Nesting of colon and ovarian cancer cells in the endothelial niche is associated with alterations in glycan and lipid metabolism

The metabolic phenotype of a cancer cell is determined by its genetic makeup and microenvironment, which dynamically modulates the tumor landscape. The endothelial cells provide both a promoting and protective microenvironment – a niche for cancer cells. Although metabolic alterations associated with cancer and its progression have been fairly defined, there is a significant gap in our understanding of cancer metabolism in context of its microenvironment. We deployed an in vitro co-culture system based on direct contact of cancer cells with endothelial cells (E4+EC), mimicking the tumor microenvironment. Metabolism of colon (HTC15 and HTC116) and ovarian (OVCAR3 and SKOV3) cancer cell lines was profiled with non-targeted metabolic approaches at different time points in the first 48 hours after co-culture was established. We found significant, coherent and non-cell line specific changes in fatty acids, glycerophospholipids and carbohydrates over time, induced by endothelial cell contact. The metabolic patterns pinpoint alterations in hexosamine biosynthetic pathway, glycosylation and lipid metabolism as crucial for cancer – endothelial cells interaction. We demonstrated that “Warburg effect” is not modulated in the initial stage of nesting of cancer cell in the endothelial niche. Our study provides novel insight into cancer cell metabolism in the context of the endothelial microenvironment.

late 1920s 7 . Proliferating tumor cells rely on increased aerobic glycolysis to produce energy and to enable the supply of building blocks that are essential for highly proliferating cells. It has been shown that abnormal vascularization of the tumor is promoting hypoxic conditions, which in turn can cause an increase in glucose uptake and lactate production 8 . The "Warburg effect" is one of the numerous of metabolic switches, identified in wide variaty of in vitro 9 and clinical studies 10 , which may potentially be regulated by the tumor microenvironment. Those interactions, although highly important from the perspective of identification of novel druggable targets and/or recognition of causal resistance mechanisms, are so far poorly defined.
The monitoring of interactions between cancer cells and their microenvironment is limited in vivo due to the numerous of systemic (organismal) features, which might dominate the signal. We previously established an in vitro model consisting of co-culture of cancer and endothelial cells [11][12][13] . We selected endothelial cells E4 + EC previously created by transfection of the Primary Endothelial Cells (PECs) with the adenoviral E4ORF1 gene 14 . The E4 + EC cells exhibit chronic, low activation of Akt signaling 14 , which is a known feature of tumor endothelial cells 15 . Using this model we can avoid the supplementation of the media with serum and cytokines in our co-culture experiments, which is an essential condition to unbiased approach to metabolomics changes.
In the present study we used a co-culture system to investigate the impact of endothelial niche on cancer cell metabolism. In a previous study, we observed significant metabolic differences between colon and ovarian cancer cells 16 . Here, we ask whether the endothelial environment modulates cancer cell metabolism in a consistant manner, independent of the cell line specific features. We deployed non-targeted metabolomics platforms of Metabolon providing a broad coverage of metabolites from eight main metabolic pathways including amino acid, carbohydrate, cofactors and vitamins, energy, lipid, nucleotide, peptide and xenobiotics. The metabolic alterations were monitored in four different cancer cell lines, including two from colon and two from ovarian origin over a period of two days at different time points (6 h, 18 h, 24 h and 48 h) after establishing co-culture with endothelial cells. We identified metabolites displaying coherent and non-cell line specific changes over time pointing toward glycerophospholipid, fatty acid metabolism and glycosylation as pathways impacted by the endothelial niche and involved in cancer-endothelium interactions. Our findings reveal a new insight into cancer cell metabolism in context of its microenvironment.
Co-culture and sample processing. For the experiments, cells were trypsinized, washed with PBS and cultivated in M199 medium supplemented with 1% penicillin-streptomycin solution (Sigma), 4 mM L-glutamine (Sigma), 50 μ g/mL heparin (Sigma) and 20 μ g/mL β ECGF (Sigma), to avoid the effect of culture milieu on metabolic profile. The E4 + EC cells were seeded in the T25 flask 24 h prior the co-culture with cancer cells, to form a dense monolayer. The cancer cells were seeded on the top of E4+ EC monolayer. Samples were collected by trypsinization at different time points (6 h, 18 h, 24 h, and 48 h) and washed with PBS. Cells were resuspended in 50 μ L of staining media (PBS1X/FBS 5%/EDTA 2 mM) and incubated with fluorochrome-conjugated antibody (CD326, EpCAM, alexa fluor, Biolegend) for 30 minutes at 4 °C and washed with PBS. The E4 + EC and cancer cell suspension was separated using Fluorescence Activated Cell Sorting (FACS). Cancer cells cultivated in the absence of E4 + EC, were used as control (T0). Five replicates for each conditions were prepared. The experimental design is illustrated in Fig. 1A.
