Endocan Expression and Relationship with Survival in Human Non–Small Cell Lung Cancer Free

Lung cancer is the leading cause of cancer mortality with, for example, >170,000 diagnosed cases, resulting in >160,000 deaths in the United States in 2005 (1) and a corresponding 28,000 diagnosed cases in France. Of these, 80% represent cases of non–small cell lung cancer (NSCLC). Despite multimodal therapy, prognosis is still dismal with an overall 5-year survival of 15% (1). Thus, efforts toward finding new potential therapeutic targets are important research directions.

Recently, we have shown that endocan, a lung- and kidney-selective endothelial cell-specific dermatan sulfate proteoglycan, binds through its glycan moiety to hepatocyte growth factor/scatter factor. In vitro, binding to endocan amplifies the growth factor mitogenic effect on cells of epithelial origin (2). In vivo, endocan overexpression by nontumorigenic epithelial cells induces tumor formation, whereas overexpression by tumorigenic cells sharply increases the growth rate of resulting tumors (3). Although the tumorigenic effect requires the presence of the glycan chain, it also depends on the protein structure, as specific protein mutation results in loss of the tumor growth promotion effect (3).

Endocan also binds to the integrin lymphocyte function-associated antigen-1 and has been implicated in inhibiting the lymphocyte function-associated antigen-1/intercellular adhesion molecule-1 interaction (4), an important step in the firm adhesion of leukocytes to the endothelium, and thus could be involved in regulating leukocyte migration into tissues. Interaction between intercellular adhesion molecule-1 and lymphocyte function-associated antigen-1 has also been implicated in the binding of cytotoxic lymphocytes and natural killer cells to tumor cells, resulting in cell lysis (5). Intercellular adhesion molecule-1 expression plays an important role in tumor development and its overexpression in tumor xenografts is associated with slower growth and higher tumor infiltration by lymphocytes (6). Thus endocan, by its dual roles in modulating cell growth activity and leukocyte function, could potentially be involved in tumor growth control.

Endocan is overexpressed during in vitro angiogenesis (7), but antisense knockout of endocan does not prevent this process, suggesting only an accessory role in angiogenesis.

Microarray studies have shown an increased endocan mRNA expression in various tumor locations, such as kidney (8), breast (9), and lung (10), as well as in highly invasive melanoma cell lines (11). In kidney tumors, endocan expression has been correlated with vascular endothelial growth factor (VEGF) expression (7). These studies revealed that elevated transcription of the endocan gene represents one of the molecular signatures of bad prognosis in both lung (10) and breast (9) cancers. Consistently, elevated blood levels of endocan were observed in patients with lung cancer (3). However, little is known about the expression of endocan protein in human tumors and the significance of endocan levels in the blood.

The human endocan gene is transcribed into two alternatively spliced variants, which either retain (i.e., endocan) or exclude (i.e., endocanΔ2) exon 2 (7). No data exists to date about relative expression of the two isoforms in human tumors. As endocan is preferentially expressed in lung tissue, a major question for evaluation is its involvement in the development of NSCLC. We show here that (a) endocan is strongly expressed in the tumor vessels, (b) that serum endocan levels have prognostic value in NSCLC, (c) that VEGF and endocan levels are linked in NSCLC tissues, and (d) that VEGF represents a powerful inducer of endocan secretion.

Patients and samples. Two groups of patients were recruited. The first consisted of 24 patients undergoing surgery for NSCLC between October 2003 and January 2005 in one of the surgical centers participating in the Thoracic Oncology Group (see Appendix A). Immediately after surgical resection, the pathologist sampled both the tumor and the lung distant from the tumor. Each sample was then cut into two pieces. One was immediately frozen at −20°C in Trizol reagent (Invitrogen, Cergy Pontoise, France) for later RNA extraction. The other one was fixed in alcohol-formaldehyde-acetic acid (Labonord, Templemars, France) for 2 hours and then processed for paraffin embedding. The second group comprised 30 previously untreated patients with NSCLC who had undergone an initial evaluation for tumor staging in the Department of Thoracic Oncology at the Centre Hospitalier Régional Universitaire de Lille (Lille, France) in 1996. Five milliliters of serum derived from the blood drawn for routine biochemical analysis were immediately frozen and conserved at −20°C until assayed for endocan. The physicians in charge of the patients were not aware of the results of the endocan assay. Appropriate antitumoral treatment was chosen according to tumor staging. Patients were then followed up as either in- or out-patients, and survival data were recorded. Staging was compliant with the fifth International Union Against Cancer guidelines (12).

