Serum Endostatin Levels Are Elevated and Correlate with Serum Vascular Endothelial Growth Factor Levels in Patients with Stage IV Clear Cell Renal Cancer

4628 Vol. 6, 4628–4634, December 2000 Clinical Cancer Research

Advances in Brief Serum Endostatin Levels Are Elevated and Correlate with Serum Vascular Endothelial Growth Factor Levels in Patients with Stage IV Clear Cell Renal Cancer

Andrew L. Feldman, Lawrence Tamarkin, Introduction Giulio F. Paciotti, Byron W. Simpson, Renal cell carcinoma accounts for ϳ2% of all cancers, and W. Marston Linehan, James C. Yang, its incidence is increasing in North America and northern Eu- William E. Fogler, Ewa M. Turner, rope (1). Five-year survival rates, although gradually improving, 1 remain around 50–60% (1) because of the high resistance of H. Richard Alexander, Jr., and Steven K. Libutti metastatic disease to systemic therapy (2). Renal cancers gen- Surgery Branch [A. L. F., J. C. Y., E. M. T., H. R. A., S. K. L.] and erally are highly vascular tumors known to secrete the proan- Urologic Oncology Branch [W. M. L.], National Cancer Institute, giogenic cytokine VEGF2 in vitro and in vivo (3–7). Tumor Bethesda, Maryland 20892; CytImmune Sciences, Inc., College Park, Maryland [L. T., G. F. P., B. W. S.]; and EntreMed, Inc., Rockville, VEGF expression is correlated with the severity of disease in Maryland [W. E. F.] patients with renal cell carcinoma (3, 4, 8), and some authors have suggested using circulating VEGF as a prognostic factor or tumor marker (3, 5, 9). Abstract In addition to producing proangiogenic cytokines, recent Clear cell renal carcinoma (CCRC) is a highly angio- data demonstrate that tumors can produce antiangiogenic cyto- genic tumor known to secrete vascular endothelial cell kines as well (10, 11). It has been suggested that, in humans, the growth factor (VEGF). Endostatin is an endogenous antian- generation of antiangiogenic compounds in the presence of a giogenic agent with antitumor activity in mice. The purpose primary tumor suppresses the growth of distant metastases (12). of this study was to evaluate serum levels of endostatin in This phenomenon has been demonstrated in mice (11, 13, 14). normal subjects and in patients with CCRC and to examine However, the presence of endogenous antiangiogenic cytokines the relationship of these levels to circulating VEGF levels. in patients with renal cell carcinoma has not been reported. In Fifteen patients (mean age, 48 years) on a clinical protocol this study, we sought to determine whether circulating levels of for stage IV CCRC at the National Cancer Institute were endostatin, an antiangiogenic cleavage product of C18 (10), included in the study. Archived prenephrectomy serum sam- were elevated in patients with stage IV CCRC. ples were analyzed for endostatin and VEGF concentrations. Endostatin and VEGF levels were compared with those of Patients and Methods (18 ؍ an age-matched group of volunteer blood donors (n Patients. Medical records and the patient serum archives using a competitive enzyme immunoassay. Data were ana- of the Surgery Branch and Urological Oncology Branch, NCI, lyzed using the Mann-Whitney U test and the Spearman were reviewed to identify patients who met the following cri- rank correlation. Median serum endostatin levels were 24.6 teria: (a) clinical and histological diagnosis of stage IV CCRC; ng/ml (range, 15.1–54.0 ng/ml) in CCRC patients versus 14.1 (b) enrollment in an institutional review board-approved clinical ng/ml (range, 1.0–19.3 ng/ml) in healthy controls (P < protocol at the NCI; and (c) multiple aliquots of a prenephrec- 0.0001). Median VEGF levels were 3.4 ng/ml (range, 0.1– tomy serum sample archived. Tumor staging was based on the 11.2 ng/ml) and 2.5 ng/ml (range, 0.1–4.2 ng/ml), respec- Robson staging system (15). Demographic and clinical data on -A highly significant correlation was ob .(0.065 ؍ tively (P these patients were recorded. Tumor volumes were calculated served between endostatin and VEGF levels among the using the formula: width ϫ length ϫ height ϫ␲/6. but not among (0.0003 ؍ P ,0.81 ؍ CCRC patients (r In addition, we obtained multiple aliquots of human serum -Endostatin levels are detect .(0.37 ؍ ؊0.22, P؍ controls (r from randomly selected volunteer blood donors (controls) from able in serum from healthy subjects as well as from CCRC the Department of Transfusion Medicine, NIH. All serum sam- patients. Levels are significantly elevated and correlate with ples were negative by ELISA for HIV as well as hepatitis B and VEGF levels in CCRC patients. Elucidating the nature of C. We were provided with the age and gender of each control this correlation may lend insight into the regulation of tu- but received no other identifying information. mor angiogenesis in patients with renal cancer. Methods. Archival serum samples collected between 1988 and 1995 from patients with CCRC who participated in an institutional review board-approved NCI study were stored in Received 4/18/00; revised 8/22/00; accepted 10/10/00. 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. 2 The abbreviations used are: VEGF, vascular endothelial growth factor; 1 To whom requests for reprints should be addressed, at Surgery Branch, CCRC, clear cell renal carcinoma; C18, collagen XVIII; NCI, National National Cancer Institute, NIH, 9000 Rockville Pike, Bethesda, MD Cancer Institute; EIA, enzyme immunoassay; CV, coefficient of varia- 20892. Phone: (301) 496-6457; Fax: (301) 402-1788. tion; EC, endothelial cell. Downloaded from clincancerres.aacrjournals.org on September 26, 2021. © 2000 American Association for Cancer Research. Clinical Cancer Research 4629

