Megakaryocytes Mediate Hyperglycemia-Induced Tumor Metastasis

The tumor metastasis cascade can be divided into steps of tumor invasion, intravasation, survival through circulation, extravasation, and regrowth in distal organs (1). During this life-threatening cascade, platelets help circulating tumor cells (CTC) to overcome the damage by blood shear forces, to avoid the attack by innate immune cells, and to facilitate tumor cell extravasation (2). Genetic and epigenetic changes in the tumor cells often lead to elevated blood coagulation to shield CTCs from anoikis-mediated cell death (3). The CTC-platelet interplay is driven by various signals such as tissue factor (4), integrin (5), and selectins (6). As a result, CTCs promote platelet aggregation, which, in turn, enhances CTC survival and movement via a range of platelet-released factors such as TGFβ (7). Indeed, the correlation of thrombocytosis and negative prognosis of cancer is reported in various cancer types, including melanoma (8). Antithrombotic drugs such as aspirin and warfarin can improve survival in patients with cancer (9). Platelets are nonnucleated cell fragments derived from the fractionation of megakaryocytes (MK). The relation between MKs and tumor cells has not been fully elucidated, partially due to technical challenges and limited clinical studies in MK research. It is recognized that MKs may (i) produce a variety of bone formation factors and directly inhibit tumor cell growth in the bone marrow metastatic niche (10); (ii) activate the bone marrow vascular niche for bone metastasis (11); (iii) mediate prometastatic phenotypes of platelets for distant CTCs. For the non-bone metastasis, the latter is more likely to be the dominating mechanism, suggesting an MK–platelet–CTC axis in the body. Indeed, Nf-E2^−/− mice, in which MKs cannot produce platelets, exhibits significantly reduced metastasis burden (12). However, the molecular details of this MK-platelet-CTC axis are still unknown.

Hyperglycemia may be induced by various conditions such as diabetes mellitus (13), obesity (14), pancreatitis (15), certain types of cancer (16, 17), and inappropriate diets (18). Although diabetes induces hyperglycemia, the prevalence of hyperglycemia is higher than, and independent of, the diagnosis of diabetes (19). Hyperglycemia is strongly associated with poor clinical outcomes in patients with cancer, and increases tumor progression, metastasis, and drug resistance in preclinical models (20, 21). Tumor cells adopt a metabolism based on inefficient aerobic glycolysis known as the Warburg effect because glycolysis can rapidly produce ATP and release metabolites that may support biogenesis (22). Overexpression of glucose transporters such as GLUT1 is commonly observed in tumor cells to meet their energy demands. Various evidence has demonstrated that glucose directly affects cancer cells (21). However, the nontumor cell targets of glucose in the metastasis cascade have been largely overlooked. Recent studies show that glucose–GLUT1 signaling might induce systemic alterations that modulate the function of various tissues and organs (23). Considering the intense crosstalk between tumor cells and various host cell types in the metastatic cascade, targeting glucose signaling in host cells, but not in tumor cells, might also have a beneficial antimetastatic effect. Indeed, clinical reports show that systemic control of patient glucose levels is beneficial for patients with certain types of cancer (24). The role of glucose signaling in host cells in the metastasis cascade needs further clarification.

Glucose-regulated protein 75 (GRP75), a mitochondrial molecular chaperone, belongs to the heat shock protein 70 family (25). Unlike other members of this family, it does not respond to heat but to stress such as glucose modulation, oxidative injury, and radiation. Studies demonstrated that GRP75 is located not only in mitochondria but also in the endoplasmic reticulum (ER), cytosol, and cytosolic vesicles (26). Besides its role as a chaperone within mitochondria (27), GRP75 is recently identified as a gatekeeper for mitochondria-ER docking and controls IP₃R for Ca²⁺ release (28). Overexpression of GRP75 reduces the ER Ca²⁺ release and changes the mitochondrial parameters. The newly discovered correlation between GRP75 and Ca²⁺ homeostasis suggests an essential role of GRP75 in cell fate determination and in mediating pathophysiologic changes within the cell. To the best of our knowledge, there is no literature on the role of GRP75 in platelets or MKs.

In this study, we explore a host cell target for glucose, and its role in facilitating tumor metastasis. In preclinical models, we show that glucose, but not fructose, drives MK GRP75 expression and facilitates tumor metastasis via Ca²⁺-PKCα-mediated platelet activation. Genetic and pharmacologic interventions identified GLUT1–MYC–GRP75 signaling in MKs that triggers the prometastatic phenotype of platelets. In human volunteers, drinking a 5% glucose solution significantly activates human platelets and these platelets promote tumor metastasis when infused into mice. These findings uncover a novel concept of host cell metabolism-assisted tumor metastasis, and identify GRP75 as a potential target for antimetastasis therapy.

Murine B16-F10 melanoma and T241 fibrosarcoma cell lines were kindly provided by Dr. Yihai Cao from the Karolinska Institute. B16-F10, T241, and primary megakaryocytes were cultured in 10% FBS-DMEM (catalog no. 40130ES76, YEASEN; catalog no. TBD10569, TBD Science) containing 100 U/mL penicillin, 100 μg/mL streptomycin (catalog no. MA0110, Meilunbio). Isolated platelets were cultured in PBS with 0.25 μmol/L prostaglandin E1 (catalog no. MA1009, Meilunbio). All cell lines used in our study were regularly tested for Mycoplasma (catalog no. LT07–318; Lonsa).