Both the protocol and the use of human tissues were approved by the local ethics committee, and the patients gave informed consent. Only patients with a clear histologic classification as NSCLC and also without any previous treatment were included in the study.

Cell isolation and culture. Primary human umbilical vascular endothelial cells (HUVEC) were derived from the veins of umbilical cords as described previously (13). Second passage HUVECs were cultured on fibronectin-coated, 48-well culture plates (Becton Dickinson, Le Pont de Claix, France) in RPMI 1640 containing 20% FCS, endothelial cell growth supplement (Sigma-Aldrich, Lyon, France), 5 IU/mL heparin (Sanofi-Synthelabo, Gentilly, France), and 2 mmol/L l-glutamine (Life Technologies, Cergy Pontoise, France). At confluency, the cells were washed once and incubated in fresh RPMI 1640 containing 2% FCS and 2 mmol/L l-glutamine for 18 hours and then washed once again with the same medium just before stimulation. Recombinant human VEGF¹⁶⁵, fibroblast growth factor-2, or tumor necrosis factor-α (R&D Systems, Abingdon, United Kingdom), all at final concentrations of 20 ng/mL, were then added to appropriate wells. In inhibition experiments, VEGF¹⁶⁵ was incubated with an anti-VEGF monoclonal antibody (R&D Systems) for 1 hour before addition to culture medium. The cell supernatants were centrifuged and stored at −20°C until endocan quantitation by ELISA. Cell viability was evaluated by a colorimetric method using the reduction of 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyl-tetrazolium bromide (Sigma-Aldrich). At the end of the experiment, cell number was quantified by assaying lactate dehydrogenase release after cell lysis using the Cytotox 96 lactate dehydrogenase release assay (Promega, Charbonieres, France). In some experiments, cells were retrieved in 1 mL Trizol reagent and then processed for total RNA extraction.

Reverse transcription-PCR. Total RNA extraction was done on Trizol-treated samples according to the manufacturer's recommendations. Two micrograms of the resulting total RNA were converted to first strand cDNA by use of random hexamer primers and Moloney murine leukemia virus reverse transcriptase (Invitrogen) according to the manufacturer's instructions. PCR was done using 1 unit HotMaster Taq DNA Polymerase (Eppendorf, Le Pecq, France) and 1:20 of the reverse transcription reaction, with an initial hotstart of two minutes at 94°C followed by 30 seconds annealing and 1 minute extension at 65°C. Primers, annealing temperatures, and number of cycles used (chosen for being in the exponential phase of PCR amplification) were as follows:

• (a) Glyceraldehyde-3-phosphate dehydrogenase (sense): GTCTTCACCACCATGGAGA, 55°C, 22 cycles, 204-bp amplicon;

• (b) Glyceraldehyde-3-phosphate dehydrogenase (antisense): CCAAAGTTGTCATGGATGACC;

• (c) RS9 (sense): CAGGCGCAGACGGTGGAAGC, 60°C, 29 cycles, 411-bp amplicon;

• (d) RS9 (antisense): CGCGAGCGTGGTGGATGGAC;

• (e) Endocan (sense): GCTACCGCACAGTCTCAGG, 60°C, 32 cycles, bands of 633 (endocan) and 463-bp (endocanΔ2);

• (f) Endocan (antisense): ATTGCATTTTTAGTTCTTGAGTGT;

• (g) Platelet/endothelial cell adhesion molecule-1 (sense): AGGGGACCAGCTGCACATTAGG, 55°C, 30 cycles, 450-bp amplicon;

• (h) Platelet/endothelial cell adhesion molecule-1 (antisense): AGGCCGCTTCTCTTGACCACT;

• (i) VEGF (sense): TGGATCCATGAACTTTCTGCTGTC, 59°C, 32 cycles; and

• (j) VEGF (antisense): TCACCGCCTTGGCTTGTCACTT.

