically enriched ontology terms (GO/KEGG terms, canonical pathways, hall mark gene sets). Accumulative hypergeometric p-values and enrichment factors are calculated and used for filtering. Remaining significant terms are then hierarchically clustered into a tree dendrogram based on Kappa-statistical similarities among their gene memberships. The term with the best p-value are selected within each cluster as a representative term to be displayed in a hierarchical tree dendrogram. The heatmap cells are colored by their p-values (see color legend). Along the same line, Metascape enrichment analysis of all statistically enriched transcription factor (TF)-target interaction networks is dermined by the TRRUST database (Han et al., 2018). Protein-protein interactions (PPI) among all input gene lists are extracted from PPI data source to form a PPI network. GO enrichment analysis is applied to the network to assign biological “meanings” of sub-protein networks. GO enrichment analysis is applied to each MCODE network to assign “meanings” to the network component, where top three best p-value terms were retained. MCODE components were identified from the merged network. Each MCODE network is assigned a unique color.

Total RNA (700 ng) isolated by RNeasy Mini Kit (Qiagen), was reverse transcribed to cDNA by the GoScript™ Reverse Transcriptase kit (Promega, United States) according to the manufacturer’s protocol. RT-qPCR was performed using GoTaq^® qPCR master mix (Promega, United States) on Rotor-Gene Q (Qiagen, Germany). Gene expression was calculated with the ΔΔCt method after normalization against the glyceraldehyde-3-phosphate dehydrogenase (GAPDH), actinB (ACTB) or beta-2-microtubulin (B2M) housekeeping genes, as further specified. Following gene specific primersets were used for array validation experiments of C-C Motif Chemokine Ligand 3 (CCL3), Collagen Type VI Alpha three chain (COL6A3), C-X-C Motif Chemokine Ligand 9 (CXCL9) and Metalloproteinase 13 (MMP13); and the following gene specific primer sets: CCL3 forward primer: 5′-CAG​CCA​GGT​GTC​ATT​TTC​CTG-3′ and reverse primer: 5′-AGG​TCT​CTT​TGG​AGT​CAG​CG-3´. COL6A3 forward primer: 5′-GGA​ACC​ACG​GAA​GAG​AGC​AA-3′ and reverse primer: 5′-CAGGGAACTGACCCAAGACA-3´.CXCL9 forward primer: 5′- TGT​GGA​GTT​CGA​GGA​ACC​CT-3′ and reverse primer: 5′- AGT​CCG​GAT​CTA​GGC​AGG​TT-3´. MMP13 forward primer: 5′- GGA​GCC​CTG​ATG​TTT​CCC​AT-3′ and reverse primer: 5′- GTC​TTC​ATC​GCC​TGG​ACC​ATA-3´. The experiment was carried out in duplicate in three independent biological replicates. The thermal cycling conditions included initial denaturation at 95°C for 2 min and subsequent 40 cycles of two-step protocol: denaturation at 95°C for 15 s and annealing/extension at 60°C for 60 s.

One-way or two-way ANOVA with Dunnett post hoc test (comparison every mean to control mean) were performed to evaluate statistical significance, as indicated in Figure legends. Survival analysis was done with the log-rank (Mantel-Cox) statistical test. All statistical analyses were performed using GraphPad Prism^® (GraphPad Software, La Jolla California, United States) statistical package. P value < 0.05 was considered significant.

The CT26 syngeneic colorectal cancer tumor model was used to investigate the effect of MGF on tumor growth in vivo. CT26. WT cells were grown for 14 days after subcutaneous injection. Changes in tumor volume by MGF treatment were measured during 2 weeks of treatment protocol. As shown in Figure 1A, allgrafts treated with MGF show dose dependent tumor regression, comparable to the volume reduction observed upon treatment with the reference chemotherapeutic drug CDDP. Significant treatment antitumor effects start to be observed from day 22, following 8 days of treatment (Figure 1A), reaching a maximal reduction in tumor size of 75% at 100 mg/kg. Overall, CDDP treatment revealed the strongest antitumor effect (approximately 90% tumor size reduction). Along the same line, MGF and CDDP dependent reduction of tumor volume corresponds with decreased immunohistochemical staining intensity of the cell proliferation marker CDC47 reflecting arrest of tumor growth (Figures 1B,C).

To get more insight in the molecular mechanism of the antitumor of MGF, a gene expression microarray experiment was performed. Gene expression profiles of subcutaneous tumor tissue obtained from mice treated for 2 weeks with 5 mg/ kg CDDP, 10 mg/ kg MGF, 50 mg/ kg MGF and 100 mg /kg were compared to untreated control mice with tumors. In general, MGF showed mild transcriptional changes as compared to CDDP treatment. Using a fold change difference of at least 1.25 and a p-value below 0.05, 454 genes were found to be differentially expressed in the 10 mg/ kg MGF condition, 263 genes in the 50 mg/ kg MGF condition, 812 genes in the 100 mg/ kg MGF condition and 3,340 genes after 5 mg/ kg CDDP treatment (Supplementary Table S1). Highly significant positive correlations could be observed upon comparing the log2 microarray gene expression values and the ΔCt qPCR expression values of CCL3, COL6A3, CXCL9 and MMP13 target genes, supporting the validity our microarray results (Supplementary Figure S1).

