A549, Raw 264.7, B16-F10, L929, SVEC4-10, and MS1 murine endothelial cells were purchased from the American Type Culture Collection (ATCC). A549 cells were cultured in RPMI-1640 (Wisent, St-Bruno, QC, Canada) supplemented with 10% FBS. RAW 264.7, B16-F10, L929, SVEC4-10, and MS1 cells were maintained in DMEM supplemented with 10% FBS. All cells were cultured under the condition of 37°C and 5% CO₂ in the cell incubator (Thermo Fisher Scientific). The recombinant murine M-CSF and IL-4 were purchased from Sino Biological Inc (Beijing, China). The recombinant murine IL-13 was purchased from PeproTech (Rocky Hill, NJ, United States). Antibodies such as p38, p-p38, Erk1/2, p-Erk1/2 and Hsp90 were from Santa Cruz Biotechnology (Santa Cruz, CA, United States). Akt, p-Akt, 4E-BP1, p-4E-BP1, and Arginase-1 antibodies were from Cell Signaling Technology (Beverley, MA, United States). Integrin α5 and nucleolin antibodies were from Abcam (Cambridge, United Kingdom). Rabbit polyclone antibody for endostatin was from Oncogene Research Products (Cambridge, MA, United States). CD31 and F4/80 antibodies were from BD Bioscience (San Jose, CA, United States). CD11b-FITC and CD11b-PE-Cy7 antibodies for flow cytometry were from eBioscience (San Diego, CA, United States). F4/80-FITC and CD206-APC antibodies were from Biolegend (San Diego, CA, United States). HIF-1α antibody was from GeneTex (Irvine, TX, United States). FITC-linked anti-rat, FITC-linked anti-rabbit, TRITC-linked anti-mouse and Dylight 649-linked anti-rabbit IgG antibodies were from Beijing Cowin Biotech (Beijing, China).

Cells were fixed with 4% paraformaldehyde and washed with ice-cold PBS twice. For tumor tissues, slides of cryo-section were fixed in ice-cold acetone and washed in PBS. Then these samples were blocked with 10% goat serum and stained with primary antibodies. At last, fluorescein-conjugated secondary antibodies were added and fluorescent images were captured by Nikon A1 laser scanning confocal microscope. These images were then analyzed with Nikon image software NIS-Elements AR 3.0.

Bone marrow-derived macrophages (BMDMs) were treated with 20 and 40 μg/mL E-M for 24 h. Then these cells were harvested and washed in cold PBS. After centrifugation, 100 μL binding buffer was added to these cells. Then 5 μL Annexin V-FITC and 5 μL PI staining solution (Biotool, China) were added to the cell suspension. These cells were incubated for 15 min at room temperature. At last, 400 μL binding buffer was added to the cell suspension and cell apoptosis was analyzed by BD Calibur (BD Biosciences, San Jose, CA, United States).

Total RNA from cells was isolated with TRIzol (Life Technologies, Carlsbad, CA, United States) and subjected to cDNA synthesis with the First Strand cDNA Synthesis Kit (Fermentas, Hanover, MD, United States). Quantitative RT-PCR (qRT-PCR) was used to analyze mRNA expression levels. It was performed with 2×TransStart Green qPCR SuperMix (TransGen Biotech, Beijing, China). The reaction was run on the Mx3000P system (Stratagene, La Jolla, CA, United States). GAPDH or 18S was used as the control. Relative quantification was analyzed with the ΔΔCt method. The primers for qRT-PCR were listed in Supplementary Table 1.

Western blot was performed following the previous report (Ding et al., 2010). In brief, cells were harvested, boiled, and subjected to 10% or 12% SDS-PAGE. Proteins in the gel were transferred to the PVDF membrane (Millipore, Bedford, MA, United States) and the membrane was blocked in 5% fat-reduced milk and incubated with primary antibodies overnight at 4°C. Then the membrane was incubated with HRP-conjugated goat anti-rabbit or goat anti-mouse secondary antibodies for 1 h at room temperature. At last, proteins on the membrane were detected with enhanced chemiluminescence system (Pierce Biotechnology, Radford, IL, United States) following the manufacturer’s protocol.

4 × 10³ cells were seeded in each well of 96-well plate. Cells were treated with E-M for 72 h. After the treatment, 10 μL MTT solution (Beyotime Biotechnology, China) in 100 μL DMEM was added to each well and the whole plate was incubated at 37°C for 4 h. Then 100 μL formazan solution was added to each well and incubated for another 4 h. At last, the absorbance of each well at 570 nm (OD₅₇₀) was measured under Varioskan Flash (Thermo Fisher Scientific).

