Fructose Metabolism in Tumor Endothelial Cells Promotes Angiogenesis by Activating AMPK Signaling and Mitochondrial Respiration

Blood vessels are vital for tumor progression by supplying nutrients and oxygen for tumor growth and providing routes for tumor metastasis (1). Endothelial cells (EC) can rapidly form new vessels from preexisting vasculature via angiogenesis, which has long been considered to be primarily orchestrated by proangiogenesis factors, such as VEGFA (2, 3). Until recently, the pivotal role of metabolism in regulating EC phenotype is becoming recognized (4, 5). It is shown that ECs rely on glycolysis for ATP and biomass synthesis, which are necessary for the key events of angiogenesis, such as proliferation and migration of ECs (5, 6). Blocking the critical glycolytic enzyme PFKFB3 in ECs inhibits angiogenesis in ocular disorders and melanoma (6, 7), suggesting that metabolic reprogramming in ECs is a potential target for antiangiogenesis therapy.

Fructose, together with glucose, are two major types of monosaccharides (8). Glucose is a primary energetic and synthetic fuel for most tissues and cell types in the body. Dietary fructose is rapidly absorbed in the intestine, then concentrated and metabolized in the liver (9, 10). It is transported into hepatocytes by solute carrier family 2 member 5 (SLC2A5, also called GLUT5) and SLC2A2, then metabolized by ketohexokinase (KHK) to fructose-1-phosphate (F1P), which is catalyzed to dihydroxyacetone phosphate (DHAP) and glyceraldehyde (GA) by Aldolase (ALDO). GA is then phosphorylated into glyceraldehyde-3-phosphate (G3P) by triose-kinase (TKFC; ref. 9). Both DHAP and G3P can be metabolized to acetyl coenzyme A (CoA) to fuel tricarboxylic acid (TCA) cycle or de novo lipogenesis (DNL; refs. 11–13), or used as glycolytic and gluconeogenic carbon pools (14, 15).

Consumption of fructose has risen markedly in recent decades owing to the use of high-fructose corn syrup in beverages and processed foods, and excessive fructose consumption is increasingly considered as a contributor to the emerging epidemics of obesity and the associated cardiometabolic disease (8, 16). As the major site to metabolize fructose, the liver may be mostly affected by high fructose consumption. Three recent studies revealed that excessive fructose stimulated DNL and hepatic fat deposition, and thus caused nonalcoholic fatty liver disease (NAFLD), steatohepatitis and even liver cancer (11, 12, 17). Besides, colorectal cancer cells that metastasized to the liver rapidly grew by enhancing fructose metabolism to adapt to the fructose-enriched liver microenvironment (15). To date, whether fructose can be utilized by tumor endothelial cells (TEC), and the roles of fructose metabolism in angiogenesis and blood-borne metastasis, have not been addressed.

Hepatocellular carcinoma (HCC) is the dominant type of liver cancer with increasing incidence worldwide (18). It is characterized by active angiogenesis, rapid blood-borne metastasis, and high mortality. Sorafenib and lenvatinib, which repress angiogenesis by targeting VEGF signaling, are the first-line systemic therapies for advanced HCC (19). However, their efficacy is much less than anticipated (20, 21). Exploring the metabolic alteration in TECs and their roles in HCC angiogenesis may provide new therapeutic targets. This study disclosed that enhanced fructose metabolism resulting from hypoxia microenvironment boosted the mitochondrial respiration activity by activating the AMPK pathway in TECs, which promoted the migration and proliferation of ECs and therefore stimulated angiogenesis in liver cancer. These findings identify the importance of fructose metabolism in regulating tumor angiogenesis and indicate that restricting fructose intake or inhibiting fructose metabolism may represent a potential strategy to suppress hepatic tumor angiogenesis.

HCC and adjacent nontumor liver tissues were collected from patients who underwent HCC resection at Sun Yat-sen University Cancer Center. Both tumor and nontumor tissues were confirmed histologically. Written informed consent was obtained from each patient, and the protocol was approved by the Institutional Research Ethics Committee (GZR2019–079). The patients were anonymously coded in accordance with ethical guidelines, as instructed by the Declaration of Helsinki.