Three bands were detected at 452, 584, and 656 bp corresponding to the VEGF isoforms 121, 165, and 189, respectively (14).

PCR products were migrated on 1.5% agarose gels with 0.01% GelStar (Cambrex Bioscience, Rockland). Band intensities were measured under UV light using Gel Analyst software (Clara Vision, Orsay, France). For semiquantitative analysis, the band intensity for endocan was normalized to platelet/endothelial cell adhesion molecule-1 as an endothelial-specific gene and VEGF was normalized to RS9.

Antibody production and purification. Production of anti-endocan antibodies was done as described previously (15). Two different hybridoma cell clones MEP14 and MEP08 directed against antigenic determinants derived from endocan exon 3 and exon 1, respectively, were selected. Hybridoma cell lines were grown in conditioned medium without FCS (Hybridoma serum-free medium, Life Technologies). Supernatants were purified on a protein G-Sepharose column (Amersham, Orsay, France). Antibodies were eluted in 3 mol/L MgCl₂, dialyzed against PBS, and then concentrated by centrifugation over Amicon Ultra 30-kDa molecular weight cutoff filters (Millipore, Molshein, France). Protein was quantified using the Bio-Rad protein assay (Bio-Rad, Marne la Coquette, France), its purity by Coomassie blue staining after SDS-PAGE, and endocan reactivity by ELISA.

ELISA. Serum endocan levels were quantified using a sandwich immunoassay described previously (15, 16). The monoclonal antibodies used recognize both endocan isoforms. Successive incubations with biotin-conjugated anti-mouse IgG1 antibody (PharMingen, Le Pont-de-Claix, France) and streptavidin-conjugated horseradish peroxidase (Clinisciences, Montrouge, France) were followed by visualization using 3,3′,5,5′-tetramethylbenzidine (Sigma-Aldrich).

Immunohistochemistry. Tissues from lung carcinoma and the corresponding nontumoral lung were analyzed under standard procedures. Staining was conducted using an automated immunostainer (Benchmark, Ventana Medical Systems, Illkirch, France) on 3-μm-thick sections of paraffin-embedded tissue, mounted on Superfrost adhesive slides. Cell conditioning solution (CC1), which is a Tris-based buffer with a slightly basic pH, was used as a pretreatment step at 98°C for 48 minutes. Sections were then incubated with primary anti-endocan monoclonal antibody MEP08 (2 mg/mL; 1:500 dilution) or MEP14 (1 mg/mL; 1:200 dilution) for 32 minutes. The immunohistochemistry method used is a three-step indirect process based on the biotin-streptavidin complex. Slides were counterstained with hematoxylin. Endogeneous peroxidase activity was suppressed by initially incubating specimens in 3% hydrogen peroxide. Positive and negative controls, excluding primary and/or secondary antibodies, were included in each automated immunohistochemistry run.

Statistical analysis. All data are presented as median values with 25th and 75th centiles reported in parentheses. Data from semiquantitative evaluation of endocan expression levels in tumors were compared with those in normal lung parenchyma from the same patient using an exact Wilcoxon test. A two-sided P < 0.05 was considered significant. The log-rank test statistic was used to compare differences in survival between groups. Patients still alive at the end of the follow-up period were considered censored observations. A stepwise Cox proportional hazard model was then used to examine associations between the various prognostic factors and survival. Age was introduced as a continuous variable, whereas node (grouped as N₀ and N₁ or N₂ and N₃) and metastasis status were introduced as binary variables. Tumor extension was taken into account either by stage (I-IV) or by using separate tumor, necrosis, and metastasis descriptors. Serum endocan was used either as a continuous variable or as a binary variable. The cutoff value for serum endocan, which best differentiate survivors and nonsurvivors, was determined by an algorithm of maximization of hazard ratio (17, 18). Statistical analyses were done with SPSS 12.0 software (SPSS, Inc., Paris La Defense, France) and the SAS system (version 8.2; SAS Institute, Inc., Cary, NC).