Primary MKs were isolated from mouse bone marrow using a Mouse Megakaryocyte Isolation Kit (catalog no. MAG2014M, TBD Science). Primary mouse platelets were isolated from peripheral blood using a Mouse Platelet Isolation Kit (catalog no. PLA2011M, TBD Science). Primary human platelets were isolated using an established protocol. In brief, acid-citrate-dextrose (ACD) solution was made by dissolving 85 mmol/L sodium citrate (catalog no. MB2492, Meilunbio), 71.38 mmol/L citric acid (catalog no. MB8820, Meilunbio), and 27.78 mmol/L glucose (catalog no. MB2510, Meilunbio) in PBS. Peripheral blood was collected into tubes containing ACD with a volume ratio of 6:1. Platelet-rich plasma was made by centrifugation at the 200 × g for 10 minutes at 20°C. Further centrifugation at 800 × g for 2 minutes at 20°C was performed for collecting platelets. Platelets were resuspended and immediately prepared for further investigations.

All animal studies were approved by the Animal Experimental Ethical Committee of Fudan University. Male C57BL/6 mice (RRID:IMSR_JAX:000664) at 6- to 8-week-old ages were obtained from the Model Animal Research Center of Nanjing University. Male CAnN.Cg-Foxn1^nu/Crl nude mice (RRID:IMSR_CRL:194) were purchased from SLAC laboratory animal, Shanghai, China. All animals were randomly assigned to groups before experiments. The experimenter was not blind to the assignment of the groups and the evaluation of the results. No samples, animals, or data were excluded.

All human studies were approved by the Ethical Review Committee in Jinling Hospital, Medical School of Nanjing University. 5% glucose solution (catalog no. H19994045, Chimin Health Management) was given to healthy volunteers with written informed consent and peripheral blood samples were collected for blood cell count and blood glucose detection. After collection, platelets were immediately isolated for further in vitro and in vivo studies.

For CTC–platelet interaction studies, a coinjection protocol was applied. Approximately, 1 × 10⁵ mouse or human tumor cells with or without 1 × 10⁹ platelets in 100 μL PBS were intravenously injected into each C57Bl/6 mouse. About 2 weeks after tumor injection, mice were sacrificed. To mimic the full metastatic cascade, two xenograft tumor models were used. In an intrasplenic injection model, a left subcostal surgical incision was created and tumor cells at the density of 1 × 10⁶ in 25 μL PBS were injected into the spleen of each mouse, followed by the closure of incisions with the sterile surgical suture (catalog no. CR436, Jinhuan Medical). About 10 days after tumor injection, mice were sacrificed. In a subcutaneous model, tumor cells at the density of 1 × 10⁶ in 50 μL PBS were subcutaneously injected into each mouse. About 20 days after tumor injection, mice were sacrificed. Tumor sizes were measured with calipers and tumor volumes were calculated according to a standard formula (29, 30). Once the mice were sacrificed, the livers and spleens were excised for the detection of visible tumor nodules on the surface of these organs and subsequently validated by histologic examination. Primary and metastatic tumor masses were monitored ex vivo under the green fluorescent channel with an VISQUE InVivo Elite system (Vieworks). At the end of each experiment, fresh tumor tissues were immediately collected for RNA extractions, protein extractions, and histologic investigation.

Isolated or cultured cells were lyzed in RIPA buffer and a proteinase and phosphatase inhibitor cocktail (catalog no. MA0151, Meilunbio; catalog no. MB2678, Meilunbio; 1:100). For immunoblotting, an equal amount of protein samples from each group and a standard molecular weight marker (catalog no. WJ102, EpiZyme) were loaded on a 10% SDS-PAGE gel (catalog no. PG112, EpiZyme), followed by transferring onto a polyvinylidene difluoride (PVDF) membrane (catalog no. IPVH00010, Millipore), which was subsequently blocked with 5% skimmed milk for 2 hours. Membranes were incubated overnight at 4°C with a rabbit anti-GRP75 antibody (catalog no. 3593S, Cell Signaling Technology; RRID:AB_2120328; 1:1,000), a rabbit anti-phospho-p38 antibody (catalog no. 4631S, Cell Signaling Technology; RRID:AB_331765; 1:1,000), a rabbit anti-p38 antibody (catalog no. 9212S, Cell Signaling Technology; RRID:AB_330713; 1:1,000), a rabbit anti-phospho-PKCα antibody (catalog no. AP0559; RRID:AB_2771421; 1:1,000), a rabbit anti-phospho-Erk1/2 antibody (catalog no. 9101S, Cell Signaling Technology; RRID:AB_331646; 1:1,000), a rabbit anti-Erk1/2 antibody (catalog no. 4695S, Cell Signaling Technology; RRID:AB_390779; 1:1,000), a rabbit anti-MYC antibody (catalog no. A1309, Abclonal; RRID:AB_2759938; 1:1,000), and a mouse anti-β-actin antibody (catalog no. A5316, Sigma-Aldrich; RRID:AB_476743; 1:1,000) in 5% skimmed milk. After rigorous washing with PBS containing 0.1% Tween-20 (catalog no. T8220, Solarbio), membranes were incubated at room temperature for 1 hour with a goat anti-mouse HRP-conjugated IgG antibody (catalog no. AS003, ABclonal; RRID:AB_2769851; 1:5,000) and a goat anti-rabbit HRP-conjugated IgG antibody (catalog no. AS014, ABclonal; RRID:AB_2769854; 1:5,000). Target proteins were visualized using a super sensitive ECL luminescence reagent (catalog no. MA0186, Meilunbio) with a Molecular Imager ChemiDoc XRS System (Bio-Rad).

