STP3725 was previously developed by site-specific conjugation of a tetradeoxycholic acid to the end-site of enoxaparin.32 The synthesis of STP3725 for all animal experiments stated were carried out by ST Pharm in a previous study. For in vivo experiments, STP3725 was dissolved in distilled water, followed by the addition of poloxamer 188 (2.16 mg/kg) and labrasol (10 µL/mg STP3725) as solubilizers for the administration dosage of 5 mg/kg. The resulting solution (200 µL) was administered to each mouse by oral gavage.

Values are presented as mean±SEM unless otherwise indicated. For in vivo studies, replicates were performed three times, each time containing at least 8–10 individuals. For all data, statistical significance was determined either by Student’s t-test between two groups or one-way analysis of variance followed by Turkey’s post hoc analysis for multiple-group comparisons. Statistical analysis was performed with GraphPad Prism V.7. Statistical significance was set to p<0.05.

Other detailed methods are provided in online supplemental methods.

In highly thrombotic B16F10.OVA melanoma (Ovalbumin-transfected B16F10 melanoma)-bearing C57BL/6 mice, an induced thrombosis model was established by injecting fluorescently labeled fibrinogen. Fibrinogen-Cy5.5 dosage was varied from low, moderate to high amounts that showed corresponding patterns of fluorescence by IVIS (In vivo imaging system) imaging (figure 1A) and this was also quantified (figure 1B) that shows 12.30-fold increase in the ‘high’ group compared with ‘low’. Detection of tumor hypoxia by pimonidazole showed an increase in pattern in correlation to fibrinogen-Cy5.5 amount by immunofluorescence staining (figure 1C) and flow cytometry (figure 1D,E) that showed 2.58-fold increase in the ‘high’ group compared with ‘low’.

Subsequently, detection of intratumoral CD8⁺ T cells showed an inverse correlation to injected fibrinogen-Cy5.5 (figure 1F). Quantified flow cytometry analysis showed that number of T cells decreased by 2.76-fold in the ‘high’ group compared with ‘low’ (figure 1G). Similarly, exogenously injected CD45.1⁺ congenic T lymphocytes were detected harvested tumors of melanoma-bearing wild-type mice to show a decreasing pattern of newly infiltrating lymphocytes (figure 1H) and quantified values showed 4.56-fold decrease in the ‘high’ group compared with ‘low’ (figure 1I).

Further, linear regression analysis was performed of the relationship between injected fibrinogen and tumor hypoxia (pimonidazole) (figure 1J) that showed a positive correlation and the relationship between injected fibrinogen and infiltration of CD45.1⁺ congenic lymphocytes that showed negative correlation (figure 1K).

Overall, these results showed that this induced thrombosis model displays tumor hypoxia exerted by CAT, and that this fibrinogen-mediated limited tumor perfusion shows a corresponding pattern of tumor hypoxia and tumor infiltration of both existing and newly introduced lymphocytes.

In the same melanoma model, various experiments were performed to investigate the effect of anticoagulation therapy on tumor perfusion and hypoxia. P selectin knockout (Psel^-/-) mice were used for negative control as P selectin is essential for the formation of CAT. Control, P sel^-/-, enoxaparin- and STP3725-treated mice were evaluated for blood clots (figure 2A) that showed that its incidence was highest in the wild-type mice compared with all other groups (figure 2B). Vessel occlusion caused by CAT was quantified by H&E staining (fully occluded >80%, partially occluded <80%). Enoxaparin-administered mice showed diminished thrombosis compared with control and STP3725 treated group presented even less incidence of clots similar to P sel^-/- group (online supplemental figure 1A).

Further, three‐dimensional (3D) Doppler analysis was used in this model to visualize and quantify perfusion in the tumor tissue in B16F10 (figure 2C,D). We found that both sectional blood flow and total blood volume in the tumor tissue were increased by STP3725 administration, also apparent in the 3D tumor images of control (online supplemental video 1 and STP3725-treated mice online supplemental video 2).

Also, similar patterns of perfusion were observed in other tumor types (AsPC-1, CT26) following STP3725 treatment (online supplemental figure 1B–E).

