Gastric Cancer-Induced Neutrophil Extracellular Traps: A Potent Mediator of Cancer Associated Thrombosis

Background: Development of venous thromboembolism (VTE) is associated with high mortalities among gastric cancer (GC) patients. Neutrophil extracellular traps (NETs) have been reported to correlated to procoagulant and prothrombotic in some diseases. We aimed to clarify that NETs participates in the development of cancer-associated thrombosis in GC. Method: The level of NETs in blood and tissue samples of patients were analyzed by ELISA and ow cytometry. NETs generation in vitro were observed by immunouorescence (IF). The NETs procoagulant activity (PCA) was performed by brin formation and thrombin-antithrombin complex (TAT) assays. Hypercoagulation of platelets and endothelial cells (ECs) stimulated by NETs were measured by IF and ow cytometry. Thrombosis in vivo was measured in an established mice model of VTE induced by ow stenosis in the inferior vena cave (IVC). Result: NETs are likely to form in blood and tissue samples of GC patients compared with healthy individuals. In vitro studies that GC cells and their conditioned medium (CM), but not gastric mucosal epithelial cell can stimulate NETs releasing from neutrophils. In addition, NETs induced hypercoagulation of platelets by up-regulating the expression of phosphatidylserine (PS) and P-selectin on the cells. Furhter, NETs stimulate adhesion of normal platelets on glass surfaces. Similarly, NETs trigger the conversion of ECs to hypercoagulable phenotypes by down-regulating the expression of their intercellular tight junctions but up-regulating that of tissue factor (TF). Treatment of normal platelets or ECs with NETs augmented the level of plasma brin generation and TAT complex. Meanwhile, in the models of IVC stenosis, tumor-bearing mice demonstrate stronger ability to form thrombi and NETs were abundantly accumulated in the thrombi compared with control mice. Notably, combination of DNase-1, activated protein C (APC) and Sivelestat markedly abolished the PCA of NETs. Conclusion: Our ndings demonstrate that GC-induced NETs strongly


Background
Gastric cancer (GC) is one of the most prevalent gastrointestinal tumors and the third most fatal cancer world [1,2] . Meanwhile, venous thromboembolism (VTE) is a common complication in GC patients, relative to the healthy individuals [3][4][5] . Notably, VTE is associated with high mortalities of GC patients [6,7] . Several factors such as tumor stage and increased concentration of D-dimer contribute to the development of VTE in GC patients [8,9] . However, the molecular mechanism underlying procoagulant activity (PCA) in GC patients are poorly understood. Uncovering molecular targets associated with VTE in GC patients can help in the development of appropriate therapy, which can improve the clinical outcomes of these patients.
Neutrophil extracellular traps (NETs) are web-like structures composed of lamentous DNA and antimicrobial proteins such as citrullinated histone-3 (citH3), neutrophil elastase (NE), myeloperoxidase (MPO), matrix metalloproteinase-9 (MMP-9) and Cathepsin G (CatG) [10,11] . They result from interaction between neutrophils and bacterial or other stimulating factors [12,13] . Overall, they protect the host from pathogens related damages. Besides the prime protective functions, undesirable effects of NETs in autoimmune diseases have been reported [14][15][16] . Recent researches have linked NETs to the development of metastasis and cancer-associated thrombosis [17,18] . Particularly, NETs results in arterial and venous thrombosis, both mediated by neutrophils [19] . Furthermore, using mouse models, it was shown that thrombosis in breast cancer tissues was closely linked to formation of NETs [20] . This underlines the probable relationship between NETs and cancer-related thrombosis.
A recent study revealed that priming of metastatic pancreatic cancer cells with platelets stimulates neutrophils to release NETs, which promote thrombosis in both static and dynamic state [21] . Additionally, platelets derived from GC can stimulate neutrophils to release NETs. Meanwhile, NETs enhance PCA in GC patients, which positively correlate with expression of thrombin-antithrombin (TAT) complex and level of serum D-dimers [22] . Moreover, both human and animal studies suggest that enhanced thrombosis may result from over-expression of activated platelets [23,24] . Nevertheless, little is known about the interaction between NETs and platelets activation in GC patients. Meanwhile, the injury of venous ECs in cancer patients is also closely related to venous thrombosis [25,26] . Interestingly, cytotoxicity of NETs against ECs enhances PCA in oral squamous cancer, even in patients with obstructive jaundice and in ammatory bowel disease [27][28][29] . Even so, the potential mechanism underlying ECs injury in GC patients is poorly understood.
Our central hypothesis is that GC-induced NETs participate in the VTE responses by platelets activation and endothelium injury. Therefore, we rst explored the complex relationship between NETs formation and platelets activation as well as ECs injury. Further, we showed the effect of NETs on thrombosis in IVC stenosis mice models. In general, our results may offer that NETs are potential therapeutic targets in the prevention and treatment of VTE in GC patients.

