Plasma Matrix Metalloproteinase-9 Level Is Better than Serum Matrix Metalloproteinase-9 Level to Predict Gastric Cancer Evolution

Matrix metalloproteinases (MMP) are a family of zinc-dependent enzymes that breakdown the extracellular matrix (1). MMPs not only play important roles in physiologic extracellular matrix remodeling but are also involved in pathologic conditions, including tumor progression, invasion, and metastasis (2–5). Because tissue remodeling is often reflected in body fluids, measurements of MMPs in blood or urine have been suggested as useful tools for characterizing processes that occur in tissue (6). Many commercial kits, such as ELISA or fluorimetric assays and zymographic methods, measure MMPs in blood and urine. The diagnostic performance of these MMP measuring kits has been proven in previous reports (7, 8). Among the MMP family, MMP-9 (also known as 92-kDa gelatinase) is a promising new noninvasive marker (6).

Elevated levels of serum or plasma MMP-9 have been found in a variety of malignant tumors, such as breast cancer, colon cancer, lung cancer, head and neck squamous cell carcinoma, hepatocellular carcinoma, and gastric cancer (9–15). Although the usefulness of MMP-9 as a tumor marker has been established, several studies measuring MMP-9 in the peripheral blood of cancer patients, using serum or plasma samples, have produced controversial results (10–12, 15–17). Differences in enrolled populations and study designs most likely contribute to the discrepancies. However, blood sampling and processing may also influence the concentration of MMP-9. Several reports have highlighted the influence of blood specimen collection methods on MMP-9 concentrations (18–20). MMP-9 concentration has been found to be 3-fold higher in serum than in heparin plasma (21). Platelet activation or neutrophil mobilization during clotting could produce such results (22). Although measurements of MMP-9 in blood have been suggested to be done in heparin plasma, rather than in serum (23), recent studies have used serum MMP-9 to investigate the correlation between MMP-9 and tumor progression (10, 11, 24–26). It is important to investigate how the differences between plasma and serum samples influence the diagnostic and prognostic performances of MMP-9. In the present study, we compared the effectiveness of plasma and serum MMP-9 levels as diagnostic and prognostic markers in gastric cancer.

Study subjects. This nested case-control study is based on two national projects to investigate the risk factors of gastric cancer in Taiwan. Since January 1998, blood samples have been collected prospectively from patients with newly diagnosed gastric cancer in the inpatient and outpatient cancer clinics of four major medical centers in Taiwan. All patient-derived specimens were collected and archived under protocols approved by the institutional review boards of the parent institutions. A full verbal explanation of the study was given to all participants. They consented to participate on a voluntary basis. Inclusion and exclusion criteria were detailed previously (27, 28). Gastric adenocarcinoma was histopathologically confirmed by endoscopic biopsy or surgical specimens. No patients had been prescribed chemotherapy before surgery, and there was no evidence of any other malignancy.

In the present study, we investigate 114 gastric cancer patients, for whom both plasma and serum samples were available. Control blood samples (n = 87) were obtained from subjects who visited health examination clinics with minimal gastritis or normal appearance of the gastric mucosa on gastroscopic examination. The controls were matched by age (±3 years) and date of blood collection (±3 months). Among the 114 gastric cancer patients, 106 received operation and had clinicopathologic data available. Clinicopathologic characteristics were studied, and survival analyses were conducted for the 106 patients who received operation. All recruited patients had been followed up for at least 5 years.

ELISA for MMP-9. To minimize inter-individual variability, we collected peripheral blood from the same individuals and placed them in different sample preparation tubes. Blood samples were collected by venipuncture and centrifuged at 3,000 rpm for 15 min at 4°C. Plasma MMP-9 values were measured in samples prepared in tubes coated with lithium heparin (BD Vacutainer, containing sodium heparin 143 USP units; BD, Franklin Lakes, NJ). Serum MMP-9 values were measured in samples prepared in Monovette tubes without additive (BD Vacutainer, BD). All samples were preserved at −80°C.