TECs and NECs were isolated from HCC tissues and the matched adjacent nontumor livers, respectively. Fresh surgical samples were rinsed in PBS containing 1% penicillin/streptomycin (15140163, Thermo Fisher Scientific) and then sliced into small pieces. The necrotic tissues were removed. The sliced tissue was incubated in a digestion buffer, which contained collagenase II (1 mg/mL; 17101015, Thermo Fisher Scientific), collagenase IV (1 mg/mL; 17104019, Thermo Fisher Scientific), DNase I (200 μg/mL; DN25–100 mg, Sigma-Aldrich), dispase (2.5 U/mL; D4693–1g, Sigma-Aldrich), and 1% penicillin/streptomycin, for 40 minutes at 37 °C with periodic shaking. Afterward, the tissue lysate was passed through a sterile 70-μm-mesh filter to collect single-cell suspension, which was subjected to EC isolation using anti-CD31 antibody-conjugated magnetic beads (130–091–935, Meltenyi Biotec). The magnetic-activated cell sorting was repeated again. Purity of isolated ECs was confirmed by flow cytometry using anti-CD31 antibody (303106, BioLegend) and by RT-PCR showing selective expression of EC-enriched surface markers (PECAM1 and VEGFR2) but not of markers of smooth muscle cells and fibroblasts (ACTA2), liver epithelial cells (ALB), and leukocytes (PTPRC).

Human umbilical vein endothelial cells (HUVEC) were isolated as described previously (22) and cultured in serum-free medium (SFM; 11111044, Thermo Fisher Scientific) for ECs supplemented with 20% fetal bovine serum (FBS, 10099, Thermo Fisher Scientific) and 0.03 mg/mL of endothelial cell growth supplement (ECGS, 02–102; Merck Millipore). Human liver sinusoidal endothelial cells (HLSEC; JNO2136, JENNIO Biological Technology) were maintained in endothelial cell medium (1001-b, ScienCell) supplemented with 5% FBS and 0.03 mg/mL ECGS. Human brain microvessel endothelial cells (HBMEC, HTX2403, Otwo BioTech), a mouse pancreatic islet endothelial cell line MS1 (ATCCCRL-2279, ATCC), and a mouse hepatoma cell line Hepa1–6 (ATCCCRL-1830, ATCC) were maintained in DMEM (11965126, Thermo Fisher Scientific) supplemented with 10% FBS. All cells were tested negative for mycoplasma contamination and authenticated by short tandem repeat fingerprinting at Cyagen Biosciences Company.

Small-interfering RNA (siRNA) duplexes were purchased from GenePharma and their sequences are provided in Supplementary Table S1. The siRNAs targeting human KHK (GenBank accession no. NM_000221.2) and HIF1A (NM_001530), and mouse KHK (NM_001310524.1) were indicated as si-KHK, si-HIF1α, and si-mKHK. The siRNA targeting GFP (Aequorea Victoria) served as a negative control and was nonhomologous to any human or mouse genome sequences.

The EC-targeted siRNA delivery nanoparticles were prepared as previous reports (23, 24). Briefly, the 1,2-Epoxydodecane modified PEI600 (named 7C1; ref. 23) was combined with C₁₄PEG₂₀₀₀ (880150P, Avanti Polar Lipids) at a molar ratio of 4:1, and then applied to form siRNA-loaded liposomes (siRNAs vs. 7C1 = 5: 1 of mass ratio). The size of nanocomposites was under 100 nm, as detected by dynamic light scattering using a Zetasizer Nano ZS machine (Malvern Panalytical Ltd.), and the siRNA concentration was determined by NanoDrop measurement (Thermo Fisher Scientific).

All animal experiments were performed according to the institutional ethical guidelines for animal experiments and the protocol was approved by the Institutional Animal Care and Use Committee at Sun Yat-sen University (SYSU-IACUC-2019-B586).

For xenograft model, Hepa1–6 cells (1 × 10⁶) were resuspended in 25 μL Matrigel/DMEM (mixed at 1:1 volume ratio; Matrigel; 3432–005–01, R&D Systems) and then inoculated under the capsule of the left hepatic lobe of male C57BL/6J mice. Tumors, livers, and lungs were collected 4 weeks after inoculation.

For autochthonous liver tumor model, a plasmid mixture of 20 μg pX330-U6-sgTP53-CBh-hspCas9 (a gift from Dr. Bin Zhao, Zhejiang University, Hangzhou, China; ref. 25), 20 μg pT3EF1aH-c-Myc (a gift from Dr. Junfang Ji, Zhejiang University; refs. 26, 27), and 1.6 μg transposase-encoding vector (pPGK-SB13, BioVector NTCC Inc.) that was dissolved in saline (0.1 mL/g mouse body weight) was hydrodynamically injected into the lateral tail veins of male C57BL/6J mice. Tumors, livers and lungs were collected 7 weeks after plasmid injection.