liquid nitrogen. After retrieval from liquid nitrogen, they were centrated ϳ4-fold compared with the original serum. Ten ␮lof stored at Ϫ80°C until ready for use. Serum samples from more each sample were run under reducing conditions on a SDS- recent CCRC patients and from volunteer blood donors were polyacrylamide gel (NuPAGE; Novex) and transferred to a stored at Ϫ80°C until ready for use. Sera were thawed at room nitrocellulose membrane (Hybond; Amersham Pharmacia Bio- temperature, and VEGF and endostatin levels were measured tech, Buckinghamshire, England). The membrane was incubated using a competitive EIA (ACCUCYTE; CytImmune Sciences, with 510 ng/ml purified rabbit antihuman endostatin polyclonal Inc., College Park, MD). Each assay was developed with its IgG antibody (Cytimmune Sciences), followed by horseradish respective recombinant protein. Recombinant VEGF was ob- peroxidase-conjugated goat antirabbit IgG antibody (Santa Cruz tained from PeproTech, Inc. (Princeton, NJ), whereas recombi- Biotechnology, Santa Cruz, CA). Images were visualized using nant endostatin was obtained from EntreMed, Inc. (Rockville, a chemiluminescent detection kit (ECL and Hyperfilm ECL; MD). These recombinant proteins were the antigens for the Amersham). generation of the rabbit polyclonal antiserum and were biotiny- Parallelism of the EIA was evaluated using a serum and lated to serve as the competitive ligands. After obtaining plasma sample from a normal volunteer assayed at varying sufficient titers of antibody, sera were fractionated by high- dilutions. The initial dilution of the sample was 1:4. The sam- performance liquid chromatography to obtain enriched immuno- ples were further diluted using 2-fold serial dilutions. The raw globulin preparations. Briefly, the assay was run by making 4- potency estimates were determined and then corrected for the or 5-fold dilutions of samples, which were added to a 96-well dilution factor. plate coated with goat antirabbit polyclonal IgG antibody. After Quantitative recovery was assessed by adding a known addition of the respective competitive ligands, rabbit antihuman amount of recombinant endostatin protein to a normal serum or VEGF or endostatin polyclonal IgG antibody was added, and the plasma sample. A 0.5-ml aliquot of each sample received a 25 plates were incubated at room temperature for 3 h. After thor- ␮l “spike” of a 2000 ng/ml stock endostatin solution, increasing ough washing, streptavidin-conjugated alkaline phosphatase the endogenous immunoreactivity by 100 ng/ml. A second and was added and incubated for 30 min at room temperature to a third aliquot of the spiked and unspiked samples underwent dephosphorylate NADPH to NADH. After further washing, one or two acute (i.e., within the same day) freeze (Ϫ80°C)/ color reagents containing alcohol dehydrogenase and diaphorase thaw (25°C) cycles. All samples were then analyzed by the were added. These reagents use NADH as a cofactor to generate ACCUCYTE endostatin EIA. formazan. Absorbances were measured at 492 nm when the A492 Cross-reactivity of the assay was tested against collagen I, for the negative control well was between 1.5 and 2.0. Each collagen IV, vitronectin, fibronectin, bFGF, and Angiostatin and sample was analyzed in triplicate, and concentrations were compared with the immunoreactivity of an equivalent amount of calculated with reference to a standard curve using Microplate recombinant endostatin and heat-inactivated endostatin. To Manager III (Bio-Rad, Hercules, CA). Each EIA was run heat-inactivate endostatin, a 5 mg/ml solution of endostatin was blinded to the origin of the serum samples. heated to 90°C for 5 min. Each cross-reactive test molecule was Rabbit antibodies against recombinant human endostatin run in the assay at a maximum concentration of 400 ng/ml. for use in the EIA and Western blotting were generated by the Two-fold serial dilutions of each test molecule to 50 ng/ml were method described by Shiosaka et al. (16). Briefly, 0.5 mg of also run to determine the specificity of any potential cross- recombinant human Endostatin (EntreMed, Inc., Rockville, reactivity. The potency estimates were determined based on the MD) was bound to 2 ml of 32-nm colloidal gold (CytImmune recombinant endostatin standard curve. Sciences, Inc., College Park, MD) at pH 8. The solution was Intraassay variance was measured by analyzing three rep- emulsified in Freund’s Complete Adjuvant and administered s.c. licates of each of 26 serum samples. The mean of the individual into a New Zealand White rabbit. Two weeks later, the rabbit potency estimates for each replicate was calculated, and the CV was boosted with endostatin/colloidal gold emulsified in for the sample was determined. To determine interassay varia- Freund’s Incomplete Adjuvant. Six weeks after the initial im- tion, aliquots of a single sample were analyzed over 5 days using munization, a 30-ml blood sample was collected. The serum was 24 different plates. fractionated on a Bakerbond AbX mixed ion exchange HPLC The range of detection for the VEGF EIA was 0.195 to column (Baker) according to manufacturer’s specifications. 50.0 ng/ml, whereas the range for the endostatin EIA was 1.95 Fractions corresponding to the rabbit antibody peak were pooled to 500 ng/ml. Calculated concentrations exceeding the upper and dialyzed against TBS. limit of detection for each EIA were reassayed using appropriate The ACCUCYTE endostatin EIA was validated by West- dilutions. Concentrations below the lower limit of detection ern blot analysis, parallelism, quantitative recovery, and cross- were set at the midpoint between 0 and this lower limit (i.e., reactivity studies. Additionally, intra- and interassay variability 0.098 and 0.98 ng/ml for VEGF and endostatin, respectively). was assessed. Two or three serum aliquots from the same venipuncture For Western blotting, 800-␮l serum samples from CCRC were analyzed for each subject. Subjects for whom only one patient 15 (Table 2; patient selected because of sample avail- aliquot was available were excluded from the study. The CV ability) and a healthy, 45-year-old male control were diluted among the samples from each subject was calculated. For sub- 5-fold in 0.5% SDS and heated at 56°C for 5 min. The diluted jects from whom only two samples were available, the subject samples were centrifuged in Microcon 100,000 MWCO col- was excluded if the CV exceeded 40%. For subjects with three umns (Millipore, Bedford, MA), and the filtrates were concen- samples and a CV exceeding 40%, the outlying value was trated ϳ20-fold in Microcon 10,000 MWCO columns. Thus, discarded, and the CV was recalculated using the two remaining serum proteins between Mr 10,000 and Mr 100,000 were con- values. If the CV still exceeded 40%, the subject was excluded.