Chromatin immunoprecipitation (ChIP) assay was performed using a ChIP Assay Kit (catalog no. p2078, Beyotime). DNA–bound proteins were fixed using 4% paraformaldehyde (PFA). Chromatin was purified and sonicated to generate fragments of approximate sizes between 500 and 1,000 bp. Twenty microliters of the sonicated chromatin was collected for input. One hundred and eighty microliters of the sonicated chromatin was immunoprecipitated by a rabbit anti-Myc antibody (catalog no. A1309, ABclonal; RRID:AB_2759938; 1:200), and a rabbit non-immune IgG antibody (catalog no. AC005, ABclonal; RRID:AB_2771930; 1:200). Five M NaCl was added to the protein–DNA complexes and incubated at 65°C for 6 h. The purified DNA fraction was used for qPCR analysis. The mouse Grp75 promoter forward primer was 5′- TGCAATGTCTGTCCTGTTTAACTT-3′, the mouse Grp75 promoter reverse primer was 5′- AGAACGTGTGGGGAAATTAGGA-3′. The mouse Grp75 exon 2 forward primer was 5′- AGACAGGGGTTGATTTGACCA-3′, the mouse Grp75 exon 2 reverse primer was 5′- ACAGTCTGCTGAACCTTGGG-3′. Data were normalized with the nonimmune rabbit IgG values and was presented as mean determinants of percentages of input.

Total RNA was extracted from tumor tissues and cultured cells using RNAsimple Total RNA Kit (catalog no. DP419, TIANGEN). Total RNA from each sample was reversely transcribed using an All-in-One cDNA Synthesis SuperMix (catalog no. B24408, Bimake). Reverse transcription was performed at 42°C for 60 minutes, followed by 70°C for 5 minutes to inactivate the enzyme activity. Samples were stored at −20°C and subjected to qPCR using a StepOnePlus Real-Time PCR System (Applied Biosystems). Each qPCR measurement was performed in a triplicate using 10 μL reaction containing 2× SYBR Green qPCR Master Mix (catalog no. B21202, Bimake), 50 nmol/L forward and reverse primers, and 2 μL cDNA. The qPCR protocol was executed for 40 cycles and each cycle consisted of denaturation at 95°C for 15 seconds, annealing at 60°C for 1 minute, and extension at 72°C for 1 minute. The primer pairs specific for various genes used in our experiments included: mouse Grp75 forward: 5′-ATGGCTGGAATGGCCTTAGC-3′; mouse Grp75 reverse: 5′-ACCCAAATCAATACCAACCACTG-3′; mouse Actb forward: 5′-GGCTGTATTCCCCTCCATCG-3′; mouse Actb reverse: 5′-CCAGTTGGTAACAATGCCATGT-3′; mouse Slc2a1 forward: 5′-CAGTTCGGCTATAACACTGGTG-3′; mouse Slc2a1 reverse: 5′-GCCCCCGACAGAGAAGATG-3′; mouse Myc forward: 5′-ATGCCCCTCAACGTGAACTTC-3′; mouse Myc reverse: 5′-CGCAACATAGGATGGAGAGCA-3′. mouse Pik3ca forward: 5′-CCACGACCATCTTCGGGTG-3′; mouse Pik3ca reverse: 5′-ACGGAGGCATTCTAAAGTCACTA-3′. mouse Akt forward: 5′-ATGAACGACGTAGCCATTGTG-3′; mouse Akt reverse: 5′-TTGTAGCCAATAAAGGTGCCAT-3′. For a qPCR array, primer pairs for the detection of 37 glucose metabolism-related genes and 13 transcription factor genes are listed in the Supplementary Information (see Supplementary Tables S1 and S2).

Whole-mount staining was performed according to our standard method (31). After decalcification for 3 days, PFA-fixed femur tissue samples were projected for hard-tissue section (catalog no. CryoJane, Instrumedics). Sections with 50 μm thickness were blocked in PBS with 10% horse serum for 1 hour, followed by staining overnight at 4°C with a goat anti-mouse CD105 antibody (catalog no. AF1320, R&D systems; RRID:AB_354735; 1:200). After rigorous rinsing in PBS, blood vessels and MKs were detected with an anti-goat Alexa Fluor 647–labeled secondary antibody (catalog no. A32849; Invitrogen; RRID:AB_2762840; 1:400), a rat anti-mouse CD41 antibody conjugated with FITC (catalog no. 11–0411–85, Invitrogen; RRID:AB_763483; 1:400), and DAPI (catalog no. D9542, Sigma-Aldrich; 1:400) mounted in Vectashield mounting medium (Vector Laboratories, Inc.), and stored at −20°C in darkness before examination under a Nikon C1 confocal microscope. Captured images were further analyzed using an Adobe Photoshop CS software program.