Cell sorting. Separation of cancer cells from E4 + EC was performed using Activated Cell Sorting (FACS) on an SORP FACSAria II (BD Biosciences). The cancer cells were separated from the E4 + ECs based on AF 647 fluorescence (APC) by 640 nm red laser and 670/14 nm emission and GFP fluorescent by 488 nm blue laser and 525/50 nm emission. Doublets were excluded by FSC-W × FSC-H and SSC-W × SSC-H analysis; Data were processed with FACSDiva 7 software (BD Biosciences). Fluorescent minus one (FMO), showing the level of background that is present in a particular fluorescent detector, was used for gating. In total 100 samples were collected (25 samples/cell line).
Metabolic profiling. The sample processing and metabolite measurements were performed as previously described 16 . Briefly, 100 of frozen pellets were sent to Metabolon Inc. (Durham, NC, USA) for metabolic profiling based on ultra-high-performance liquid chromatography-mass spectrometry (UPLC-MS) and gas chromatography-mass spectrometry (GC-MS) approaches 17 . The cell pellets were extracted using a series of organic and aqueous solvents to remove the protein fraction. Each sample extract was divided into equal parts to be measured using two methods -UPLC-MS and GC-MS. Samples were evaporated, frozen and vacuum-dried.
Sample measurements on UPLC-MS platform. Before measurement with UPLC-MS, samples were reconstituted in an acidic or basic solution compatible with LC-solvents. The samples were assayed using a Waters ACQUITY UPLC system and a Thermo Fisher Scientific Orbitrap Elite high resolution/accurate mass spectrometercomposed of a heated electrospray ionization (HESI) source and an Orbitrap mass analyzer operating at 30,000 mass resolution. Two different extract reconstitutions were performed for sample measurements in positive and negative mode. Extracts reconstituted in acidic conditions were gradient eluted using water/methanol solvent containing 0.1% formic acid followed by measurements in acidic positive ion optimized conditions. Extracts reconstituted in basic conditions were gradient eluted using water/methanol solvent containing 6.5 mM ammonium bicarbonate followed by measurements in basic negative ion optimized conditions. Sample measurements on GC-MS platform. Before measurement with GC-MS samples extracts were re-dried under vacuum for at least 24 h and derivatized using N, O-Bis(trimethylsilyl)trifluoroacetamide (BSTFA) under dried nitrogen conditions. The samples were assayed using Thermo-Finnigan Trace DSQ fast-scanning single-quadrupole mass spectrometer instrument. Sample separation was performed on a 5% diphenyl/95% dimethyl polysiloxane GC column, with helium as the carrier gas. The electron impact ionization was performed prior to mass spectrometric analysis and detection. The instrument was tuned and calibrated daily for mass resolution and mass accuracy.
Metabolomics data. The metabolite identification was performed as described previously 16 . Briefly, the metabolite peaks obtained after measurements from UPLC-MS and GC-MS platform were identified using Metabolon's propriety peak integration software. Identification of metabolites was performed by comparison of the obtained data to the library entries of purified standards or unknown recurrent entities. The combination of chromatographic properties and mass spectra provided an indication of a match to the specific compound on known or unknown entity as well as isobaric entity. The unknown metabolites, labeled as "X -number" (for example X -14577), are reproducibly detectable molecules, for which the chemical structure has not been identified yet. The isobaric entity is representing the total ion counts of components having the same retention time, mass-to-charge ratio (m/z), and GC-MS or UPLC-MS spectra, that the individual contributions of the compounds to the peak cannot be determined. The series of quality control (QC) and curation processes were conducted to ensure data quality. Library matches for each compound were checked for each sample, and manually corrected if necessary. Sample measurements were obtained during a 5-day period, and data was normalized to correct variations resulting from inter-day tuning differences in the instrument. Each compound was corrected in a run-day.
