Highly Sensitive Detection of Melanoma at an Early Stage Based on the Increased Serum Secreted Protein Acidic and Rich in Cysteine and Glypican-3 Levels

The incidence rates for melanoma have steadily increased worldwide, and the mortality rates have increased as well. Several molecules have been evaluated as tumor markers to detect melanoma (1–3). Recently, several investigators reported the 5-S-cysteinyldopa and melanoma-inhibitory activity to be useful as a marker for melanoma progression or for monitoring metastatic melanoma (4–9). However, current methods are not sensitive enough to detect melanoma in its early stages. Thus, there is a need for new tumor markers that can detect primary melanoma in the early stages. We recently reported that glypican-3 (GPC3), which is overexpressed in melanoma, is a novel tumor marker for melanoma (10).

Secreted protein acidic and rich in cysteine (SPARC), also called osteonectin or BM-40, is a matricellular glycoprotein that modulates cellular interaction with the extracellular matrix during tissue remodeling (11). SPARC was overexpressed in primary and metastatic melanomas, and an overexpression of SPARC by melanoma cell was associated with an invasive phenotype in vivo (12, 13). In this study, we detected SPARC in the sera of melanoma patients at higher concentrations than in healthy donors. Indeed, SPARC was detected in the sera of 33% of all melanoma patients, irrespective of the clinical stages and even in the sera of patients with stage 0 in situ melanoma. Moreover, the combined use of SPARC and GPC3 will thus make it possible to diagnose melanoma, especially in the early stages (0-II).

Tissues, blood samples, and cell lines. After receiving their informed written consent, we obtained tissue, serum, and plasma samples from the patients with melanoma and large congenital melanocytic nevus treated at the Department of Dermatology, Graduate School of Medical Sciences, Kumamoto University from 1997 to 2004. We also obtained 61 serum samples and 21 plasma samples from age-matched and sex-matched healthy donors from Hiraki hospital (Kumamoto, Japan) after receiving their informed written consents. All samples were anonymized, numbered at random, and stored at −80°C until use. We collected the patient profiles from medical records to determine the clinical stages, according to the Unio Internationale Contra Cancrum/American Joint Committee on Cancer tumor-node-metastasis classification (14). The subjects consisted of 113 consecutive and preoperative patients with melanoma comprising 52 male and 61 female patients with an average age of 67 years (range, 22-91 years): 15 had stage 0 melanoma (in situ); 30 had stage I melanoma; 30 had stage II melanoma; 19 had stage III melanoma; and 19 had stage IV melanoma. Five patients with large congenital melanocytic nevus consisted of four male and one female patients with an average age of 21 years (range, 4-38 years). All patients were of Japanese nationality.

Melanoma cell lines CRL1579, G361, HMV-I, and SK-MEL-28 were kindly provided by the Cell Resource Center for Biomedical Research Institute of Development, Aging, and Cancer, Tohoku University (Sendai, Japan), and 888mel and 526mel were provided by Dr. Yutaka Kawakami (Keio University, Tokyo, Japan). The cell lines were cultured in DMEM or RPMI 1640 supplemented with 10% FCS. Human epidermal melanocytes, neonatal (HEMn), in culture medium 154S supplemented with human melanocyte growth supplements, were purchased from KURABO (Osaka, Japan).

Quantitative reverse transcription-PCR. The SPARC mRNA levels were analyzed using real-time reverse transcription-PCR, as described previously (15). We designed SPARC gene–specific PCR primers to amplify the fragments of 374 bp; SPARC PCR primer sequences were sense, 5′-CGAAGAGGAGGTGGTGGCGGAAAA-3′ and antisense, 5′-GGTTGTTGTCCTCATCCCTCTCATAC-3′. Reaction mixtures contained 2 μL of DNA Master SYBR green I, 1 mmol/L MgCl₂, 0.4 μmol/L of each primer, and 1 μL of cDNA in a total volume of 20 μL. The PCR cycles were 95°C for 10 minutes followed by 35 cycles of 95°C for 1 second, 68°C for 1 second, and 72°C for 16 seconds. For β-actin, we used LightCycler-Primer Set Human β-actin (Search LC, Heidelberg, Germany). The quantification was achieved by comparisons with an internal standard curve containing 10-fold dilutions of HEMn cDNA probe. The relative expressions of SPARC mRNA were calculated as the ratio of the SPARC/β-actin expressions from three replicate reverse transcription-PCR experiments.

