Monocyte Induction of E-Selectin–Mediated Endothelial Activation Releases VE-Cadherin Junctions to Promote Tumor Cell Extravasation in the Metastasis Cascade

Hematogenous metastasis is a multistep process in which diverse interactions between tumor cells and their microenvironment allow the malignant cells to cross physical boundaries, disseminate, and colonize distant organs. Specifically, cell–cell interactions between tumor cells and blood constituents, such as platelets, leukocytes, and endothelial cells, are initially mediated by selectins at different steps of the metastatic cascade (1–3).

Selectins are vascular cell adhesion receptors that are responsible for initial rolling and attachment of leukocytes to the endothelium, enabling leukocyte homeostasis (2, 4). Selectins bind to sialylated and fucosylated lactosamine terminal glycan structures displayed on leukocytes, platelets, and endothelium, or tumor cells. It is accepted that malignant transformation is associated with altered carbohydrate structure presentation on tumor cells, which are potential ligands for selectins, and correlates with poor prognosis due to metastasis (5, 6). E-selectin is the major leukocyte adhesion receptor that is present only on endothelial cells upon endothelial activation and requires de novo expression. E-selectin has been investigated as the primary receptor mediating tumor cell metastasis through facilitating adhesion of tumor cells on the endothelium shown in vitro (1, 7, 8). In addition, E-selectin upregulation was observed during metastatic liver colonization (9, 10). E-selectin was detected in the tumor cell microenvironment several hours after their arrest, indicating an inflammatory-like endothelial activation (11–13). E-selectin expression in the premetastatic lungs correlated with increased tumor cell homing to these tissues and with enhanced recruitment of myeloid cells (14). Although there is accumulating evidence that selectins contribute to the metastatic microenvironment (12, 15), the mechanism of E-selectin contribution to cancer progression requires further in vivo studies.

The endothelium in blood vessels controls the extravasation of cells, for example, leukocytes, and the egress of soluble factors from the plasma. Most leukocytes extravasate from circulation through the paracellular route by opening endothelial junctions at sites of inflammation (16, 17). The maintenance of the vascular barrier function is dependent on the stability of endothelial adherence junctions mediated by vascular endothelial cadherin, VE-cadherin (18). Phosphorylation of Tyr residues of VE-cadherin regulates vascular permeability and the capacity of leukocytes to transmigrate through the endothelium (19, 20). Breast cancer cells were shown to induce disruption of endothelial adherence junctions by inducing phosphorylation of VE-cadherin (21). Yet, tumor cell extravasation is significantly promoted by myeloid cells recruited to the metastatic sites through chemokines such as CCL2 and CCL5 (11, 15, 22–26), albeit the cellular and molecular mechanism of this process remains unclear.

As E-selectin expression upon tumor cell injection has been frequently observed, the involvement of E-selectin in metastasis could be anticipated. The current study describes the mechanism of E-selectin–dependent recruitment and activation of monocytes, which drives the dissociation of VE-cadherin junctions and thereby promotes tumor cell extravasation required for metastasis.

Mouse colon carcinoma cell line, MC-38 was originally provided by Dr. J. Schlom (NIH, Bethesda, MD). MC-38 cells stably expressing GFP (MC-38GFP) were characterized as described previously (27). B16-BL6 melanoma cells provided by Dr. I. Vlodavsky (Technion Haifa, Israel) were grown in DMEM/10% FCS as described previously (24). Lewis lung carcinoma cells (3LL) were grown in RPMI/10% FCS (12). All cells were kept at low passages and were not further authenticated. Lewis lung carcinoma-LLC1 cells (ATCC) were grown in DMEM/10% FCS. Endothelial cells bEnd.3 (ATCC) were grown in DMEM/10% FCS.

Animal experiments were performed according to the guidelines of the Swiss Animal Protection Law, and approved by Veterinary Office of Kanton Zurich. C57BL/6, Ccl2-deficient (Ccl2^−/−), and E-selectin–deficient mice (E-sel^−/−) were purchased from The Jackson Laboratory. Fucosyltransferase 7–deficient mice (Fuc-TVII^−/−) and Fucosyltransferase 4 and 7 double deficient mice (Fuc-TIV/VII^−/−) were used as described in the previous studies (15).