Paraffin-embedded tissues were sectioned at a thickness of 5 μm, mounted onto glass slides, baked for 1 hour at 60°C, deparaffinized in Xylene (catalog no. 10023418, SCR), and sequentially rehydrated in 99%, 95%, and 70% ethanol (catalog no. 10009218, SCR). Tissue slides were counterstained with hematoxylin (Mayer's; catalog no. MB9897, Meilunbio) and Eosin (catalog no. MA0164, Meilunbio) before dehydration with 95% and 99% ethanol, and were mounted with neutral balsam (catalog no. 1004160, SCR). Stained tissues were analyzed under a light microscope (Leica DM IL LED). For immunofluorescence staining, paraffin-embedded tissue sections or cells on glass coverslips were stained with a rabbit anti-CD42b antibody (catalog no. 12860-1-AP, Proteintech; RRID:AB_10644481; 1:50); a rabbit anti-GRP75 antibody (catalog no. 3593S, Cell Signaling Technology; RRID:AB_2120328; 1:100), and a rabbit anti-SLC2A1 antibody (catalog no. A11170, ABclonal; RRID:AB_2758445; 1:100). After rinsing, tissue samples were further stained for 30 minutes with the secondary antibodies, which included a donkey anti-rabbit Alexa Fluor 488 antibody (catalog no. A21206, Invitrogen; RRID:AB_2535792; 1:400); and a donkey anti-rabbit Alexa Fluor 647 antibody (catalog no. A31573, Invitrogen; RRID:AB_2536183; 1:400). Nuclei were counterstained with DAPI (catalog no. MA0128, Meilunbio) and mounted with antifade mounting medium (catalog no. MA0221, Meilunbio). Positive signals were captured using a fluorescence microscope (Olympus BX51). Captured images were further analyzed using the Adobe Photoshop CS software (RRID:SCR_014199). For the platelets adhesion assay, the platelet cell membrane was prelabeled by a PKH26 Cell Linker Kit (catalog no. MX4021, Maokangbio) for visualization.

For inhibiting platelet aggregation, aspirin (catalog no. J20171021, Bayer) at 30 mg/kg or warfarin (catalog no. H31022123, SINE) at 30 mg/kg was given orally once per day to each mouse for 10 days. Distilled water was used as a control. For metastasis inhibition, a selective GRP75 inhibitor MKT-077 (catalog no. A12388, AdooQ Bioscience) at 15 mg/kg was intraperitoneally injected once per day into each mouse. DMSO (catalog no. MB2505, Meilunbio) was used as a control. For in vitro platelet activation, freshly isolated mouse or human platelets were treated with ADP (catalog no. MB1706, Meilunbio), collagen (catalog no. MB5213, Meilunbio), thrombin (catalog no. MB1368, Meilunbio), ionomycin (catalog no. MB7511, Meilunbio), and arachidonic acid (AA, catalog no. MB6012, Meilunbio). In some experiments, a GRP75 inhibitor MKT-077 (catalog no. A12388, AdooQ Bioscience) at 20 μmol/L was pretreated for 30 minutes prior to agonist stimulation.

Peripheral blood was collected and transferred to an anti-coagulation tube containing ACD buffer (150 μL/mL blood). Isotonic HEPES Tyrode buffer was used to dilute the blood. Platelets challenged with or without ADP (10 μmol/L) for 3 minutes were incubated with antibodies including: an FITC-conjugated rat anti-mouse CD62p antibody (catalog no. 561923, BD Biosciences; RRID:AB_10896149), and an FITC-conjugated mouse anti-human CD62p antibody (catalog no. 11–0628–42. Invitrogen; RRID:AB_10668715) at room temperature for 30 minutes without stirring. Cell suspension was washed with ice-cold PBS and was then transferred to a FACS system (FACSCanto II, BD Biosciences). FlowJo software (Version 10, BD Biosciences; RRID:SCR_008520) was used to analyze the FACS result.

B16-F10 melanoma cells were grown in 24-well plates to about 80% confluency. Freshly isolated platelets were labelled using the PKH26 Cell Linker Kit (catalog no. MX4021, Maokangbio) according to the manufacturer's instructions for 10 minutes. Prelabeled platelets at the density of 1 × 10⁵ per well were seeded in each well. After 30-minute incubation, unattached platelets were rinsed out. Adhesive platelets were then analyzed under a fluorescent microscope (Leica DM IL LED).

Calcium flux was visualized by a BBcellProbe Fluo-3/AM Kit (catalog no. BB-481122, BestBio) according to the manufacturer's protocol. In brief, isolated platelets were washed in Hank's Balanced Salt Solution (HBSS; catalog no. MA0039, Meilunbio). Platelets were resuspended in Fluo-3 solution for 30 minutes at 37°C. After washing with HBSS, the platelets were resuspended in HBSS and incubated for 30 minutes. The cell solution was then detected by FACS (FACSCanto II, BD Biosciences) or fluorescent microscopy (Leica DM IL LED).