Western blot analysis and immunohistochemical examination. SDS-PAGE and Western blotting were done as described previously (16). The membranes were incubated with anti-SPARC monoclonal antibody AON-5031 (Haematological Technologies, Inc., Essex Junction, VT). Immunohistochemical examination was done using the DakoCytomation EnVision+ System according to the manufacturer's instructions with minor modifications, as described previously (17). Briefly, 4-μm-thick paraffin sections were cut and stained with AON-5031 at a dilution of 1:2,000 (0.216 μg/mL). After washing, the sections were incubated for 60 minutes with polymer/horseradish peroxidase–labeled goat anti-mouse IgG at room temperature. 3,3′-Diaminobenzidine tetrahydrochloride was used as the chromogen. The intensity of staining was classified as weak; weaker than the adjacent epidermis, moderate; same as the adjacent epidermis and strong; and stronger than the adjacent epidermis. These samples were estimated independently by two observers in a blinded manner (T.K. and S.F.).

Double-determinant (sandwich) ELISA. The SPARC concentrations in the culture supernatants of melanoma cell lines, sera, and plasma were measured by ELISA in duplicated wells in each plate assay. ELISA was done as described previously (16). All samples were tested in a blinded manner. We used mouse anti-human SPARC monoclonal antibody ON1-1 (Zymed Lab, South San Francisco, CA) with 0.05 μg/well and biotinylated polyclonal goat anti-human SPARC antibody EYR01 (R&D Systems, Minneapolis, MN) with 0.01 μg/well. To obtain a serum-free culture supernatant, cells were grown to near confluence, washed twice with PBS, and kept in a serum-free medium. After 24 hours, the medium was collected and centrifuged for 10 minutes at 375 × g to remove debris. The samples were divided and diluted at 1:4 with 10% Block Ace (Dainippon Pharmaceutical, Osaka, Japan) to serve as samples for ELISA. The serum and plasma samples were diluted at 1:200 with 10% Block Ace as described above. In this ELISA system, human SPARC HON-3030 (Haematological Technologies) was used to estimate the standard curve to quantify the SPARC protein based on absorbance data.

Statistical analysis. We analyzed all of the data using the StatView statistical program for Macintosh (SAS, Cary, NC) and then evaluated the statistical significance using Student's t test, χ², and Fisher's exact test. Because the SPARC concentration values exhibited a normal distribution in each group, the values were analyzed using Student's t test. We considered P < 0.05 to be statistically significant.

Expression of SPARC mRNA and protein in human melanoma. The expression levels of SPARC in various melanoma cell lines were determined by quantitative reverse transcription-PCR (Fig. 1A) and Western blot (Fig. 1B). SPARC was expressed in all cell lines tested, except for HMV-1, in both mRNA and protein levels. SPARC proteins in the human tissue specimens were examined by Western blotting (Fig. 1C) and an immunohistochemical analysis (Fig. 1D). The vertical growth phase of primary melanomas and lymph node metastasis expressed a large amount of SPARC, whereas large congenital melanocytic nevi and radial growth phase of primary melanomas showed a moderate expression. Normal skin, including a few melanocytes, showed a weakly positive expression (Fig. 1C). Hence, all the examined tissue samples of melanomas and large congenital melanocytic nevi were positive for SPARC protein. An immunohistochemical analysis of SPARC was made on primary melanomas (33 cases), metastatic melanomas (seven cases), and melanocytic nevus (14 cases) tissue specimens. The results obtained from primary melanoma are summarized in Table 1, and the representative strong staining of primary melanoma patient is shown in Fig. 1D. SPARC was detected immunohistochemically in all 33 independent primary melanoma lesions (weak, 7; moderate, 14; strong, 12) and in all seven metastatic lesions tested (weak, 0; moderate, 2; strong, 5). SPARC was predominantly located in the cytoplasm of malignant cells. All of the 14 melanocytic nevi lesions also showed a positive expression (weak, 4; moderate, 6; strong, 4).