Patients and tissue samples
Sixty-three patients newly diagnosed with primary GC and thirteen healthy donors (HD) attending the Second A liated Hospital of Harbin Medical University between October 2019 and April 2021 were enrolled in this study. GC diagnoses were performed based on pathological examinations. Pathologic tumor-node-metastasis (TNM) stage and histological classi cation of GC were performed according to the 7 th American Joint Committee on Cancer (AJCC) guidelines [30] . Underage (<18 years), pregnant, those on anti-tumor or anticoagulant treatment before surgical treatment or patients with underlying complications such as endocrine system, cardiovascular, hematological system, infectious and other separated below the left renal vein plane. After 5-0 (1mm) suture passed through the IVC, 3-0 (2mm) suture was placed at the parallel part of the IVC as the blocking line. The IVC was ligated and 3-0 suture was carefully extracted. This procedure has been shown to decrease vascular lumen by about 90%. The other branches of IVC were ligated to the level of iliac vein. Thereafter, the abdominal incision was closed, mice were sacri ced after 6 or 48 h and thrombi formed in the IVC were harvested. The IVC stenosis mice in the treatment groups were injected with DNAse-1 (50µg/mouse, Roche) intraperitoneally every 12 h until the time of sacri ce.

Quanti cation of plasma NETs marker
Plasma cell-free DNA (cf-DNA), MPO-DNA and citH3-DNA complex were quanti ed using capture ELISA as previously described [22,27] . The quanti cation of cf-DNA was performed by the Quant-iT PicoGreen dsDNA assay kit (Invitrogen, USA). For detection of NETs complexes, MPO ELISA kit (Jianglaibio, Shanghai, China), and citH3 ELISA kit (Jianglaibio, Shanghai, China) were combined with Quant-iT PicoGreen dsDNA assay kit (Invitrogen) respectively to detect MPO-DNA and citH3-DNA.
. Fibrin formation and TAT complex assay Fibrin formation of platelets and ECs were detected by turbidity as previously described [33] . Platelets and ECs monolayers were stimulated by NETs with or without DNase-1, APC and Sivelestat treatment alone or together for 4 h, and then 150ul PFP from HD were co-cultured with them at 37℃ for 2 min, followed by the addition of 50 ul of prewarmed 25 mmol/l CaCl 2 . The brin formation was detected by measuring OD at 405 nm. For detection of TAT complex level, Human and Mouse TAT complex ELISA kit (Jingkbio, Shanghai, China) were performed as previously described [34] .

Flow cytometry
Circulating NETs were measured using ow cytometry. Here, whole blood from HD and GC patients was rst diluted with 1×PBS and incubated in the dark at RT for 30min with FITC-conjugated-citH3  . Thrombi in the IVC of tumor models or control mice were stained with primary rabbit anti-histone H3(1:500, ab5103, Cambridge, UK) and rat anti-Ly6G (1:500, Novus, USA), and then it was incubated with the with Alexa Fluor 594 conjugated goat anti-rabbit (1:200, proteintech, China) and Alexa Fluor 488 conjugated goat anti-rat (1:200, proteintech, China) secondary antibodies as previously described [35] . The tissues images were captured using a confocal microscope.
Neutrophils (5×10 5 cells) isolated from healthy individuals were seeded and incubated in glass-based poly-L-lysine-coated 24-well plate for 1h at 37℃ under 5% CO 2 chamber. Thereafter, cell suspension of KATO-, MKN-45, AGS, GES-1 cells (2×10 5 cells) or CM from GC cells were co-cultured with neutrophils for 4h at 37℃ under 5% CO 2 . To detect and quantify NETs, the samples of CM group were incubated with primary rabbit anti-Histone H3 and mouse anti-MPO antibodies and thereafter uorescent secondary antibodies. The samples of cell-cell contact groups, NETs were stained with Sytox® Green (Solelybio, Beijing, China) for 10min in the dark at RT. All experiments were analyzed by confocal microscope.