The data were found to pass the D'Agostino–Pearson normality test. Differences between two groups were performed using the standard two-tailed Student t test, and P < 0.05 was considered statistically significant. Differences among multiple groups were evaluated using a one-way ANOVA test. The data are presented as means ± SEM.

The data generated in this study are available upon request from the corresponding author.

To investigate the impact of glucose on tumor metastasis, a murine metastatic melanoma cell line B16-F10 was stably transfected with enhanced GFP for visualizing the metastasis foci. Glucose, at the most commonly used clinical dose of 5% w/v, was given ad libitum through the drinking water to the mice. As expected, in a subcutaneous tumor implantation model, glucose increased primary tumor volume, without altering the food intake (Supplementary Figs. S1A–S1C). Similarly, in a spleen tumor implantation model, glucose administration promotes primary tumor growth (Supplementary Fig. S1D). Interestingly, liver metastasis increased three folds in the glucose-treated group (Supplementary Fig. S1E), indicating glucose promotes tumor metastasis. To exclude the interference of primary tumors on the metastatic burden, we omitted the tumor invasion and intravasation steps by directly injecting the tumor cells into the tail vein. Ex vivo visualization and histologic examination revealed a significant increase in the number and size of pulmonary metastatic foci, and a reduction of survival in glucose-treated mice (Fig. 1A and B; Supplementary Fig. S1F), suggesting glucose promotes CTCs for metastasis.

Glucose transporters, such as GLUT1, mediate glucose uptake and regulate the energy metabolism in cancer cells. In glucose-treated tumor-bearing mice, GLUT1 expression in the primary tumor dramatically increased (Supplementary Fig. S1G). Stable knockdown of Glut1 using shRNA effectively inhibited in vitro tumor cell proliferation and in vivo primary tumor growth (Supplementary Figs. S1H–S1J). Surprisingly, detailed gross and histologic detection showed that loss-of-Glut1 experiments did not alter the number of visible and microscopic pulmonary metastatic nodules, although the average size of metastasis foci decreased (Fig. 1C–E). These surprising findings suggest that glucose-promoted metastasis foci number is tumor cell GLUT1-independent.

In the circulation, the CTC microenvironment is unique for (i) lack of extracellular matrix, (ii) presence of hemodynamic shear forces, and (iii) continuous surveillance of immune cells. Platelets are major players in the CTC microenvironment, protecting CTCs from the above-mentioned adversities and promoting metastasis (2). We hypothesized that the glucose-promoted metastasis foci number is platelet-associated. To investigate the role of platelets in facilitating metastasis, we collected the platelets from glucose- or vehicle-treated mice on day 7, and mixed them with two types of tumor cells before injection into the tail vein of healthy recipient mice. In both models, platelets isolated from glucose-treated donor mice markedly increased the number, but not the size of pulmonary metastasis (Fig. 1F–H; Supplementary Figs. S1K and S2A–S2C), indicating platelets from glucose-treated mice is metastasis-facilitating.

Next, we investigated the physiopathologic functions of platelets from glucose-treated mice. Interestingly, a glucose dose gradient experiment showed that 5% glucose solution in mice significantly potentiated platelet activation (Fig. 1I). Administration of 5% glucose or 10% sucrose, but not fructose, in mice, promoted CD62p expression in ADP-stimulated isolated platelets, without altering the mouse insulin sensitivity (Fig. 1J; Supplementary Fig. S2D). These results show that oral administration of 5% glucose primes circulating platelets in vivo.

Taking streptozotocin-induced diabetic mice as a positive control, 5% glucose treatment for 7 days induced nonfasting blood glucose level to a similar extent (Fig. 1K) without changing fasting glucose level and insulin sensitivity (Supplementary Figs. S2E and S2F), suggesting short-term 5% glucose administration induces hyperglycemia. Next, we investigated the platelet phenotype in these groups. Both hyperglycemic and diabetic mice exhibited high levels of ADP-induced platelet activation, without altering resting CD62p expression levels (Fig. 1L). As a result, mice with these primed platelets exhibited a shortened tail bleeding time (Fig. 1M). Of note, this diet-induced hyperglycemia model did not alter the levels of water intake, urine glucose, and urine protein in mice, compared with healthy controls (Supplementary Figs. S2G–S2I). These findings demonstrate that hyperglycemia primes platelets in mice.

To further explore the functional effect of hyperglycemia on circulating platelets in vivo, we administered ADP, AA, thrombin, and collagen on freshly isolated circulating platelets. Surprisingly, Hyperglycemia increased platelet sensitivity to all of these activators (Fig. 2A). Platelets from hyperglycemic mice aggregated more rapidly as tested by aggregometry, and led to a more significant clot retraction (Fig. 2B and C). Mean platelet volume (MPV), platelet distribution width (PDW), were not disrupted in hyperglycemic mice, while platelet count (PLT) and plateletcrit (PCT) showed a 60% reduction as compared with controls (Fig. 2D), indicating that hyperglycemia reduces platelet number without affecting the platelet morphology. Notably, the parameters of red blood cells, which originate from megakaryocyte/erythroid progenitors same as platelets, are not altered in hyperglycemic mice (Supplementary Fig. S3A). The physiologic role of hyperglycemia in arterial thrombosis was investigated in a ferric chloride (FeCl₃)-induced injury model. In the mesenteric artery, the rate and extent of thrombus formation are significantly increased in the hyperglycemic mice (Fig. 2E and F). These in vivo results demonstrate platelet hyperactivity in hyperglycemic mice.