Presence of soluble SPARC protein in the culture supernatants of melanoma cell lines and sera from melanoma patients. We detected soluble SPARC protein using ELISA. Soluble SPARC protein could be detected in the culture supernatants of all human melanoma cell lines tested, with the exception of HMV-1, and cultured melanocyte HEMn (Fig. 2A). The concentration of soluble SPARC secreted from each cell line into the culture supernatant did not always correlate with the expression levels of SPARC mRNA and protein (Fig. 1A and B and Fig. 2A).

Serum SPARC concentrations for 109 preoperative melanoma patients, five patients with large congenital melanocytic nevus, and 61 healthy donors are shown in Fig. 2B and Table 1. Figure 2C shows the standard curve for ELISA detection of SPARC that confirmed the linearity of the ELISA determination of SPARC concentration. According to these data, we were convinced that the range for accurate detection of serum SPARC was between 0 and 16 μg/mL by using 200-fold diluted serum samples. We could obtain reproducible results in three independent ELISA assays, and the representative results were shown. The mean ± SD serum SPARC concentration in 109 preoperative melanoma patients (2.02 ± 1.02 μg/mL) was significantly greater than that in the 61 healthy donors (1.62 ± 0.36 μg/mL; P = 0.001, Student's t test). When the cutoff value was fixed at 2.34 μg/mL, which was the mean SPARC concentration plus 2 SD in the healthy donors, 36 of 109 (33.0%; 3.21 ± 0.70 μg/mL) melanoma patients were positive for increased serum SPARC. Thereby, the sensitivity of this assay was 33.0%. On the other hand, three (4.9%; 2.46 ± 0.08 μg/mL) and two (40%) positive cases were found in 61 healthy donors and five melanocytic nevi patients, respectively. Thus, the specificity of this assay was 92.4%. The prevalence of increased SPARC protein in the sera of melanoma patients was higher than that in healthy donors. The presence of a significant amount of SPARC in the sera from melanoma patients suggested that melanoma cells secrete SPARC in melanoma patients.

Among the 30 cases of melanoma patients in which both immunohistochemical staining and ELISA detection of serum SPARC were done, increased serum SPARC was detected in nine patients (30%; Table 1). In six of nine patients (patients 39, 50, 42, 76, 90, and 112), strong SPARC protein expression was immunohistochemically detected in their melanoma cells, and moderate SPARC protein expression in melanoma cells was observed in two patients (patients 44 ad 49). Patient 96 expressed weak SPARC protein in melanoma cells. However, 7 of 13 cases expressing strong SPARC protein did not secrete SPARC in the sera. Because we were not able to prepare serum and tissue samples from the same patients with congenital melanocytic nevus, we could not evaluate the correlation between the serum SPARC levels and the expression levels of SPARC in melanocytic nevi tissue in these patients.

Human platelets contain and secrete SPARC protein in the sera of healthy donors (18). Thus, we measured SPARC in plasma to eliminate the influence of the SPARC secreted from the platelets. The plasma concentrations of SPARC in 11 preoperative patients with melanoma and 21 healthy donors were shown in Fig. 2D. The mean SPARC value in 11 preoperative melanoma patients (0.61 ± 0.65 μg/mL) was significantly greater than that in the 21 healthy donors (0.14 ± 0.14 μg/mL; P = 0.003). When the cutoff value was fixed at 0.43 μg/mL (a mean SPARC concentration plus 2 SD in the healthy donors), 4 of 11 melanoma patients (36.4%) were positive for an increased plasma SPARC as observed in the serum samples of the melanoma patients. In addition, one (4.8%) positive case of 21 healthy donors was observed.