Preparation of cell-free NETs
Cell-free NETs were isolated from neutrophils of GC patients as previously described, but with slight modi cations [36] . Brie y, neutrophils (1×10 7 cells/ml) were cultured for 4h at 37 °C under 5% CO 2 in media supplemented with 500 nM PMA (HY-18739, MCE, USA). The supernatant was abandoned, then ice-cold 1×PBS were added to wash down the cell layer of neutrophils to obtain the NETs medium and centrifuged at 1,500×g for 10 min at 4 °C to remove cell debris. Thereafter, 1.5ml of the supernatant (sterile DNAprotein complex) was centrifuged at 15,000×g for 15min at 4℃. The resultant pellets were suspended in ice-cold 1×PBS and quanti ed using spectrophotometry. The medium containing the NETs was stored at -80℃ for subsequent experiments.

Platelet activation and adhesion assay
Platelets activation and adhesion assays were performed as previously described [21] . Brie y, glass-based wells of 24 well plate were coated with cell-free NETs after overnight incubation with corresponding medium at 4℃ in a humidi ed chamber. For controls, 1% of denatured bovine serum albumin (dBSA) was used. Platelets suspension (1×10 7 cells/ml) was then seeded in the wells, cultured for 1 h at 37℃ under 5% CO 2 , xed for 15 min at RT with 4% paraformaldehyde and washed three times using 1×PBS before 20 min permeabilization using 0.1% Triton-X 100. The platelets were then incubated for 30 min with Alexa Flour 594 conjugated phalloidin primary antibodies (1:300, ThermoFisher, Waltham, USA). To assess PS and P-selectin expression, the platelets were stimulated by NETs for 1 h, and the cells were rstly stained with FITC-conjugated-lactadherin (Haematologic, EssexJunction, VT) for 30min, and then stained with primary rabbit anti-P-selectin (1:200, Proteinteck, China), and mouse anti-CD41 (1:500, Novus, USA) antibodies, imaging were observed and photographed using confocal microscopy.

HUVECs stimulation assay
HUVECs were incubated for 4 h in 24 well plates with cell-free NETs or PBS. The cells were xed in 4% paraformaldehyde for 15 min at RT, washed three times using 1×PBS and thereafter blocked for 1h using 10% goat serum with 1% BSA solution in PBS. For detection of TF expression, ECs were incubated overnight at 4℃ with rabbit anti-TF (1:500, ab228968, Cambridge, UK) and mouse anti-CD31 (1:500, ab9498, Cambridge, UK) primary antibodies. Then cells were washed with PBS and re-incubated for 1h at RT with Alexa Fluor 594 conjugated (Proteinteck, China) goat anti-rabbit and Alexa Fluor 488 (Proteinteck, China) conjugated goat anti-mouse secondary antibodies. For detection of intercellular junctions of cells, incubated overnight at 4℃ with rabbit anti-VE-cadherin (1:500, ab33168, Cambridge, UK) primary antibodies followed the Alexa Fluor 488 (Proteinteck, China) conjugated goat anti-rabbit secondary antibodies and further incubated with Alexa Flour 594 conjugated phalloidin primary antibodies (1:300, ThermoFisher, Waltham, USA). They were then stained with 4',6-diamidino-2-phenylindole (DAPI) and xed with mounting medium (Solarbio, Beijing, China) for 5min at RT in the dark. The cells were then observed and photographed using a confocal microscope. The photos were analyzed using Image J software.

Statistical analysis
Comparisons between groups were performed using Student's t-tests, paired t-tests and analysis of variance (ANOVA). Continuous data was expressed as mean ± standard deviation (SD). All analyses were performed using GraphPad Prism software, V. 8.0. P< 0.05 was considered statistically signi cant.

Gc Patients Display Greater Nets Formation
The level of plasma cf-DNA, citH3-DNA and MPO-DNA complexes between GC patients (n = 63) and healthy donors (HD, n = 13), which re ects the concentration of NETs, were measured using capture ELISA. In general, the levels of NETs marker were signi cantly higher in patients with stage / / GC, relative to those with HD ( Fig. 1A-1C). There was also a signi cant difference in preoperative and postoperative plasma NETs marker levels in GC patients (Fig. 1D-1F). Moreover, the levels of NETs marker positively correlated with that of serum D-dimer. This suggests that NETs are associated with hypercoagulation and VTE development in GC patients ( Table 1). The level of NETs (MPO + /citH3 + neutrophils) in circulation in GC patients and HD were measured using ow cytometry. We found circulating NETs were higher in blood of patients with either of GC stages (II, III, and IV), relative to HD counterparts ( Fig. 1G-H). Furthermore, based on MPO and citH3 levels, immuno uorescence staining revealed that NETs were signi cantly higher in tumor microenvironment (Fig. 1I) relative to paratumor tissue of the same patients (Fig. 1J).