Next, bioinformatic analysis of transcriptome changes of the different treatments was performed by correlating the -log10 p-values for every treatment combination (Figure 2A). Whereas strong positive correlations were observed for transcriptional responses for the different MGF doses (r = ≥ 0.5), only weak overlap could be observed for MGF and CDDP treatments (r = 0.13), suggesting a different mechanism of action (Figure 2B). Next, we focused on pathway enrichment analysis of disease and biological pathways of the MGF treatment specific transcriptional changes using Ingenuity Pathway Analysis (IPA^®, QIAGEN Redwood City) (Figures 2C,D; Supplementary Table S1). Differentially expressed genes for all MGF doses are enriched for regulation of cancer and colon cancer related disease functions, including cell death and survival and cell motility, gastrointestinal neoplasia, gastrointestinal carcinoma, malignant neoplasm of large intestine, and digestive system cancer (Figure 2C).

Along the same line, significant transcriptional changes could be observed at the gene expression level in cytokine signaling (acute phase response, oncostatinM), mitochondrial metabolic stress (oxidative phosphorylation, fatty acid oxidation, Nrf2, Glutathione detoxification, SIRT, and AMPK/mTOR signaling), cell cycle (cyclins, checkpoints, mitotic PLK kinases), metalloproteases (MMPs), cell death (death receptors), iron homeostasis (heme degradation) DNA damage (ATM, BRCA1, GADD45 signaling), cell death (ubiquitination), differentiation (Wnt/catenin) and xenobiotic-metabolizing hormone receptor signaling (Aryl hydrocarbon and aldosterone receptor signaling) (Figure 2D; Supplementary Figure S3; Supplementary Table S1).

Next, we used the IPA regulation z-score algorithm to identify biological functions and pathways that are most significantly activated (Z-score >1) or repressed (Z-score < -1) by MGF based on direction of gene expression changes (increase/decrease). Interestingly, MGF dose dependently suppresses expression of multiple genes involved in tumor cell motility (invasion), angiogenesis, vasculogenesis and development of vasculature functions (Figure 3A). Furthermore, when analyzing possible protein interaction networks of MGF responsive gene targets by the STRING algorithm (https://string-db.org/), we identified various interconnected protein network clusters involved in cell motility, angiogenesis and cell death/autophagy/proliferation, or dysfunctional mitochondrial oxidoreductase metabolism, in line with the results of our IPA analysis (Figures 2D, 3A and Supplementary Figure S2).

To further characterize MGF anticancer effects on cell viability and cell motility in vitro, we next performed MTT and wound scratch assays in mouse CT26 and human HT29 colon cancer in vitro cell models. First, CT26 and HT29 were treated with MGF for 24, 48, and 72 h with a MGF concentration range from 10 to 400 μM, followed by a colorimetric MTT assay, as previously described (Hernandez-Balmaseda et al., 2021). As can be observed from Figure 3B, MGF does not significantly decrease cell viability in the dose range tested and shows no adverse toxicity. Next, MGF effects on prometastic and proangiogenic cell motility were analyzed for 24 h in a wound scratch assay in VEGF treated HT29 cells by realtime Incucyte Live Cell imaging in IncuCyte^® ImageLock 96-well microplates, after scratching cells with the Incucyte Woundmaker tool (Kobelt et al., 2021). Time dependent changes in wound closure of the different treatments were quantified by the Incucyte Scratch Wound Cell Migration software module of the ZOOM imaging system. In line with our gene expression pathway analysis in mouse CT26 cells, MGF treatment was also found to significantly suppress VEGF induced prometastatic and proangiogenic cell motility in human HT29 cells (Figures 3C,D).

Next, we further verified whether inhibition of vasculogenesis by MGF could also be experimentally confirmed in a CT26. WT cells engrafted matrigel plug assay, allowing to measure effects on neovascularization of tumor cells in vivo (Kastana et al., 2019). Angiogenesis is an attractive cancer therapeutic target, not only because it supplies oxygen and nutrients for the survival of tumor cells, but also provides the route for metastatic spread of these cancer cells.

Tumor formation in the subcutaneous matrigel plug tissue, with abundant blood supply, was observed in the CT26. WT engrafted control animals. In contrast, the plug in control animals colors yellow lacking blood irrigation in the matrigel subcutaneous implant (Figure 4A). Clearly, MGF treatment promotes a reduction of peritumoral and intratumoral blood vessel supply and tumor size in a dose-dependent manner (Figure 4A from left to right). Moreover, quantitative analysis of Hb content shows that MGF dose dependently inhibits angiogenesis, i.e. 58% and 81% decrease of Hb levels at respectively 100 and 200 μg/ml MGF treatment (p < 0.05) (Figure 4B). As expected, CT26. WT tumor cells show an adequate proangiogenic response in matrigel (81% increase in Hb content) as compared to the control setup (Figure 4B). Besides, the antiangiogenic effect of MGF also results in dose dependent reduction in tumor volume (84% reduction at 200 μg/ ml MGF, Figure 4C) or tumor weight (73% reduction at 200 μg/ ml MGF, Figure 4D). Besides, a histological study was performed using the hematoxylin-eosin and Masson’s trichrome staining. This allows clear differentiation of the collagen fibers (colored in blue) from the blood vessels (colored in red). As can be observed in the histological staining (Figure 4E) or blood vessel count (Figure 4F), MGF dose dependently inhibits the formation of blood vessels in the tumor periphery of the CT26. WT engrafted Matrigel (p < 0.001). Altogether, these results support our gene expression pathway enrichment analysis which identified various MGF responsive target genes involved in suppression of angiogenesis-vasculogenesis.