STP3725 enhanced tumor perfusion and subsequently diminished tumor hypoxia, analyzed by flow cytometric analysis that showed hypoxic (pimonidazole⁺) cells decreased by 44.39±7.64% in STP3725-treated mice compared with control (figure 2E,F). This was also evident by staining the tumor tissue for GLUT-1, overexpressed in hypoxic conditions, that stained significantly less by 35.06±4.32% for tumors in the STP3725-treated mice (figure 2G,H). Subsequently, live imaging of hypoxia in tumor was observed by PET/MR, using11 FMISO dye, that showed STP3725 diminished the SUVR (Standardized uptake value/region) value by 27.27±5.55% compared with control (figure 2I,J).

Collectively, we conclude that prevention of CAT using anticoagulation therapy can be an effective strategy to enhance tumor perfusion while alleviating hypoxia.

The alleviation of hypoxia rendered by STP3725 treatment, and its effects on the tumor immune microenvironment was investigated by analyzing hypoxia-related immunosuppressive molecules: HIF-1α; transforming growth factor β; vascular endothelial growth factor A; chemokine ligand 28. STP3725 directly reduced the expression of these proteins (figure 3A–F) compared with control group and this evidently verifies that STP3725 can contribute to the relief of immunosuppression.

Tumor whole slide showed that total level of hypoxia was also lower in the STP3725 group compared with control, but number of infiltrated CD8⁺ cells were higher (figure 3G). Magnified images showed that CD8⁺ cells were almost depleted in the pimonidazole⁺ area in both control and STP3725 groups (figure 3H). This indicates that STP3725 can expand the spatial distribution of lymphocytes within the tumor tissue by alleviating hypoxia.

Further, the effect of STP3725 on the infiltration of new lymphocytes was investigated using a congenic (CD45.1⁺) mouse model, where harvested lymphocytes were injected into B16F10.OVA tumor-bearing syngeneic wild-type (CD45.2⁺) mice after immunization of lymphocytes with tumor-specific antigen (figure 3I). Infiltration of new (CD45.1⁺) lymphocytes in the tumor tissue of receiving mice were higher in STP3725-treated group by 3.96-fold compared with control (figure 3J,K) suggesting that consistent anticoagulation therapy promotes the entry of new lymphocytes into tumor tissue.

Excised tumors were also analyzed for hypoxia following rivaroxaban and STP3725 treatment. Tissue immunofluorescence staining showed a similar pattern of decrease in tumor hypoxia following oral anticoagulation therapy by rivaroxaban and STP3725, and this was quantified by flow cytometry to show the same pattern (online supplemental figure 2A,B). In this tumor model, rivaroxaban or STP3725 treatment did not affect tumor growth compared with control (data not shown). Further, tumor CD8⁺ T cells were detected by flow cytometry to show increased numbers of T cells for rivaroxaban and STP3725-treated mice (online supplemental figure 2C).

Collectively, the data indicate that STP3725 can enhance the infiltration and spatial distribution of new lymphocytes in the tumor tissue by alleviating hypoxia (online supplemental figure 3).

After confirming the immune-favorable microenvironment changes induced by STP3725 treatment, this was combined with αPD-1 (anti-programmed cell death protein 1) antibody to show inhibition of tumor growth by 77.26±2.78% in the combination group (figure 4A) in a B16F10.OVA model; STP3725 monotherapy showed marginal effects and antibody monotherapy presented 46.60±7.88% inhibition compared with control without severe toxicity (figure 4B–D). Effective tumor growth inhibition was also confirmed in a mouse colorectal CT26.CL25 model (figure 4E–H).

STP3725 treatment also significantly improved both the distribution and the accumulation of αPD-1 antibody-Cy5.5 in the tumor tissue by 1.92-fold compared with control (online supplemental figure 4) indicating that STP3725 administration promotes αPD-1 antibody delivery and distribution in the tumor tissue possibly by promoting tumor perfusion. Furthermore, higher CD8⁺ TIL numbers and diminished hypoxia was observed in the combination therapy compared with αPD-1 group (online supplemental figure 5). These results correlated with the enhanced anti-tumor efficacy and indicate synergistic benefits of the combination therapy.

A highly immune-depleted mouse pancreatic Pan02 model was also investigated, and the results showed similar patterns of promotion of αPD-1 antibody efficacy in inhibiting tumor growth (online supplemental figure 6).