Endocan mRNA levels are increased in lung tumors. A total of 24 patients were recruited for the mRNA study. As endocan is an endothelial cell–specific molecule, the endocan band intensity was normalized to that of platelet/endothelial cell adhesion molecule-1 (CD31 antigen), used as an endothelial cell marker. By semiquantitative PCR analysis, we found that the level of the exon 2-containing endocan mRNA was increased in lung tumor tissue, with an average 2.5-fold increase (P = 0.029) over that in lung tissue sampled at a distance from the tumor (Fig. 1A and B). Across the 24 lung tumor samples, endocan mRNA increased in 20 cases did not change in two and slightly decreased in two. No differences in endocan expression were seen between squamous cell carcinomas and adenocarcinomas. Contrary to the endocan mRNA data, we observed only a small increase in the tumor levels of the exon 2-deleted endocanΔ2 mRNA, difference that did not reach statistically signification in our series (P = 0.09).

Thus, there is an overexpression of the endocan mRNA in NSCLC compared with distant lung parenchyma.

Intratumoral and peritumoral vessels exhibit strong endocan protein expression. To understand expression pattern of endocan in the tumor microenvironment and to determine the specificity of endocan to tumor-associated vessels, we did immunohistochemical studies. These showed that endocan protein was detected in the endothelium of all tumors studied (example in Fig. 2A). The intratumoral vessels within the cancer nests exhibited endothelial staining for endocan, with expression present in ∼80% of the vessels identified by anti-CD34 staining. Interestingly, endocan expression was more intense in the tumor periphery and in the lung parenchyma immediately lining the tumor, where most vessels expressed endocan (Fig. 2B). Moreover, endocan was strongly expressed by intratumoral vessels in areas of pleural invasion, bronchial infiltration, and lymph node metastasis. Only rarely vessels in the healthy visceral pleura were stained by the anti-endocan antibody. Vessels in normal bronchial mucosa distant from the carcinoma also lacked staining. No staining was observed in tumoral cells in all cases studied.

The same staining pattern was observed whichever antibody (MEP08 or MEP14) was used. These two antibodies, which recognize epitopes within exon 1 or exon 3 respectively, thus recognize both endocan and endocan Δ2, and their reactivity is also independent of endocan glycanation. Thus, the staining confirms that both epitopes are equally distributed within NSCLC tissues. No differences were seen between adenocarcinomas and small cell carcinomas, either in intensity of staining or the number of stained vessels.

Despite the fact that endocan is normally expressed at the mRNA level by endothelial cells in the normal lung, immunohistochemistry showed only a weak signal for endocan in very few vessels in the normal lung (Fig. 2C). Therefore, in normal lung vessels, endocan protein is not expressed at sufficient levels to give a clear signal in immunohistochemistry.

Thus, intratumoral and peritumoral endothelial cells represent the main source of endocan in the vicinity of the tumor. In addition, the low expression of endocan by normal lung vessels confers on this molecule a special interest as a potential marker of intratumoral vasculature.

VEGF stimulates endocan secretion from endothelial cells and correlates with endocan expression in NSCLC. As endocan is expressed in vessels from tumor periphery, we searched potential factors involved in up-regulation of endocan expression at the endothelial level. VEGF is one of the main mediators involved in the formation and activation of tumor vessels (19, 20). Previous results indicated a correlation between endocan and VEGF mRNA expression in kidney tumors (7). We therefore focused on VEGF as an angiogenic mediator that might up-regulate endocan mRNA in situ. To prove a direct effect of VEGF on endocan production by tumor endothelial cells, we studied whether VEGF induced endocan secretion in HUVECs in vitro. HUVECs spontaneously secrete low levels of endocan. On stimulation with 20 ng/mL VEGF¹⁶⁵, HUVEC supernatants exhibited 3-fold increased levels of endocan (Fig. 3A). An increase was already observed after 24 hours and continued for the 4 days of stimulation. By comparison, VEGF stimulation was at least as efficient as fibroblast growth factor-2 (another known angiogenic growth factor) or tumor necrosis factor-α, which has been already reported to induce an increase in endocan secretion (21). Consistently, VEGF-stimulated HUVECs also showed a substantial increase in endocan mRNA expression, mostly of the larger transcript (Fig. 3C). We also tested whether blocking anti-VEGF antibodies could inhibit VEGF-induced endocan expression. Addition of an anti-VEGF antibody induced a dose-dependent inhibition of VEGF-induced endocan secretion by endothelial cell with an almost complete (95%) inhibition at 0.1 μg/mL and a 50% neutralization dose of 0.02 μg/mL, which is the range of the antiproliferative effect reported by the manufacturer for this antibody (Fig. 3B).

