Eotaxin (CCL11) Induces In Vivo Angiogenic Responses by Human CCR3⁺ Endothelial Cells¹

Chemokines are multifunctional mediators that can promote inflammation, enhance immune responses, and foster angiogenesis. Chemokines have a dual role in angiogenesis because some can function as angiogenic and others as angiostatic factors. Although several members of the CXC chemokine family, including IL-8 (CXCL8), neutrophil-activating peptide-2 (CXCL7), epithelial cell-derived neutrophil-activating protein-78 (CXCL5), melanoma growth-stimulatory activity-α (CXCL1) and stroma-derived factor-1α (SDF-1α; CXCL12) (1, 2, 3, 4, 5), have been shown to be angiogenic, the sole CC chemokine member which has been shown to play a direct role in angiogenesis is monocyte-chemotactic protein-1 (MCP-1³; CCL2) (6). Here, we report that eotaxin, which will be referred as CCL11, another CC chemokine, can also directly mediate angiogenic responses via CCR3 receptor expressed by human microvascular endothelial cells (HMECs).

The CCL11 gene is well conserved in several species including human, mouse, and guinea pig (7). CCL11 has ∼50% homology with MCPs (8) and is constitutively expressed in several tissues including intestine, lymphoid node, thymus, skin, heart, kidney, and mammary gland (9). CCL11 is a potent chemoattractant for eosinophils (7), basophils (10), and Th2 lymphocytes (11). The major receptor that mediates the biological effects of CCL11 is CCR3, a seven-transmembrane receptor coupled to heterotrimeric G proteins. It is not known whether CCL11 is chemotactic for noncirculating cells such as endothelial cells, but CCR3 has been reported to be expressed on human brain endothelial cells (12, 13).

CCL11 has a number of important biological functions in disease processes. It plays a critical role in allergic and nonallergic inflammatory reactions, such as mycobacterial and schistosomal induced granulomatosis (8). In chronic inflammatory diseases that are characterized by an eosinophilic-rich inflammatory infiltrate such as nasal polyposis, sinusitis, rhinitis, ulcerative colitis and Crohn’s disease, CCL11 is up-regulated (8). Hodgkin’s disease constitutes another example in which the level of CCL11 protein was correlated directly with the extent of tissue eosinophilia (14). CCL11 is also important in the recruitment of eosinophils into the cornea during experimental helminth-mediated keratitis (9) and in experimental lung inflammation induced by ozone inhalation (8). In addition, CCL11 participates in ischemia-induced vessel wall remodeling, as demonstrated by its release into the medium of normal rat aorta cultures after exposure to prolonged ischemia (15). Thus, CCL11 seems to play a potential role in a number of diseases.

Because CCL11 is widely recognized as an eosinophil chemoattractant and because eosinophilic products including TGF-α and -β, have been shown to induce angiogenesis (16), this raises the question whether eosinophil products are responsible for angiogenesis or are due to a direct effect of CCL11 on endothelial cells. On the basis of our data, CCL11 has the potential to be a direct mediator of angiogenesis, as measured by its ability to induce in vitro endothelial cell migration and in vivo angiogenesis in the Matrigel plug and in CAM assays. Furthermore, CCL11 induced endothelial cell sprouting from aortic rings in the absence of an eosinophil infiltrate. These results demonstrate that CCL11 directly mediates angiogenesis.

Recombinant human CCL11, recombinant human vascular endothelial growth factor (VEGF), and recombinant human basic fibroblast growth factor (bFGF), recombinant human SDF-1α (CXCL12), recombinant human MCP-1 (CCL2) and recombinant human IL-8 (CXCL8) were purchased from PeproTech (Rocky Hill, NJ). Endothelial cell growth supplement was purchased from Sigma (St. Louis, MO). Mouse anti-human CCL11 and monoclonal rat anti-human CCR3 was purchased from R&D Systems (Minneapolis, MN), and mouse IgG and rat IgG (Coulter, Miami, FL) were used as the negative controls.