Analysis of immune cell changes after STP3725 administration in a B16F10.OVA model showed increases in the population of TIL (CD45⁺), cytotoxic T cells (CD3⁺CD8⁺) and helper T cells (CD3⁺CD4⁺) compared with control, and the combination group showed highest populations (figure 5A–C). There were no significant differences in the regulatory T cells (CD4⁺foxp3⁺) numbers, however, the increase in ratio of cytotoxic to regulatory T cells was meaningful only in the combination group (figure 5D,E). Interestingly, STP3725 treatment reduced the myeloid-derived suppressor cell (MDSC) (CD11b⁺Gr-1⁺) populations compared with control, and the pattern was the same for combination group compared with αPD-1 monotherapy (figure 5F). Here, αPD-1 monotherapy exerted no changes and it can be deduced that STP3725 was solely responsible for the reduction of MDSCs. Similarly, only in the combination group did the ratio of cytotoxic T cells to MDSCs show significant increase compared with both the control and αPD-1 monotherapy groups (figure 5G).

Evaluation of cytotoxic T cell function showed that PD-1⁺ mean fluorescence intensity (MFI) were lowered after STP3725 treatment regardless of αPD-1 treatment (figure 5H); also, the population of Ki67⁺ cells were highest in the combination group (figure 5I). T cell functions in tumor were further investigated by showing that STP3725 and the combination groups substantially enhanced interferon gamma (IFNγ) secretion of OVA-stimulated B16F10.OVA tumor-derived lymphocytes (figure 5J). Similarly, numbers of IFNγ, tumor necrosis factor alpha (TNFα)-releasing CD8⁺ T cell and ‘polyfunctional T cell’ (IFNγ⁺TNFα⁺) showed the same patterns (figure 5K,L). Based on the data that show reduction of PD-1-expressing T cells and increase in proliferating and cytokine-secreting T cells in tumor, we contend that STP3725 also improved the function of cytotoxic T cells. All data was analyzed in comparison to the corresponding IgG antibody data (online supplemental figure 7).

Additionally, in a CT26.CL25 model, STP3725 also exerted increase in TILs, cytotoxic T cells and helper T cells, while reducing MDSCs and PD-1^- CD8⁺ T cells in tumor (online supplemental figure 5); these results collectively confirm again, the alteration of the tumor immune microenvironment by STP3725.

Overall, STP3725 significantly enhanced the efficacy of αPD-1 antibody by reducing MDSC population while enhancing both the numbers and function of T cells.

Additionally, tumor (OVA)-specific immune responses were assessed by measuring IFNγ levels after stimulating tumor-derived lymphocytes and splenocytes with tumor specific antigen from B16F10.OVA-bearing mice (figure 6A). The results showed highest IFNγ levels in the combination group analyzed by ELIspot as depicted in the representative images and quantitative analysis of the spotting in tumor derived lymphocytes (figure 6B,C). Splenocytes also presented similar results, in that the combination group showed highest IFNγ levels in ELIspot as well as and ELISA (figure 6D,E). It is important to note that STP3725 monotherapy showed negligible results on IFNγ levels, which implies that anticoagulant alone is not enough to exert a strong local or systemic immune memory response. Collectively, these data suggest that combination of anticoagulant and immunotherapy can strongly elicit tumor specific memory responses both for local and systemic manner.

The efficacy of STP3725 combination with αPD-1 antibody was previously verified in various syngeneic/orthotopic tumor models and the same regimen was also applied in the highly thrombotic K-Ras mutant spontaneous lung cancer mouse model. After treatment of drugs, mice were sacrificed and the lungs were harvested and imaged (figure 7A,B). We found that while STP3725 monotherapy mediated negligible changes compared with control, the combination with αPD-1 antibody substantially reduced nodule formation, lung weight and tumor area fractions in the H&E slides (figure 7C–F). The MSB staining showed that STP3725 significantly attenuated clot incidence in both intratumoral and non-tumoral areas of the lung tissue (figure 7G,H). In conclusion, the treatment of STP3725 considerably diminished the thrombotic incidences and the combination of STP3725 and αPD-1 antibody substantially attenuated spontaneous lung cancer development in this model. Based on these results, we postulate that the potentiation of anticancer effects of αPD-1 antibody in this K-Ras mutant model is also attributable to the improved tumor immune-microenvironment resulted from reduced hypoxia and incidence of CAT.