In our lung tumor samples, we found that VEGF mRNA is increased compared with the distant lung tissue (Fig. 1A), especially VEGF¹⁶⁵, which is one of the dominant VEGF isoforms expressed in lung tumors (20). Statistical comparisons between the mRNA levels of the endocan and VEGF¹⁶⁵ revealed a significant positive correlation (Spearman ρ = 0.568; P = 0.006; Fig. 1C).

Thus, from these results, VEGF, a major angiogenic factor involved in tumoral angiogenesis, induces a sustained increase in both endocan mRNA expression and subsequent endocan protein secretion by primary cultured human endothelial cells, which could explain the similar endocan overexpression observed in tumor vessels in vivo.

Prognostic value of serum endocan in patients with NSCLC. In an earlier preliminary study, we found elevated serum levels of endocan in patients with lung cancer (3). Moreover, endocan now seems to be a marker of tumor vessels. Bearing in mind the fact that tumor vessel density represents one key determinant of tumor aggressiveness and poor prognosis (22), we may speculate that elevated serum endocan might reflect tumor angiogenesis and thus might be a potential prognostic factor.

Our first approach to test this hypothesis was to look for any correlation between serum endocan levels and survival. We followed up 30 patients with NSCLC for a median of 10 months (25th-75th centiles of 4-24 months). All but four patients died during the follow-up period. The demographic characteristics of the patients are reported in Table 1. The endocan median serum titer (25th-75th centiles) was significantly higher (P < 0.001, Mann-Whitney U test) for cancer patients [1.74 ng/mL (1.21-4.3 ng/mL)] compared with that for a group of 20 healthy volunteers [0.77 ng/mL (0.51-0.95 ng/mL)] reported previously by us (16). By univariate analysis, overall survival was statistically significantly associated with serum endocan levels (P = 0.013), N₂ or N₃ status (P = 0.03), M₁ status (P = 0.002), and tumor stage (P = 0.011) but not with sex, age, or histologic subtype (adenocarcinoma/squamous cell carcinoma). The latter variables were subsequently not introduced into the multivariate analysis (Table 2). Significant correlations were found between endocan levels and the presence of metastasis (P = 0.025) but not with tumor grade (P = 0.26) or node status (P = 0.09). However, our study included only 30 patients, and eventually in a larger series, significant correlations could be shown especially with node status. The best cutoff level for endocan that differentiated between survivors and nonsurvivors was 1.3 ng/mL. The Cox regression analysis revealed that endocan of >1.3 ng/mL was a significant predictor of overall survival only when tumor-node-metastasis classification of tumor extension was not taken into account. When only tumor and node descriptors for tumor extension were taken into account, endocan was still predictive of overall survival but not after introduction of metastasis status into the model. As expected, the tumor tumor-node-metastasis stage descriptor remained the strongest predictor of survival (P = 0.002).

Based on the above indicators, we divided the patients into two groups according to a serum endocan cutoff level of 1.3 ng/mL. Median survival was significantly different (P = 0.016, log-rank test) between the two groups: 6 months (95% confidence interval, 1-11 months) for the high endocan group [endocan median, 3.78 ng/mL (1.62-4.99 ng/mL)] versus 33 months (95% confidence interval, 9-56 months) for the low endocan group [endocan median, 1 ng/mL (0.55-1.2 ng/mL); Fig. 4A]. Similar results were obtained when considering the time to tumor progression [i.e., a significantly longer time (11 months; 95% confidence interval, 9-56 months) in the low endocan group compared with that in the high-endocan group (4 months; 95% confidence interval, 3-9 months; Fig. 4B)].

Therefore, high serum endocan levels seem to be a predictor of bad prognosis in NSCLC, albeit it is not an independent prognostic factor.

