Endocan Is Upregulated on Tumor Vessels in Invasive Bladder Cancer Where It Mediates VEGF-A–Induced Angiogenesis

Bladder cancer is the fifth most common cancer in the developed world. It has been estimated that in 2008, 386,300 persons were diagnosed with bladder cancer worldwide and 150,200 succumbed to the disease (1). Bladder cancer has been classified into superficial (pTa, pT1, and CIS) and muscle-invasive (pT2-4) cancer based on whether tumor infiltration extends to the muscular bladder wall (2). However, based on genetic (3) and expression data (4) of bladder cancers of different stages, the revised World Health Organization classification proposes a distinction of noninvasive (pTa) and invasive (pT1-4) bladder cancers. Noninvasive bladder cancers (pTa) have a low risk of progression but a high risk of recurrence. To date, there are no clinical parameters that reliably predict tumor recurrence or progression. Despite the treatment of invasive bladder cancer with radical cystectomy (5), up to 50% of patients develop metastases and 5-year survival is low (6). Increased microvessel density (7) and high expression of VEGF-A correlate with poor prognosis (8), progression (9, 10), and shorter survival of patients with bladder cancer (11). We aimed at elucidating differences at the molecular level between tumor-associated blood vasculature and normal vasculature, which might serve as therapeutic targets or as prognostic biomarkers. Therefore, we isolated blood vascular endothelial cells (BEC) from bladder cancer and normal bladder tissue, using immunolaser capture microdissection (i-LCM), and conducted a comparative microarray expression analysis.

We found that endocan (endothelial cell-specific molecule 1) was one of the most strongly upregulated genes in tumor BECs. Endocan is a secreted proteoglycan (12) that is upregulated by growth factors and chemokines in vitro (13, 14) and on tumor vasculature in several types of cancer (15–17), with diverse suggested biologic roles (18). Using quantitative reverse-transcription PCR (qRT-PCR) and immunohistochemistry, we found that endocan was strongly upregulated on tumor vascular endothelium and that its expression correlated with the invasiveness of bladder cancer. Significantly elevated endocan levels were measured in the plasma of patients with invasive bladder cancer compared with healthy individuals. Importantly, we found that VEGF-A induces endocan expression in vitro and in vivo via VEGF receptor-2 (VEGFR-2) and that endocan knockdown in BECs inhibited VEGF-A–induced tube formation, migration, and VEGFR-2 phosphorylation. Therefore, endocan seems to be an important mediator of tumor angiogenesis induced via the VEGF-A–VEGFR-2 axis and might serve as a novel biomarker in bladder cancer.

Clinically annotated frozen and paraffin-embedded tissue samples of bladder cancers and normal bladder tissue as well as plasma samples were obtained from the University Hospital Zurich (Zurich, Switzerland). The tissue collection was approved by the University Hospital Zurich Ethics Committee (SPUK GGU-USZ Ethics Committee KEK-StV-Nr 02/09) and informed written consent was obtained from each patient. Control plasma samples were obtained from healthy blood donors. In addition, we used a bladder cancer tissue microarray (TMA) containing 143 samples from 91 patients (67 male, 24 female; age range, 29–97 years) from the Institute of Pathology, University Hospital Erlangen (Erlangen, Germany). The TMA consisted of 89 papillary noninvasive [pTa and papillary urothelial neoplasms of low malignant potential (PUNLMP)] and 54 invasive (pT1-4) tumors. Blood vessels were identified on the basis of their typical morphology by a board-certified pathologist (P. Wild), and only vessels within tumors were analyzed to reduce heterogeneity. Endocan expression on blood vessels was semiquantitatively assessed by grading the intensity of the staining on the majority of vessels as absent (0), weak (1+; endocan staining weaker than nuclear staining), or strong (2+; endocan staining stronger than nuclear staining). The statistical significance was computed using Fisher's exact test. The Kaplan–Meier log-rank test was used to analyze recurrence-free survival of patients with noninvasive bladder cancer (pTa; n = 40) with strong endocan expression versus those with weak or absent endocan expression.