MC-38GFP cells (300,000 cells) were intravenously injected into the tail vein and metastatic foci analyzed on day 28. 3LL and B16-BL6 cells (150,000 cells) were intravenously injected and lungs analyzed on day 14. LLC1 cells (200,000 cells) were subcutaneously injected into the right flank and primary tumors were removed 18 days later. Mice were terminated at day 30 and lung metastasis was analyzed.

Mouse lungs perfused with PBS were minced and digested with Collagenase D and A (1 mg/mL each, Roche) in 2 mL for 1 hour at 37°C. Single-cell suspension using 40-μm cell strainers was prepared; erythrocytes were lysed with PharmLyse (BD Biosciences; ref. 15). Cells were incubated with anti-mouse CD16/32 mAb and stained with directly labeled antibodies against CD45, CD11b, F4/80, Ly6G, Ly6C, and CD31 (BD Biosciences). Blood samples were treated with PharmLyse and stained as described above. Data were acquired on a FACSCanto II (BD Biosciences) and analyzed using Flow Jo software (Tree Star).

Cryosections (6 μm) were stained with following antibodies: CD11b-biotin, Ly6G, CD62E (BD Biosciences), and F4/80 (AbD Serotec). Goat anti-rat-Alexa 568 Ab or Streptavidin-Alexa 647 (Life Technologies) were used for detection using fluorescence microscope (Zeiss). The percentage of tumor cells associated with leukocytes was determined. The analysis of selectin and myeloid cell detection in lung sections was performed with a SP5 confocal microscope (Leica). Images were acquired of a total of 5-μm stacks and analyzed with Imaris Software (Bitplane).

Pulmonary endothelial cells were isolated using a positive immunomagnetic selection as described previously (24). Femur and tibia were crushed in PBS containing 2% FCS and 2.5 mmol/L EDTA. After red blood cell lysis, monocytes were enriched by magnetic-activated cell sorting using biotinylated anti-CD115 (M-CSFR) antibody (Biolegend) and streptavidin-conjugated magnetic beads (Miltenyi Biotec).

Primary lung microvascular cells (175,000 cells) were plated on gelatin-coated μ-Slide I^0.2 Luer (ibidi GmbH) and grown to confluence for 2 days. The endothelial monolayer was manually perfused five times with 2 × 10⁵ of MC-38GFP cells in 100-μL media. After 4 hours, slides were perfused with CellTrace calcein-red-orange, AM (Life Technologies) stained bone marrow monocytes (2 × 10⁶ cells/mL) in HEPES/Ringer solution supplemented with 25% washed red blood cells. Using air pressure pump system (Ibidi), we applied a flow rate of 2 dyne/cm². Mosaic images of slides were acquired using an inverted fluorescence microscope (Zeiss Axio Observer Z.1). Pictures were analyzed with ZEN software (Zeiss).

Vascular permeability in the lungs was determined with Evans blue assay (24). Briefly, 24 hours after intravenous injection of tumor cells, 2 mg of Evans blue was intravenously injected and mice were terminated 30 minutes later. Monocyte depletion with intravenously injected clodronate liposomes (1.8 mg; ref. 28) 24 hours prior to tumor cell intravenous injection was followed by Evans blue injection.

Total RNA from perfused and snap-frozen lungs was isolated using TRI Reagent (Sigma-Aldrich) and reversely transcribed into cDNA using Omniscript RT Kit (Qiagen). Real-time PCR was performed with the SYBR Green JumpStart Taq ReadyMix (Sigma-Aldrich) using the MX300P light cycler (Agilent). Intron-spanning primers were used (Supplementary Table S1) and data normalized to GAPDH.

