Identification of Matrix Metalloproteinase 11 as a Predictive Tumor Marker in Serum Based on Gene Expression Profiling Free

Purpose: Prognostic markers discovery is a strategy for early diagnosis and individualization therapy for human cancer. In this study, we focus to integrate different methods to identify specific biomarker and elucidate its clinical significance.

Experimental Design: A powerful tool named Digital Gene Expression Display online was applied to isolate differentially expressed genes correlated with gastric cancer. Matrix metalloproteinase 11 (MMP11) was selected and confirmed at both mRNA and protein level in 10 cell lines, 123 cases of tumor tissues, and 305 cases of gastric cancer serum specimen by semiquantitative PCR, immunohistochemistry staining, and ELISA techniques, respectively.

Results: Our data showed that overexpression of MMP11 at mRNA and protein level was consistently detected in cell lines and primary tumors compared with matched normal tissues. Importantly, serum MMP11 levels were also significantly elevated in gastric cancer patients compared with those of the control subjects (P < 0.001), and the positive expression was well correlated with metastasis in gastric cancer patients (P = 0.009). Furthermore, we have shown that overexpression of MMP11 was associated with the malignant proliferation of AGS cells.

Conclusions: Combination of gene expression profiling and specific clinical resource is a promising approach to validate gene expression patterns associated with malignant phenotype. As a secreted protein, MMP11 may play an important role in carcinogenesis and has potential implication as a biomarker for the diagnosis and prognosis of human cancers including gastric cancer.

Gastric cancer is the second most common malignancy and one of the leading causes of death in Chinese people. Survival can be dramatically increased if the disease is diagnosed at an early stage. However, early detection is hampered by the lack of highly sensitive and specific biomarkers. In recent years, powerful techniques have been developed that allow comprehensive analysis of gene expression. Thousands of genes can be monitored simultaneously for changes in their expression levels using cDNA microarray analysis.

Cancer Genome Anatomy Project has become the largest contributor of tags and sequences to the serial analysis of gene expression (SAGE) and expressed sequence tags (EST) databases, most of which are free and open access to raw data (1). SAGE is a developing technique that allows us to quantitatively analyze overall gene expression profile by providing absolute transcript tags numbers in digital format and statistical comparisons (2). In contrast to microarray methodologies, SAGE does not require a prior knowledge of the expressed genes in the starting material, so it can lead to an unbiased comprehensive representation of the transcripts present in a sample (3, 4). Digital gene expression display (DGED) is a tool that compares gene expression between two pools of libraries. It evaluates the statistical significance of the differences using the sequence odds ratio and a Bayesian test. According to the number of libraries that contain this tag or EST and the sequence frequency in either pool A or pool B, the odds ratio uses a simple mathematical formula to provide a measure of the relative amount of a tag or EST in pool A or pool B.

In this study, we have compared the gene expression patterns of gastric cancer and normal gastric tissue by using DGED analysis, and numerous differentially expressed genes in gastric cancer have been identified. Matrix metalloproteinase 11 (MMP11) was selected as an example to evaluate its biosignature and clinical significance from our gene expression profiling. MMP11 is a member of the MMP family which can degrade extracellular matrix components and may play a central role on the enhancement of tumor-induced angiogenesis, cell migration, proliferation, apoptosis, and connective tissue degradation (5, 6). MMP11 was first cloned by differential screening of a cDNA library prepared from a human breast carcinoma subtracted with a breast fibroadenoma (7). MMP11 seems unable to degrade major extracellular matrix component (8, 9); furthermore, it is processed intracellularly and secreted as an active form (10). MMP11 thus differs from other MMPs that are expressed as proenzymes and processed to active forms through proteolytic cleavage activated extracellularly, indicating that MMP11 may have a unique role in tumor development and progression (11). Based on this hypothesis, we have integrated gene expression profiling, tissue microarray, and regular laboratory analysis to validate MMP11 as a biomarker for cancer development and progression.

Databases and bioinformatic analysis. DGED was used to search SAGE and EST libraries housed at the National Center for Biotechnology Information Web site⁵

and to screen differentially expressed genes between gastric tumors and normal gastric mucosa. The cutoff value was set to an F value of 3 (expression factor) and a P value of 0.01 (signification filter), which has strong probability that the tag or EST occurs thrice as frequently in one pool as in the other. F value in conjunction with P value determines which results are reported. Differentially expressed transcripts are defined as up-regulated genes if the F value is significantly >3 and the P value is <0.01; an F value of <1/3 and a P value of <0.01 are considered as down-regulated genes.