Quality check of metabolomics data. We applied principal component analysis (PCA), an unsupervised statistical analysis that converts high-dimensional data into fewer dimensions with maintenance of the variance from the original data, prior to the statistical data analysis to inspect the quality of the metabolic data. The PCA enables visual examination of sample distribution in the PC space. The PCA was carried out using software SIMCA version 14. The 100 samples clustered into four groups based on the cell line, among which we identified two outliers in SKOV3 cell line, clustered with HCT15 cell line as indicated in supplementary Figure S1. The outliers were excluded from the dataset prior statistical analysis.
Statistical analysis. Statistical analysis was performed in R version 3.1.3 and R-Studio version 0.97.551. We applied a mixed effects linear regression model for each metabolite, with dependent variable metabolite intensity and fixed effect time, and a random effect identifying the cell line. Dependent variables (metabolite intensities) were log-scaled and then z-scored before computing the statistics. Metabolites with > 20% missing values in at least one cell line were removed from the dataset. In total 143 metabolites were used for the analysis. Significance was defined at a stringent Bonferroni level of p < 3.49 × 10 −4 (p < 0.05/143).
We tested two different models of time-dependence: (1) treating time as a quantitative independent variable and (2) pair-wise comparisons of all time-points post-co-culture to time point zero.
We showed previously that including all ratios between metabolites into the analysis can reduce the overall variance and allow for the identification of biologically relevant pathways, provided conservative p-gain statistics are used 18 . We therefore also examined the time-resolved effect of endothelial cells on metabolite ratios, using mixed effects linear regression model for each pair of metabolites, time as well as the fixed effects and cell line. In total, we calculated 10,153 metabolic ratios (143 * 142/2). We considered metabolite ratios as significantly altered if they matched following conditions: p-value < 4.92 × 10 −06 (p < 0.05/10,153) and simultaneously p-gain > 1.01 × 10 05 (p-gain > 10 * 10,153).

Metabotype of cancer cell remains unique after contact with endothelial cells. We identified 143
metabolites present in all four cancer cell lines including two from ovarian (OVCAR3 and SKOV3) and two from colon (HCT15 and HCT116) cancer. The PCA results showed clear separation into four groups reflecting four cell lines as illustrated in the Fig. 1B. This observation suggests a dominant effect of the cell line on metabotype compared to a time-resolved contact with endothelial cells. We further ask whether the endothelial cells trigger metabolic changes in cancer cells which would be common among all the cell lines.
Endothelial cells are triggering coherent and non-cell line specific changes in cancer cell metabolism. We applied mixed effect linear regression model that enables testing for metabolic alterations occuring over the entire experimental period, which will be coherent for all four cancer cell lines. Five metabolites, including two polyunsaturated fatty acids (PUFA) (docosapentaenoate (DPA 22:5n3) and eicosapentaenoate (EPA 20:5n3)), one dipeptide (gamma-glutamylglutamate) and two nucleotide sugars (isobar: UDP-acetylglucosamine, UDP-Acetylgalactosamine and UDP-glucuronate) were significantly altered in all cancer cell lines ( Table 1). The PUFA's and gamma-glutamylglutamate were increased and nucleotide sugars decreased over the experimental period. Detailed information on metabolic patterns for all metabolites and cell lines are presented in Figure S1, and the computed p-values in Table S1A. Additionally, significant increase in five pairs of metabolites was observed over the whole experimental period (Fig. 2), when we used computed metabolic ratios (p-value ≤ 2.4*10 −6 and p-gain ≥ 1.02*10 5 ). The significantly altered metabolic pairs are predominantly fatty acids (DPA, EPA, DHA and palmitoleate). The full set of associations for metabolic ratios is provided in Table S2.
This data demonstrate that endothelial cells are triggering changes in cancer cell metabolism coherently and independently of the cancer cell line origin. co-culture with endothelial cells suggest alterations in pathways described below. Data interpretation for the metabolites altered as single compounds on the pathway is provided in the supplementary material ( Figure S7).

Hexosamine biosynthetic pathway but not "Warburg effect" is altered due to the contact with endothelial cells. Previous reports suggested regulatory role of tumor microenvironment on Warburg effect.