For both xenograft and autochthonous tumor models, male C57BL/6J mice were administered with drinking water without or with 10% (weight/volume) fructose (F3510–500g, Sigma-Aldrich) during the experiment period. The daily intake of calories was provided by diet (12.355 kcal/day) with or without 10% fructose (1.875 kcal/day) in drinking water. PBS (vehicle control) or 2,5-anhydro-d-mannitol (2,5-AM; 25 mg/kg; A649000, Toronto Research Chemicals) were intraperitoneally administered daily, beginning on the 8th day after tumor cell implantation or hydrodynamic injection. The length (L) and width (W) of the dissected tumors were measured, and the tumor volume (V) was calculated using the formula: V = (L × W²) × 0.5. Tumor tissues were further applied to IHC staining for vessels. Thirty serial sections from the liver or lung were examined by hematoxylin–eosin staining, and the metastatic foci were scanned under a microscope independently by two researchers, who were blinded to the treatment. Metastasis rate was defined as the number of mice displaying metastasis relative to the total number of tumor-bearing mice. The total number of metastatic foci from 30-serial sections and the average diameter of the top three largest metastatic foci among 30-serial sections were recorded.

For Matrigel plug assay, fructose (2.5 μL, 800 mmol/L) or an equal volume of PBS was resuspended in 500 μL Cultrex Matrigel (Matrigel with growth factor: Matrigel with reduced growth factor = 1:1; 3432–005–01 and 3433–005–01; R&D Systems) and then injected subcutaneously into either side of the posterior flank of the same BALB/c nude mouse. The Matrigel plugs were harvested 7 days after inoculation, embedded in OCT compound (4583, Sakura), and cut into 4-μm sections at −20°C. The frozen sections were IHC stained for vessels.

For the endothelial-specific loss-of-function assay, Hepa1–6 cells (4 × 10⁵) were resuspended in 100 μL Matrigel/DMEM (mixed at 1:1 volume ratio) and then implanted subcutaneously into the posterior flank of male C57BL/6J mice. On the 7th, 10th and 14th day postimplantation, fructose (at a final concentration of 4 mmol/L in tumor) or an equal volume of PBS (vehicle control), together with siGFP (control) or simKHK (1 mg/kg) coupled with EC-targeted polymeric nanoparticles (EC-nanoparticles) were injected into the xenografts. Tumors were dissected three days after the last injection.

Tissue sections were prepared as described (28). Sections were incubated at 4°C overnight with rat mAb against mouse CD34 (mCD34, 119302, BioLegend) or mouse mAb against KHK (sc-377411, Santa Cruz Biotechnology), followed by a goat anti-rat secondary antibody kit (PV-9004, ZSGB-Bio) or two-step EnVision System (Dako Cytomation). Sections were counterstained with hematoxylin (ZLI-9610, ZSGB-BIO) and mounted in a nonaqueous mounting medium. All runs included a no primary antibody control. The stained sections were then scanned using a digital scanner (Aperio VERSA 200, Leica). The mCD34 staining area relative to the total tissue area was evaluated and the expression of KHK was determined by a Histo-score (H-score) using the Aperio software (Leica).

To mimic the low-glucose tumor environment and evaluate the role of fructose, ECs were cultured in a basic medium (no glucose, no sodium pyruvate DMEM; 11966–025) supplemented with 0.6 mmol/L glucose (D9434, Sigma-Aldrich), 1 mmol/L sodium pyruvate and 10% FBS. And fructose was added at a concentration of 4 mmol/L.

For tube formation assays, ECs treated with or without 4 mmol/L fructose, glucose, mannose (D813082, Macklin), sorbose (L818022, Macklin), galactose (D810318, Macklin), or allose (D800985, Macklin) for 48 hours were resuspended in 1% FBS-containing basic medium, reseeded to a 96-well plate coated with growth-factor-reduced Matrigel (R&D Systems), then incubated at 37°C for 6 hours. The formation of capillary-like structures was examined under a microscope. The number of the formed tube network in the whole field, which represented the ability of angiogenesis in vitro, was quantitated.

For migration assays, ECs treated with or without 4 mmol/L fructose or glucose for 48 hours were resuspended in serum-free basic medium, placed into the upper chamber of the 24-well Boyden insert with an 8-μm pore size polycarbonate membrane (3422, Corning), whereas the lower chamber was filled with 600 μL of 10% FBS-containing medium. Six hours after incubation, cells were fixed and stained with 0.1% crystal violet, and the cells remaining on the upper surface of the membrane were removed. All cells that migrated to the lower surface of the membrane were counted under a light microscope.

For proliferation assays, ECs were seeded in a 12-well plate for 24 hours, then cultured with or without 4 mmol/L fructose or glucose for another 3 days, followed by cell counting with a Count Star cell counter (12–0005–50, CountStar).

Reverse transfection of RNA duplex (20 nmol/L) was performed with Lipofectamine-RNAiMAX (Life Technologies) or EC-nanoparticles.

Gene expression was analyzed by qPCR or Western blotting. Total RNA was extracted using TRIzol reagent (15596026, Thermo Fisher Scientific) and reversely transcribed using M-MLV reverse transcriptase (M1701, Promega). The mRNA levels of target genes were detected using 2×SYBR qPCR Mix (B21202, Biomake). β-Actin was used as a reference gene. All reactions were run in duplicate. The cycle threshold (Ct) values did not differ by more than 0.5 between the duplicates. The level of the target gene was normalized to that of β-actin, which yielded a 2^−ΔΔC_t value. The sequences of primers are provided in Supplementary Table S1.