Frozen sections of invasive bladder cancer (5 patients; pT1-pT4) and matched normal (tumor-adjacent) bladder tissue were fixed in acetone for 1 minute at 4°C. Blood vessels were stained using biotinylated rabbit anti-human von Willebrand factor (vWF) antibody (0.25 mg/mL; Dako Cytomation) and Cy3-streptavidin (1:100) in a mixture of buffers A and B (all Life Technologies) containing Protector RNase inhibitor (2U/μL; Roche). Slides were dehydrated in 75% ethanol, 95% ethanol, 100% ethanol, and xylene. Immediately afterward, i-LCM was conducted using near infrared-laser–based Arcturus Veritas LCM (Life Technologies). Up to 1 mm² of immunostained blood vessels were isolated per sample and immediately lysed in Buffer RLT Plus (Qiagen) containing 3% β-mercaptoethanol.

RNA was isolated using the RNeasy Plus Micro Kit (Qiagen), amplified and converted to cDNA using the whole transcriptome amplification kit Ovation Pico WTA System (Nugen). Then, cDNA was purified using the QIAquick PCR Purification Kit (Qiagen). Biotin-labeled cDNA targets were hybridized to Human Exon ST 1.0 arrays (Affymetrix) and arrays were scanned according to the manufacturer's protocol. The microarray data are available at Gene Expression Omnibus (GEO) under accession number GSE41614.

Paraffin-embedded tissue sections and the TMA were dewaxed in xylene and rehydrated using graded percentages of ethanol. After incubation in 3% H₂O₂ for 10 minutes, sections were boiled in Tris–EDTA buffer, pH 9 for 20 minutes. Staining for endocan and VEGF-A was conducted using mouse anti-human endocan MEP08 antibody (10 μg/mL, Lunginnov) and rabbit anti-human VEGF-A A-20 antibody (4 μg/mL, Santa Cruz). Biotin-labeled horse anti-mouse (5 μg/mL) and biotin-labeled goat anti-rabbit (15 μg/mL) antibodies were used as secondary antibodies. Immunoreactive signals were amplified by formation of avidin–biotin peroxidase complexes and visualized using 3-amino-9-ethylcarbazole (AEC) or 3,3′-diaminobenzidine (DAB). Nuclear counterstaining was conducted with hematoxylin. A rabbit anti-human vWF antibody (15.5 μg/mL) and Alexa Fluor 594-labeled donkey anti-rabbit antibody (1:200) were used for immunofluorescence stainings, followed by Hoechst (10 μg/mL) nuclear staining.

Immunofluorescent double staining was conducted on 80-μm thick sections according to ref. (19), using the endocan antibody MEP08 and the vWF antibody, followed by Alexa Fluor 594-labeled donkey anti-mouse (1:500) and Alexa Fluor 488-labeled donkey anti-rabbit (1:500) antibodies. Images were acquired with a Leica SP2 confocal microscope using a 63 × 1.4 NA Plan Apochromat objective with oil immersion lens. Three-dimensional reconstruction was conducted using the Imaris software (Bitplane Scientific Software). The resulting image represents a stack of 47 sections (Z step of 0.244 μm) with a total physical length of 11.23 μm. The voxel height was 0.173 μm with a zoom of 2.693.

Plasma samples from 60 healthy donors (median, 58 years; range, 36.8–75 years; male, 87%; female, 13%) and 53 patients with invasive bladder cancer (median, 70.8 years; range, 45.3–84 years; male, 79%; female, 21%) were analyzed by ELISA for endocan (Diyek endomark H1; Lunginnov) and VEGF-A (Platinum; eBiosciences), using a Sunrise Tecan microplate reader (Tecan). For BEC lysates, the protein concentration of samples was adjusted, using a bicinchoninic acid assay (Thermo Scientific).

K14/VEGF-A transgenic (Tg) mice that express mouse VEGF-A164 under control of the K14 promoter (20, 21) were bred and housed in the animal facility of ETH Zurich. Untreated FVB wild-type mice were used as controls. Experiments were carried out in accordance with animal protocol 149/2008 approved by the Kantonales Veterinaeramt Zurich (Zurich, Switzerland). Mouse ears were harvested from 24-week-old VEGF-A Tg mice (n = 6) and wild-type mice (n = 6) and homogenized in RLT buffer (RNeasy lysis buffer, Qiagen) containing 1% β-mercaptoethanol. VEGF-A Tg mice (15-weeks old) were also treated with 800 μg of rat anti-mouse VEGFR-2–blocking antibody DC101 (ImClone Systems) or control rat immunoglobulin G (IgG; Sigma) intraperitoneally every other day for 7 days.