Platelet hyperactivity requires a faster clearance in the body and may explain the decreased number of circulating platelet in hyperglycemic mice (32). Indeed, the spleen of hyperglycemic mice exhibited more CD42b⁺ signals than controls (Fig. 2G), indicating increased clearance of activated platelets. The spleen weight was not changed (Supplementary Fig. S3B). Splenectomy in the hyperglycemia group restored the platelet count to those seen in the splenectomized control group (Fig. 2H), suggesting PLT reduction in hyperglycemic mice is triggered by splenic platelet clearance.

Although GLUT1 is expressed on platelets (33), we surprisingly found that the isolated healthy platelet did not respond to glucose administration, whereas platelets from hyperglycemic mice were strongly hyperactive (Fig. 3A). These results suggest that glucose indirectly facilitates platelet priming. The platelet response to glucose could be predetermined by MKs. In hyperglycemic mice, size, number, and distance to the adjacent vessel of MKs were not altered (Fig. 3B). However, almost all glucose metabolism-related genes were upregulated, indicating a strong stimulating effect of glucose on MKs (Fig. 3C and D). Surprisingly, a mitochondrial chaperone, Grp75, ranked among the top 10 genes (Fig. 3C and D). This unexpected mitochondrial protein aroused our interest. GRP75 is recently identified as a gatekeeper for mitochondria-ER docking and plays an important role in cell fate determination. We hypothesized that it might be related to platelet priming. These top genes were validated (Fig. 3E and F; Supplementary Fig. S4A) and a robust GRP75 protein expression was observed in the MK cytoplasm (Fig. 3G). Compared with other bone marrow cells, MKs express high levels of GLUT1 (Supplementary Fig. S4B). As a positive control, GLUT1 was also upregulated in MKs from hyperglycemia mice (Supplementary Fig. S4C). Unlike isolated healthy platelets, isolated healthy MKs strongly responded to 12.5 mmol/L glucose mimicking hyperglycemia, and promoted expressions of Grp75 (Fig. 3H) and other top genes (Supplementary Fig. S4D), whereas Glut1 shRNA impeded this effect (Fig. 3H; Supplementary Fig. S4D). In platelets, similar to the platelet activation results, GRP75 expression was not altered upon direct glucose challenge, whereas platelets isolated from hyperglycemic mice or diabetic mice showed a strong expression of GRP75 (Fig. 3I). These data suggest that hyperglycemia upregulates platelet GRP75 through an MK-GLUT1-dependent mechanism.

Considering GRP75 is a mitochondrial protein, we tested mitochondria in platelets. Ultrastructurally, mitochondria number and granule number increased in platelets from the hyperglycemic mice (Fig. 3J; Supplementary Fig. S4E), indicating platelet GRP75 upregulation is accompanied by increased mitochondria.

To assess the role of GRP75 in platelet activation, a specific GRP75 inhibitor MKT-077 was administered on isolated platelets from control and hyperglycemic mice. Interestingly, GRP75 inhibition completely blocked the platelet activation, aggregation, and clot retraction (Fig. 4A–C). Notably, GRP75 inhibition also blocked the platelet aggregation in control groups, suggesting the role of GRP75 in healthy platelet activation (Fig. 4B). In the previously described in vivo thrombosis model, platelets pretreated with MKT-077 showed decreased rate and extent of thrombosis (Fig. 4D and E).

To test whether these platelets interact with tumor cells, we detected the adhesion of platelets on cultured B16-F10 cells using a PKH26 dye (Fig. 4F). Platelets from hyperglycemic mice adhered to tumor cells, whereas GRP75 inhibitor treatment blocked this effect (Fig. 4G). Scanning electron microscopy was performed to assess the morphology of the adherent platelets to tumor cells. Platelets from hyperglycemic mice increased tumor adhesion with a morphology of activation and formation of aggregates, whereas MKT-077 markedly blocked this pattern (Fig. 4H). Together, these results suggest that GRP75 plays a vital role in platelet activation, thrombosis, and tumor cell adhesion.

To investigate the potential mechanism by which glucose induces GRP75 expression in MKs, we analyzed the 2000 bp Grp75 promoter and found 13 transcription factors potentially bind to it, using a prediction tool PROMO (34). QPCR array of these transcription factors showed that Myc increased nearly ten-fold in hyperglycemic MKs (Fig. 5A). MYC expression was validated and found GLUT1-dependent, similar to Grp75 expression (Fig. 5B and C; Supplementary Fig. S5A). To investigate whether GRP75 expression is associated with MYC, we treated isolated MKs from hyperglycemic mice with MYC inhibitor 10058-F4. Ex vivo experiments showed that Grp75 expression in MKs was blocked by the MYC inhibitor (Fig. 5D). Next, ChIP assay in isolated MKs demonstrated that MYC binds to the Grp75 promoter (Fig. 5E). These results pinpoint the upstream mechanism responsible for regulating GRP75 expression in MKs. The proposed mechanism is summarized (Fig. 5F).