PBS-perfused lungs were minced and digested in Collagenase D and A (1 mg/mL each, Roche) for 1 hour at 37°C. A single-cell suspension was prepared by passing the digested lungs through 40-μm cell strainers (BD Biosciences). Cells were incubated with antibodies against CD31, CD45, CD11b, Ly6C, and Ly6G (eBiosciences). Endothelial cells (CD45⁻CD11b⁻CD31⁺) and inflammatory monocytes (CD45⁺CD11b⁺Ly6C^highLy6G⁻) were sorted with a FACSAria III sorter (BD Biosciences). Total RNA was isolated using the RNeasy Mini Kit (Qiagen).

Perfused lungs were homogenized and concentration of the chemokine CCL2 was assessed in the supernatant using the cytometric bead array for mouse CCL2 (BD Biosciences). CCL2 levels were normalized to the total protein amount.

Transendothelial migration assay was performed as described previously (24). Briefly, primary lung endothelial cells (25,000) were seeded on gelatin-coated 24-well transwell inserts (8-μm pores; BD Biosciences) and were grown to confluence. MC-38GFP cells (25,000) were seeded into transwell inserts with/without monocytes (100,000) purified from bone marrow (untreated or PFA-fixed) in 3% FCS/RPMI and 10% FCS/RPMI was added into the bottom chamber.

Endothelial F-actin was detected with phalloidin staining (29). Briefly, 40,000 pulmonary endothelial cells grown on gelatin-coated chamber slides for 36 hours were coincubated with MC-38GFP cells (40,000) stained with a PKH26 red fluorescent dye (Sigma-Aldrich) and bone marrow monocytes (200,000 cells) for 8 hours. Cells were fixed with 2% PFA, permeabilized (0.1% saponin), and stained with Phalloidin-FITC (6.67 mg/mL, Sigma-Aldrich). Nuclei were stained with DAPI and mounted in ProLong Gold (Life Technologies).

Confluent bEnd.3 cells were stimulated with 40 ng/mL rmIL-1β (R&D Systems) for 2 hours. Freshly isolated primary CD115⁺ monocytes from C57BL/6 or FucTIV/VII^−/− double deficient mice bone marrow (1 × 10⁷) with or without MC-38GFP tumor cells (5 × 10⁶) in a T75 flask were cocultured with activated bEnd.3 cells for 1 hour at 37°C. Monocytes were removed by washing with prewarmed PBS (5 times). Alternatively, the cells were incubated with 10 μg/mL anti-E-selectin antibody (BD Biosciences) for 1 hour at 37°C. After washing with prewarmed PBS, anti-E-selectin Ab crosslinking was induced with an addition of goat-anti-rat polyclonal IgG for 15 minutes at 37°C. bEnd.3 cells were lysed with 20 mmol/L Tris pH 7.5, 150 mmol/L NaCl, 1 mmol/L EDTA, 1 % Triton X-100, 2% phosphatase inhibitors, and 20× Complete EDTA-free protease inhibitors (Roche). Lysates were precleared with Protein G Mag beads (GE Healthcare) for 1 hour at 4°C and incubated overnight at 4°C with Protein G Mag beads precomplexed with 5 μg/sample VE-cadherin antibody (Abcam). Beads were washed with PBS (3 times) and boiled in Laemmli buffer and supernatants separated on 7.5 % SDS-PAGE gel. After transfer to a nitrocellulose membrane and blocking with 5% milk in Tris-buffered saline (TBS, pH 7.5), the membrane was incubated with mAb to phosphorylated Tyr731 of VE-cadherin (a generous gift from Dietmar Vestweber; ref. 19) diluted in 5% BSA in TBS overnight or with VE-cadherin antibody. After incubation with secondary antibody, the proteins were detected with ECL method (GE Healthcare).