Table 1 Demographic data for CCRC patients and healthy controls Median serum VEGF levels were 3.4 ng/ml (range, 0.1–11.2 n Mean age (range)a Male (%) Female (%) ng/ml) in CCRC patients and 2.5 ng/ml (range, 0.1–4.2 ng/ml) in controls (P ϭ 0.065). As defined in “Patients and Methods,” CCRC 15 48 (26–69) 12 (80) 3 (20) Control 18 48 (26–68) 9 (50) 9 (50) the upper limits of normal for serum concentrations of endostatin and VEGF were 22.1 and 4.3 ng/ml, respectively. By these a In years. criteria, endostatin levels were abnormally elevated in 8 of 15 (53%) CCRC patients, and VEGF levels were abnormal in 6 of 15 (40%). To determine whether the difference in gender distribution VEGF and endostatin levels were represented as the mean of contributed to the difference in cytokine levels in the two sample values considered concordant using this method. groups, we analyzed endostatin and VEGF levels in CCRC The renal cancer cell lines 1581-RCC, 1764-RCC, UOK- patients and controls stratified by gender (Fig. 2). In the control 125, and UOK-131 (developed in our laboratories), as well as group, median endostatin levels were 13.8 ng/ml (range, 11.1– the transformed human embryonic kidney cell 293 (American 15.4 ng/ml) and 15.5 ng/ml (range, 1.0–19.3 ng/ml) in men and Type Culture Collection, Manassas, VA), were assayed for women, respectively (P ϭ 0.67). Median VEGF levels were 2.5 supernatant endostatin and VEGF concentrations. Cells were ng/ml (range, 0.3–4.2 ng/ml) and 2.4 ng/ml (range, 0.1–3.7 ϫ 5 plated in 12-well plates at a density of 5 10 cells/well in 0.5 ng/ml) in men and women, respectively (P Ͼ 0.99). In the ml of DMEM containing 10% FCS, 100 units/ml penicillin, 100 CCRC patients, median endostatin levels were 21.9 ng/ml ␮ ␮ ␮ g/ml streptomycin, 50 g/ml gentamicin, 0.5 g/ml Fungi- (range, 15.1–54.0 ng/ml) and 40.6 ng/ml (range, 19.2–44.5 zone, and 4 mM glutamine (Biofluids, Rockville, MD). After ng/ml) in men and women, respectively (P ϭ 0.63). Median 24 h incubation at 37°C, supernatants were harvested and cen- VEGF levels were 3.2 ng/ml (range, 0.1–10.8 ng/ml) and 4.8 ϫ trifuged at 2000 g for 5 min. All experiments were performed ng/ml (range, 3.4–11.2 ng/ml) in men and women, respectively in triplicate. Each sample was assayed in duplicate for endosta- (P ϭ 0.18). tin and VEGF concentrations by EIA as described above. Correlation between Serum Endostatin and VEGF The upper limit of normal for endostatin and VEGF levels Levels. There was no evidence of correlation between serum was defined as 2 SD above the mean. Comparisons between endostatin and VEGF levels among the healthy controls (r ϭ groups were performed using the Mann-Whitney U test to Ϫ0.22, P ϭ 0.37; Fig. 3). There was, however, a highly signif- compare groups according to their median values with no as- icant correlation between endostatin and VEGF levels among sumption about the scatter of the data. Correlations were per- the CCRC patients (r ϭ 0.81, P ϭ 0.0003; Fig. 3). Neither formed using the Spearman rank correlation. All calculations endostatin levels nor VEGF levels were correlated with the size Ͻ were done using Instat 2.01 (GraphPad Software), and P 0.05 of the primary tumor (r ϭ 0.05, P ϭ 0.82; and r ϭ 0.26, P ϭ was considered significant. 0.16, respectively; n ϭ 14). Similarly, there was no significant difference in median levels of endostatin or VEGF in CCRC Results patients with or without renal vein thrombus (endostatin, P ϭ Demographic Data for CCRC Patients and Healthy 0.36; VEGF, P ϭ 0.94). Controls. As described in “Patients and Methods,” subjects Supernatant Concentrations of Endostatin and VEGF for whom serum endostatin and/or VEGF levels were not con- in Renal Carcinoma Cell Lines. There was no detectable cordant among aliquots were excluded from the study. Of 18 endostatin in the supernatants of any of the cell lines tested patients with CCRC who met the other inclusion criteria for the (Table 3). VEGF levels varied from below the limit of detection study, 3 patients (17%) were excluded because of nonconcor- (0.195 ng/ml) to 7.3 ng/ml. Neither endostatin nor VEGF was dance of values. Of 19 controls, 1 (5%) was excluded because of detectable in the supernatant of 293 cells. nonconcordance. Thus, 15 patients with CCRC and 18 controls Validation of ACCUCYTE Endostatin Competitive were analyzed in this study (Table 1). Mean age was 48 years in Enzyme Immunoassay. Western blot analysis of the serum both groups with similar age ranges. Of the 15 patients with from patient 15 (endostatin concentration, 44.3 ng/ml by EIA) CCRC, 12 (80%) were men, whereas 9 (50%) of the 18 controls revealed a band of endogenous endostatin immunoreactivity were men. with mobility equivalent to that of recombinant human endosta- Clinical Characteristics of Patients with CCRC. Of tin (Fig. 4). This band was not visualized in the lane containing the 15 CCRC patients, the size of the primary tumor was the normal human serum sample (endostatin concentration, 13.4 documented in 14 and ranged from 16 to 2246 cm3 (median, 305 ng/ml by EIA). cm3; mean, 460 cm3; Table 2). Tumor thrombus in the renal The validation data derived from studies of parallelism are vein was present in 3 (20%) of the 15 patients, and the inferior presented Fig. 5. Dilution of both the serum and plasma sample vena cava was uninvolved in all patients. All but one patient had led to diminished endostatin immunoreactivity paralleling the solid organ metastases, with the lung being the most common dilution curve of the recombinant endostatin standard. Quanti- site (12 of 14 patients). The remaining patient was classified as tative recovery validation of the ACCUCYTE endostatin EIA