HMECs were obtained from CSC (Kirkland, WA). Endothelial cell cultures were tested for their expression of CD31 and von Willebrand factor by flow cytometry, and preparations containing <2% contaminating cell types were selected for further studies. Endothelial cells were cultured on collagen type I-coated plastic wells (Biocoat; BD Biosciences, Mountain View, CA), in endothelial growth medium (Clonetics, Walkersville, MD) containing 5% FCS, VEGF (10 ng/ml), bFGF (10 ng/ml), glutamine (2 mM), and gentamicin (100 U/ml). All experiments were performed using subcultures between the second and seventh passages. Human eosinophils were isolated from Leukopacks by Percoll gradient centrifugation followed by MACs CD16 negative selection (Miltenyi Biotec, Auburn, CA).

Indirect immunofluorescence was performed on HMECs by exposing cells to saturating amounts of rat Abs to human CCR3. Fluorescein-conjugated F(ab′)₂ fragments of goat anti-rat (Sigma) diluted 1/100 were used as the secondary Ab. After staining, cells were analyzed using a FACScan flow cytometer (BD Biosciences).

HMEC and human eosinophil chemotaxis was performed using microBoyden chambers. Briefly, polycarbonate filters of 5 μm pore size (Nuclepore; NeuroProbe, Cabin John, MD) were coated with fibronectin or collagen I (10 μg/ml; Sigma) overnight at 4°C. Binding medium (BM) containing 1.0% BSA in RPMI 1640 with or without various amounts of chemokine including CCL11, SDF-1α (CXCL12), MCP-1 (CCL2), and IL-8 (CXCL8) was placed in the lower compartment of the chamber, and 50 μl HMECs resuspended at a concentration of 0.5 × 10⁶ cells/ml in BM were then added to the upper compartment. The chambers were incubated for 2 h for eosinophil chemotaxis and 3–4 h for HMEC chemotaxis at 37°C. After the filters were removed, the upper surface was scraped, fixed with methanol, and stained with Leukostat (Fisher Scientific, Pittsburgh, PA). Membranes were analyzed using the BIOQUANT program (R & M Biometrics, Nashville, TN), and the results were expressed as the mean number of migrated cells/ten fields at ×10 magnification. For inhibitory assays, human CCL11 Ab was added together with CCL11 in the lower compartment of the chamber. Anti-human CCR3 Ab was added to HMECs or human eosinophils 10 min before chemotaxis was performed. Each sample was tested in triplicate. Chemotaxis and inhibition of chemotaxis experiments were performed five times.

HMECs were resuspended at 1 × 10⁶ cells/ml of proliferation medium (RPMI, 0.5% FCS, 2 mM glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin). 50 μl of the cell suspension/well was placed in a gelatin-coated 96-well plate. Cells were stimulated with different concentrations of CCL11, bFGF (100 nM), or VEGF (300 nM). Plates were incubated at 37°C in 5% CO₂ for 48 h. To determine cell proliferation, cells were incubated with [³H]thymidine (0.5 to 1 μCi/well) 18 h before uptake determination. After the incubation, plates were kept at −70°C overnight; finally, the plates were thawed at room temperature and harvested, and [³H]thymidine incorporation was determined using a beta counter.

HMECs were grown in EBM medium containing 5 ng/ml recombinant human epidermal growth factor (EGF). RNA was isolated by the TRIzol method as directed (Life Technologies, Gaithersburg, MD) and thereafter used for analysis of mRNA expression by the Riboquant RNase Protection Assay System (human CR6 probe set; PharMingen, San Diego, CA) according to the manufacturer’s instructions. Briefly, ³³P-labeled antisense RNA probes were synthesized from the human chemokine receptor 5 (CC chemokine receptor) template by T₇ RNA polymerase. The probe (1.5 × 10⁶ cpm) was hybridized in solution overnight in excess to target RNA (10 μg total RNA/treatment) in a total reaction volume of 10 μl. The free probe and other single-stranded RNA were digested with RNases A and T₁ per instructions provided by the manufacturer. The remaining RNase-protected probes were precipitated, dissolved in 3 μl sample buffer (PharMingen), and resolved on denaturing polyacrylamide gels followed by autoradiography for 1–7 days at −70°C.