A glass bottom dish (Mattek) was treated with 0.01% poly-l-lysine for 20 minutes at room temperature, washed with water, and treated with 0.25% glutaraldehyde for 15 minutes at room temperature. The dish was washed with DMEM/20% FCS (3 times) and put at 4°C for 30 minutes. Ice-cold collagen type I (Sigma) was mixed with EC medium and PBS, loaded on the glass dish, and incubated for 1 hour at 37°C. Primary lung endothelial cells (3.7 × 10⁵) were seeded on the collagen matrix and stained with CellTracker Red CMTPX Dye (Life Technologies) 24 hours later. MC-38GFP cells (2.5 × 10⁵) and CellTracker Deep Red Dye–stained CD115⁺ bone marrow–derived monocytes (1 × 10⁶) were added and the imaging was performed on a Leica SP5 confocal microscope. Data were analyzed using the Imaris 7.3.1 software (Bitplane).

Statistical analysis was performed with the GraphPad Prism software (version 5.01). All data were analyzed by ANOVA with the post hoc Bonferroni multiple comparison test and are presented as mean ± SEM. Analysis of two samples was performed with Mann–Whitney test unless stated otherwise.

It has been postulated that E-selectin binding to tumor cells facilitates tumor cell lodging in the microvasculature and thereby metastasis (1, 30). To assess whether a direct interaction of tumor cells with E-selectin is required for metastasis, we tested mouse colon carcinoma cells (MC-38GFP) in an experimental metastasis model. MC-38GFP cells express no E-selectin ligands while having P- and L-selectin ligands (Supplementary Fig. S1A). Intravenous injection of MC-38GFP into C57BL/6 and E-selectin–deficient (E-sel^−/−) mice revealed a significant reduction in the number of pulmonary metastatic foci and total tumor burden in E-sel^−/− mice compared with C57BL/6 mice (Fig. 1A and B). The intravenous injection of Lewis lung carcinoma cells (3LL) and melanoma cells (B16-BL6), both without E-selectin ligands (Supplementary Fig. S1A), also resulted in reduced metastasis in E-sel^−/− mice (Supplementary Fig. S1B–S1E). As in the experimental metastasis model, tumor cells are present in the lungs prior to endothelial activation (12), and the fact that E-selectin is expressed only on activated endothelium, these findings indicate that E-selectin promotes metastasis without directly interacting with tumor cells.

Next, we tested the role of E-selectin in a spontaneous lung metastasis model using Lewis lung carcinoma cells (LLC1) that were subcutaneously injected. LLC1 cells do not have any E-selectin ligands (Supplementary Fig. S1A). Lung metastasis was significantly decreased in E-sel^−/− mice compared with C57BL/6 mice (Fig. 1C and D).

To assess whether E-selectin facilitates metastasis through recruitment of leukocytes, we analyzed lungs of mice injected with tumor cells intravenously that were terminated 24 and 48 hours after tumor cell injection (p.i.) by flow cytometry (Supplementary Fig. S2A-C). Total leukocyte infiltration was significantly increased in C57BL/6 mice 48 hours p.i. but remained unchanged in E-sel^−/− mice. Specifically, a significant increase in the number of inflammatory monocytes (CD11b⁺F4/80⁻Ly6G⁻Ly6C^hi) and macrophages (CD11b⁺F4/80⁺) in C57BL/6 mice compared with E-sel^−/− mice (Fig. 2A and B) was observed. No differences in granulocyte levels (CD11b⁺F4/80⁻Ly6C^medLy6G⁺) were detected (Fig. 2C). The reduced numbers of macrophages at 16 and 24 hours p.i. in E-sel^−/− mice compared with C57BL/6 mice (Fig. 2D and Supplementary Fig. S2D) was further confirmed by IHC. We detected a transient increase (16 hours p.i.) of granulocytes in C57BL/6 compared with E-sel^−/− mice (Fig. 2E). The analysis of peripheral blood cells from naïve mice showed no difference in numbers of CD11b⁺, F4/80⁺, and Ly6G⁺ cells between E-sel^−/− and C57BL/6 mice (Supplementary Fig. S2E).