stage IV because of N2 nodal disease. yielded recovery of the spike ranging from 83 to 96%, with an Serum Endostatin and VEGF Levels in Patients with average recovery of 90%. Of the molecules tested (collagen I, CCRC and Controls. Median serum endostatin levels were collagen IV, vitronectin, fibronectin, basic fibroblast growth 24.6 ng/ml (range, 15.1–54.0 ng/ml) in CCRC patients and 14.1 factor, and angiostatin), angiostatin had the highest cross-reac- ng/ml (range, 1.0–19.3 ng/ml) in controls (P Ͻ 0.0001; Fig. 1). tivity, which was 0.04% as compared with an equivalent dose of

Table 2 Clinical characteristics of 15 patients with stage IV CCRC Size of primary Renal vein Patient Age Sex tumor (cm3) thrombus Metastatic sites 1 30 F 2246 ϩ Lung, bone 2 26 M 1014 Ϫ Iliac lymph nodes 3 38 M 277 Ϫ Lung 4 40 M 513 Ϫ Lung 555F 42 Ϫ Lung 648M NDa Ϫ Lung 7 52 M 333 Ϫ Lung 8 69 M 112 Ϫ Spinal cord, liver 9 54 M 250 ϩ Lung, adrenal 10 60 M 29 Ϫ Lung, liver 11 54 M 379 ϩ Brain 12 50 M 757 Ϫ Lung 13 51 F 16 Ϫ Lung 14 47 M 140 Ϫ Lung 15 44 M 337 Ϫ Lung, bone, heart a ND, not documented.