The thoracic aorta was obtained from 100- to 150-g male Sprague Dawley rats (Taconic Farms, Germantown, NY). Excess perivascular tissue was removed, transverse sections (1–2 mm) were made, and the resulting aortic rings were then washed in medium 199 (Life Technologies). The rings were then embedded in Matrigel (BD Biosciences) in eight-well chamber slides (Nalge Nunc International, Milwaukee, WI) so that the lumen was parallel to the base of the slide. After the Matrigel gelled, serum-free medium (endothelial basal medium supplemented with antibiotics) with or without different concentrations of CCL11 (0.1–100 nM) was added to each well, and the slides were incubated at 37°C, with 5% CO₂, for 3 days. (n = 6 per dose). Endothelial cell growth supplement or bFGF were used as the positive controls at concentrations of 10 μg/ml or 60 nM, respectively. For inhibition experiments, Abs to CCL11 or control mouse IgG (10 μg/ml) were added simultaneously with CCL11. After the incubation period, the rings were fixed, stained, and photographed. The ring assay was repeated four times.

OVA (4 ml) was removed from 3-day-old embryonated eggs (Truslow Farms, Charlestown, MD). Thereafter, windows were opened for each egg, coated with tape to prevent drying, and eggs were incubated at 37°C. On day 10, 5 μl distilled water containing different amounts of CCL11 were applied in the center of quartered 13-mm-diameter plastic coverslips (Thermanox; Nalge Nunc International) and dried for 10 min at 37°C. Each coverslip was placed on the CAM of the chick, and the eggs were incubated at 37°C for 3 days. The assay was scored by a blinded observer and photographed on the 13th embryonic day. Thereafter, the CAMs were cut, fixed with 4% paraformaldehyde, and stained with either hematoxylin and eosin or Giemsa to evaluate the inflammatory infiltrate. bFGF (10 ng) and water were used as positive and negative controls, respectively. Twenty eggs were used in total for each data point. A positive score for angiogenesis was made when vessels appeared to radiate from the spot in the coverslip to which the stimulant was applied. The scores are reported as a percentage of positive CAMs at each dilution. The experiment was repeated twice.

Matrigel (9 mg/ml; 0.3 ml/mouse) alone or mixed with different concentrations of CCL11 was injected s.c. into the flank of C57BL/6 mice. On day 7, mice were sacrificed, and plugs were removed, fixed in 3.7% formaldehyde-PBS, paraffin embedded, and hematoxylin-eosin- or Giemsa-stained slides were photographed. The experiment was repeated twice with eight mice per group in each experiment. For quantification of angiogenesis, the procedure described by Hoffman et al. (16) was used, with modifications. Briefly, 1% of FITC-conjugated dextran (100 mg/kg; Sigma) in PBS (0.2 ml/mouse) was injected i.v. in the tail vein of mice, 20 min before the extraction of the Matrigel. Matrigel sections were weighed and boiled in 5 N HCl, and the fluorescence was read in a fluorometer by excitation at 485 nm and emission at 530 nm. Some Matrigel sections were fixed in buffered formaldehyde and photographed using a fluorescence microscope.

We evaluated the ability of human CCL11 to induce in vitro chemotaxis of HMECs. As shown in Fig. 1, we observed a dose-dependent chemotactic response of HMECs toward CCL11. The maximal chemotactic response for each of the HMECs and human eosinophils was observed at 10 nM CCL11 (Fig. 1,A). The chemotactic response of human eosinophils, which were used as the positive control, was at least 2-fold higher than the response of HMECs (Fig. 1,A). A checkerboard analysis indicated that the migratory effect of CCL11 on HMECs is chemotactic rather than chemokinetic (Table I). The specificity of this chemotactic response was determined using a blocking mAb to human CCL11. This assay indicated that the chemotactic response of HMECs toward CCL11 was blocked by CCL11 mAb at concentrations of 20 μg/ml (Fig. 1,B). Both human and murine CCL11 elicited similar chemotactic responses by HMECs (Fig. 2,C) and heat inactivation completely abolished the activity of either human or murine CCL11 (Fig. 2,C). Because several chemokines have been shown to induce responses by endothelial cells, we compared the chemotactic response induced by CCL11 with other chemokines including SDF-1α, (CXCL12), MCP-1 (CCL2), and IL-8 (CXCL8). As shown in Fig. 1,D, CCL11 responses were comparable with CCL2 responses, but were 3- and 1.5-fold lower than those induced by CXCL12 and CXCL8, respectively. Because several angiogenic factors also induce endothelial cell proliferation, we tested whether CCL11 has a proliferative effect. In contrast to VEGF and bFGF, CCL11 did not induce in vitro proliferation of HMECs (Table II).