Dermal BECs were isolated from the ears of VEGF-A Tg mice treated with DC101 (n = 5) or control IgG (n = 3), and of wild-type FVB mice (n = 6) using high-speed cell sorting as described (22).

Primary human dermal BECs (23) and human umbilical vein endothelial cells (HUVEC; Promocell) were seeded onto fibronectin-coated culture dishes (10 μg/mL; BD Biosciences) and were cultured in endothelial cell basal medium (EBM; Lonza) supplemented with 20% FBS (Invitrogen), 2 mmol/L l-glutamine (Fluka), 10 μg/mL hydrocortisone (Fluka), and antibiotic/antimycotic solution (Invitrogen). The BEC media also contained endothelial cell growth supplement (Promocell). Cells of passage 6 to 8 were used.

Twelve hours before stimulation, media were replaced with EBM containing 1% FBS. Cells were then treated with recombinant human VEGF-A (20 ng/mL) or fibroblast growth factor (FGF)-2 (20 ng/mL; R&D Systems) or both. Cells treated with vehicle were used as controls. For some experiments, cells were preincubated with human monoclonal antibody IMC-1121B against VEGFR-2 (20 μg/mL; ImClone Systems) or with control human IgG for 4 hours, followed by incubation with VEGF-A (20 ng/mL) for 24 hours. Then, cells were washed and lysed in RLT buffer containing 3% β-mercaptoethanol.

RNA was isolated from mouse ear homogenates and from BECs using the RNeasy Mini Kit and treated with RNase-free DNase (all from Qiagen). cDNA was generated from 1 μg RNA using the High-Capacity cDNA Reverse Transcription Kit (Life Technologies).

The expression of mouse and human endocan was quantified by TaqMan RT-PCR with the AB 7900 HT Fast Real-Time PCR System and the |$2^{- {\rm \Delta \Delta}C_t}$| method (24). TaqMan probe/primer sets for mouse (Mm00469953_m1) and human endocan (Hs00199831_m1) and for mouse CD34 (Mm00519283_m1) were predesigned by Life Technologies. Each reaction was multiplexed with β-actin (4326315E) or B2M (4326319E) for human endothelial cells, or β-actin (4352341E, all from Life Technologies) for mouse samples. Expression of endocan and CD34 was normalized to the expression of the reference gene.

siRNA electroporation of HUVEC was conducted using the Amaxa Basic Nucleofactor Kit for primary endothelial cells (Lonza). The knockdown efficiency of 3 different siRNA constructs against endocan was evaluated (ID: 136192, ID: 19124, ID: 19216). Knockdown using siRNA ID: 136192 showed the strongest reduction of endocan mRNA (data not shown). Therefore, siRNA ID: 136192 and the no. 1 siRNA control nontargeted siRNA (all Ambion) were used. After 24 or 36 hours, supernatants were collected and cells were lysed in RLT buffer containing 3% β-mercaptoethanol, or in a hypotonic PBS solution containing 1 mmol/L MgCl₂, 1 mmol/L CaCl₂, 1 mmol/L phenylmethanesulfonyl fluoride, and protease inhibitor cocktail (Roche) as described (25).