In our studies neither levels of glucose nor lactate were significantly altered in cancer cells after culture with endothelial cells. The alterations in TCA cycle molecules were also not observed in all the cell lines. The individual cell line levels of lactate remain unchanged in all cell lines and glucose exhibit minor cell line specific decrease over the entire experiment ( Figure S1). This results show that in cancer cells, metabolites associated with "Warburg effect" are not modulated by the endothelial cells in the early stage of niche formation. In turn, the endothelial cells induce significant alteration in carbohydrates level (Table 1), which can be synthetized in the hexamine biosynthetic pathway (UDP-GlcNAc, UDP-GalNac, UDP-glucuronate and Neu5Ac). All carbohydrates exhibiting significant alteration can serve as substrates for glycoconjugates formation and glycosylation. Figure 3 illustrates alteration patterns as well as potential interactions between the molecules in metabolic network related to hexosamine biosynthetic pathway. The UDP-GlcNAc can be utilized as a substrate for Neu5Ac synthesis or together with the UDP-glucoronate can serve as monosaccharide building blocks for hyaluronan synthesis, which is glycosaminoglycan, composed of 2,000-25,000 disacharides of glucuronic acid and N-acetylglucosamine 19 . Increase in lactose can also be linked to changes in glycosylation patterns, since glucose together with galactose are involved in assembling lactosylceramide 20 , ubiquitously present glycosphingolipid structures in mammals involved in cell proliferation, adhesion, migration and angiogenesis 21 . These results suggest potential involvement of hexamine biosynthetic pathway and glycosylation in the formation of cancer niche.

Endothelial cells modulate glycerophospholipids metabolism of cancer cells. Among the 21 sig-
nificantly altered metabolites we identified 9 lipids including three omega-3 (n-3 FA) and three omega-6 (n-6 FA) fatty acids as well as choline and its intermediates. These metabolites are closely connected, since glycerophospholipids such as phosphatidylcholines or lysophosphatidylcholines are composed of fatty acid chains (one or two), a phosphate group, and choline molecule. The n-3 and n-6 FA cannot be synthetized de novo in mammals, but can be modified (elongated or desaturated), once the primary molecule of n-3 (α -linolenic acid) or n-6 FA (linoleic acid (LA)) are delivered into the cell. As shown in Fig. 4A linoleic acid significantly increases and α -linolenic acid exhibits a trend of increase after the co-culture was established. These molecules can serve as a building blocks for the remaining n-6 FA's and n-3 FA's molecules. Alternatively, n-6 FA's and n-3 FA's can be released from the cellular membrane; simultaneous significant increase in choline and its intermediates (Fig. 4B) as well as increase of EPA and DPA over the entire experiment followed by the increase of DHA after 18 h, might suggest possible release of   EPA and its elongation and desaturation to form DHA. These results indicate that contact with endothelial cells induces time resolved alteration in glycerophospholipids metabolism as reflected by the alterations in fatty acid as well as choline and its intermediates.

Discussion
In this study we characterized metabolic changes associated with nesting of cancer cell in the endothelial niche by deploying in vitro co-culture system. Although the cell line specific metabotype exhibited dominant signal, we identified coherent metabolic alterations in all cancer cell lines triggered by the contact with endothelial cells. While our study shows that glycerophospholipid metabolism, hexosamine biosynthetic pathway, and glycosylation have significant role in the first 48 hours of the co-culture establishment, we do not observe alteration in the "Warburg effect" as well as TCA cycle metabolism. The elevated "Warburg effect", is considered as one of the metabolic features of cancer metabolism regulated by the microenvironment, as it is a a consequence of hypoxic conditions caused by the abnormal vascularization 8 . In this study the levels of metabolites associated with "Warburg effect" remained unchanged at different time points within the first 48 hours after the co-culture was established. This, suggest that "Warburg effect" is not a critical metabolic event in the initial phase of co-culture formation, and the alteration initially reported as caused by microenvironment 8 , might be rather associated with cancer proliferation. Generally, increase in glucose metabolism and lactate production correlates with the increased proliferation rate of the cancer cells. Previously we reported that endothelial cells enhance proliferation capacity of cancer cells 4 days after co-culture was established 12 . Accordingly, the metabolic alterations identified in the first 48 hours might be solely associated with the cancer-endothelial-cells interactions and not with proliferation of cancer cells.