Because the responsiveness of HMECs to CCL11 suggests the expression of CCL11 receptor on endothelial cells, we used the RNase protection assay which allows for the detection of multiple chemokine receptor mRNAs in a single RNA preparation. As shown in Fig. 2,A, HMECs expressed low levels of CCR3 mRNA. Additionally, expression of CCR1, CCR2, and CCR4 mRNA was observed. The expression of CCR3 mRNA was not up-regulated by VEGF or bFGF stimulation (data not shown). We further analyzed the expression level of CCR3 protein, the receptor for CCL11, by HMECs. Using immunofluorescence, we detected CCR3 on the cell surface of HMECs (Fig. 2,B). The mean fluorescence intensity was 48 (±16) for HMECs vs 158 (±32) for human eosinophils. The expression of CCR3 on HMECs was at least 4-fold lower than that found on human eosinophils (Fig. 2,C). Blocking experiments indicated that the CCL11-induced chemotactic responses by HMECs were partially inhibited (75%) by anti-human CCR3 (Fig. 2 D), whereas the chemotactic responses by eosinophils were completely abolished by anti-CCR3, suggesting the existence of other receptors(s) for CCL11 on HMECs.

To evaluate whether CCL11 could exhibit angiogenic activities in vivo, we tested different doses of CCL11 ranging from 1 to 100 ng, using the CAM assay. As shown in Fig. 3, CCL11 induced angiogenesis in the CAM assay. The maximal angiogenic responses were obtained at 100 ng (78%) (Fig. 3, C and D) and 10 ng (63%) of CCL11 (Fig. 3, B and D). No significant angiogenic response was obtained at a dose of 1 ng CCL11 (Fig. 3,D). The negative control showed <20% positivity (Fig. 3, A and D). As expected, an inflammatory infiltrate composed of heterophils and eosinophils was observed in association with the angiogenic effect induced by CCL11 as shown by the histological sections of the CAMs (Fig. 3,F). An absence of this inflammatory response was observed in the negative control (Fig. 3 E). These data demonstrate that CCL11 has angiogenic effects in vivo and suggest the existence of a chick receptor functionally homologous to the human CCR3.

We also evaluated the effect of CCL11 using the in vivo Matrigel plug assay. Mice were injected with Matrigel alone or with Matrigel containing CCL11, s.c. in the flank region. Histological sections of the Matrigel plugs indicated a significant angiogenic effect induced by CCL11 when used at a concentration between 10 and 100 nM (Fig. 4,B, C, F, G, and I), in contrast to Matrigel alone (Fig. 4, A, E, and I). Furthermore, as with the CAM assay, an inflammatory reaction was also observed in the Matrigel plugs containing CCL11, which consisted predominantly of eosinophils (Fig. 4,D). Quantitative studies using human (Fig. 4, F and I) and murine (Fig. 4, G and I) CCL11 indicated similar angiogenic responses, that were about one-half as potent as the angiogenic response induced by human bFGF in vivo (Fig. 4, H and I).

We used the ex vivo rat aortic ring-sprouting assay, which allows the detection of angiogenesis in the absence of an inflammatory response to rule out the possibility that the angiogenic responses mediated by CCL11 were a result of eosinophilic products. Transverse sections of rat aorta tissue embedded in Matrigel were cultured with different concentrations of CCL11 as described in Materials and Methods and thereafter examined for the degree of sprouting vessels. As shown in Fig. 5, CCL11 stimulated numerous capillary sprouts at 1 nM (Fig. 5,E) and 10 nM (Fig. 5, B and E). Thus, CCL11 can induce endothelial cell sprouting at nanomolar concentrations from rat aortic rings in the absence of eosinophils, favoring a direct effect in promoting angiogenesis. Additionally, the angiogenic response induced by CCL11 was markedly inhibited by a mAb to CCL11 (Fig. 5, C and E).

Eotaxin (CCL11) is abundantly expressed in the mucous membranes during allergic inflammation, in which angiogenesis also occurs. In this process, it is thought that CCL11-mediated angiogenesis is dependent on eosinophil and mast cell products including TGF-α and -β (17) and chymase, respectively (18, 19), which are also potent angiogenic factors. Additionally, CCL11 expression during endometriosis correlated with eosinophil degranulation and wound healing (20). CCR3, the major receptor for CCL11, has been reported to be expressed by CNS endothelial cells (12, 13).