Cell lines and specimens. Cell lines PAMC82, MKN45, AGS, SNU5, SNU16, RF1, and RF48 were from American Type Culture Collection, whereas BGC823, MGC803, and SGC7901 cell lines were established in China. Tumor cells were routinely maintained as previously described (12, 13). Thirty paired sets of tissues and adjacent normal mucosa including 18 female and 12 male, ages 47 to 76 years, were obtained from the tumor bank of Beijing Cancer Hospital. A total of 305 serum samples from gastric cancer patients (211 men and 94 women ages 26-87 years; mean, 59; median, 60); 155 from gastritis, intestinal metaplasia, and dysplasia patients (80 men and 75 women ages 20-85 years; mean, 55; median, 54); 90 from breast cancer patients (90 women ages 38-79 years; mean, 57; median, 58); 40 from colorectal cancer (28 men and 12 women ages 39-78 years; mean, 63; median, 66); 33 from lung cancer (21 men and 12 women ages 39-78 years; mean, 62; median, 64); and 302 common controls (167 men and 135 women ages 20-90 years; mean, 53; median, 54) from noncancerous patients were obtained from the clinical laboratory of the Beijing Cancer Hospital, the basic medical center of Beijing Chao-Yang Hospital and Qinghai Medical College. The controls of the breast cancer (92 women ages 22-78 years; mean, 52; median, 53), colorectal cancer (21 men and 19 women ages 26-75 years; mean, 58; median, 60), and lung cancer (21 men and 12 women ages 26-75 years; mean, 58; median, 61) were chosen from the 302 cases of common control group with matched case numbers, gender, and age. The samples were leftovers from routine biochemical testing. All sera were stored at −70°C until use.

Analysis of tumor markers. The tumor markers CEA, CA199, CA72.4, and CA242 were analyzed on the Elecsys immunoassay analyzer (Roche Diagnostics). The upper limit of normal for these tumor markers is 5 ng/mL (CEA), 27 units/mL (CA199), 6.7 units/mL (CA72.4), and 20 units/mL (CA242). MMP9 was tested by the Human MMP9 Immunoassay kit (Roche Diagnostics).

Reverse transcription-PCR. Total RNA was extracted from cell lines and tissue blocks using the RNeasy Mini kit (Qiagen). RNA samples (5 μg) were subjected to reverse transcription for 60 min at 37°C using 200 units of Moloney murine leukemia virus reverse transcriptase (Promega). The primer pairs for MMP11 were 5′-GAGCAGGTGCGGCAGACGA-3′ (F) and 5′-CGAAAGGTGTAGAAGGCGGACA-3′ (R). “Hot start” PCR reaction was initiated by 5-min incubation at 95°C, ended after a 10-min extension at 72°C, 32 to 34 cycles for denaturation at 94°C for 45 s, annealing at 60°C for 45 s, and extension 72°C for 1 min. We did reverse transcription-PCR (RT-PCR) twice to confirm the reproducibility of the experiment, and the internal control RT-PCR of β-actin was done on all specimens simultaneously.

Western blot. Equal amounts of protein were electrophoresed on 12% SDS-PAGE and transferred to polyvinylidene difluoride membrane using a standard protocol. Immunoreactivity was tested with the anti-MMP11 (Ab-5; Neomarker) at a 1:500 dilution. Nonspecific binding was blocked by 5% fat-free milk solution, and MMP11 protein was detected by the enhanced chemiluminescence system (Amersham Pharmacia Biotech).

Tissue microarray and immunohistochemistry staining. TMA blocks were constructed in our laboratory as previously described (14); for each case, we sampled five tissue cores at 0.6 mm in diameter, including three tumors and two matched-adjacent normal mucosa tissue to construct the tissue microarray. A total of 210 human gastric specimens including 123 gastric cancer and 87 normal cases were obtained from the tumor bank of Beijing Cancer Hospital. The patients were fully informed and given the consents for collection of clinical samples. Immunohistochemistry analysis was carried out using avidin-biotinylated horseradish peroxidase complex method. The section was incubated with anti-MMP11 (Ab-5; 1:1,000) at 4°C overnight. More than 5% stained cells in the tissue was defined as positive reaction in this experiment.