The observed difference in leukocyte infiltration to the lungs indicates that E-selectin promotes the early phase of metastasis. We detected maximal E-selectin expression in the lungs 6 hours p.i., which was reduced by 12 hours p.i. (Fig. 3A). E-selectin was detected only in the vicinity of tumor cells 6 and 14 hours p.i. in lungs of C57BL/6 mice as determined by IHC (Fig. 3B). Importantly, we detected myeloid cells (CD11b⁺) mostly in the E-selectin–positive areas (Fig. 3C). To assess whether E-selectin facilitates leukocyte recruitment to metastatic tumor cells we analyzed the tumor cell–leukocyte association. We observed a significant reduction in both F4/80⁺ and Ly6G⁺ cell interactions at 16 hours p.i. in E-sel^−/− mice compared with C57BL/6 mice, and only F4/80⁺ cell interactions remained significantly different also 24 hours p.i. (Fig. 3D). To confirm that E-selectin contributes to the leukocyte recruitment, we analyzed this process using a microfluidic system in vitro (Supplementary Fig. S3A). We determined the monocyte recruitment to MC-38GFP cells adherent on primary lung endothelial monolayers derived from C57BL/6 and E-sel^−/− mice under physiologic post-capillary flow conditions. Less monocytes were adherent to endothelial cells around tumor cells on E-sel^−/−–derived compared with C57BL/6–derived endothelial monolayers (Fig. 3E and F). In addition, the number of adherent monocytes in the vicinity of tumor cells was also reduced on E-sel^−/− endothelial monolayers (Supplementary Fig. S3B). These findings indicate that E-selectin promotes the recruitment of monocytes to metastasizing tumor cells.

E-selectin is an established marker of endothelial activation that is observed both in inflammatory and cancer-related situations (2, 4). Hence, we analyzed the activation status of the lung endothelium in response to tumor cell injection. Expression levels of the activation markers vascular adhesion molecule-1 (VCAM-1) and intercellular adhesion molecule-1 (ICAM-1) were increased 6 and 12 hours p.i. in lungs of C57BL/6 mice but remained unchanged in lungs of E-sel^−/− mice compared with untreated mice (Fig. 4A and Supplementary Fig. S4A). In addition, we observed significant increase in expression of a CCL2 chemokine, and its receptor CCR2 in the lungs of C57BL/6 mice 6, 12, and 24 hours p.i., but no changes were detected in E-sel^−/− mice. CCL2 is produced both by MC-38GFP cells and by stromal cells and the CCL2/CCR2 axis has been associated with promotion of metastasis (22–24). Of note, neither LLC1 nor MC-38GFP cells express CCR2 receptor (31) thereby excluding any direct effect of CCL2 on tumor cells.

Next, we assessed the role of tumor cell–derived CCL2 on the metastatic microenvironment using MC-38GFP cells expressing reduced amounts of CCL2 (MC-38GFP CCL2KD) and compared with parental MC-38GFP cells 6 hours p.i. (Supplementary Fig. S4B). Significantly reduced endothelial activation was observed in mice injected with MC-38GFP CCL2KD as measured by E-selectin and VCAM-1 expression; and reduced expression of CCL2 and CCR2 (Fig. 4B and Supplementary Fig. S4B), indicating a direct involvement of CCL2 in endothelial activation.

To test how tumor cell injection changes the CCL2 levels within the metastatic lungs, we analyzed CCL2 protein concentrations in the lungs of C57BL/6, E-sel^−/−, and Ccl2-deficient mice (Ccl2^−/−) 6 and 12 hours p.i. Compared with untreated mice (3 pg/mg), CCL2 protein levels were significantly elevated in C57BL/6 mice (27 pg/mg) while a reduced increase in lungs of E-sel^−/− mice (11 pg/mg) was detected (Fig. 4C). The concentration of CCL2 remained high in the lungs of C57BL/6 mice (8 pg/mg) also 12 hours p.i. (Supplementary Fig. S4C). In lungs of Ccl2^−/− mice we observed minimal amounts of CCL2 (2 pg/mg), which likely corresponded to the tumor cell–derived CCL2. These findings suggest that the majority of CCL2 in metastatic lungs is derived from the local microenvironment. To address the source of CCL2 in the metastatic tissue, we sorted endothelial (CD31⁺) and myeloid (CD11b⁺/Ly6C^high) cells from lungs of mice 12 hours p.i. and analyzed CCL2 expression. We observed significant increase in CCL2 expression in endothelial cells from C57BL/6 compared with untreated mice or E-sel^−/− mice (Fig. 4D). Interestingly, CCL2 expression was also increased in sorted monocytes only from C57BL/6 mice, indicating that E-selectin–mediated activation of monocytes induces CCL2 expression. Indeed, the absolute CCL2 expression levels were mostly derived from monocytes (Supplementary Fig. S4D). These data show that E-selectin–mediated activation of both endothelial cells and monocytes induces CCL2 expression and thereby contribute to the chemokine pool in the metastatic lungs.