To mechanistically link GRP75 to platelet activation, we checked platelet activation pathways (35). Interestingly, a Ca²⁺-PKCα pathway is upregulated in hyperglycemic platelets (Fig. 5G and H). Moreover, hyperglycemia leads to potent phosphorylation of ERK and p38 (Fig. 5I), which is known as the downstream effectors of PKCα (36). Consistent with these results, a higher concentration of thromboxane A2 (TxA₂) was detected in hyperglycemia primed platelets, confirming the hyperactivity of these platelets (Fig. 5J). In contrast, GRP75 inhibition blocks Ca²⁺-PKCα-ERK/p38 axis (Fig. 5K and L), supporting a GRP75-associated calcium release (37). To investigate the role of calcium in GRP75 signaling, we performed calcium rescue experiments. As expected, in isolated hyperglycemia-primed platelets, direct calcium ion supplementation at the second messenger level using calcium ionophores (ionomycin) completely reversed the MKT-077 inhibited activation (Fig. 5M). Our results are in accordance with GRP75′s role in calcium homeostasis (Fig. 5N).

To explore whether platelet-mediated metastasis is dependent on platelet activation, we used two clinical antiplatelet drugs, warfarin and aspirin, for treating the platelet donor mice. Interestingly, both warfarin and aspirin markedly reduced the number but not the size of the metastatic lesion in recipient mice (Fig. 6A–C). These findings indicate that platelet-mediated metastasis depends on platelet activation. Next, we tested whether the GRP75 inhibitor MKT-077 may reduce platelet-mediated metastasis. In a platelet transfer model, by treating isolated platelets with MKT-077, pulmonary metastasis was inhibited (Fig. 6D–F). Again, MKT-077 reduced the foci number, but not the size of the metastatic lesion (Fig. 6E and F), indicating the antimetastasis effect of the platelet-specific GRP75 inhibition. Finally, to link our findings to clinical relevance, we administered MKT-077 to the recipient mice, prior to intravenous tumor cells injection (Fig. 6G). Following previous findings, the pulmonary metastatic foci number dramatically decreased in the MKT-077-treated group (Fig. 6H–J). Interestingly, the metastatic lesion size also shrank (Fig. 6J), probably due to the previously reported antitumor effect of GRP75 inhibition (38). These results demonstrate the antimetastatic potential of GRP75 inhibition.

Glucose solutions, commonly at 5% w/v concentration, are applied in the clinic for supplying carbohydrates or diluting compatible drugs. Although hyperglycemia activates platelets in human (39, 40), the impact of glucose-induced GRP75 on human platelets is unknown. For clinical relevance, we tested the blood samples from volunteers receiving 1 L 5% glucose solution per day for 7 days (Fig. 7A). On day 0, 3, and 7, no insulin sensitivity changes were observed (Fig. 7B). However, the GRP75 expression and the platelet activity were significantly elevated on day 7 (Fig. 7C and D). Administering the GRP75 inhibitor significantly reduced platelet activity on the isolated platelets, recapitulating our findings in mouse models (Fig. 7D). Notably, platelet activity did not increase on day 3 (Fig. 7D), which might reflect the 7- to 10-day platelet turnover time in humans. The average PLT and PCT significantly decreased on day 7 in glucose-uptaking volunteers (Fig. 7E). Ultrastructural analysis showed significant increases of platelet mitochondria (Fig. 7F; Supplementary Fig. S6A). These results suggest a glucose–GRP75–platelet axis in humans.

To test the prometastasis effect of these hyperactive human platelets, we exploited an adoptive platelet transfer model, in which platelets from glucose-drinking volunteers were coinjected with tumor cells in immunodeficient mice (Fig. 7G). Interestingly, these primed human platelets significantly increased B16-F10 pulmonary metastasis in mice, whereas pretreatment of these platelets with MKT-077 markedly blocked these effects (Fig. 7H–J). Histologic full scans demonstrated that the number of the metastatic lesion significantly increased in the hyperglycemia donor platelet group, and this effect was inhibited by pretreating platelets with GRP75 inhibitor (Fig. 7J). Altogether, these results show the antimetastatic effect of GRP75 inhibition in human platelets.

Targeted cancer therapies are important for improving the survival of patients. Off-tumor targets on host cells may indirectly affect tumor development (41, 42). As a strong driving force for tumor development, glucose and its downstream signaling pathways are considered as potential targets for treating cancer. Are there host cell targets for glucose-signaling that are important for promoting tumor development and metastasis? Although hyperglycemia stimulates tumor cell proliferation through various signals, the off-tumor target for hyperglycemia is not well understood in the context of metastasis. As summarized (Fig. 7K), we show that glucose stimulates MKs in the bone marrow and promotes MK-CTC communication. Blocking this host cell target of glucose is beneficial for combating tumor metastasis as evidenced by a broad range of complementary experiments. By a platelet adoptive-transfer approach, we exclude the insulin effect in the animal models used in this study. Our findings thereby provide a novel piece of evidence for tumor cell-MK communication involved in the metastatic cascade.