Lung cancer is the leading cause of death by cancer in developed countries (1). The increasing incidence, the absence of blood markers of the disease, and the poor prognosis underline the need to search for new strategies for diagnosis, follow-up, and treatment. The development of solid tumors involves early angiogenesis and reorganization of the microenvironment that is pivotal for tumor growth. Both stromal cells and extracellular matrix components contribute to host-tumor relationships and to the successful growth of the tumor (23–25). It is now accepted that the growth of solid tumors is dependent on their capacity to acquire a blood supply, and much effort has been directed toward the development of agents that disrupt this process (26).

Endocan is a secreted endothelial cell–restricted dermatan sulfate proteoglycan (27), which is preferentially expressed in lung and kidney tissues (21), and exhibits growth-promoting activities both in vitro and in vivo (2, 3). Endocan interacts with growth factors, such as hepatocyte growth factor/scatter factor, through its glycan chain, thereby eliciting epithelial cell growth in vitro (2). The crucial role of the glycan chain was further shown when 293 cells transfected with wild-type endocan, but not when transfected by its nonglycanated mutant, formed tumors in the skin of severe combined immunodeficient mice. However, these in vivo results also revealed that growth-promoting activity is also dependent on the protein core of endocan, in particular, a phenylalanine-rich region centered on amino acid residues 115 and 116 (3). So, endocan that is a molecule sustaining tumorigenesis (3) seems to be a potential target for antitumor therapy.

VEGF is a major player in angiogenesis by virtue of its ability to induce vascular permeability, to induce and sustain endothelial cell division and survival, and to initiate the sprouting of new blood vessels to supply tumors with oxygen and nutrients. VEGF expression has been described as an important prognostic factor in NSCLC (28). As already described in the literature (29), we found that VEGF is also overexpressed in the NSCLC samples studied but also that it correlates with endocan expression. To confirm a direct link between VEGF and endocan, we looked at the possible up-regulation of endocan by VEGF. We have shown in vitro, that VEGF induces endocan expression in endothelial cells at both mRNA and protein levels. Moreover, anti-VEGF antibodies inhibit VEGF-induced endocan secretion by endothelial cells. With this data in mind, it could be interesting to assess serum endocan kinetics in patients treated with various combination schemes especially with antiangiogenic therapy.

An important point concerns the potential role of endocan in the process of tumor angiogenesis. Inhibition of endocan expression by antisense oligonucleotides does not modify in vitro angiogenesis (7). In addition, endocan is not involved in fibroblast growth factor-2–induced endothelial cell proliferation as it is not inhibited by anti-endocan antibodies (3). Therefore, the cellular target of endocan seems to be the malignant epithelial cell rather than the endothelium itself. We could speculate that endocan acts locally in a paracrine positive feedback loop and could be an important player in supporting tumor growth (i.e., angiogenic growth factors stimulate endothelial cells to secrete endocan), which in turn acts by amplifying the effect of protumoral growth factors (2) and by inhibiting effective immune cell migration into the tumor (4). This putative role of endocan is also sustained by data obtained from mouse models, in which anti-endocan antibodies prevent or slow down the growth of endocan-expressing xenografts in severe combined immunodeficient mice (3) without any direct in vitro antiproliferative effect. The fact that endocan is highly expressed at the protein level, as detected by immunohistochemistry, in the immediate vicinity of the tumor as well as in areas of pleural and nodal invasion, also argues for a local paracrine effect of endocan. Moreover, this role of endocan would not be restricted solely to bronchopulmonary tumors but would be shared by other tumoral localizations, and indeed, it has been identified and linked to prognosis in cancers other than those of the lung (9, 10).

The involvement of endocan in tumor development and its preferential expression in lung has led us to explore the expression of endocan in lung cancer. Recent data suggested that endocan mRNA is intensely overexpressed in several tumors (8–11). In lung cancer, endocan overexpression has been linked to worst prognosis as part of a 30-gene cluster predicting shorter survival (10). In breast cancer, endocan overexpression has been identified along with another 70-gene signatures as a marker of early (<5 years) development of metastasis (9). Moreover, in human melanoma cell lines, the overexpression of endocan was associated with a highly invasive phenotype (11).