Increased levels of CCL2 correlate with metastatic progression in various mouse models and contribute to tumor cell extravasation (23–25, 32). To test whether E-selectin expression is required for tumor cell extravasation, we analyzed lung vascular permeability using the Evans blue assay (24). Tumor cell injection induced lung vascular permeability in C57BL/6 mice, which was almost absent in E-sel^−/− mice (Fig. 5A). We hypothesized that the E-selectin–mediated recruitment and activation of monocytes contribute to vascular permeability. Depletion of circulating monocytes with clodronate liposomes 24 hours prior to MC-38GFP injection resulted in significantly decreased vascular permeability (Fig. 5B), suggesting that monocyte recruitment facilitates tumor cell extravasation by E-selectin–mediated leukocyte–endothelial activation.

Next we analyzed the capacity of monocytes to promote transendothelial migration of tumor cells (Supplementary Fig. S5A). We used lung microvascular endothelial cell monolayers derived from C57BL/6 and E-sel^−/− mice and studied the migration of MC-38GFP cells (Fig. 5C and Supplementary Fig. S5B). While tumor cells have an intrinsic ability to migrate through endothelial cells, the presence of monocytes significantly promoted this process. However, monocytes did not increase tumor cell migration through E-selectin–deficient endothelial cells, suggesting that E-selectin binding to monocytes is essential for transendothelial migration. To test this hypothesis, we used fixed monocytes, which only present ligands on their surfaces (Fig. 5D). Notably, fixed monocytes increased the tumor cell migration albeit not to the same level as unfixed cells. Next, we tested monocytes derived from Fucosyltransferase-7–deficient mice (Fuc-TVII^−/−) lacking most of E-selectin ligands (15). While Fuc-TVII^−/− monocytes only partially promoted tumor cell migration, fixed Fuc-TVII^−/− monocytes showed no effect (Fig. 5D). Thus, the interaction between endothelial E-selectin and selectin ligands on monocytes supports tumor cell transmigration, while soluble factors from monocytes further promote this process. To analyze the role of monocytes, we used a live imaging microscopy to follow the transendothelial migration of tumor cells in vitro (Fig. 5E and Supplementary Fig. S5C; Supplementary Video S1). MC-38GFP cells migrated through the endothelial cell layer in a close contact with monocytes.

Activation of selectins by ligand-binding is known to trigger “outside-in” signaling in endothelial cells and leukocytes, which induces migration of leukocytes (33, 34). To elucidate the mechanism how monocyte-E–selectin interaction contributes to tumor cell migration, we analyzed the cytoskeletal retraction of endothelial cells cocultured with tumor cells and monocytes. While the combination of tumor cells and leukocytes strongly increased the retraction of lung endothelial cells, the individual cells alone only slightly affected retraction (Fig. 6A and Supplementary Fig. S6A). However, we observed no changes in E-selectin–deficient endothelial cells after coculture with tumor cells or monocytes alone nor with the combination of both cells (Fig. 6A and Supplementary Fig. S6B).