The significant alterations in metabolites linked to glycosylation, was a common response to the co-culture for all tested cancer cell lines. The remodeling of glycan structures in cancer has been already associated with cross-talk between cancer cells and their microenvironment, which in turn supports its progression, invasiveness, and metastasis 20 . A direct impact of glycosylation on the cell-cell adhesion, contributing to cancer invasiveness and metastasis was reported in several studies [22][23][24][25] . For instance, significantly altered UDP-glucuronate and UDP-glucosamine can be deployed for glycosylation as separate units or can serve as substrates for hyaluronan biosynthesis, which also is involved in cellular membrane glycosylation 26 . Furthermore, increased synthesis of hyaluronan was associated with malignant progression in several cancers, including ovarian and colon cancer 27 ; its role in tumor growth and progression in vivo as well as in malignant transformation was previously reported 28 . Thus, modifications in glycosylation patterns triggered by the endothelium might be considered as a metabolic switch crucial for both: implanting of cancer cell in the endothelial niche as well as its progression and transition towards more aggressive phenotype.
The phosphatidylcholine and phosphatidylethanolamine, which can be de novo synthetized in the Kennedy pathway 29 , are the major building blocks of the cellular membrane and have direct impact on membrane structure, its integrity and signaling function 30 . Thus, the significant increase observed in both glycerophospholipid metabolism and PUFA's could be associated with remodeling of cancer cellular membrane. These changes, induced by the endothelial cells, might improve cellular cross-talk or modulate signaling pathways promoting cancer progression 31 . The composition of the fatty acid chains in the glycerophospholipid residue can be rearranged to modulate membrane fluidity and permeability in order to meet cellular demands 32 . Furthermore, increase in choline metabolism was associated with malignant transformation 33 . Increased choline and cytidine-5′ -diphosphocholine levels can be linked with the choline kinase (ChoK) displaying oncogenic activities 34 . It has been shown that increase in choline metabolism is strongly regulated by the RAS and PI3K-Akt oncogenic pathways 33,35 . This observation may suggest that endothelial niche is modulating malignant phenotype subsequent to RAS and PI3K-Akt pathway activation, resulting in increased choline metabolism.
Increase in n-6 PUFA's such as AA is associated with metastatic behaviors in ovarian cancer 36 , as well as in colon cancer 37 . Additionally, in accordance to our study, increase in lipid metabolism over the metabolism of carbohydrates and amino acids was associated with tumor metastasis 6 . The increased metastatic potential of tumor cells associated with AA can be directly linked with its derivatives namely eicosanoids (prostaglandins, leucotrines, hydroxyeicosatraenoic acids), promoting angiogenesis, and subsequently metastatic spread 38 . These might suggest that PUFA's are released from the membrane to stimulate endothelial cell proliferation, which in turn underscores the cross-talk between cancer and endothelial cells previously showed in our studies 13 . In contrast, n-3 PUFA's were associated with anti-inflammatory, anti-carcinogenic and anti-angiogenic effects 39 . It has been shown that n-3 PUFA's compete with n-6 PUFA's for incorporation in cellular membrane as well as for the COX-2 and 5-LOX enzymes, involved in eicosanoid production 39 . Thus, the simultaneous elevation of both n-3 and n-6 PUFA's may suggest tight regulation of angiogenesis stimulated by cancer cells.

Conclusions
The metabolic events accompanying the implantation of cancer cell into the endothelial niche are poorly understood. Using the in vitro co-culture system, mimicking cancer-endothelial-cell interaction, we identified metabolic alterations coherent in all cancer cell lines (independent of cancer origin), which were triggered by the endothelial cells. We have shown that contact with endothelial cells has a significant impact on the cellular membrane composition as pointed by the alterations in glycerophospholipid and fatty acid metabolism. Significant alteration in the metabolites involved in glycoconjugates formation and glycosylation might be associated with the cellular adhesion crucial for cancer -endothelial cells interaction as well as malignant transformation. Additionally, increased PUFA's level, involved in angiogenesis modulation, may suggest that cancer cells tightly control the endothelial cells progression. We demonstrated that "Warburg effect" is not modulated by the endothelial cells in the initial stage of cancer -endothelial cell interaction. Our study provides a novel input into the cancer cell metabolism in the context of endothelial microenvironment, which was poorly defined so far. However, further studies monitoring the impact of endothelial cells on cancer cell metabolism over increased time period and with larger cell line number and cancer origin would be required to provide a broad overview on formation, preservation and progression of cancer cell in endothelial niche.