In the present work, it has been shown that endocan mRNA is overexpressed in lung tumors compared with normal lung tissue from the same individuals, in agreement with the data obtained from microarray analysis (10). This overexpression concerned mainly the exon 2-containing isoform of endocan, which has been shown to specifically exert a protumoral effect (3). On the contrary, the increase of the endocanΔ2 isoform in tumors, compared with the corresponding normal lung, was relatively small and did not reach statistical significance. Similarly, in vitro stimulation of endothelial cells by VEGF results mainly in an elevation of expression of endocan mRNA. Thus, both in vitro and in vivo data consistently show that endocan mRNA overexpression predominantly involves the full-length form of endocan.

We also showed that the elevated mRNA expression translates into endocan protein production by endothelial cells in all tumors studied. The intense expression of endocan at the invasion front (adjacent to lung nontumoral parenchyma and visceral pleura) and in lymph node metastasis also suggests that endocan is expressed preferentially in areas of intense tumor growth. The strong expression of endocan in nontumoral lung adjacent to carcinoma could reflect the role that vessels in nontumoral lung, but located in the relatively close vicinity of carcinoma cells, could have on tumor growth, and argues for a potential paracrine effect of endocan on tumoral cells.

Endocan could therefore represent a marker of endothelial cell activation in response to proangiogenic signals in tumors, its overexpression being correlated with a more rapid tumor growth. Interestingly, overexpression of endocan mRNA, as evaluated by microarray technology, has already been described as belonging to the molecular signature of bad prognosis in lung (10) and breast cancers (9). As endocan is a secreted protein, it can be quantified in blood by ELISA (15). We show that, in a small group of NSCLC patients who have not received any previous therapy, serum endocan values were significantly elevated compared with healthy controls. High endocan values were significantly correlated with the presence of metastasis and with limited survival. However, in this small group, endocan was not an independent prognostic factor. A trend exists, also linking high serum endocan levels and distant nodal invasion (N2-N3), a known factor of bad prognosis, but due to the limited number of cases in our patient series it fails to reach statistical significance.

Contrary to tumor status, the presence of distant nodal invasion and metastasis is not always easy to evaluate in daily practice, even if they represent critical determinants of survival in NSCLC (12). Moreover, in NSCLC, up to 40% of those patients who are subjected to surgery with curative intent die of distant metastases in the first 24 months after surgery (30). As serum endocan is correlated with distant spread of the tumor, this blood marker might be useful as an adjunct in the evaluation of the extension of lung cancer or possibly even for early detection of recurrence after attempted treatment. However, our patient sample was not large enough and a more thorough investigation in a larger series of patients will be necessary to confirm the use of this new variable.

In conclusion, we show that endocan is overexpressed in NSCLC, both at mRNA and protein levels, and might reflect angiogenesis mediator-induced activation of tumor endothelium. Serum endocan level is a prognostic factor, albeit not an independent one, and might reflect the intensity of the tumoral angiogenic stimulation.

The Thoracic Oncology Group includes Arnaud Scherpereel as the coordinator; Benoît Wallaert, Charles-Hugo Marquette, Jean-Jacques Lafitte, Philippe Ramon, Isabelle Tillie-Leblond, and André-Bernard Tonnel (Clinique des Maladies Respiratoires, Centre Hospitalier Régional Universitaire de Lille); Alain Wurst, Henri Porte, and Massimo Conti (Service de Chirurgie Thoracique, Centre Hospitalier Régional Universitaire de Lille); Bernard Gosselin and Marie Christine Copin (Service d'Anatomie et Cytologie Pathologiques, Centre Hospitalier Régional Universitaire de Lille); Eric Mensier and Michel Debaert (Polyclinique La Louvière, Lille, France); Jean-Marie Faillon and Sophie Jaillard (Clinique du Bois, Lille, France); Nicolas Just and François Steenhouver (Hôpital V Provo, Roubaix, France); and Jean-Paul Grignet (Centre Hospitalier de Denain, Denain, France).

We thank Anne Tsicopoulos, Catherine Duez, and Philippe Gosset (Institut National de la Santé et de la Reserche Médicale U774, Lille, France) for their helpful critical reading and suggestions; Han Vorng and Genevieve Marchandise for their kind technical help in these investigations; and Dr. Malcolm Lyon (Patterson Institute for Cancer Research, Manchester, United Kingdom) for his helpful suggestions.