To further evaluate suppression of cancer cell motility-migration in response to MGF, as predicted in our transcriptome analysis, we next investigated whether MGF could reduce number of lung metastasis nodules upon intravenous puncture of colorectal cells in the mouse tail veins. The number of CT26 metastasis nodules in lungs were quantified following 17 days treatment in the different experimental setups as shown in Figure 5. A significant reduction (p < 0.001) in the number of pulmonary surface metastatic nodules could be observed by increasing doses of MGF, respectively 46, 77, and 88% decrease upon treatment with 10, 50 or 100 mg/ kg b. w. MGF as compared to the vehicle control group (Figure 5A). Similarly, metastatic nodules could be reduced 83% upon treatment with 5 mg/kg b. w. CDDP (83%) (Figure 5A). Treatments with 50 and 100 mg/ kg b. w. MGF, or 5 mg/ kg b. w. CDDP also resulted in a significant improvement of longterm survival rate as compared to the vehicle control group (MGF 50 mg/ kg: p = 0.027; MGF 100 mg/ kg: p = 0.020; CDDP: p = 0.022) (Figure 5B). Remarkably, gene expression analysis of pulmonary metastatic colon tumor tissue biopsies from mice treated with MGF revealed similar gene expression patterns and pathway enrichment as in the primary CT26 tumor model (Figure 5C; Supplementary Table S2). Accordingly, MGF treatment targets mitochondrial energy metabolism (fatty acid oxidation, sirtuin, TCA) NFκB, PPAR, Wnt and GP6 signaling to suppress cancer cell migration-metastasis processes (Figure 5C). Since tumor tissue biopsies are a heterogenous mixture of CT26 tumor metastasis and healthy lung cells, tumor gene expression changes by MGF treatment at 100 mg/kg are less pronounced than at 50 mg/kg since the fraction of CT26 tumor cells is strongly decreased at the highest MGF concentration, which significantly reduces the MGF effect size in the mixed cell population.

To further characterize molecular mechanism of action of MGF, we further crosscompared MGF responsive genes in the primary CT26 tumor model and the metastatic cancer models by Metascape systems biology freeware (https://metascape.org/) (Zhou et al., 2019) (Supplementary Figure S4; Supplementary Table S3). As can be observed in the circos plot (Figure 6A), we observe significant overlap between gene expression changes as well as biological pathways, as both experiments probably target similar biological processes. This is also clear in the overlapping heatmap representation (Figure 6B), showing highly significant pathway enrichment related to blood vessel development, lipid metabolism, chemotaxis, locomotion, wound healing, generation of energy metabolites, hormone stimulus, growth factor response and leukocyte activation among others (Supplementary Table S3). Further TRRUST analysis of all statistically enriched TF-target interaction networks (Koundouros and Poulogiannis, 2020) identified major regulatory roles for STAT3, PPARA/G, HIF1A and NFκB1/RelA transcription factor families, which are key players in lipid metabolism, hypoxia and inflammatory tumor signaling (Figure 6C). Furthermore, Metascape protein-protein-interaction (PPI) analysis of all MGF responsive target genes reveals a global protein network involved in reprogramming of mitochondrial energy metabolism in CT26 tumor cells (Figures 6D,E). More specifically, 4 major protein interaction subnetworks could be identified (Figure 6F) involved in fatty acid β-oxidation and toll like receptor signaling (MCODE1), generation of energy metabolites (oxidation organic compounds, ribose phosphate metabolism) (MCODE2), mitochondrial respiratory chain complex 1 (MCODE3) and NFκB signaling (MCODE4) (Figure 6G).

To verify whether MGF directly targets fatty acid β-oxidation energy metabolism in colon cancer cells, we measured ketone body concentrations (β-hydroxybutyrate) upon exposure of human HT29 colon cancer cells to free fatty acids (to mimick free fatty acids present in the tumor microenvironment in vivo) in presence or absence of MGF. Figure 7A shows that 24 h treatment of HT29 cells with 0.2 mM free fatty acids (FFA) increases intracellular concentrations of β-hydroxybutyrate as compared to untreated HT29 cells. Of special note, cotreatment with 400 µM MGF reverses this effect to levels observed in untreated cells. Along the same line we measured changes in protein expression of carnitine palmitoyltransferase (CPT1), a key metabolic enzyme involved in fatty acid β-oxidation. Similarly, FFA treatment increases CPT1 expression levels whereas, MGF reverses the effect to control levels (Figure 7B).