Supernatants were concentrated using a Centricon Ultracel YM-10 membrane filter (Millipore). Proteins were precipitated as described (26). Supernatant samples containing 100 μg of protein and cell lysate samples containing 10 μg of protein (assessed using the BCA protein assay) were subjected to SDS-PAGE using 10% acrylamide separating gels, 1.0 mm. Proteins were transferred to Immobilon-P transfer membranes (Millipore) and stained using a biotinylated goat anti-human endocan antibody (0.2 μg/mL; R&D Systems) or rabbit anti-human (p1175)VEGFR-2 (D55B11; 1:1,000), VEGFR-2 (55B11; 1:1,000), or extracellular signal–regulated kinase (ERK) 1/2 (137F5; 1:1,000; all Cell Signaling) antibodies and a streptavidin-bound horseradish peroxidase conjugate (1:5,000) or horseradish peroxidase–coupled anti-rabbit antibody (1:5,000). Stained proteins were detected using the ECL Plus Western Blotting Detection System (GE Healthcare). Equal loading was confirmed by Ponceau S (Sigma) staining and by staining for vWF (6 μg/mL; ref. 27) or ERK 1/2.

The capillary tube formation assay was conducted as described (28) with minor modifications. Matrigel (BD Biosciences) was added to 96-well plates (40 μL per well) and let solidify for 30 minutes at 37°C. HUVECs were electroporated with endocan siRNA or nontargeting siRNA (100 μL of cell suspension containing 15,000 cells/well), seeded on top of Matrigel in sextaplicates and incubated for 12 hours. Images were acquired with a digital camera (AxioCam MRm, Carl Zeiss) mounted on an inverted microscope (Axiovert 200M, Carl Zeiss). The total length of tube-like structures per well was measured using ImageJ. Endothelial cell migration was assessed using a Scratch monolayer wound-healing assay (29). HUVECs were electroporated using endocan siRNA or nontargeting siRNA and seeded into fibronectin-coated wells in quadruplicates (120,000 cells/well). At confluency, 2 cross-shaped scratches were made in each well using a sterile 200-μL pipette tip. Cells were washed and VEGF-A (20 ng/mL) or plain medium were added. Images of crosses were taken immediately and 25 hours later. The surface areas of the cell-free zones were measured and the percentage scratch closure was determined using TScratch software (29).

Statistical analyses were conducted using Prism version 5.00 (GraphPad Software, Inc.). For most comparisons, a 2-tailed unpaired Student t test was conducted. The differences in plasma concentrations of VEGF-A and endocan were analyzed with the Mann–Whitney U test. Spearman rank correlations were made to test for the association between endocan and VEGF-A concentrations in plasma. The area under curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and likelihood ratio were calculated from the receiver operating characteristics (ROC) curve. The comparison between endocan expression in noninvasive and invasive cancers was calculated using Fisher's exact test. Survival data were plotted using the Kaplan–Meier method and analyzed using the log-rank (Mantel–Cox) test. Differences were considered statistically significant at P < 0.05.

We isolated endothelial cells from bladder cancer-associated vessels and from vessels in normal surrounding bladder tissue by i-LCM. TaqMan-based qRT-PCR showed that endocan mRNA expression was 1,000- to 100,000-fold higher in tumor-associated BECs than in normal BECs (Fig. 1A). Immunohistochemical staining revealed that the endocan protein was present on tumor-associated blood vessels but not on the blood vessels of normal bladder tissue (Fig. 1B). Confocal imaging of immunofluorescently stained sections of invasive bladder cancer revealed that endocan was associated with the surface of the plasma membrane of BECs. Staining was particularly intense at the filopodia of endothelial tip cells, which typically are vWF-negative (ref. 30; Fig. 1C).

No staining for endocan was detected in normal bladder tissue (Fig. 2A). We next analyzed a TMA consisting of noninvasive (n = 89) and invasive bladder cancers (n = 54). Of these, 116 samples contained a sufficient number of vessels for analysis of endocan expression, which was semiquantitatively assessed by grading the intensity of the staining as absent (0; Fig. 2B), weak (1+; Fig. 2C), or strong (2+; Fig. 2D). We observed that endocan expression was heterogeneous in individual punches (Fig. 2E). Vessels outside the tumors were negative (Fig. 2F), therefore only vessels within tumors were analyzed to reduce heterogeneity (Fig. 2G). Fifty-three of 70 samples of noninvasive bladder cancer had no or weak staining for endocan, whereas 17 samples had strong staining. From 46 invasive samples, 23 had no or weak staining and 23 had strong staining. The difference in frequencies between these groups was significant (P = 0.005; Fig. 2H), indicating that endocan expression on tumor blood vessels increases with invasiveness of bladder cancer. Kaplan–Meier analysis of 40 samples of pTa cancers showed that patients with strong endocan expression (n = 12; N_events = 9) had a reduced recurrence-free survival compared with those with no or weak endocan expression (n = 28; N_events = 14; P = 0.079; Fig. 2I).