During inflammation, the induction of vascular permeability is dependent on dephosphorylation of Tyr731 on vascular endothelial adhesion molecule VE-cadherin (19). We tested whether enhanced endothelial retraction induced by tumor cell monocytes is dependent on VE-cadherin modification. Monolayers of endothelioma bEnd.3 cells expressing E-selectin were incubated with either tumor cells, monocytes, or both cell types combined. Immunoprecipitated VE-cadherin was analyzed with an antibody specific for phosphorylated Tyr731 (p-Tyr731). While incubation of wt monocytes significantly reduced the p-Tyr731 signal, the addition of MC-38GFP cells alone had no effect (Fig. 6C). The combination of wt monocytes with MC-38GFP cells induced Tyr731 dephosphorylation of VE-cadherin to a similar extent as wt monocytes alone. Next, we used monocytes from Fucosyltransferase-4 and -7 double deficient mice (Fuc-TIV/VII^−/−, DKO), which contain no selectin ligands (35). DKO monocytes induced only partial dephosphorylation of Tyr731 either alone or with MC-38GFP cells, indicating that selectin ligands on monocytes contribute to the E-selectin–mediated signaling. To prove that the presence of selectin ligands on monocytes is sufficient to induce Tyr731 dephosphorylation of VE-cadherin, we used wt monocytes, which were fixed prior to addition to bEnd3 cells. Indeed, reduced p-Tyr731 phosphorylation was observed, albeit not to the same level as unfixed cells (Fig. 6C), indicating a contribution of monocyte-secreted factors to permeability induction. To test whether the activation of E-selectin is sufficient to induce Tyr731 dephosphorylation we applied anti-E-selectin antibody that was previously shown to activate E-selectin and thereby mimic monocyte adhesion (36, 37). Indeed, E-selectin crosslinking resulted in reduction of Tyr731 phosphorylation albeit not to the same extent as the addition of monocytes (Fig. 6D). Taken together, monocyte-assisted transendothelial migration of tumor cells involves E-selectin–induced loosening of VE-cadherin tight junctions.

The cross-talk between tumor cells and their environment is essential in all steps of the metastatic cascade (3). E-selectin has been previously studied only as a direct receptor for tumor cell–endothelial adhesion (12, 38). However, initial arrest of tumor cells is also likely caused by physical constrictions in capillaries and therefore independent of active adhesion (39). As we used tumor cells without endogenous E-selectin ligands, a direct tumor cell binding to E-selectin and thereby endothelial adhesion could be excluded. The observed reduction of monocyte recruitment to tumor cells in the lungs of E-sel^−/− mice suggest that E-selectin mediates host cell interactions within the tumor microenvironment.

Markers of endothelial activation and inflammation are often upregulated during the initial hours of intravascular tumor cell arrest in lung and liver experimental metastasis models (12, 38, 40, 41). Conversely, reduction of endothelial activation diminished myeloid cell recruitment, tumor cell survival, and metastasis (15, 41–43). In line with these observations, we observed upregulation of E-selectin shortly after tumor cell arrest in the lungs that correlated with increase in the endothelial activation markers VCAM-1 and ICAM-1 that were detected in the lung vasculature of C57BL/6 mice, but not of E-sel^−/− mice. Tumor cell–derived factors together with physical factors like shear forces may trigger endothelial expression of E-selectin (40, 44). Tumor cells with reduced CCL2 expression elicited weaker E-selectin expression and correspondingly lower local endothelial activation. This observation is in line with previous findings that endothelial CCR2 expression in the lungs triggers increased vascular permeability and thereby assists tumor cell extravasation (24). However, whether CCL2-induced endothelial activation is directly linked to E-selectin expression or whether tumor cell–derived factors are involved requires further analysis. Chemokine production has been linked to E-selectin expression only in few studies using inflammatory models (45, 46). In an atherosclerotic rat model, the expression of CCL2 and adhesion molecules, including E-selectin, were upregulated on the endothelium and associated with monocyte recruitment (45). The inhibition of p38 MAPKs in TNF-α–stimulated endothelial cells resulted in reduced expression of E-selectin and cytokines including CCL2, IL8, and IL6 (46). Our data indicate that tumor cells induce endothelial activation and E-selectin expression (Fig. 6E), which in turn contributes to enhanced presence of CCL2 in the metastatic lungs.