Comparison of serum concentration of SPARC, GPC3, and 5-S-cysteinyldopa in patients with melanoma classified by stage. The above results clearly indicate that SPARC is a novel tumor marker for melanoma. We next compared the serum concentrations of SPARC, GPC3, and 5-S-cysteinyldopa in patients with melanoma classified by clinical stage (Tables 1 and 2). We fixed the cutoff level at 1 unit/mL in GPC3 and at 10 nmol/L in 5-S-cysteinyldopa as reported (6, 10). Figure 2E shows the serum concentrations of SPARC quantified by ELISA in 109 melanoma patients classified by stage. Although the serum concentrations of 5-S-cysteinyldopa increased markedly in patients at stage IV (10), the percentages of serum SPARC positive patients were almost equal among the five clinical stages as seen in GPC3 (10). To our surprise, we detected an increase of SPARC in the sera of patients with very small lesion of melanoma such as stage 0 or I. No significant correlation was observed between the patients positive for each of three markers (Table 1). More importantly, 18 of 36 SPARC-increased patients were negative for both GPC3 and 5-S-cysteinyldopa, and many were classified as cases of relatively early Unio Internationale Contra Cancrum stages 0, I, and II (Table 1). The positive rate of these three tumor markers in patients with melanoma, as classified by stage, is shown in Table 2. The total positive rates of increased SPARC (36 of 109, 33.0%) and GPC3 (48 of 113, 42.5%) were significantly higher than the rate for 5-S-cysteinyldopa (25 of 110, 22.7%). The positive rates of increased SPARC (8 of 15, 53.3%) and GPC3 (7 of 15, 46.7%) at stage 0 were significantly higher than that for 5-S-cysteinyldopa (0 of 15, 0.0%; P < 0.001). In addition, when we use SPARC and GPC3 in combination, the positive rates at stage 0 (13 of 15, 86.7%), stage I (14 of 28, 50.0%), stage II (20 of 28, 71.4%), and stage III (12 of 19, 63.2%) were all significantly higher than that of 5-S-cysteinyldopa (P<0.05). In all, the positive rate of increased SPARC or GPC3 in patients at stages 0 to II (47 of 71, 66.2%) was significantly higher than that of 5-S-cysteinyldopa (4 of 71, 5.6%; P< 0.001). On the other hand, the positive rate of 5-S-cysteinyldopa in stage IV patients (16 of 19, 84.2%) was significantly higher than that of SPARC or GPC3 in combination (12 of 19, 63.2 %). Finally, we were able to detect 78 of 107 (72.9%) cases of preoperative melanoma patients by the combined use of SPARC, GPC3, and 5-S-cysteinyldopa. This is an extremely high positive rate.

Decrease of serum SPARC protein in post-operative melanoma patients. The changes in the serum levels of three tumor markers (SPARC, GPC3, and 5-S-cysteinyldopa) before and after surgical treatments in SPARC-positive 13 patients are shown in Table 3. In 10 of 13 patients, the serum SPARC levels decreased to below cutoff levels after the surgical treatments, although GPC3 and 5-S-cysteinyldopa values were negative before and after the operation in the majority of these patients. It should be noted that SPARC is useful tumor marker to follow the efficacy of surgical treatments. In the case of patients 87 and 92, whose melanoma recurred, the serum SPARC values once decreased to below negative levels and then later increased again, although the serum SPARC level in patient 69 did not increase again when tumor recurrence was identified. The 5-S-cysteinyldopa level increased in all patients whose melanoma recurred. Further examinations are needed to elucidate whether the serum SPARC is useful for detecting recurrent tumors.

SPARC is a matricellular glycoprotein that modulates cellular interaction with the extracellular matrix during tissue remodeling. Although the specific functions of SPARC still remain unclear, it also plays an important role in wound repair, cell proliferation, cell migration, morphogenesis, cellular differentiation, and angiogenesis (11, 19–21). SPARC was first identified in 1981 as a major noncollagenous constituent of bovine bone (22) and is expressed abundantly in the bone and platelets (23). Recently, many publications have described a high expression of SPARC in a variety of human malignancies (12, 24–35). Tumor-derived SPARC was reported to stimulate tumor progression in many types of cancers. The expression levels of SPARC correlated with the histologic grade of tumor cells (25, 26, 30, 33). A higher SPARC expression was associated with local tumor invasion (24, 29–31, 33); metastasis to the lymph nodes, liver, and bone (24, 28, 32, 33); and poor prognosis and survival (24, 32, 33, 35). Conversely, the expression levels of SPARC was inversely correlated with the degree of malignancy in ovarian cancer (34). In most cancers, SPARC protein is overexpressed in the stromal cells of tumor tissue but is rarely expressed in cancer cells themselves (24, 31). In contrast, melanoma cells by themselves have been shown to express a high level of SPARC, and such increased levels are associated with an invasive phenotype in vivo (12, 13).