We next tested whether the increased endocan expression in invasive bladder cancers would also be apparent in plasma. The mean concentration of endocan was significantly higher (P < 0.001; Mann–Whitney U test) in the plasma of patients with invasive bladder cancer (0.79 ng/mL; range, 0.23–3.53 ng/mL; n = 53) than of healthy volunteers (0.43 ng/mL; range, 0–1 ng/mL; n = 60; Fig. 3). Using ROC curve and AUC, we determined the ability of endocan levels to allow for distinction between patients with invasive bladder cancer and healthy subjects. The AUC, sensitivity, and specificity were 0.76, 64%, and 80%, respectively, with a cutoff of 0.63 ng/mL. The PPV and the NPV were 74% and 72%, respectively, with a likelihood ratio of 3.208.

It has been reported that in vitro, endocan expression is induced by VEGF-A (13, 16). We therefore asked whether the enhanced endocan expression on tumor blood vessels might be due to increased VEGF-A production. Immunostaining of invasive bladder cancers revealed VEGF-A expression by tumor cells (Fig. 4A, right, white arrows). VEGF-A was also present on tumor-associated blood vessels, likely reflecting VEGF-A bound to its receptors (31; Fig. 4A, right, black arrows). In contrast, VEGF-A was not detected in normal bladder tissue (Fig. 4A, left) and in noninvasive bladder cancer (Fig. 4A, middle). VEGF-A levels in plasma of healthy blood donors were significantly lower (mean, 0.12 ng/mL; range, 0.01–0.4 ng/mL; n = 60) than in patients with invasive bladder cancer (mean, 0.28 ng/mL; range, 0.04–0.89 ng/mL; n = 53; P < 0.001; Mann–Whitney U test; Fig. 4B). The AUC, sensitivity, and specificity were 0.83, 68%, and 90%, respectively, with a cutoff of 0.1895 ng/mL. The PPV and the NPV were 86% and 76%. The likelihood ratio was 6.72. Combining the 113 samples analyzed, we found a significant correlation of VEGF-A and endocan levels (P = 0.004; Fig. 4C).

After treatment of human BECs with VEGF-A (20 ng/mL) for 24 hours, endocan expression was significantly increased at the mRNA (1.58–1.80-fold; Fig. 5A) and protein level in cell lysates (mean, 10.17; range, 9.48–10.62 ng/mL; control treated: mean, 7.79 ng/mL; range, 6.21–8.73 ng/mL; Fig. 5B). Increased endocan levels were also detected in culture supernatants (mean, 41.2 ng/mL; range, 36.45–44.1 ng/mL; control treated: mean, 33.8 ng/mL; range, 32.52–34.59 ng/mL; Fig. 5C). FGF-2 alone or in combination with VEGF-A did not alter endocan mRNA expression. Pretreatment of BECs with a VEGFR-2–blocking antibody inhibited induction of endocan by VEGF-A at the mRNA (0.74–0.76-fold compared with control IgG; Fig. 5D) and protein level (mean, 30.6 ng/mL in supernatants; range, 29.07–32.55 ng/mL; control IgG treated: mean, 39.42 ng/mL; range, 37.74–40.8 ng/mL; Fig. 5E).

To investigate whether VEGF-A might also enhance endocan expression by BECs in vivo, we analyzed skin samples obtained from homozygous VEGF-A Tg mice, which have elevated VEGF-A levels in the skin (21). Endocan mRNA expression was strongly increased (27; 2–67.6-fold; Fig. 5F) in the skin of transgenic mice (n = 6) compared with wild-type mice (n = 6). To investigate whether this was solely the consequence of increased blood vessel numbers, we specifically isolated blood vessel-derived endothelial cells from the ear skin by fluorescence-activated cell sorting (FACS). We found upregulation of endocan in endothelial cells derived from VEGF-A Tg mice (2.91–5.34-fold; Fig. 5G), whereas expression of the vascular marker CD34 was unchanged (Fig. 5H). Treatment of VEGF-A Tg mice with the VEGFR-2 receptor-blocking antibody DC101 resulted in reduced endocan expression in ex vivo isolated endothelial cells compared with the control IgG group (0.2–0.46-fold; Fig. 5I); CD34 expression was not affected (Fig. 5J).