Gc Cells Stimulates Formation Of Nets By Neutrophils
To assess whether gastric cancer cells can directly stimulate NETs formation, we analyzed the expression of NETs in a co-culture of GC cell lines and neutrophils. Immuno uorescence analysis revealed that compared with GES-1, the rate of NETs formation was signi cantly high in GC cells ( Fig. 2A). However, formation of NETs was greater in metastatic than the non-metastatic GC cell line (Fig. 2B). Moreover, we measured the NETs formation when normal neutrophils were co-cultured with CM from KATO-, MKN-45, AGS and GES-1. Immuno uorescence analysis further revealed that CM of KATO-and MKN-45 exerted greater neutrophils activation for NETs formation, relative to CM of AGS cells. However, the CM of GES-1 had no effect on NETs formation (Fig. 2C-2D). Overall, these ndings demonstrate that GC cells stimulate NETs formation through both intercellular contact and non-contact mechanisms.

Nets Contribute To Hypercoagulation Of Platelets
To examine the effect of NETs on platelets activation, we measured the levels of PS and P-selectin expression on these cells. Flow cytometry revealed that compared to HD, the expression of PS and Pselectin was signi cantly high on platelets of patients with GC ( Fig. 3A-3D). In addition, platelets isolated from HD were co-cultured with NETs medium or PBS before analyzing PS and P-selectin expression. We found NETs stimulates PS and P-selectin expression on platelets by confocal microscope (Fig. 3E-3G). Flow cytometry also demonstrated this hypercoagulable phenotype of platelets which was stimulated by NETs ( Fig. 3H-3I).

Nets Promote Platelets Adhesion And Prothrombotic State
Previous studies have shown that NETs promote thrombosis in murine late-stage breast cancer models and deep vein thrombosis models [18,37] . However, whether NETs derived from GC neutrophils have ability to stimulate platelets adhesion under static condition was unknown. In order to determine the effect of NETs on platelets adhesion, platelets isolated from HD were further seeded in NETs-coated wells to measure the effects of NETs on sticking of platelets on blood vessels. Confocal microscopy revealed that NETs enhanced sticking of platelets on to glass slides (Fig. 4A-4B), indicating that NETs induce development of thrombosis. Moreover, results revealed that brin formation and TAT complex level were obviously increased when control plasma were co-cultured with platelets activated by NETs (Fig. 4C-4D).
Inhibition assay revealed that digestion of NETs-DNA using DNase-1 modulated adhesion of platelets on glass surfaces. Even so, few platelets still adhered to the NETs-coated well pretreated with DNase-1 (Fig. 4A). This suggests that other protein components other than NETs-DNA participate in the adhesion of platelets. Further, NETs were treated with DNase-1, APC, Sivelestat or a combination of the three inhibited 67.1%, 56.6%, 38.9% and 91.8% of platelets adhesion (Fig. 4B). The degree of platelets adhesion in the combination group were comparable to that of control. Further, we found DNase-1, APC, Sivelestat or a combination of the three reduced 81.1%, 73.9%, 64.3% and 90.7%, respectively, of brin formation and inhibited 78.9%, 57.4%, 51.2% and 91.9%, respectively, of TAT complex level at the highest concentration of NETs (Fig. 4C-4D). Taken together, these ndings demonstrate that NETs play a role in the development of thrombosis.