Fig. 1 Serum endostatin (A) and VEGF (B) levels in CCRC patients and healthy controls. ‚, individual subjects; ----, median values. En- dostatin levels were significantly elevated in the CCRC patients (P Ͻ 0.0001, Mann-Whitney U test). VEGF levels were higher in CCRC patients than in controls, but this difference was not significant (P ϭ 0.065).

endostatin. Heat-inactivated endostatin retained 1% immunoreactivity. No visible precipitates were observed in the boiled preparation; therefore, the diminished immunoreactivity of heat- inactivated protein did not appear related to a change in solu- bility. Under less damaging conditions, such as preparing the molecule for Western blot analysis under reducing conditions, Fig. 2 Effect of gender on endostatin and VEGF levels. ‚, individual the endostatin retained its immunoreactivity (data not shown; subjects; ----, median values. In healthy controls, endostatin (A) and the endostatin was not heated for the Western blot in Fig. 4). VEGF (B) levels were not significantly different between men and Intraassay evaluation of three replicates of each of 26 serum women (P ϭ 0.67 and P Ͼ 0.99, respectively; Mann-Whitney U test). samples yielded an average CV of 6.4%. Evaluation of interas- In CCRC patients, no significant differences were observed (endostatin: C, P ϭ 0.63; VEGF: D, P ϭ 0.18), although there were only three say variation yielded a SD of 0.83 ng/ml and a CV of 13%. women in this group.

Discussion In this study, we demonstrated that serum endostatin levels are elevated in the majority of patients with stage IV CCRC. urable levels of endostatin in their supernatants (10). This ob- Endostatin is one of a number of endogenously generated anti- servation includes the evaluation of renal cancer cell lines in the angiogenic protein fragments that have been shown to have present study. antitumor activity in murine models (10). Endostatin originally The nucleotide sequence encoding endostatin resides was identified in the supernatant of the murine hemangioendo- within the COOH-terminal, noncollagenous domain of C18, thelioma cell line, EOMA. Interestingly, we are not aware of termed NC1 (10). C18 has been shown to be a member of the any other murine or human tumor cell lines that produce meas- unique collagen family, the multiplexins, that resides in base-

Fig. 4 Western blot of concentrated serum from CCRC patient 15 (see Table 2) and a healthy 45-year-old male control. Lanes 1 and 12, molecular weight marker. Lanes 2–5, recombinant endostatin standard at concentrations of 250, 125, 62.5, and 31.25 ng/ml. Lane 7, CCRC patient 15 (serum endostatin concentration, 44.4 ng/ml). Lane 9, healthy control (serum endostatin concentration, 13.4 ng/ml). Lanes 6, 8, 10, and Fig. 3 Correlation between serum endostatin and VEGF levels in 11, blank. CCRC patients and controls. Although there was no evidence of correlation among the healthy controls (r ϭϪ0.22, P ϭ 0.37; Spearman rank correlation), the correlation between serum endostatin and VEGF levels among CCRC patients was highly significant (r ϭ 0.81, P ϭ 0.0003).

Table 3 Endostatin and VEGF concentrations in renal cancer cell line supernatants Cell line Endostatin (ng/ml) VEGF (ng/ml)a 1581-RCC NDb 6.5 Ϯ 2.1 1764-RCC ND ND UOK-125 ND 0.8 Ϯ 0.1 UOK-131 ND 7.3 Ϯ 0.8 293c ND ND a Means Ϯ SE. b ND, not detected. c Included as a reference, although not a renal cancer cell line.

Fig. 5 Dilution of human serum and plasma plotted against the recombinant endostatin standard curve for the ACCUCYTE endostatin assay. ment membranes and particularly in the liver (17–19). Most Higher dilutions corresponded to higher absorbances, which when plotted against the recombinant standard fell on the recombinant standard collagens undergo cleavage of their noncollagenous domains dilution curve, indicating that serum and plasma demonstrate immuno- after secretion from the cell (19, 20). This phenomenon would reactive behavior parallel to that of recombinant endostatin. explain the lack of endostatin secretion from most tumor cell lines in vitro, including those from renal cell carcinomas. More likely, endostatin is cleaved from C18 extracellularly. Recently, elastase and cathepsin activities have been shown to cleave the larger, ill-defined areas of intensity on the Western blot are endostatin from NC1 in vitro (21, 22). The mechanisms of most likely attributable to nonspecific binding to ubiquitous endostatin generation in vivo remain unknown. serum proteins. John et al. (26) have reported the presence in Our findings suggest that endostatin is detectable in healthy human plasma of a number of endostatin antibody-immunore- subjects. The validation data evaluating parallelism and quanti- active protein fragments; however, these are relatively small

tative recovery using the ACCUCYTE endostatin EIA show that molecules with molecular weights ranging from Mr 16,000 to the immunoreactivity detected in normal human serum and Mr 22,000. plasma is essentially identical to that of recombinant endostatin. In this study, patients with CCRC had higher serum VEGF