We next investigated whether and how endocan might contribute to VEGF-A–induced angiogenesis. The biologic functions of proteoglycans often depend on the interactions of their glycosaminoglycan chains with protein ligands, such as cytokines and growth factors (32, 33). Therefore, we investigated whether silencing of endocan would affect VEGF-A effects on endothelial tube formation or migration. Electroporation of HUVECs with endocan siRNA strongly reduced endocan expression at the mRNA (Fig. 6A) and protein levels (Fig. 6B). The total length of tube-like structures formed by HUVECs on growth factor-rich Matrigel was significantly reduced upon endocan knockdown compared with control siRNA (−45%; P = 0.012; Fig. 6C). In a monolayer wound-healing (“scratch”) assay, migration of HUVECs was not affected by endocan knockdown (data not shown). However, when cells were stimulated with VEGF-A, silencing of endocan abolished the migration-inducing effect of VEGF-A (reduction of wound closure by 53% compared with control siRNA knockdown; P = 0.007) to levels observed in non-VEGF-A-treated controls (Fig. 6D and E). Importantly, knockdown of endocan, followed by incubation of HUVECs with VEGF-A for 10 minutes, resulted in a strong reduction of VEGFR-2 phosphorylation, whereas the total amount of VEGFR-2 was unchanged (Fig. 6F). These findings indicate that endocan is necessary for VEGF-A–induced signaling and VEGFR-2 phosphorylation.

Here, we established a novel method for efficient isolation of tumor-associated blood vessels from invasive bladder cancers and matched normal bladder tissue using i-LCM. Transcriptional profiling revealed endocan as a gene whose expression was strongly increased on tumor-associated blood vessels, in line with gene expression profiling studies in breast (34), lung (35), and thyroid (36) cancers.

Endocan mRNA and protein expression was strongly increased on blood vessels of invasive bladder cancers compared with normal bladder tissue and noninvasive bladder cancers. Importantly, noninvasive bladder cancer patients with strong endocan expression had a shorter recurrence-free survival than those with absent or weak endocan expression. Currently, the standard follow-up for noninvasive bladder cancer includes cystoscopy combined with cytologic examination at intervals of 3 to 6 months, depending on tumor malignancy and previous recurrence rate. Cystoscopic examinations are unpleasant, time-consuming, expensive, and may have serious side effects, such as infections and damage to the urethra (37). Cytology is characterized by a high specificity but a low sensitivity. So far, there are only few promising molecular markers that might predict recurrence-free survival (38). Our study suggests that endocan might be an additional marker that could allow some prediction how often patients need to undergo cystoscopic examination.

The increased endocan concentrations in the plasma of patients with invasive bladder cancers indicate that endocan might serve as an additional biomarker to evaluate the overall prognosis of invasive bladder cancers. However, additional studies are needed to investigate whether there are consistent differences of plasma endocan levels between invasive and noninvasive cases.

As endocan is a soluble proteoglycan, we asked where it may be localized within the tumor-associated blood vessels. Using high-resolution confocal imaging, we found that endocan was associated with cell membranes of vascular endothelial cells. Particularly high amounts of endocan were associated with filopodia of endothelial cells. Filopodia are typically present on tip cells and are needed for their guidance and motility during angiogenesis (39). In agreement with our findings, expression of endocan was previously also found on tip cells in developing mouse retinas (40, 41).