Nets Drive Hypercoagulation Of Ecs
To detected the effect of NETs on ECs thrombogenicity, HUVECs were co-cultured with NETs medium. Confocal microscopy revealed that NETs destroyed the normal intercellular junctions between ECs ( Fig. 5A, 5C). Moreover, NETs treatment up-regulated the expression of tissue factor (TF) on surface membrane of ECs (Fig. 5B, 5D). In addition, plasma brin formation and TAT complex level were signi cantly increased when control plasma were incubated with ECs monolayers activated by NETs (Fig. 5E-5F). Further inhibition assays were performed to assess the effect of NETs on ECs after pretreatment with DNase-1, APC, Sivelestat or a combination of the three. We found NETs treatment after incubation with DNase-1, APC, Sivelestat or all the three returned 47.5%, 36.3%, 33.4% and 86.5%, respectively, of VE expression on ECs whereas similar treatment inhibited 60.4%, 44.8%, 38.6% and 95.5%, respectively, of TF expression (Fig. 5H-5I). Meanwhile, we found NETs treatment after incubation with above inhibitors inhibited 59.7%, 54.4%, 54.0% and 91.5%, respectively, of brin formation level and also inhibited 51.2%, 35.6%, 25.9% and 84.3%, respectively, of TAT complex level. Taken together, these ndings suggest that NETs promote hypercoagulation of ECs, thus inhibiting NETs function can protect against venous injury. 6. NETs promote the formation of thrombi in IVC ow restriction models of tumor-bearing mice.
Based on these ndings in vitro and pivotal role of NETs in thrombosis, we hypothesized that GC-induced NETs can also promote thrombosis in vivo. Here, in mice IVC ow stenosis models, we found tumorbearing mice demonstrate more capacity to form thrombi and showed heavier weight and longer length of thrombi compared to control mice (Fig. 6A-6D). In the 6 h models, three of 9 of control mice showed thrombi, whereas seven of 9 of tumor-bearing mice formed thrombi. In the 48 h models, all mice demonstrated thrombi in IVC. In addition, confocal imagines of thrombi formed in the tumor-bearing mice after 48h IVC stenosis showed that NETs were signi cantly accumulated compared to control mice ( Fig. 6E-6H). The thrombi of control mice included some neutrophils (Ly6G + ) but wasn't activated to form NETs (Fig. 6G). Furthermore, tumor-bearing mice showed higher brin formation and TAT complex level compared to control mice (Fig. 6I-6J).
In inhibition assay, we infused DNase-1 in mice immediately after IVC stenosis and examined thrombosis after 6 or 48h of surgery. We found treatment with DNase-1 signi cantly inhibited the thrombi formation in IVC stenosis models of tumor-bearing mice (Fig. 6A-6D). Further, brin formation and TAT complex level of tumor-bearing mice were signi cantly decreased by DNase-1 treatment (Fig. 6I-6J). These data potent suggest that NETs play a role in thrombosis in vivo which was induced by GC and inhibit NETs by DNase-1 have a protective effect on thrombosis in this mice model.