1 × 10⁶ cells in 100 μL PBS containing 3% FBS were stained with primary antibodies on ice for 30 min and washed with ice-cold PBS once. Then these cells were stained with fluorescein-conjugated secondary antibodies for 30 min and washed with ice-cold PBS twice. For fluorescein-labeled primary antibodies, cells were just stained with these antibodies and washed with PBS once. At last, 500 μL PBS was added to these cells and cells were analyzed with FACSAria III (BD Biosciences).

Raw 264.7 cells or BMDMs were pre-treated with endostatin or different ATPase mutants for 1 h and 1 × 10⁵ cells were seeded in the upper chamber of 5 μm Millicell (Millipore). Recombinant proteins including WT endostatin, E-M, E176A, K96A, and K96R were added to the medium containing concentrated A549 conditioned medium (CM) or 2% FBS in the lower chamber. Cells were allowed to migrate for 6 h at 37°C. Then cells were fixed with 4% paraformaldehyde and stained with crystal violet (Sigma–Aldrich, St. Louis, MO, United States). Migrated cells were counted randomly in five independent fields under the Olympus IX71 optical microscope.

SVEC4-10 cells were plated in the 6-well plate and allowed to grow to confluence. Then 10 μL pipette tips were used to scratch lines. BMDM CM from different treatment groups was added to these wells and cells were incubated at 37°C for 6 h. Images were taken with optical microscope and five independent were analyzed with Image-Pro Plus 6.0 software (Media Cybernetics, Silver Spring, MD, United States).

The tubule formation assay was performed as previously described (Huang et al., 2006). In brief, 100 μL Matrigel (Corning, NY, United States) was added to wells of 48-well plate and the plate was incubated at 37°C for 30 min. Then 3 × 10⁴ MS1 cells in the medium containing 2% FBS were added to each well and cultured at 37°C. After 3.5 h, tubule formation was observed and five independent fields were captured under the optical microscope. The tubule length was analyzed with Image-Pro Plus 6.0 software.

Cells in the tibia and femur of C57BL/6 mice were flushed out with 10 mL DMEM. After red blood cells lysis, myeloid cells were incubated in medium containing 10% FBS and 100 ng/mL M-CSF or L929 CM. At the 4th day, fresh medium was added to these cells and incubated for another 3 days.

All animal experiments were approved by the Institutional Animal Care and Use Committees of Tsinghua University (Approval No.14-LYZ4). 1 × 10⁶ A549 cells were inoculated subcutaneously into the right axilla of female BALB/c nude mice (Vital River, China). When these tumors grew to 100 mm³, mice were separated randomly into two groups (n = 5 mice/group). For the E-M treatment group, 12 mg/kg E-M was administered to mice intravenously every other day and the treatment was lasted for 12 days. For liposome treatment group, 200 μL PBS or clodronate liposomes were injected to mice (n = 5 mice/group) 6 days before the tumor implantation. The injection was given every 3 days and tumors were allowed to grow for 12 days. Tumor growth was monitored and tumor volumes were calculated by the formula: volume = 0.52ab² (a represents the long diameter and b indicates the short diameter).

PBS and clodronate liposomes were prepared following the previous report (Van Rooijen and Sanders, 1994). Under the protection of argon, clodronate was encapsulated in liposomes that consisted of phosphatidylcholine (Lipoid, Germany) and cholesterol (Sigma–Aldrich).

All experimental data were presented as mean ± SD or SEM. A two-tailed Student’s t-test was used to assess the statistical differences for two-group comparisons. For comparing three or more groups, an one-way analysis of variance (ANOVA) was applied to assess the statistical differences. P < 0.05 was regarded to be significant.

As the integrin α5β1-nucleolin-uPAR co-receptor complex was identified to mediate endostatin internalization and nuclear translocation of endostatin in endothelial cells (Rehn et al., 2001; Shi et al., 2007; Song et al., 2012), we examined the expression of nucleolin, integrin α5 and uPAR in BMDMs and Raw 264.7 cells, and found that all these proteins were expressed on the cell surface (Figure 1A and Supplementary Figure 1A). Flow cytometric analysis confirmed these results (Figure 1B and Supplementary Figure 1B). We also isolated TAMs from A549 tumor tissues and found that both nucleolin and integrin α5 were expressed on the surface of TAMs (Supplementary Figure 1C). To test whether ATPase activity is crucial for the bioactivities of endostatin on macrophages, we treated BMDMs and Raw 264.7 cells with WT endostatin, E-M and other mutants with lower ATPase activity (K176A, K96A, and K96R) in in vitro migration assay. Compared to endostatin, E-M exhibited a much stronger inhibitory effect on macrophage migration, whereas these low-ATPase mutants had no or only minor effects on macrophage migration (Figures 1C,D and Supplementary Figures 1D,E). Therefore for the rest of this study we mainly focused on the recombinant endostatin derivative E-M. As no pronounced changes of the tertiary structure happened in E-M (Wang S. et al., 2015), we assumed that nucleolin and integrin α5β1 could interact with E-M and they were also E-M receptors. The immunoprecipitation result showed that nucleolin and integrin α5 both directly interacted with E-M in BMDMs (Figure 1E). We also linked E-M to CNBr-activated Sepharose 4B column and incubated it with BMDM lysates. After elution, nucleolin and integrin α5 were detected in the eluted solution which further confirmed the interaction between E-M and nucleolin or integrin α5 (Figure 1F). Taken together, the ATPase activity is necessary for the biofunctions of endostatin on macrophage functions, and nucleolin and integrin α5β1 expressed on macrophage surface were also E-M receptors.