In the metastasis cascade, tumor cell survival in circulation is a crucial step for allowing dissemination. Related to this step, our data brings four points that need further discussion: (i) Reports show strong associations between thrombocytosis and tumor burden. However, our work suggests that platelet activation, but not platelet amount, promotes tumor metastasis. Circulating platelets are exposed to trace amounts of agonists. Under hyperglycemia conditions, MKs produce hypersensitive platelets that lead to increased spleen clearance, as shown in our data. These primed platelets, even less than those under healthy conditions, facilitate CTC metastasis. (ii) In our human studies, the human-to-mouse platelet transfer model shows that hyperglycemic platelets increased both the number and size of metastatic foci, indicating hyperactive platelets in nude mice not only protect the CTCs in the circulation, but also increase metastatic tumor growth. The latter was not seen in similar models using immune-competent C57BL/6 mice. This inconsistency could be due to the lack of T, B cells in nude mice, as many studies have demonstrated the importance of immune-mediated clearance (43). (iii) Hyperglycemia is one of the most significant hazards on platelets, and our work demonstrates that the impact of hyperglycemia on the platelets is not direct but through activating MK metabolism. Hyperglycemia may alter the turnover period of platelets, and increase the number of mitochondria and granules. (iv) It should be noted that hyperglycemia can be achieved by drinking 5% glucose or 10% sucrose solution. The latter is comparable to the sugar content of commercially available carbonated beverages. Our study thereby suggests additional risks of excessive sugar intake.

Previous studies suggest that platelets absorb glucose in vitro and glucose in the culture medium is required for clot retraction (44). Recent work using genetically modified mice shows that GLUT1 and GLUT3 are important for glycolysis in platelets in vivo (33, 45). However, these in vivo studies did not clarify whether platelets respond to glucose directly. In our study, GRP75 responds to glucose only in MKs but not in platelets, suggesting platelets and MKs respond to glucose differently. As polyploid cells, MKs have a comprehensive, or even more powerful sugar metabolism capacity than normal host cells. Until now, MK metabolism is incompletely understood. We show that the glycolysis-related genes such as Pklr and Bpgm are highly upregulated in MKs from hyperglycemic mice. Targeting these glycolysis key enzymes may regulate MK glucose metabolism. This view requires further investigation.

Antiplatelet drugs are associated with a lower tumor burden and, in some types of cancer patients, better survival. However, due to the multitargeting characteristics and bleeding side effects of these drugs, identifying novel specific targets for platelet–CTC interaction is needed. Our findings suggest that GRP75 could be an antiplatelet target for inhibiting platelet-associated metastasis. To our knowledge, this is the first time that anti-GRP75 is proposed as antiplatelet therapy. Of note, other drugs targeting MK glucose metabolism might have similar effect (46, 47). Importantly, GRP75 blockade in tumor cells leads to increased apoptosis and chemosensitivity (48). It is plausible that the GRP75 blockade may have both antitumor cell growth and antimetastatic effects, making it more attractive for cancer therapy. Generally, due to the broad and important role, targeting chaperones could exhibit severe toxicity (49). Blocking GRP75, however, shows selective cancer cell toxicity and is well tolerated, as evident from a phase I human toxicity study (50). Our data support targeting GRP75 for antimetastasis therapy.

In conclusion, we demonstrate for the first time that a mitochondrial protein, GRP75, is upregulated in glucose-stimulated MK and is an essential regulator of platelet activation. Hyperglycemia potentiates platelet activation via the Ca²⁺-PKCα axis and increases platelet-mediated tumor metastasis through an off-tumor MK target. A GRP75 inhibitor significantly blocks platelet activation and tumor metastasis in mice and in a human-mouse adoptive platelet transfer model. These results provide molecular insight into the role of GRP75 on hyperglycemia-associated platelet activation and tumor metastasis, and suggest that GRP75 blockade might be beneficial for cancer patients.

No disclosures were reported.

B. Wu: Investigation. Y. Ye: Resources, investigation, writing–original draft. S. Xie: Investigation. Y. Li: Resources, investigation. X. Sun: Investigation. M. Lv: Investigation. L. Yang: Resources. N. Cui: Resources. Q. Chen: Resources. L.D. Jensen: Writing–review and editing. D. Cui: Resources. G. Huang: Resources. J. Zuo: Resources. S. Zhang: Resources, writing–review and editing. W. Liu: Resources, supervision. Y. Yang: Conceptualization, resources, supervision, writing–original draft, writing–review and editing.

The authors thank Ms. Kaijing Gao for her artistic input in schematic diagrams. They thank Mr. Xin Huang and Dr. Xue Lv for his and her technical support in some of the experiments. The authors thank Prof. David Saffen at the Fudan University and Prof. Mien-Chie Hung at the MD Anderson Cancer Center for reading of the manuscript and giving helpful suggestions. Y. Ye was supported by the National Natural Science Foundation of China (Project No. 81600839). Y. Li was supported by the General Program of Shandong Natural Science Foundation (Project No. ZR2019MH086). L. Yang was supported by the National Natural Science Foundation of China (Project. No. 81773203). Y. Yang was supported by the National Natural Science Foundation of China (Project No. 81773059), the Shanghai Pujiang Program (Project No. 18PJ1400600), Original Research Program of Fudan University, Innovation Research Team of High-level Local Universities in Shanghai, and the Program for Professor of Special Appointment in Shanghai (Eastern Scholar, Project No. TP2018007).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.