For Western blotting, the target proteins were detected with the specific antibody and visualized using an ECL Kit (Merck Millipore) and a TANON-5200 system (Bio-tanon). The antibodies used included: rabbit polyclonal antibodies against AMPK (2532, Cell Signaling Technology), Thr172-phosphorylated AMPK (2535, CST), β-actin (4970, CST); rabbit mAbs against p70S6K (9202, CST), Ser371-phosphorylated p70S6K (9208, CST); mouse mAb against SLC2A5 (sc-271055, Santa Cruz Biotechnology), KHK (sc-377411, Santa Cruz Biotechnology).

After incubation with blocking buffer (PBS containing 1% Tween-20 and 5% BSA) for 30 minutes, the mouse tissue section was incubated with antibody against mouse CD31 (mCD31, AF3628, R&D Systems), Thr172-phosphorylated AMPK or KHK at 4°C overnight, followed by incubation with donkey against goat IgG (H+L) Alexa Fluor 647, rabbit IgG (H+L) Alexa Fluor 555 or mouse IgG (H+L) Alexa Fluor 555 (A-21447, A-31572, and A-31570, Thermo Fisher Scientific) at 37°C for 30 minutes, and then counterstained with 4′,6‐diamidino‐2‐phenylindole (DAPI; D9542, Sigma-Aldrich) for nuclei.

The levels of fructose and glucose in the culture medium of ECs or in the sera and tissues of mice were measured by a colorimetric method, using EnzyChromTM Fructose Assay Kit (EFRU-100, BioAssay Systems) and Glucose (HK) Assay Kit (GAHK20–1KT, Sigma-Aldrich), respectively.

The oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) of ECs were examined in a Seahorse XF24 Analyzer (Seahorse Bioscience) by using a XF Cell Mito Stress Test (103015–100, Seahorse Bioscience) and XF Glycolysis Stress Test (103020–100, Seahorse Bioscience), respectively.

The cellular ATP level was measured by ENLITEN ATP Assay System (FF2000; Promega) according to the manufacturer's instructions.

ECs were seeded in a 12-well plate and treated without or with 4 mmol/L fructose and/or 4 mmol/L glucose for 48 hours. The culture medium was then collected and subjected to lactate detection by a colorimetric method (A019–2-1, Jiancheng). Lactate concentration was normalized to the cell number.

For culture in a hypoxia environment, ECs were placed in an incubator chamber (RS Biotech) with a gas mixture of 1% O₂, 5% CO₂, and 94% N₂.

The transcriptome profiles of TECs and NECs from HCC and adjacent nontumor livers (accession no.: GSE51401), and from colorectal liver metastases and distant healthy liver tissues (GSE77199) were downloaded from GEO datasets. Gene set enrichment analysis (GSEA) was used to analyze the enrichment of the predefined sets of genes (molecular signature database, MSigDB) in TECs and NECs. The q and P values were determined by Kolmogorov–Smirnov statistic with GSEA v4.0. The heat map was produced with Morpheus (https://software.broadinstitute.org/morpheus/).

Data were expressed as the mean ± SEM from at least three independent experiments. Student t test was used to compare the differences between two groups, one-way ANOVA was used for comparisons of more than two groups, and two-way ANOVA was applied when two independent variables on one dependent variable were assessed. Analyses were performed with GraphPad Prism (version 8, GraphPad Software, Inc.). All statistical tests were two-sided and P < 0.05 was considered statistically significant.

The data generated in this study are available within the article and its supplementary data files or are available upon request from the corresponding author. The transcriptome profiles of TECs and NECs in this study were obtained from GEO datasets (Gene Expression Omnibus, https://www.ncbi.nlm.nih.gov/gds/) under accession numbers GSE51401 and GSE77199.

To identify novel metabolic pathways that regulated HCC angiogenesis, we first performed GSEA on the transcriptome profiles of endothelial cells from 16 paired HCC tumor tissues (TEC) and their adjacent nontumor livers (NEC). As shown, genes in some metabolic pathways, like glutathione, fructose, and amino acid metabolism, were significantly enriched in TECs compared with NECs (Fig. 1A; Supplementary Figs. S1A and S1B,). We then focused on fructose metabolism, considering that fructose is highly concentrated in the liver and its role in tumor angiogenesis has not been reported yet. As shown, most of the enzymes in the canonical fructose metabolism pathway (Supplementary Fig. S1C) were dramatically upregulated in TECs (Fig. 1B). The two most upregulated genes, fructose transporter SLC2A5/GLUT5 and fructose-metabolizing enzyme KHK (Fig. 1B), were confirmed to be elevated in freshly isolated TECs from HCC tissues (Fig. 1C; Supplementary Fig. S1D). Consistently, TECs from liver metastatic nodules of colorectal cancer also showed significantly higher activity of fructose metabolism, compared with NECs from distant healthy liver tissues (Supplementary Fig. S1E).