We confirmed the expression of SPARC at the mRNA and protein levels in human melanoma cell lines and melanoma tissue specimens. In addition, we proved that SPARC was secreted and detected in the culture supernatants of melanoma cell lines and the sera of melanoma patients. In this study, increased serum levels of SPARC were observed in 33% of the melanoma patients. A correlation between the serum levels of tumor markers and tumor progression has been reported (6), as is the case of 5-S-cysteinyldopa in their study. However, no significant correlation was observed between the serum SPARC levels and the progression levels of melanoma in this study. We wonder why only 33% of the melanoma patients showed increased levels of SPARC in the sera, and such SPARC concentrations did not correlate with tumor progression, although most melanoma tissues specimens express SPARC protein and almost all examined melanoma cell lines secreted SPARC protein. In melanoma cell lines, the expression level of SPARC did not completely correlate with the levels of soluble SPARC secreted in the culture supernatant (Fig. 1B and Fig. 2A). The same phenomena were also observed in melanoma patients in this study. Sixteen of 24 patients who expressed moderate or strong SPARC protein in melanoma cells did not show elevated SPARC protein in their sera (Table 1). Therefore, not all but some subpopulations of melanoma cells might secrete SPARC protein in the sera of melanoma patients. The concentration of soluble SPARC secreted into the culture supernatant positively correlated with the cell number in the culture dish (data not shown). However, the serum SPARC concentration was not positively associated with the tumor size or stage. Our evaluation of stage IV individuals revealed the SPARC concentrations to be comparatively low in this group compared with those observed in the early stages. This suggested that SPARC may be a useful tumor marker for detecting relatively early-stage melanoma, although the mechanism of secretion of SPARC from melanoma remains to be elucidated. We are now undertaking further studies to answer these questions.

Positive correlation between the serum SPARC level and platelet count have been reported previously (36). In our work, the serum SPARC concentrations closely correlated with platelet counts in healthy donors (data not shown). In addition, no significant difference was observed between the platelet counts of healthy donors and melanoma patients [213,000 ± 9,000/mm³ versus 216,000 ± 8,000/mm³ (mean ± SE)]. In this study of SPARC measurements in sera, there were three (4.9%) false-positive cases, although those increased values were lower than those in most of the positive patients with melanoma. We thought the elevated SPARC concentrations in healthy donors might thus have been due to confounding conditions, such as thrombocytosis; however, no thrombocytosis cases were identified. Although we considered whether the influence of the platelets might be suppressed by measuring SPARC in plasma, we also had one false-positive case (1 of 21, 4.8%). The positive rate of increased plasma SPARC did not significantly differ from that of the increased serum SPARC in melanoma patients (36.4% versus 33.0%). Thereby, it is unlikely that increased SPARC levels in sera of melanoma patients are due to secretion of SPARC from platelets.

Melanocytic nevi tissues expressed both SPARC and GPC3 protein (10). The SPARC levels in sera were increased in two of five (40%) patients with large congenital melanocytic nevus, although the GPC3 level did not increase. We have to thus pay close attention to cases, which are GPC3 negative and SPARC positive, because of the risk of making a false-positive diagnosis.

We used sera from Japanese patients only. The incidence of acral lentiginous melanoma in the Japanese population is much higher than that in Caucasians, whereas superficial spreading melanoma and lentigo maligna melanoma are frequent types observed in Caucasians. Some groups have reported that acral lentiginous melanoma differs from other types of melanomas in its clinical, histopathologic, and genetic characteristics (37–40). We compared the positive rate of increased serum SPARC and GPC3 among patients classified by these clinical types. Thus, no significant correlations were observed between the positive rate and melanoma types (data not shown). Therefore, it seems likely that the usefulness of SPARC and GPC3 for diagnosis of melanoma is not restricted to only Japanese patients.

In conclusion, SPARC was found to be a useful tumor marker for melanoma particularly at the early stage of the disease, and the addition of the other two markers (GPC3 and 5-S-cysteinyldopa) had added benefit in diagnosis. On the other hand, the serum levels of these three markers are still unknown in a population of patients with atypical nevi syndrome or other high-risk population in this study. Further investigations are needed to consider future applications of serum SPARC and GPC3 for the mass screening of melanoma.