MMP-9 Quantikine ELISA kits were purchased from R&D Systems Europe (Abingdon, Oxon, United Kingdom; product code DMP900). Plasma and serum samples were diluted according to the manufacturer's specifications for MMP-9 assay. All measurements were done in duplicate. Similar standard curves for each of the proteins were required in each plot. The immunoassays were carried out following the manufacturer's instructions. MMP-9 immunoassay recognizes both active (82 kDa) and pro-MMP-9 (92 kDa). The sensitivity limit of the assay was 0.156 ng/mL. All assays were done in duplicate. A subset of samples has been reassayed for five times in every ELISA plate for quality control. The intra-assay variability is 5.2%. The inter-assay variability is 4.8%.

Measurements of MMP-9 activity by zymography. Quantitative gelatin zymography for MMP-9 was carried out as described before (29, 30). Protein concentration for each sample was determined by Bio-Rad Protein Assay (Bio-Rad, Richmond, CA). Briefly, plasma and serum containing 25 μg protein were diluted in sample loading buffer [5% SDS, 20% glycerol in 0.4 mol/L Tris (pH 6.8), containing 0.02% bromophenol blue without 2-mercaptoethanol] and loaded for zymography on a 10% polyacrylamide gel containing 1 mg/mL gelatin (from procine skin, Type A; Sigma Co., St. Louis, MO). Afterwards, gels were washed for 30 min in 2.5% Triton X-100 and incubated for 24 h at 37°C in 50 mmol/L Tris (pH 7.5), 10 mmol/L CaCl₂, 0.02% NaAzide. The buffer was decanted, and the gels were stained with 0.5% Coomassie blue G in 30% methanol and 10% acetic acid for 15 min at room temperature on a rotary shaker. Stain was washed out with destaining buffer (10% acetic acid, 20% methanol) until clear bands were seen. Areas where proteolytic activity degraded the gelatin were identified as showing an absence of staining. MultiGauge version 1.01 (Fuji Photo Film Co., Ltd., Tokyo, Japan) was used to evaluate the semiquantitative levels of MMP-9 forms. For active-form MMP-9, the intra-assay variability is 8.8%, and the inter-assay variability is 6.9%.

Statistical analysis. The demographic characteristics of patients and controls were compared using the χ² test and Student's t test. The relative measures of MMP-9 levels in plasma and serum were compared using Wilcoxon signed rank test. In view of the non-Gaussian distribution of data for the MMP-9 levels in study subjects, the differences in MMP-9 values between gastric cancer patients and controls and among gastric cancer patients with different clinicopathologic features were compared using nonparametric Mann-Whitney U test. To evaluate the association between plasma MMP-9 and serum MMP-9 values and between MMP-9 levels and zymography, we used nonparametric Spearman's correlation coefficients method. The diagnostic performances of serum and plasma MMP-9 for differentiating between gastric cancer patients and controls were determined by receiver-operating characteristics (ROC) analysis. The cutoff values between normal and elevated levels of plasma MMP-9 and serum MMP-9 were set at 60 and 293 ng/mL, respectively, according to the most accurate values obtained from nonparametric ROC analyses. Logistic regression analyses were conducted to evaluate the associations between plasma MMP-9 levels, serum MMP-9 levels, and clinicopathologic features of gastric cancer patients. Survival analyses were done with the Kaplan-Meier method, and the differences between the curves were tested with the two-tailed log-rank test. To determine whether plasma OPN concentration is an independent prognostic factor for survival, and whether the inclusion of MMP-9 improve the predictive power, hazards ratios (HR) were calculated using the Cox proportional hazards model with backward stepwise calculation. Most data were analyzed via the SPSS program for Windows 11.0 (SPSS, Inc., Chicago, IL). Nonparametric ROC analyses were done via the STATA program for Windows 8.0 (Stata Corp., College Station, TX).

Associations between MMP-9 levels and zymography. There were no statistically significant differences between age and gender. The demographic and clinicopathologic data of gastric cancer patients and controls were shown in Table 1. For all enrolled study subjects, the median concentration of serum MMP-9 was significantly higher than plasma MMP-9 (292.2 versus 83.4 ng/mL, P < 0.001). There was significant correlation between serum MMP-9 values and plasma MMP-9 values (r = 0.438, P < 0.001).

To identify different activities of MMP-9 forms, we conducted quantitative zymography. Table 2 shows the correlation coefficients and Ps by comparing plasma and serum MMP-9 levels with the results from zymography. Both plasma and serum MMP-9 levels measured by ELISA correlated significantly with active MMP-9 isoform identified by the very sensitive zymography in the plasma and serum samples (P = 0.002 and P = 0.048, respectively). A representative zymography has been shown in Fig. 1.