Previous studies showed that internalization of endostatin via its receptors was essential for its biological functions in endothelial cells and adipocytes (Dixelius et al., 2000; Wang H. et al., 2015). We hypothesized that E-M could also be internalized into macrophages. After treating BMDMs and Raw 264.7 cells with E-M, we found that E-M was uptaken by these cells (Figure 2A). Flow cytometric results showed that rhodamine-conjugated E-M (Rh-E-M) was internalized into BMDMs and Raw264.7 cells in a time-dependent manner (Figure 2B and Supplementary Figure 2A). After acid buffer washing, almost no E-M existed on the cell membrane (Figure 2C and Supplementary Figures 2B–D). After blockade of nucleolin and integrin α5 with antibodies, the internalization of E-M was significantly inhibited in BMDMs (Figures 2D,E). Our group previously reported that blockade of caveolae/lipid raft pathway accelerated endostatin internalization through the more efficient clathrin pathway (Chen et al., 2011). Similar results were obtained in BMDMs. The internalization of E-M was switched to clathrin pathway after nystatin treatment and the internalization was enhanced. However, blockade of clathrin pathway with chlorpromazine (CPZ) reversed this effect (Figures 2F,G and Supplementary Figure 2E). TAMs express genes typical of M2 macrophages and closely resemble M2 type macrophages. Our findings observed that after the treatment with A549 CM, the amount of nucleolin and integrin α5 on the surface of M2 type BMDMs or Raw 264.7 cells was increased (Figures 2H,I and Supplementary Figure 2F). The internalization of E-M was also enhanced (Figure 2J and Supplementary Figures 2G,H). Moreover, we induced M2 macrophages with IL-4 and IL-13 and similar results were obtained in these M2 macrophages (Figure 2K and Supplementary Figure 2I). In conclusion, E-M is internalized into macrophages via nucleolin and integrin α5β1 in caveolin- and clathrin-dependent pathways, and up-regulation of these two receptors on the surface of M2 macrophages enhances the internalization of E-M.

In malignant tumor tissues, the recruitment of TAMs is driven by chemotactic factors mainly derived from tumor cells. Therefore we induced macrophage motility with A549 CM in the cell migration assay and found that enhanced cell migration was significantly suppressed by E-M (Figures 3A,B). In the Matrigel plug assay, A549 CM and B16-F10 CM recruited a large number of F4/80⁺ macrophages into plugs but E-M blocked the macrophage recruitment (Figures 3C–E and Supplementary Figure 3A). Moreover, the decreased recruitment was not due to cell apoptosis induced by E-M (Supplementary Figure 3B). Activation of Erk1/2 and p38 MAP kinase pathways reportedly mediates the chemotactic migration in macrophages (Chiou et al., 2004). We found that activation of p38 and Erk1/2 in BMDMs and Raw 264.7 cells were both inhibited after E-M treatment (Figure 3F and Supplementary Figure 3C). To further confirm that activation of p38 and Erk1/2 mediated the BMDM migration process induced by A549 CM, we treated BMDMs with inhibitors SB203580 and U0126 in the presence of A549 CM. Both inhibitors remarkably suppressed BMDM migration induced by A549 CM (Figure 3G and Supplementary Figure 3D). In summary, E-M blocks the recruitment of macrophages both in vitro and in vivo through inhibiting the activation of p38 MAP kinase and Erk1/2 signaling pathways.