Next, we used an orthotopic liver xenograft model to examine the importance of fructose metabolism in the angiogenesis of liver tumors (Supplementary Fig. S2A). Compared with the Hepa1–6 xenografts of the water-fed group, liver tumors from the fructose-fed mice displayed a significant increase in tumor vessels and tumor growth (Fig. 1D and E, bar 2 vs. bar 1), and also showed an enhanced incidence of hepatic and pulmonary metastasis (water vs. fructose: 1/6 vs. 6/8 in the liver; 4/6 vs. 8/8 in the lung) and a larger number and size of metastatic foci (Fig. 1F and G, bar 2 vs. bar 1). These promoting effects of fructose were attenuated when tumor-bearing mice were administered with 2,5-AM, a specific SLC2A5 inhibitor that blocks fructose ingestion (Fig. 1D–G, bar 3 vs. bar 2). As shown, the body weight, the glucose and fructose levels in serum, and the glucose level in hepatic tumors were similar between water- and fructose-fed mice (Supplementary Figs. S2B–S2D), but the fructose level in hepatic tumors was significantly increased in the fructose-treated group (Supplementary Fig. S2D), suggesting fructose intake increased fructose level in tumor without affecting glucose level.

We further ruled out the possibility that the promotive effects of fructose were simply due to more calories being provided in the fructose group. Fructose and mannose have a similar structure and both are important hexoses that cells use for energy production (29). Therefore, we further used mannose as a monosaccharide control for fructose in the orthotopic liver xenograft model (Supplementary Fig. S3A). Compared with the Hepa1–6 xenografts of the water- or mannose-fed group, liver tumors from the fructose-fed mice displayed a significant increase of tumor vessel areas, whereas the xenografts from the mannose- and water-fed mice had similar vessel areas (Supplementary Fig. S3B), indicating a specific promotive effect of fructose on tumor angiogenesis.

Next, we also employed an autochthonous model of liver cancer developed by hydrodynamic injection with plasmids expressing c-Myc, Cas9, and a single guide RNA targeting p53 (sgp53, Supplementary Fig. S4A; refs. 30, 31) to validate the observed role of fructose. As shown, the fructose-fed group had higher tumor incidence (Fig. 2A), more vessels in tumor tissues (Fig. 2B, bar 2 vs. bar 1) and enhanced pulmonary metastasis (Fig. 2C and D, bar 2 vs. bar 1), compared with the water-fed group. And these promoting effects of fructose were abolished when tumor-bearing mice were treated with 2,5-AM (Fig. 2B–D, bar 3 vs. bar 2). However, the tumor growth was similar among different groups (Fig. 2A, middle and right), which might be due to the strong tumor-promoting effects of c-Myc and sgp53. Again, fructose intake did not significantly affect the body weight, the glucose levels in serum and hepatic tumors, and the serum fructose level (Supplementary Figs. S4B–S4D), but increased the fructose level in hepatic tumors (Supplementary Fig. S4D).

Taken together, the enhanced fructose intake may promote angiogenesis and then facilitates the growth and metastasis of liver cancer.

To verify that fructose promotes angiogenesis by acting directly on ECs, two in vivo models were employed. First, Matrigel was mixed with fructose or PBS and then injected subcutaneously into nude mice. Compared with the PBS-Matrigel, the fructose-Matrigel showed much more vessels (Fig. 3A; Supplementary Fig. S5A), suggesting that in the absence of tumor cells, high levels of fructose in microenvironment can facilitate ECs to form vessels in vivo.

Next, an endothelial-specific loss-of-function assay was conducted. One week after subcutaneous implantation of hepatoma cells, fructose or vehicle control, together with siGFP- or simKHK-EC-nanoparticles, which specifically targeted endothelial cells (23, 24), were injected into the xenografts (Supplementary Fig. S5B). Both in vitro and in vivo analyses revealed that simKHK-EC-nanoparticles did not affect KHK level in tumor cells (Fig. 3B and C), but reduced KHK level in ECs (Fig. 3B and D). Importantly, the role of fructose in promoting tumor angiogenesis was abolished by simKHK-EC-nanoparticles (Fig. 3E). And fructose enhanced tumor growth and this effect was attenuated by EC-specific deletion of KHK (Supplementary Fig. S5C), although the differences did not reach statistical significance, which might be due to short experimental period (Supplementary Fig. S5B). These results indicate a causal contribution of enhanced EC fructose metabolism to liver cancer angiogenesis.