Associations between MMP-9 levels and occurrence of gastric cancer. To evaluate the diagnostic usefulness of plasma and serum MMP-9, we first compared the distributions of plasma and serum MMP-9 values in gastric cancer patients and controls. For plasma MMP-9, the median concentration was significantly higher in gastric cancer patients than in control subjects (122.4 versus 47.2 ng/mL, P < 0.001). For serum MMP-9, the median values did not differ significantly between gastric cancer patients and controls (261.6 versus 305.2 ng/mL, P = 0.17; Fig. 2).

We then plotted ROC curves and calculated area under the curve (AUC) values. The curve of plasma MMP-9 deviated to the left with the greatest difference from the neutrality line at AUC value as high as 0.803 (95% confidence interval, 0.741-0.865; P < 0.001). The AUC value of serum MMP-9 was 0.443 (95% confidence interval, 0.364-0.523; P = 0.170), which was significantly lower than the AUC value of plasma MMP-9 (P < 0.001; Fig. 3). This shows better overall diagnostic performance of plasma MMP-9 in comparison with serum MMP-9.

On ROC analyses, a plasma MMP-9 cutoff value of 60.0 ng/mL provided the best discrimination between gastric cancer patients and controls. The sensitivity, specificity, and accuracy for predicting occurrence of gastric cancer were 82.5%, 65.5%, and 75.1%, respectively. Serum MMP-9 cutoff value of 293.0 ng/mL provided the best discrimination between gastric cancer patients and controls. The sensitivity, specificity, and accuracy for predicting the occurrence of gastric cancer were 47.4%, 47.1%, and 47.3%, respectively.

Associations between MMP-9 levels and clinicopathologic characteristics. Among the 114 gastric cancer patients, 106 who received operation with complete staging were recruited to analyze the distributions of plasma and serum MMP-9 values and their associations with clinicopathologic characteristics of gastric cancer (Tables 3 and 4). Elevated plasma MMP-9 levels (≥60 ng/mL) were associated with lymph node metastasis (odds ratio, 3.43; P = 0.019), lymphatic invasion (odds ratio, 7.58; P = 0.009), and venous invasion (odds ratio, 4.14; P = 0.033). Serum MMP-9 levels were not associated with age, gender, stage, serosal invasion, lymph node metastasis, lymphatic invasion, or venous invasion.

Associations between MMP-9 levels and survival of gastric cancer. Kaplan-Meier survival curves are shown in Fig. 4. The survival rates for patients with elevated plasma MMP-9 levels (≥60 ng/mL) were significantly lower than the survival rates of those with normal plasma MMP-9 levels (P = 0.038). The survival rates for patients with elevated serum MMP-9 levels (≥293 ng/mL) did not differ significantly from the survival rates of those with normal serum MMP-9 levels (P = 0.614).

On Cox univariate analysis, we found that advanced stage (HR, 5.65; P = 0.004), serosal invasion (HR, 4.71; P < 0.001), lymph node metastasis (HR, 3.02; P = 0.006), lymphatic invasion (HR, 2.17; P = 0.020), and venous invasion (HR, 2.80; P = 0.002) were of prognostic significance for poor overall survival (Table 5). On Cox multivariate analysis with backward stepwise calculation, only serosal invasion (HR, 3.93; P = 0.003) and venous invasion (HR, 2.12; P = 0.028) were independent risk factors for poor survival (Table 6). Plasma MMP-9 and serum MMP-9 levels were not prognostic markers.

As in previous reports, which showed that higher MMP-9 concentrations are present in serum samples than in heparin plasma samples (21, 23), we found that serum MMP-9 values of all enrolled subjects were 3.5-fold, in control subjects were 6.5-fold, and in gastric cancer patients were 2.2-fold higher than in the respective plasma MMP-9 values. It is most likely due to additional release of MMP-9 from blood cells during platelet activation or the sampling process, as leukocytes and platelets contain high concentrations of MMP-9 (31, 32). Because circulating MMP-9 has different forms, it is crucial to characterize the activity of this gelatinase. We use zymography to identify the pro-MMP-9 and active MMP-9 forms because zymography is more sensitive to detect active MMP-9 than bioactivity assays (33), We found that ELISA level of MMP-9 correlated significantly with active MMP-9 form not only in plasma but also in serum. This observation is consistent with previous reports that ELISA level of MMP-9 is significantly correlated with pro-MMP-9 and active MMP-9 detected by zymography (33).