To induce M2-polarized macrophages, we treated BMDMs with A549 CM and observed that the expression of M2 markers, including Mgl1, Fizz1, Mr, Ccl17, and ARG-1, were up-regulated. However, pre-treatment with E-M prevented the up-regulation of these markers (Figures 4A,B). During the polarization switch process, STAT3 and STAT6 were selectively activated in macrophages (Sica and Bronte, 2007). However, E-M could block the activation of STAT3 and STAT6 in BMDMs (Figure 4B). Compared to WT endostatin, E-M was also found to have stronger inhibitory effects on M2 polarization (Supplementary Figure 4A). In BMDM-A549 (GFP⁺) cell co-culture system, Arg-1 was also up-regulated in BMDMs but E-M reversed this effect (Figure 4C). Moreover, BMDMs enhanced the proliferation of A549 cells after 12 h in the co-culture system but the proliferation rate was reduced in the presence of E-M (Figures 4D,E). To exclude the direct influence on A549 cells, MTT assay was used to examine the effects of E-M on A549 cells and we found that E-M had no effects on A549 cell proliferation (Supplementary Figure 4B). Our findings also discovered that IL-4-induced M2 polarization was partially blocked by E-M (Figure 4F). Moreover, B16-F10 CM induced alternative activation of BMDMs was suppressed by E-M (Figures 4G,H). In conclusion, E-M can indeed inhibit the alternative activation of macrophages and M2 macrophage-induced tumor cell proliferation.

Tumor-associated macrophages are well-known to promote tumor angiogenesis through production of different pro-angiogenic factors (Murdoch et al., 2008; Ruffell and Coussens, 2015). To detect whether E-M could inhibit pro-angiogenic effects of TAMs, we treated A549 CM-stimulated BMDMs with E-M. Several well-known pro-angiogenic factors such as VEGF-A, PDGF-B and placental growth factor (PlGF) were up-regulated after A549 CM treatment, whereas E-M blocked this effect (Figure 5A). After A549 CM treatment, we collected the BMDM CM in different treatment groups and treated endothelial cells with these CM. The tube formation ability of MS1 cells was enhanced in A549 CM pre-treatment group but this effect was abolished in the E-M pre-treatment group (Figures 5B,C). Moreover, consistent results were obtained in endothelial cell migration and wound healing assays. BMDM CM from A549 CM pre-treatment group enhanced the motility of endothelial cells but CM from E-M pre-treatment group had no such effect (Figures 5D–G). It was reported before that the transcription factor HIF-1α participated in regulation of pro-angiogenic factors production (Liu et al., 2008; Weidemann and Johnson, 2008). Our findings found that the expression of HIF-1α was up-regulated in mRNA and protein levels by A549 CM but it was suppressed by E-M (Figures 5H,I). PI3K/Akt and mTOR signaling pathways were known to regulate the HIF-1α expression in endothelial cells (Song et al., 2009). Then we detected whether these two pathways were also activated in BMDMs after A549 CM treatment. We found that the phosphorylation of Akt and 4E-BP1, the downstream target of mTOR, was increased after A549 CM treatment but the activation was blocked by E-M (Figure 5J). After the blockade of activation of PI3K/Akt and mTOR with their respective inhibitors LY294002 and KU0063794, the expression of HIF-1α induced by A549 CM in BMDMs was significantly down-regulated (Figure 5K). Taken together, E-M can abolish the pro-angiogenic effects of tumor cell-induced TAMs.

To examine whether E-M could inhibit the recruitment of macrophages in vivo, we firstly implanted A549 cells into mice subcutaneously and treated mice with PBS and E-M. Six days after implantation, tumor volumes in both groups displayed no significant differences so impacts of tumor volume differences were excluded (Figure 6A). However, macrophage recruitment was notably inhibited in E-M-treated group, accompanied by decreased tumor angiogenesis (Figures 6B–E). Moreover, when the tumor volumes reached 100 mm³, we treated A549 tumor-bearing mice with E-M for 12 days. After the treatment, the tumor growth was significantly inhibited by E-M (Figures 6F,G). And we observed that the recruitment of macrophages was remarkably blocked after E-M treatment (Figures 6H,I). However, the density of macrophages in other organs including liver, spleen, kidney and colon from E-M treated tumor-bearing mice was not affected (Supplementary Figure 5A). Tumor angiogenesis was also significantly inhibited in E-M treatment group compared to PBS treatment group (Figures 6J,K). We also confirmed these results in a syngeneic tumor model. We treated B16-F10-bearing mice with E-M and got the similar results. E-M inhibited B16-F10 tumor growth and the density of intratumoral macrophages and blood vessels was significantly decreased by E-M (Supplementary Figures 5B–G). To elucidate the importance of TAMs in this process, we depleted macrophages in tumor-bearing mice with clodronate liposomes (Supplementary Figure 5H). After macrophages depletion, tumor growth and angiogenesis were both remarkably inhibited (Supplementary Figures 5I–L). In summary, E-M exhibited its inhibitory effects on tumor growth via suppression of macrophage recruitment and tumor angiogenesis.