CCL2 is a potent regulator of monocyte recruitment to metastasizing tumor cells and increased CCL2 levels at metastatic sites strongly correlate with metastasis (23–25, 32). We observed decreased expression of CCL2 in lungs of E-sel^−/− mice compared with C57BL/6 mice and identified both monocytes and endothelial cells as the major source of CCL2 in the metastatic lungs. Stromal-derived CCL2 was previously linked to metastasis (15, 23, 25, 26). The unchanged CCL2 expression in monocytes isolated from lungs of E-sel^−/− mice is likely due to the lacking activation. Lukacs and colleagues previously reported that monocyte–HUVEC interactions trigger an increased expression of CCL2 and IL8 in monocytes, suggesting a need for cellular contacts (47). Our findings indicate that E-selectin binding is essential for increased CCL2 production by myeloid cells within the developing metastatic niche. In addition, reduced recruitment of myeloid cells to metastatic sites was observed in lungs of E-sel^−/− mice. The absence of E-selectin, together with reduced CCL2 levels, is likely impeding efficient capturing and firm adhesion of monocytes.

Leukocyte binding to E-selectin is known to induce signaling through MAPK, Erk/Src, or MLC/p38 pathways and linked to endothelial permeability (10, 17, 33). Leukocyte adhesion to endothelial cells induces E-selectin clustering and the attachment of E-selectin to actin cytoskeleton, which is partially dependent on Rho GTPase activity (36, 37). We showed that monocytes assist tumor cells to induce endothelial retraction in an E-selectin–dependent manner, while the depletion of monocytes prevented an increase in lung vascular permeability. Furthermore, we provided the first evidence that a direct E-selectin activation is linked to Tyr731 dephosphorylation of VE-cadherin. These observations are in line with previous findings that a disruption of VE-cadherin/β-catenin complex and/or formation of stress fibers is required for leukocyte migration through the endothelium (10, 17, 33). Thus, we propose that activation of pulmonary endothelial E-selectin shapes the metastatic lungs and effects VE-cadherin junctions (Fig. 6E) thereby promoting metastasis. Although the process of leukocyte crossing through the endothelial barrier is well characterized (17), the mechanism of loosening of VE-cadherin junctions remains to be defined. Accordingly, E-selectin is likely an additional factor to VCAM-1 inducing this process. While selectin ligands on tumor cells might induce efficient transendothelial migration through a direct E-selectin engagement (10), the presence of monocytes apparently promotes this process. The presented data suggest that monocyte recruitment and E-selectin activation are important regulators of pulmonary vascular permeability for tumor cell extravasation.

No potential conflicts of interest were disclosed.

Conception and design: M. Roblek, L. Borsig

Development of methodology: I. Häuselmann, D. Protsyuk, L. Knopfova, L. Borsig

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Häuselmann, M. Roblek, D. Protsyuk, V. Huck, L. Knopfova, S. Grässle, A.T. Bauer, S.W. Schneider, L. Borsig

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I. Häuselmann, M. Roblek, D. Protsyuk, V. Huck, S. Grässle, S.W. Schneider, L. Borsig

Writing, review, and/or revision of the manuscript: I. Häuselmann, M. Roblek, D. Protsyuk, A.T. Bauer, L. Borsig

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Grässle

Study supervision: I. Häuselmann, L. Borsig

The authors acknowledge the assistance of the Center for Microscopy and Image Analysis, University of Zurich for confocal microscopy experiments.

This study was supported by the SNF grant #310030-152901 (L. Borsig). This study was also supported in part by the DFG-SFB/Transregio 23 and SHENC-Research Unit FOR1543 (S.W. Schneider) and in part by National Program of Sustainability II LQ1605-MEYS CR; project FNUSA-ICRC (CZ.1.05/1.1.00/02.0123); and ICRC-ERA-HumanBridge/no. 316345 (L. Knopfova).

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.