Although both plasma and serum ELISA level of MMP-9 correlate significantly with active MMP9, the usefulness of plasma and serum MMP-9 levels as diagnostic markers actually varies. Jung et al. used renal cell carcinoma patients to determine the diagnostic performances of serum and plasma MMP-9 via ROC analyses (23). The AUC for plasma MMP-9 was significantly higher than the AUC for serum MMP-9. We tested the diagnostic performances of serum and plasma MMP-9 in gastric cancer patients and confirmed that plasma MMP-9 is a better diagnostic marker than serum MMP-9, at least in gastric cancer. This may partly explain why studies measuring serum MMP-9 tend to report nonsignificant results (10, 25, 34, 35), whereas those using plasma samples present more promising data (9, 13, 15, 36, 37).

In the present study, plasma MMP-9 reached a sensitivity of 82.5% and specificity of 65.5% at a cutoff value of 60 ng/mL. Previous studies of gastric cancer reported sensitivity of plasma MMP-9 as varying from 47% to 56% (14, 15). Endo et al. measured the precursor form of MMP-9 (pro-MMP-9) in plasma, and 47% of gastric cancer patients showed elevated (>63 ng/mL) pro-MMP-9 levels (14). In Torii et al.'s study, pro-MMP-9, active MMP-9, and their complex form with tissue inhibitor of metalloproteinase-1 were detected. Elevated plasma MMP-9 (>62 ng/mL) was found in 56% of gastric cancer patients (15). In the present study, we measured both pro-MMP-9 and active MMP-9. Different detecting substrates and study populations may have contributed to the discrepancies.

Although the usefulness of plasma MMP-9 as a diagnostic tool has been suggested, the role of plasma MMP-9 as a prognostic marker is still controversial. Torii et al. reported that plasma MMP-9 concentrations in patients with gastric cancer are associated with tumor size, lymph node metastasis, and tumor-node-metastasis stage (15). However, Endo et al. did not find any association between clinicopathologic characteristics of gastric cancer and elevated plasma MMP-9 level (14). In non–small cell lung cancer, plasma MMP-9 levels have been shown to be significantly higher in patients with metastasis (36). In breast cancer, plasma MMP-9 concentration has not correlated well with clinicopathologic factors (16). In the present study, plasma MMP-9 levels were associated with lymph node metastasis, lymphatic invasion, and venous invasion of gastric cancer.

Serum MMP-9 is a less effective diagnostic marker than plasma MMP-9. However, it may play a role in predicting patient survival. In head and neck squamous cell carcinoma (11) and melanoma (38), survival rates have been shown to be lower among patients with high MMP-9 serum levels. In lung cancer, serum MMP-9 levels have not correlated well with survival (10, 39, 40). In breast cancer, low serum MMP-9 level has been found to be associated with poor relapse-free survival (40). In contrast to serum MMP-9, evaluated in many clinical studies, the role of plasma MMP-9 in survival has not been widely investigated. In nasopharyngeal carcinoma (41) and ovarian carcinoma (42), plasma pro-MMP-9 levels do not influence survival. In lung cancer, patients with lower plasma MMP-9 levels survive longer than those with higher levels (43). In the present study, elevated plasma MMP-9 levels, not serum MMP-9 levels, showed significant correlation with poor survival. However, both plasma MMP-9 and serum MMP-9 level are not independent prognostic factors in Cox multivariate analysis. To the best of our knowledge, this is the first study to compare the usefulness of plasma and serum MMP-9 levels in predicting gastric cancer survival. More studies, especially using model-based strategy to compare best models with or without MMP-9 level and calculate its predictive power (44, 45), will be helpful to confirm our observations.

In conclusion, plasma MMP-9 level is a better diagnostic marker than serum MMP-9 level for detecting gastric cancer, and there is significant association between plasma MMP-9 level and gastric cancer invasive phenotypes and survival. To the best of our knowledge, this is the first study to systematically compare the usefulness of plasma and serum MMP-9 in predicting occurrence, clinicopathologic features, and survival of gastric cancer.