Discussion
In ammation is one of the hallmarks of cancer. Meanwhile, neutrophils are among the most important immune cells implicated for promoting tumor progression [38,39] . On the other hand, NETs participate in cancer progression by promoting proliferation, invasion, metastasis and angiogenesis of cancer cells as well as thrombosis in numerous tumor types [40][41][42] . A recent study using mice models through Jak 2V617F knock in revealed that most myeloproliferative neoplasms (MPN) display NETs formation and deep vein thrombosis [43] . Our previous studies revealed that NETs promote migration and metastasis of GC cells both in vitro and in vivo through epithelial mesenchymal transition (EMT) [44] . Intriguingly, inhibition of NETs promotes apoptosis and inhibits the invasion of GC cells by regulating the expression of Bcl-2, Bax and NF-κB proteins [45] . Our initial studies revealed that NETs released by neutrophils in GC patients promoted conversion of thrombin and brin [22] . Accordingly, we hypothesized that NETs promote thrombosis in GC patients. We then investigated the interactions between GC cells, neutrophils, platelets and ECs, with a keen focus on their role in cancer-associated thrombosis.
In this study, we found the level of plasma NETs marker and citH3 positive neutrophils were signi cantly higher in GC patients relative to HD. Also, the expression of NETs decreased signi cantly after resection of GC tissues. Furthermore, we found that neutrophils in ltration and NETs formation were up-regulated in tumor tissues, relative to adjacent paratumor tissues of the same GC patient. Moreover, the levels of serum D-dimer positively correlated with tumor TNM stage, consistent with previous ndings [46] . These ndings suggest that expression of NETs promotes GC development and thrombosis in the same group of patients.
Recent studies have shown that hepatocellular cancer (HCC) and hypoxic CM stimulate production of NETs by neutrophils [42] . However, the relationship between GC cells and neutrophils is poorly understood. Our experiments demonstrated that both metastasis and non-metastatic GC cancer cells directly stimulates production of NETs from neutrophils, contrary to GES-1. Further, analyses revealed that CM of GC cells, but not that of GES-1, also induce formation of neutrophils-related NETs. This suggests that NETs formation is also mediated by factors secreted by GC cells. In ammatory cytokines such as interleukin (IL)-1β, IL-6, IL-8 and tumor necrosis factor-alpha (TNF-α) as well as damage associated molecular patterns (DAMPs), all over-expressed tumor microenvironments, stimulates neutrophils to release NETs [47,48] . Even though IL-8 was the most over-expressed cytokine in GC patients, whether or not it is the main mediator of NETs formation and subsequent underlying mechanism in GC cancer remain to be validated.
A recent study showed that NETs promotes thrombosis by activating and promoting adhesion of platelets in venous walls of pancreatic cancer patients [21] . Activated platelets display expression of PS on their surface membranes [49] . Meanwhile, P-selectin expression on the surface membrane of platelets is also associated with thrombosis [50] . In this study, we found that compared to HD, PS and P-selectin expression on platelets were signi cantly high in GC patients, particularly those with / GC. Moreover, NETs treatment upregulated PS and P-selectin expression on platelets. Also, NETs stimulated adhesion of normal platelets on glass slides. However, even though DNAse-1 treatment modulated this phenomenon, some platelets still adhered to the glass slides, suggesting that other secretory factors participated in the adhesion property.
Previous studies have shown that histone in NETs promote thrombosis in CRC patients [51] . NE is another most abundant protein that binds NETs. Even though the potent protein stimulates tumor progression both in vitro and in vivo, mechanisms underlying NE-mediated cancer-associated thrombosis remain to be clari ed. A recent study on DVT using mice models showed that NE de ciency or NE inhibition alone does not completely inhibit DVT [52] . In this study, we found hypercoagulation of platelets was not completely mediated by NETs-DNA, but also by other secretory proteins in the NETs such as histone and NE. Consequently, DNase-1 treatment of NETs had no complete effect on hypercoagulation of platelets.
However, a combination of DNase-1, APC and Sivelestat treatment almost completely inhibited hypercoagulation of platelets. Although Sivelestat didn't show a strong anti-hypercoagulation effect like that of DNase-1, it nonetheless modulated activation and adhesion of platelets. This demonstrated that histones and NE also participate in activation and adhesion of platelets.
DVT can be triggered by injury to vascular ECs. Previous studies have shown that under certain malignancies, NETs can induce dysfunction and apoptosis of ECs [33] . In patients with chronic pancreatic disease and pancreatic cancer, NETs exert their cytotoxic against ECs via intercellular cell adhesion molecule-1 (ICAM-1) and vascular cell adhesion molecule-1 (VCAM-1) expression [25] . Recent studies have shown that treatment of ECs with NETs derived from patients with colorectal cancer (CRC) promotes and fasten production of brin and corresponding coagulation [51] . In this study, we found that NETs treatment inhibited secretion of intercellular junctions in ECs and promoted hypercoagulation of platelets by upregulating TF expression. Moreover, NETs treatment up-regulated the expression of PS on ECs. On the other hand, ECs activated by NETs signi cantly increase the level of TAT complexes and brin generation in plasma of HD. Given that a combination of DNase-I, APC and Sivelestat treatment completely inhibited hypercoagulation, the process is regulated through numerous mechanisms. In general, these ndings strongly suggest that NETs contribute to GC-associated thrombosis.
In the late stage of murine mammary tumor models, thrombus was found in the venous of lung and NETs was accumulated in it, indicating that cancers induced NETs contribute to the cancer-associated thrombosis [18] . Here, we demonstrated that GC-bearing mice have more ability to form thrombi than control mice and NETs was abundantly present in the thrombi of tumor-bearing mice IVC stenosis models, most of this response can be blocked with DNase-1 treatment, which was similar with previous researches. The DVT models by ow restriction of IVC may result a hypoxia microenvironment to recruit neutrophils and stimulate NETs releasing. In addition, cancer cells often secrete more in ammatory factors, which aggravate the recruitment of neutrophils to form NETs under a hypoxic condition [48] . Therefore, neutrophils exposed to two major triggers of NETs releasing, tumor hypoxia environment and IVC ow restriction, which then participate in the development of thrombi in GC.

Conclusion
Our ndings demonstrate that GC cells can directly induce NETs formation, which in turn strongly increases the risks of VTE development both in vitro and in vivo. In addition, we found that not only NETs-DNA, but also histone and NE participate in the development of cancer-associated thrombosis. Accordingly, NETs are potential therapeutic target against VTE in GC patients.    selectin. Scale bars:10um. All values are mean±SD. Statistics, one-way ANOVA. ns=no signi cant, *P 0.05, **P 0.01,***P 0.001, ****P 0.0001.  ns=no signi cant, *P 0.05, **P 0.01,***P 0.001, ****P 0.0001.