We then investigated whether fructose affected EC phenotypes. Compared with the control medium from no-cell group, fructose in the conditioned medium of ECs was significantly decreased, and this decrease was abolished by inhibiting the transporter of fructose using 2,5-AM (Supplementary Fig. S6A), indicating the ingestion of fructose by ECs. Further in vitro capillary tube formation assays revealed that fructose treatment significantly promoted ECs to form capillary-like structures (Fig. 4A), which phenocopied the proangiogenesis effect of fructose in vivo. Migration and proliferation of ECs are critical events for angiogenesis. As shown, fructose treatment significantly enhanced EC migration (Fig. 4B) and proliferation (Fig. 4C). Although glucose showed a similar effect, but fructose treatment had a stronger role in enhancing EC migration and a slightly weaker effect in EC proliferation (Fig. 4B and C).

To rule out the possibility that the observed fructose's promotive effects were simply due to more monosaccharides being provided under low-glucose (0.6 mmol/L) conditions, we first examined whether other monosaccharides in nature (32), including mannose (Man), sorbose (Sor), galactose (Gal), and allose (All) also affected EC behavior. As shown, the tube formation of ECs was not affected by both mannose and sorbose, and was slightly promoted by galactose and allose (Supplementary Fig. S6B), suggesting that fructose may have a unique impact on EC phenotypes. Furthermore, 2,5-AM or siKHK (Supplementary Fig. S6C) abrogated the promotive roles of fructose on EC migration (Fig. 4D and E). Even under a high glucose (4 mmol/L) condition, fructose could still promote EC migration, and this effect could be abrogated by silencing KHK (Supplementary Fig. S6D). However, silencing KHK did not affect the pro-migration ability of glucose (Supplementary Fig. S6E).

These findings indicate that ECs may metabolize fructose to support their migration and proliferation, consequently resulting in enhanced angiogenesis.

Next, we investigated the downstream signaling that mediated the role of fructose in regulating EC phenotypes. The downstream metabolites of fructose, DHAP and G3P, can enter the glycolytic carbon pool and produce lactic acid (9, 14, 15), or can be metabolized to acetyl-CoA and fuel the mitochondrial TCA cycle (Supplementary Fig. S1C; refs. 11, 12, 17). Therefore, the cellular energetics parameters in ECs were measured. As shown, fructose treatment significantly increased mitochondrial respiration activity (Fig. 5A–D) and total cellular ATP production (Fig. 5E), indicating that fructose may facilitate the TCA cycle in ECs. Furthermore, oligomycin, a selected inhibitor of mitochondria respiration, attenuated the promotive effect of fructose on EC migration (Fig. 5F). On the other hand, fructose treatment in ECs did not alter the levels of ECAR (Fig. 5G and H) and the production of lactic acid, the endpoint product of glycolytic energy metabolism, either under low glucose or high glucose level (Supplementary Figs. S7A and S7B). These results suggest that fructose may promote vessel formation by fueling mitochondrial TCA and ATP production in ECs.

We further exploited how fructose fueled mitochondrial respiration in ECs. The AMP-activated protein kinase (AMPK) and mammalian target of rapamycin complex 1 (mTORC1) pathways play central roles in sensing nutrients and energy, and regulating cell metabolism (33). Activation of AMPK may promote mitochondrial respiration (34, 35), whereas mTORC1 signaling has controversial roles in mitochondrial function (36, 37). To date, whether fructose affected AMPK or mTORC1 signaling remains obscure. We found that the levels of phosphorylated p70S6K, a marker of mTORC1 pathway activation (33), did not change in ECs after fructose exposure (Fig. 6A). However, the Thr172-phosphorylated AMPK, representing AMPK activation (38), was increased in the fructose-treated ECs (Fig. 6A; Supplementary Figs. S8A and S8B). In contrast, glucose treatment reduced the level of Thr172-phosphorylated AMPK (Fig. 6A; Supplementary Figs. S8A and S8B), which was consistent with previous observations in other cell types (38, 39). We further used other monosaccharide molecules as osmolality control, and found that AMPK phosphorylation in ECs was only increased by fructose, but not by galactose, sorbose, mannose, and allose (Supplementary Fig. S8C), which again confirmed the specific influence of fructose on ECs. Moreover, blockade of fructose metabolism by silencing KHK diminished the fructose-induced Thr172-phosphorylation of AMPK in ECs (Fig. 6B). In agreement with the in vitro observations, tumor vessels in the xenografts from fructose-fed mice also displayed a much higher level of Thr172-phosphorylated AMPK than the control ones (Fig. 6C; Supplementary Fig. S8D), indicating that the AMPK signaling in ECs may be activated by fructose metabolism.