Tip cells are responsive to VEGF-A and other angiogenic factors and guidance molecules (42), and tumor-associated vessels form extensive filopodia (19). Thus, we asked whether endocan may play a role in migration and sprouting of endothelial cells. Indeed, when endothelial cells were stimulated with VEGF-A, siRNA-mediated endocan silencing abolished the migration induced by VEGF-A. In contrast to a previous report (43), we found that siRNA-mediated endocan silencing also abolished endothelial tube formation. This discrepancy is likely due to different assay conditions used (HUVECs embedded between 2 layers of collagen gel vs. plating on top of a solidified Matrigel in our study). We also found that expression of endocan by BECs was upregulated after VEGF-A treatment in vitro, whereas incubation with FGF-2 had no effect on endocan mRNA or protein production. The different responsiveness of HUVECs (17) and BECs (our study) to FGF-2 likely reflects functional differences between large vessel-derived endothelial cells (HUVEC) and microvasculature-derived endothelial cells (BECs).

Endocan carries a dermatan sulfate chain attached to its serine 137 (18). Dermatan sulfate is a linear polysaccharide, which consists of repeating disaccharide units composed of sulfated N-acetylgalactosamine and either glucuronic or iduronic acid (44). Both the sulfates and the carboxylates of the uronic acids are negatively charged at physiologic pH, thus providing binding sites for signaling molecules comprising positively charged amino acids, such as cytokines and growth factors. Interactions between signaling molecules and glycosaminoglycan chains are often highly specific and may serve to create high local concentrations of signaling molecules, to prolong their half-life by protecting them from proteolytic degradation and to facilitate their binding to their cognate receptors (45). It has been shown that dermatan sulfate chains of endocan bind hepatocyte growth factor (HGF; ref. 46) and promote its mitogenic activity on HEK293 cells (47). However, we previously found that HGF only had a weak effect on the proliferation of BECs (48). In contrast, our novel finding that siRNA-mediated silencing of endocan resulted in strongly reduced phosphorylation of VEGFR-2—the major transmitter of VEGF-A's effects on endothelial cells—upon incubation with VEGF-A, indicates a major role of endocan in mediating the effects of VEGF-A. Endocan is secreted by endothelial cells upon VEGF-A stimulation and binds VEGF-A on the cell surface, facilitating its interaction with VEGFR-2 and increasing the intensity of the VEGF-A signal. It remains to be investigated whether endocan might also have paracrine effects on nonvascular cells (16).

Taken together, these results indicate that VEGF-A induces endocan expression in endothelial cells, which in turn enhances VEGF-A–induced endothelial cell migration and angiogenesis. The interrelation of endocan expression with the presence of VEGF-A on one side and the angiogenesis-inducing function of VEGF-A, supported by endocan, on the other side could make endocan an ideal biomarker for monitoring the therapeutic response to treatment with VEGF-A–targeting antiangiogenic agents. This concept is further supported by the strongly reduced endocan levels after systemic treatment of VEGF-A Tg mice with an anti-VEGFR-2 antibody.

No potential conflicts of interest were disclosed.

Conception and design: F. Roudnicky, V.I. Otto, M. Detmar

Development of methodology: F. Roudnicky

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Roudnicky, C. Poyet, P. Wild, S. Krampitz, R. Huggenberger, A. Rogler, A. Hartmann, M. Provenzano

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Roudnicky, C. Poyet, P. Wild, S. Krampitz, F. Negrini, A. Rogler, V.I. Otto, M. Detmar

Writing, review, and/or revision of the manuscript: F. Roudnicky, R. Huggenberger, A. Rogler, R. Stöhr, A. Hartmann, M. Provenzano, V.I. Otto, M. Detmar

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Roudnicky, C. Poyet, P. Wild, A. Rogler, R. Stöhr, A. Hartmann

Study supervision: M. Detmar

The authors thank Sabrina Petsch and Stefan Schick, Tumorzentrum Erlangen, for providing follow-up patient data, Dr. Martin Schulz for help with the scratch assay, Dr. Maija Hollmén and Sinem Karaman for helpful discussions, and Jeannette Scholl and Peter Camenzind for excellent technical assistance.

This work was supported by Swiss National Science Foundation grants 3100A0–108207 and 31003A_130627, Advanced European Research Council Grant LYVICAM, Krebsliga Schweiz and Krebsliga Zurich (to M. Detmar). The TMA construction was supported by a grant from the Interdisciplinary Center for Clinical Research (IZKF), University Hospital Erlangen.

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.