ELISA. By incorporating two anti-MMP11 antibodies, a sandwich-type immunofluorometric assay was developed for detecting MMP11 protein in serum. A rabbit anti-MMP11 polyclonal antibody (Biological) diluted as concentration at 1 μg/mL in coating buffer (0.05 mol/L carbonate buffer; pH 9.6) was dispensed into a 96-well plate and incubated at 4°C overnight. The plate was then washed thrice with PBS containing 0.05% Tween 20 and blocked with PBS containing 5% milk powder for 2 h at 37°C. Patient serum samples were diluted with 1:20 in block buffer and added into the plate, incubating for 2 h at 37°C. The anti-MMP11 monoclonal antibody (NeoMarker) was then added and the samples were incubated for 2 h at 37°C. Subsequently, horseradish peroxidase–conjugated goat anti-mouse antibody diluted with 1:1,500 was added and incubated for 40 min and washed six times as described above. Finally, 50 μL of 3,3′,5,5′-tetramethylbenzidine was added into each well and incubated for 20 min. Fifty microliters of 12.5% H₂SO₄ were pipetted into each well and mixed for 1 min. The fluorescence was measured with a time-resolved fluorometer, the CyberFluor 615 Immunoanalyzer (MDS Nordion) at 450 nm.

Stable transfection and malignant growth assay. The complete MMP11 cDNA was cloned into the plasmid pcDNA3.1 (Invitrogen) to obtain the transfer vector pcDNA-MMP11. Human gastric cancer cell line AGS was routinely maintained as described previously in our laboratory (12, 13). Cultured cells at 60% to 70% confluence in 60-mm plates were transfected with parental and recombinant plasmids with the Lipofectamine2000 reagent (Invitrogen) according to the manufacturer's instruction. Selective medium containing 400 μg/mL G418 was used to screen stable transfectants. We did RT-PCR and Western blotting to confirm the reproducibility of the experiment. The invasion assay was done by the BD Matrigel Invasion Chamber (Lot 17044; BD Biosciences) according to the manufacturer's instruction. 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay and soft agar colony formation assay were done as described previously in our laboratory (11).

Statistical analysis. Statistical analysis was carried out using the χ² test, Fisher's exact test, and independent samples t test. Receiver operating characteristic curves were constructed by plotting sensitivity versus (1-specificity), and the areas under the receiver operating characteristic curves were calculated. For all analyses, a P value of <0.05 was considered statistically significant. SPSS software (version 13) was used for statistical analysis.

Electronic expression profiling of differentially expressed genes in gastric cancer. SAGE DGED was used to distinguish significant differences in gene expression profiles between two pools of human SAGE libraries. SAGE–pool A contained one database of gastric cancer, which had 64,925 tags and arose from a 57-year-old male patient with moderately differentiated adenocarcinoma (T4N0M0). SAGE–pool B was composed of 32 SAGE databases of normal tissue including one normal stomach epithelium SAGE database, with a total of 1,619,006 tags.

We evaluated the statistical significance of the differences using the sequence odds ratio and a Bayesian test. According to the number of libraries that contain this tag or EST and the sequence frequency in either pool A or pool B, the odds ratio used a simple mathematical formula to provide a measure of the relative amount of a tag or EST in pool A or pool B.

In addition, we also analyzed EST–pool A and EST–pool B using the same method. EST–pool A had 27 databases of gastric cancer, and EST–pool B contained 229 databases of normal tissue including five databases of normal gastric epithelia. Differentially expressed genes containing 115 tags and 181 ESTs in gastric cancer were extracted. These candidates might be involved in tumor development and progression and were briefly classified as proteases, cell proliferation–related molecules, transcription factors, cell adhesion molecules, extracellular matrix–related molecules, and other uncharacterized genes (see Supplementary Table S1). Among these candidates, frequent overexpression of MMP11 was determined in gastric cancer, and we then investigated the expression status of MMP11 at both mRNA and protein level to validate the result of electronic expression profiling.