We next used Compound C/Dorsomorphin, a specific inhibitor of AMPK activation, to examine whether AMPK mediated the effect of fructose. As shown, Compound C blocked the fructose-induced activation of AMPK (Fig. 6D), and abolished the fructose-induced elevation of the basal and maximal oxygen consumption and ATP production in ECs (Fig. 6E and F). Consistently, the function of fructose in promoting EC migration was also abrogated by Compound C (Fig. 6G).

These findings suggest that fructose may promote angiogenesis by activating AMPK to fuel mitochondrial respiration in endothelial cells.

Finally, we investigated how the fructose-metabolizing enzyme was upregulated in TECs. Bioinformatic analysis revealed that several pathways were positively correlated with the mRNA levels of SLC2A5 and KHK (Supplementary Fig. S9), based on transcriptome profiles of TECs and NECs. We further focused on the hypoxia pathway, as hypoxia is a hallmark of tumor microenvironment and whether hypoxia can regulate fructose metabolism is still unknown. As shown, hypoxia exposure significantly elevated both mRNA and protein levels of SLC2A5 and KHK (Fig. 7A and B). Multiple genes in the fructose metabolism pathway were also upregulated by hypoxia (Fig. 7A; Supplementary Fig. S10). Notably, the roles of hypoxia in enhancing SLC2A5 and KHK were retracted by silencing HIF1α (Fig. 7C and D), the master transcription factor that mediates the response to hypoxia (40, 41). Functionally, hypoxia treatment significantly promoted ECs to uptake fructose (Supplementary Fig. S11) and intensified the role of fructose in enhancing the mitochondrial respiration activity (Fig. 7E and F) and migration (Fig. 7G) of ECs. These results suggest that the hypoxia-HIF1α signaling may promote fructose metabolism and then fructose-induced EC migration.

Collectively, our data suggest that hypoxia tumor microenvironment stimulates fructose metabolism in TECs, which enhances AMPK activity and thus mitochondrial respiration to support the migration and proliferation of ECs, consequently augmenting tumor angiogenesis.

The novelty of this study relates to the following findings: (1) Fructose can be metabolized in ECs to promote tumor angiogenesis. (2) Fructose activates the AMPK signaling in ECs. (3) The hypoxia-HIF1α signaling stimulates the fructose metabolism pathway in TECs.

It has been shown that fructose could be metabolized by hepatocytes and stimulated hepatocyte lipogenesis to induce hepatic steatosis (11), and fructose stimulated DNL in hepatocytes and induced NAFLD, steatohepatitis, and HCC (12). In addition, the expression of ALDOB was increased in the liver-metastatic colorectal cancer cells, which enhanced the utilization of liver-enriched fructose and facilitated the growth of metastatic cells (15). Nonetheless, whether fructose can be metabolized by tumor stromal cells is still unknown. We showed that fructose could be ingested by ECs, and promoted vessel formation in vitro and in vivo. In particular, the in vivo studies disclosed that in the absence of tumor cells, fructose could facilitate ECs to form vessels in the subcutaneously implanted Matrigel; and silencing KHK expression in ECs could attenuate the fructose-enhanced tumor angiogenesis, indicating a direct effect of fructose on endothelial cells and a causal relationship between fructose metabolism and tumor angiogenesis. Mechanistically, fructose promoted EC proliferation and migration by enhancing mitochondrial respiration and ATP production. Significantly, inhibitor of fructose transporter SLC2A5 or silencing of fructose-metabolizing enzyme KHK attenuated the promotive effect of fructose on angiogenesis, suggesting that limiting the ingestion of fructose or inhibiting fructose metabolism may represent a promising strategy to suppress HCC angiogenesis and tumor progression.

In this study, we found that: (i) TECs of liver cancer expressed higher levels of fructose transporter and metabolizing enzymes, compared with liver NECs; (ii) ECs in subcutaneous Matrigel plug or tumors in mice or ECs under in vitro culture could ingest and metabolize exogenously added fructose; (iii) the hypoxia stress enhanced the levels of fructose transporter and metabolizing enzymes in ECs. These observations suggest that ECs from the liver and other organs may utilize fructose. However, under in vivo physiological conditions, dietary fructose is absorbed in the intestine, then concentrated and metabolized in the liver, which express high levels of fructose transporter and metabolizing enzymes (9, 15). Therefore, the liver may provide a unique fructose-enriched microenvironment that facilitates fructose utility in liver TECs, thereby promoting angiogenesis of liver cancer. In addition, the hypoxia tumor microenvironment further augments the ability of liver TECs to metabolize fructose by enhancing the expression of fructose transporter and metabolizing enzymes. In contrast, NECs and TECs in other organs, which have very limited fructose in the tissue microenvironment, may not display high fructose metabolism. In this respect, the enhanced fructose metabolism in liver TECs may be an organotypic observation. In line with the hypothesis, we found that inhibitor of fructose transporter SLC2A5 reduced fructose levels in liver tumor tissues and attenuated fructose metabolism and angiogenesis of TECs.