The absence of a Mr 20,000 endostatin-immunoreactive band levels than healthy controls, although this difference was not Western blotting of normal human serum was to be expected, quite significant. Given previous data in the literature with because the 4-fold concentration of serum samples would yield larger numbers of patients (3), we suspect that the high VEGF an estimated endostatin concentration (53.6 ng/ml) below the levels seen in our series represent a real phenomenon. Nonethe- sensitivity of the blot (125 ng/ml). Endostatin in the circulation less, the difference in endostatin levels between CCRC patients of healthy subjects may play a role in the homeostatic regulatory and controls was highly significant. Elevations in circulating network controlling angiogenesis, termed the “angiogenic endostatin levels are not unique to patients with cancer. Per- switch” (23–25). Alternatively, it may be generated as a by- turbed regulation of angiogenesis is an important feature of product of physiological collagen turnover. rheumatoid, vascular, and other nonneoplastic diseases (25); Western blotting of the concentrated serum sample from recently, circulating endostatin levels were reported to be ele- CCRC patient 15 (calculated endostatin level after concentra- vated in patients with systemic sclerosis (27). tion, 177.2 ng/ml), however, revealed a band with mobility Most importantly, serum endostatin levels correlated sig- equal to that of recombinant endostatin (Fig. 4). We believe that nificantly with serum VEGF levels in patients with CCRC but

in activity to the process which created them to keep such pathological processes in check or as part of the “off” switch that controls physiological angiogenesis. The relationships between PEX and matrix metalloproteinase-2 and between angiostatin and macrophage-derived metalloelastase may be other examples of this phenomenon (32, 33). VEGF, in addition to promoting EC proliferation, also induces the release of multiple collagenases and other proteases from ECs (34–36). Therefore, the collagenase that cleaves endostatin from C18 in vivo may also be EC derived and its expression up-regulated by VEGF binding to its EC receptors (Fig. 6B). We believe that both of these processes may occur simultaneously, because tumor neoangiogenesis is known to Fig. 6 Two hypotheses to explain the correlation of serum endostatin involve degradation of extracellular matrix components by both and VEGF levels in CCRC. A, in addition to VEGF, tumors secrete endothelial cells and tumor cells. The expanding list of endog- multiple collagenases, one of which may cleave endostatin from C18 in enous pro- and antiangiogenic factors reveals the complexity of the extracellular matrix and basement membrane. B, alternatively, the homeostatic system controlling angiogenesis. VEGF may regulate the expression and/or secretion of this collagenase from endothelial cells. These hypotheses are not mutually exclusive. In conclusion, serum endostatin levels can be reliably determined using a competitive enzyme immunoassay, are detectable in normal subjects, and are elevated in patients with CCRC. Furthermore, endostatin levels are significantly correlated with not in controls. There was no correlation of either endostatin or VEGF levels in CCRC patients but not in healthy controls. We VEGF levels with primary tumor size, which represented the hypothesize that elevated endostatin levels represent an attempt majority of disease burden in these patients. The association at a compensatory response to the angiogenic phenotype of between endostatin and VEGF demonstrated in this study does CCRC. We are currently investigating the interrelationship of not necessarily indicate a causal link between the elevated levels endostatin, VEGF, and other related cytokines in vitro and in in of these cytokines in CCRC patients. It is possible that elevated vivo animal models, as well as in patients with other tumor serum levels of VEGF and endostatin in advanced CCRC are histologies. We believe that elucidating the nature of the ho- unrelated and attributable to some clinical factor not examined meostatic relationship between pro- and antiangiogenic, tumor- in the present study. To demonstrate the role of various clinical derived cytokines will be critical in developing and refining factors will require multivariate analysis of a larger group of treatment strategies to inhibit tumor angiogenesis. patients with more detailed prospective data collection. We are now conducting such a study. Alternatively, elevated VEGF and Acknowledgments endostatin levels may be correlated because of the homeostatic We thank S. M. Steinberg, National Cancer Institute, for help with interrelationship between pro- and antiangiogenic substances, the statistical analysis; J. Procter and S. F. Leitman, Department of which has been hypothesized but not fully elucidated. The Transfusion Medicine, NIH, for help in obtaining volunteer blood sam- interaction between VEGF and endostatin might be a direct or ples; and M. H. Klein for help in data collection. an indirect one. We do not believe that VEGF expression increases in References response to circulating endostatin levels in CCRC, because of 1. McLaughlin, J. K., and Lipworth, L. Epidemiologic aspects of renal the angiogenic phenotype of the tumor and the relatively low cell cancer. Semin. Oncol., 27: 115–123, 2000. levels of endostatin present. In vitro, the threshold concentration 2. Motzer, R. J., and Russo, P. Systemic therapy for renal call carcinoma. J. Urol., 163: 408–417, 2000. of murine endostatin for inhibition of endothelial cell prolifer- ϳ 3. Jacobson, J., Rasmuson, T., Grankvist, K., and Ljungberg, B. Vas- ation is 100 ng/ml (10, 28) and may be much higher for human cular endothelial growth factor as prognostic factor in renal cell carci- endostatin (29). Furthermore, our data indicate that VEGF is noma. J. Urol., 163: 343–347, 2000. secreted by at least some renal cancer cell lines in vitro in the 4. Tomisawa, M., Tokunaga, T., Oshika, Y., Tsuchida, T., Fukushima, absence of endostatin. Finally, in a murine model (28), adeno- Y., Sato, H., Kijima, H., Yamazaki, H., Ueyama, Y., Tamaoki, N., and viral delivery of the endostatin gene leads to high plasma Nakamura, M. Expression pattern of vascular endothelial growth factor concentrations of endostatin (mean, 1770 ng/ml) without a con- isoform is closely correlated with tumour stage and vascularisation in renal cell carcinoma. Eur. J. Cancer, 35: 133–137, 1999. comitant elevation in VEGF levels (data not shown). 5. Dirix, L. Y., Vermeulen, P. B., Pawinski, A., Prove, A., Benoy, I., De We therefore have proposed two hypotheses to explain the Pooter, C., Martin, M., and Van Oosterom, A. T. Elevated levels of the correlation of endostatin and VEGF in this study, summarized in angiogenic cytokines basic fibroblast growth factor and vascular endo- Fig. 6. In addition to VEGF, invasive tumors secrete multiple thelial growth factor in sera of cancer patients. Br. J. Cancer, 76: collagenases, including matrix metalloproteinases, which facil- 238–243, 1997. itate digestion of the extracellular matrix and basement mem- 6. Brown, L. F., Berse, B., Jackman, R. W., Tognazzi, K., Manseau, E. J., Dvorak, H. F., and Senger, D. R. Increased expression of vascular perme- brane, allowing the tumor access to the circulation. This is true ability factor (vascular endothelial growth factor) and its receptors in kidney of renal carcinomas as well (30, 31). One of these collagenases and bladder carcinomas. Am. J. Pathol., 143: 1255–1262, 1993. may cleave endostatin from C18 (Fig. 6A). Evolution may have 7. Takahashi, A., Sasaki, H., Kim, S. J., Tobisu, K., Kakizoe, T., favored the development of cleavage products that are opposite Tsukamoto, T., Kumamoto, Y., Sugimura, T., and Terada, M. Markedly