On the basis of the above evidence, we tested the hypothesis that CCL11 also directly mediates angiogenic responses. Using in vitro and in vivo assays, we demonstrated that CCL11 induces angiogenic responses by human, mouse, rat, and chick endothelial cells. In vitro, CCL11 induced human endothelial cell migration in a dose-responsive manner that was correlated with the expression of CCR3 at the RNA and protein levels. In contrast to CXCR4, (4) the expression of CCR3 by human endothelial cells was not up-regulated by classical angiogenic factors such as VEGF or bFGF (data not shown). The in vivo angiogenic effect of CCL11 was associated with an inflammatory infiltrate comprised mostly of eosinophils. However, because angiogenesis induced by eosinophil products has been proposed, we used the rat aortic ring assay to evaluate angiogenic effects in the absence of an inflammatory response. CCL11 induced rat aortic endothelial cell sprouting in a dose-responsive manner, and this angiogenic effect was specific, because Abs to human CCL11 abolished these angiogenic responses. Therefore, CCL11 can act as a direct mediator of angiogenesis. The ability of CCL11 to induce inflammatory and angiogenic responses in the chick is the first evidence of the existence of a chick CCR3 homologous to human CCR3. Studies along this line of investigation are currently in progress.

Our data on the potential contribution of CCL11 and its receptor on angiogenesis are in line with the data published by Boshoff et al. (21), in which Kaposi’s sarcoma-associated herpesvirus-encoded chemokine-like proteins (vMIP-I and vMIP-II) were found to be highly angiogenic in the CAM assay, via CCR3 but not via CCR5. Furthermore, angiogenic inhibitors, such as IL-10 (22) and IFN-γ (23, 24), inhibit the production of CCL11 by smooth muscle cells (25) and dermal fibroblasts (26). Conversely, TNF-α, a known angiogenic factor, induced CCL11 production by fibroblasts (27). Clearly, CCL11 is a key player in the angiogenic cascade.

The role of CCL11 in the progression of cancer is controversial. In certain cases, such as Hodgkin’s tumors, the production of CCL11 correlated with eosinophil infiltration (14, 28), and the degree of this eosinophil infiltration correlated with an unfavorable prognosis (29). Oral squamous cell carcinoma (30) and gastric carcinoma with squamous cell differentiation (31) constitute additional examples of eosinophil infiltration that correlate with an unfavorable prognosis. In contrast, infiltrating eosinophils in cervix (32), lung (33), colon carcinomas (34, 35), and murine mammary adenocarcinoma (36) have been considered as good prognostic indicators. Although the underlying basis of these divergent responses is unclear, perhaps it is based on heterogeneity in the cytokine repertoire of different tumor types. The contribution of CCL11 toward tumor angiogenesis should be determined to address this issue by testing the effect of inhibitors of CCL11 on tumor growth and other pathological conditions, such as allergic nasal polyposis, and allergic inflammatory states.

Although multiple chemokine receptors have been shown to be expressed by endothelial cells including CXCR-1, -2, -3 (37), CXCR4 (4, 38, 39), CCR2, CCR3 (12, 13), CCR8 (40), CCR4 and CCR5 (13), only CXCR4 plays a requisite role in angiogenesis as shown by the CXCR4 knockout (5). In contrast, disruption of the CCL11 gene apparently did not affect angiogenesis (41, 42, 43), presumably due to the redundancy of the chemokine network. The functional activities of CCL11 can also be compensated by eotaxin-2 (CCL24), eotaxin-3 (CCL26), MCP-4, or RANTES, which also interact with CCR3. Despite the plethora of identified angiogenic factors, ligands for CCR3 may make potentially some unique contributions. The expression of multiple chemokine receptors by endothelial cells might reflect a steady state of the endothelium in an environment in which the balance or imbalance of inducers and inhibitors drives vasculature to develop or regress. For example, in situations like wound healing, in which “angiogenic” chemokines such as IL-8 and MCP-1 are generated, within the first hours, and “angiostatic” IP-10 and MIG are generated after 3 days of the healing processes (44); participation of different chemokine receptors with different specificities is apparently required to bring about normal wound healing (44). Many proinflammatory chemokines including CCL11 might not only promote inflammation but also support the necessary vascularity at inflammatory sites.

We thank Dr. Ken Wasserman for critically reviewing the manuscript and helpful suggestions. We also thank Nancy Dunlop, Douglas Halverson, and Keith Rogers for technical assistance.