Increased expression of MMP11 at mRNA and protein levels in cell lines and primary tumors of gastric cancer. RT-PCR analysis was used for detecting MMP11 expression at mRNA level in the tumor cell lines. The result obtained from the amplification of MMP11 gene showed that the different gastric cancer cell lines contained various levels of MMP11 mRNA. Of 10 gastric cancer lines, SNU16 and BGC823 maintained higher expression of MMP11 (Fig. 1A). PCR product of MMP11 (449 bp) was identified by DNA sequencing. In fresh tissues, increased MMP11 mRNA expression was present in tumor samples than in normal mucosa (Fig. 1B). Eighteen of thirty (60%) cases of gastric tumor highly expressed MMP11 mRNA, whereas weak expression was detected only in 7 of 30 (23%) cases of normal controls. Subsequently, we detected the expression change of MMP11 on TMA section at protein level by immunohistochemistry assessment (Fig. 1C), and the results showed positive granular staining of MMP11 in 87 of 123 (70.7%) tumors and 40 of 87 (46%) normal gastric mucosa (P = 0.001; Table 1). Survival analysis showed that there was no significant difference between the survival time of patients with positive MMP11 expression and patients with negative expression (see Supplementary Fig. S1).

MMP11 protein frequently detected in sera of common cancer patients including gastric cancer. Based on increased MMP11 expression in both mRNA and protein level in tumors, we used ELISA to measure the expression level of MMP11 protein in the serum samples of gastric cancer, and found that the serum levels of MMP11 were significantly elevated in gastric cancer patients (n = 305; median, 1.413; range from 1.02-2.46) compared with those in common controls (n = 302; median, 1.159; range from 1.0-1.53; P < 0.001; Table 2 and Fig. 2). MMP11 was also found to be elevated in the sera of intestinal metaplasia and dysplasia patients (see Supplementary Table S2). Based on the digital Northern data, MMP11 was also overexpressed in breast, colorectal cancer, and lung cancer; thus, we also tested MMP11 protein level in these cancer sera and observed the same result (n = 90; median, 2.166; range from 1.42-3.11 for breast cancer; n = 40; median, 2.217; range from 1.35-2.73 for colorectal cancer; n = 33; median, 1.781; range from 1.05-2.52 for lung cancer; Table 2). The medians between the two groups were significantly different (P < 0.001). These data consistently showed that elevated MMP11 protein could be detected in sera of cancer patient. When we chose the value of 1.442 (the 99th percentile value for controls) as cutoff, the sensitivity for gastric cancer was 45.2%, and the specificity was 97.7%. Receiver operating characteristic curve analysis was further shown in Supplementary Fig. S2. We also tested the MMP11 protein of the sera of cancer patients and controls by Western blot. The result showed us that the protein level of MMP11 was higher in the sera of cancer patients than in controls, which confirmed the results of ELISA (see Supplementary Fig. S3).

MMP11 expression level is correlated with poor prognosis in advanced gastric cancer. To further investigate whether MMP11 serum concentration has any value in monitoring disease progression, part of the serum samples of gastric cancer patients were also analyzed with CEA, CA199, CA72.4, CA242, and MMP9 (Table 3). There was a correlation between serum MMP11 and CA199 in gastric cancer patients (P = 0.017; see Supplementary Table S3). We then further examined whether the levels of MMP11 expression were also correlated with clinicopathologic features, and the results showed that the high and low levels of serum MMP11 were significantly associated with metastasis of gastric cancer (P = 0.009). No significances were found between levels of serum MMP11 and other variables, including gender, age, tumor-node-metastasis stage, and differentiation (see Supplementary Table S4).

Overexpression of MMP11 induced proliferation and invasion in AGS cells. To study the functions and molecular mechanisms of MMP11, recombinant plasmid containing MMP11 was constructed and transfected into the AGS cell line.

Two clones exhibiting dramatically increased expression of MMP11 at both mRNA and protein levels were selected (Fig. 3A). First, the effect of MMP11 on invasion was detected by invasion chamber, and the result showed that the number of transfectants that passed the membrane was increased compared with AGS (Fig. 3B). The in vitro growth ability of MMP11 overexpressed cells was first determined by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. The results indicated that AGS overexpressed with MMP11 proliferated more quickly than AGS transfected with the vector alone and AGS (Fig. 3C). Colonies were observed in AGS cell line–overexpressed MMP11 but not in the vector-transfectants and parental AGS cells in soft agar (Fig. 3D).