It is well-recognized that AMPK is activated via phosphorylation by a lack of energy or nutrients (33, 42). Then, AMPK phosphorylates and activates the regulators in catabolic pathways to enhance ATP production, whereas phosphorylates and inactivates the regulators in anabolic pathways to reduce ATP consumption (33). Glucose deprivation can induce phosphorylation of AMPK, whereas an increase of glucose level inactivated the AMPK signaling (33, 38, 39), which was also observed in our study model. Previous studies have shown that the canonical AMPK activation is dependent on the decreased ATP-to-AMP ratio resulting from reduced glucose metabolism (43, 44). In addition, it has been disclosed that a noncanonical AMP-independent mechanism activates AMPK by sensing the absence of fructose-1,6-bisphosphate, a downstream metabolite of glucose glycolysis (38). Whether fructose, another monosaccharide with the same caloric value and molecular formula as glucose, can regulate AMPK signaling is still obscure. Only one study has shown that fructose treatment decreased the level of phosphorylated AMPK in lung cancer cells via an unknown mechanism (45). Intriguingly, we found that exposure to fructose could promote but not suppress AMPK phosphorylation in ECs, and this effect of fructose was diminished by blocking fructose metabolism with siKHK. The first step in fructose metabolism is the rapid and irreversible phosphorylation of fructose to F1P catalyzed by KHK (9). This process is specific to the fructose metabolism pathway and is not shared with glycolysis or gluconeogenesis. Therefore, fructose metabolism may activate the AMPK signaling in a manner different from glucose deprivation in ECs, which might also be a cell-context-dependent mechanism.

Previous studies have shown that enhanced fructose metabolism by high intake of fructose provides fuel for glycolysis or the TCA cycle of tumor cells (9, 15), but the mechanism remains obscure. Here, we revealed that fructose metabolism was more effective in inducing the mitochondrial TCA cycle than glycolysis in ECs. Fructose augmented mitochondrial respiration, and specific inhibition of AMPK signaling abolished the roles of fructose in enhancing mitochondrial respiration and migration of ECs. Moreover, blockade of mitochondrial function in ECs attenuated the proangiogenesis effect of fructose. It has been reported that overexpression of a constitutively active AMPK induces biogenesis and fragmentation of mitochondria (34, 35). Our data suggest that fructose may activate AMPK to facilitate the usage of downstream fructose metabolite in the mitochondria of ECs, which enriches our understanding on the relationship between AMPK signaling and mitochondrial function.

Hypoxia arises in tumors due to the rapid proliferation of cancer cells and lack of enough vessels. Under hypoxia, HIF1α is stabilized and induces an increased expression of glycolytic enzymes and suppresses the expression or activity of electron transport chain subunits in mitochondria (46, 47). Consequently, tumor cells usually increase the uptake and utilization of glucose via glycolysis, while reducing their metabolism in mitochondria. Although one study showed that the mRNA and protein levels of SCL2A5 (Glut5) were upregulated by hypoxia in breast cancer cell MCF7, but not MDA231 cells (48), the role of hypoxia on the fructose metabolism pathway has not been identified. Here, we revealed that hypoxia-induced HIF1α increased the expression of the fructose transporter and multiple enzymes in the fructose metabolism pathway, which might enhance fructose utility in TECs. Furthermore, the hypoxia-enhanced fructose metabolism in ECs preferred to fuel mitochondrial respiration but not glycolysis, highlighting the complexity of hypoxia in regulating mitochondrial function.

Collectively, we identify the stimulatory role of fructose in tumor angiogenesis, uncover its underlying mechanism, and imply that restricting fructose intake or targeting fructose metabolism may be a potential strategy to suppress tumor angiogenesis and tumor progression.

No author disclosures were reported.

J.-H. Fang: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. J.-Y. Chen: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft. J.-L. Zheng: Data curation, formal analysis, validation, investigation, visualization, methodology. H.-X. Zeng: Data curation, formal analysis, validation, investigation. J.-G. Chen: Data curation, formal analysis, investigation. C.-H. Wu: Data curation, investigation. J.-L. Cai: Data curation, investigation. Z.-Y. Wang: Investigation, methodology. S.-M. Zhuang: Conceptualization, data curation, formal analysis, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

This study was supported by grants from the National Key R&D Program of China (2019YFA0906001 to S-M. Zhuang), The National Natural Science Foundation of China (82230093 and 81930076 to S-M. Zhuang; 81972272 and 82273313 to J-H. Fang), The Guangdong Basic and Applied Basic Research Foundation (2019A1515011586 to J-H. Fang), The Fundamental Research Funds for the Central Universities, Sun Yat-sen University (22lgqb27 to J-H. Fang). The authors thank Prof. Yun-Fei Yuan at Sun Yat-sen University Cancer Center for providing human HCC tissues and clinical data.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.