increased amounts of messenger RNAs for vascular endothelial growth 23. Folkman, J., Watson, K., Ingber, D., and Hanahan, D. Induction of factor and placenta growth factor in renal cell carcinoma associated with angiogenesis during the transition from hyperplasia to neoplasia. Nature angiogenesis. Cancer Res., 54: 4233–4237, 1994. (Lond.), 339: 58–61, 1989. 8. Dosquet, C., Coudert, M-C., Lepage, E., Cabane, J., and Richard, F. 24. Rastinejad, F., Polverini, P. J., and Bouck, N. P. Regulation of the Are angiogenic factors, cytokines, and soluble adhesion molecules prog- activity of a new inhibitor of angiogenesis by a cancer suppressor gene. nostic factors in patients with renal cell carcinoma? Clin. Cancer Res., Cell, 56: 345–355, 1989. 3: 2451–2458, 1997. 25. Folkman, J. Angiogenesis in cancer, vascular, rheumatoid and other 9. Baccala, A. A., Zhong, H., Clift, S. M., Nelson, W. G., Marshall, disease. Nat. Med., 1: 27–31, 1995. F. F., Passe, T. J., Gambill, N. B., and Simons, J. W. Serum vascular endothelial growth factor is a candidate biomarker of metastatic tumor 26. John, H., Preissner, K. T., Forssmann, W. G., and Standker, L. response to ex vivo gene therapy of renal cell cancer. Urology, 51: Novel glycosylated forms of human plasma endostatin and circulating 327–332, 1998. endostatin-related fragments of collagen XV. Biochemistry, 38: 10217– 10. O’Reilly, M. S., Boehm, T., Shing, Y., Fukai, N., Vasios, G., Lane, 10224, 1999. W. S., Flynn, E., Birkhead, J. R., Olsen, B. R., and Folkman, J. 27. Hebbar, M., Peyrat, J-P., Hornez, L., Hatron, P-Y., Hachulla, E., Endostatin: an endogenous inhibitor of angiogenesis and tumor growth. and Devulder, B. Increased concentrations of the circulating angiogen- Cell, 88: 277–285, 1997. esis inhibitor endostatin in patients with systemic sclerosis. Arthritis 11. O’Reilly, M. S., Holmgren, L., Shing, Y., Chen, C., Rosenthal, Rheum., 43: 889–893, 2000. R. A., Moses, M., Lane, W. S., Cao, Y., Sage, E. H., and Folkman, J. 28. Feldman, A. L., Restifo, N. P., Alexander, H. R., Bartlett, D. L., Angiostatin: a novel angiogenesis inhibitor that mediates the suppres- Hwu, P., Seth, P., and Libutti, S. K. Antiangiogenic gene therapy of sion of metastases by a Lewis lung carcinoma. Cell, 79: 315–328, 1994. cancer utilizing a recombinant adenovirus to elevate systemic endostatin 12. Prehn, R. T. The inhibition of tumor growth by tumor mass. Cancer levels in mice. Cancer Res., 60: 1503–1506, 2000. Res., 51: 2–4, 1991. 29. Dhanabal, M., Volk, R., Ramchandran, R., Simons, M., and 13. Gorelik, E., Segal, S., and Feldman, M. Growth of a local tumor Sukhatme, V. P. Cloning, expression, and in vitro activity of human exerts a specific inhibitory effect on progression of lung metastases. Int. endostatin. Biochem. Biophys. Res. Commun., 258: 345–352, 1999. J. Cancer, 21: 617–625, 1978. 30. Kugler, A., Hemmerlein, B., Thelen, P., Kallerhoff, M., Radzun, 14. Fisher, B., Gunduz, N., Coyle, J., Rudock, C., and Saffer, E. H. J., and Ringert, R. H. Expression of metalloproteinase 2 and 9 and Presence of a growth-stimulating factor in serum following primary their inhibitors in renal cell carcinoma. J. Urol., 160: 1914–1918, 1998. tumor removal in mice. Cancer Res., 49: 1996–2001, 1989. 31. Walther, M. M., Kleiner, D. E., Lubensky, I. A., Pozzatti, R., 15. Robson, C. J., Churchill, B. M., and Anderson, W. The results of Nyguen, T., Gnarra, J. R., Hurley, K., Venzon, D., Linehan, W. M., radical nephrectomy for renal cell carcinoma. J. Urol., 101: 297–301, 1969. and Stetler-Stevenson, W. G. Progelatinase. A mRNA expression in 16. Shiosaka, S., Kiyama, H., Wanaka, A., and Yohyama, M. A new cell lines derived from tumors in patients with metastatic renal cell method for producing a specific and high titer antibody against gluta- carcinoma correlates inversely with survival. Urology, 50: 295–301, mate using colloidal gold as a carrier. Brain Res., 382: 399–403, 1986. 