In the report here, we analyzed the SAGE database and generated a gene expression profile of gastric cancer. MMP11 was selected as a cancer-related protein and its overexpression was consistently confirmed at both mRNA and protein level in cell lines and primary tumors of gastric cancer compared with matched normal tissues. Importantly, we found that MMP11 protein was significantly elevated in serum specimens of gastric cancer patients but not in those of the control subjects. Our results also showed that serum MMP11 level was well correlated with metastasis in advanced gastric cancer patients, which suggested that MMP11 might involve in gastric cancer progression and have a potential implication as a biomarker for the diagnosis and prognosis.

The MMP, with a broad spectrum of proteolytic activities toward extracellular matrix components, are believed to involve in many biological processes such as embryo implantation and morphogenesis, cell migration, metastasis, tumor invasion, and wound healing (15, 16). For example, MMP2, MMP7, and MMP9 have been believed to be related with the progression and prognosis of endometrial carcinoma (17), colorectal cancer (18), gastric cancer (19, 20), and breast cancer (21). MMP11 is an interesting member of the MMP family, and its overexpression has been shown in human cancers such as oral cancer (22, 23), desmoid tumors (24), non–small cell lung cancer (25), and esophageal adenocarcinoma (26). Furthermore, strong MMP11 gene expression is correlated with both increased aggressiveness of tumors and a poor clinical outcome (27, 28). Evidence of the participation of MMP11 in tumorigenesis has also been documented in vivo. It has been reported that the level of MMP11 expression is a critical factor in the ability of MCF7 human breast cancer cells to generate tumors in nude mice (27). Previously, our study has also reported that the BGC823 cell line–silenced MMP11 expression exhibited significantly inhibited cell proliferation, including colony formation in soft agar and tumorigenicity in nude mice (11). Our present results also showed that MMP11 could promote the proliferation and invasion of AGS cell line. These experimental evidences have been overwhelming in proving the causative role of MMP11 in tumor progression. Otherwise, previous data showed that the coexpression of Sp1 and MMP11 were observed in gastric tumors (see Supplementary Table S5 and Fig. S4), and the fact that the promoter region of MMP11 gene has the binding site of Sp1 led us to conclude that perhaps Sp1 could regulate the expression of MMP11 by binding to the promoter region of MMP11.

In this study, we have combined the methods of DGED, tissue microarray, and ELISA to identify MMP11 as a biomarker in serum specimens of gastric cancer. In fact, database mining is an effective way to profile tumor markers other than regular laboratory approaches, including mRNA differential display, representational display analysis, and two-dimensional gel electrophoresis. A number of potential tumor proteins have been identified by bioinformatics in recent years. For example, Colon Cancer Related Gene, a growth factor encoding a novel cysteine-rich motif with preferential expression in majority of colon tumors, was found by using a data-mining tool called Digital Differential Display to search the EST databases (29). Of our electronic gene expression profiling, a number of genes have been previously proven to be involved in gastric cancer from literatures, such as mucin 5B (30), prothymosin α (31), CDX1 (32), Wnt7B (33), syndecan1 (34), mucin1 (35), and HMGA1 (36), which might confirm the validity of digital gene expression profiling of gastric cancer based on public databases. Recently, we have confirmed these data, including MMP11overexpression of gastric cancer in our following gastric cancer profiling by Oligo microarray analysis (data not shown).

In this study, we used DGED to search SAGE and EST libraries of Cancer Genome Anatomy Project and constructed an expression profiling of gastric cancer. Furthermore, we selected MMP11 from the gene expression profiling as an example, and its up-regulation was consistently detected on mRNA and protein levels in gastric cancer. Most importantly, serum MMP11 levels were found to be significantly elevated in gastric cancer patients compared with those of the control subjects and were well correlated with metastasis in gastric cancer patients. These results indicate that MMP11 has great potential to be a new tumor marker in serum, and it is an effective way to identify tumor biomarkers from gene expression profiling for guiding both early detection and therapy of cancers in the future.

We thank the tissue bank of Beijing Cancer Hospital and Institute for gastric tissue and serum specimens.