1997. 17. Rehn, M., and Pihlajaniemi, T. ␣1(XVIII), a collagen chain with frequent interruptions in the collagenous sequence, a distinct tissue 32. Brooks, P. C., Silletti, S., von Schalscha, T. L., Friedlander, M., and distribution, and a homology with type XV collagen. Proc. Natl. Acad. Cheresh, D. A. Disruption of angiogenesis by PEX, a noncatalytic Sci. USA, 91: 4234–4238, 1994. metalloproteinase fragment with integrin binding activity. Cell, 92: 391–400, 1998. 18. Saarela, J., Rehn, M., Oikarinen, A., Autio-Harmainen, H., and Pihlajaniemi, T. The short and long forms of type XVIII collagen show 33. Dong, Z., Kumar, R., Yang, X., and Fidler, I. J. Macrophage- clear tissue specificities in their expression and location in basement derived metalloelastase is responsible for the generation of angiostatin membrane zones in humans. Am. J. Pathol., 153: 611–626, 1998. in Lewis lung carcinoma. Cell, 188: 801–810, 1997. 19. Musso, O., Rehn, M., Saarela, J., Theret, N., Lietard, J., Hintikka, 34. Lamoreaux, W. J., Fitzgerald, M. E. C., Reiner, A., Hasty, K. A., E., Lotrian, D., Campion, J-P., Pihlajaniemi, T., and Clement, B. Col- and Charles, S. T. Vascular endothelial growth factor increases release lagen XVIII is localized in sinusoids and basement membrane zones and of gelatinase A and decrease release of tissue inhibitor of metallopro- expressed by hepatocytes and activated stellate cells in fibrotic human teinases by microvascular endothelial cells in vitro. Microvasc. Res., 55: liver. Hepatology, 28: 98–107, 1998. 29–42, 1998. 20. Kessler, E., Takahara, K., Biniaminov, L., Brusel, M., and Green- 35. Unemori, E. N., Ferrara, N., Bauer, E. A., and Amento, E. P. span, D. S. Bone morphogenetic protein-1: the type I procollagen Vascular endothelial growth factor induces interstitial collagenase ex- C-proteinase. Science (Washington DC), 271: 360–362, 1996. pression in human endothelial cells. J. Cell Physiol., 153: 557–562, 21. Wen, W., Moses, M. A., Wiederschain, D., Arbiser. J. L., and 1992. Folkman, J. The generation of endostatin is mediated by elastase. 36. Zucker, S., Mirza, H., Conner, C. E., Lorenz, A. F., Drews, M. H., Cancer Res., 59: 6052–6056, 1999. Bahou, W. F., and Jesty, J. Vascular endothelial growth factor induces 22. Felbor, U., Dreier, L., Bryant, R. A. R., Ploegh, H. L., Olsen, B. R., tissue factor and matrix metalloproteinase production in endothelial and Mothes, W. Secreted cathepsin L generates endostatin from colla- cells: conversion of prothrombin to thrombin results in progelatinase A gen XVIII. EMBO J., 19: 1187–1194, 2000. activation and cell proliferation. Int. J. Cancer, 75: 780–786, 1998.

Downloaded from clincancerres.aacrjournals.org on September 26, 2021. © 2000 American Association for Cancer Research. Serum Endostatin Levels Are Elevated and Correlate with Serum Vascular Endothelial Growth Factor Levels in Patients with Stage IV Clear Cell Renal Cancer

Andrew L. Feldman, Lawrence Tamarkin, Giulio F. Paciotti, et al.

Clin Cancer Res 2000;6:4628-4634.

Updated version Access the most recent version of this article at: http://clincancerres.aacrjournals.org/content/6/12/4628

Cited articles This article cites 34 articles, 9 of which you can access for free at: http://clincancerres.aacrjournals.org/content/6/12/4628.full#ref-list-1

Citing articles This article has been cited by 10 HighWire-hosted articles. Access the articles at: http://clincancerres.aacrjournals.org/content/6/12/4628.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/6/12/4628. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.