SEARCH

SEARCH BY CITATION

Keywords:

  • lymph node metastasis;
  • gastric cancer;
  • protein pathway array;
  • biomarkers

Abstract

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Lymph node status remains one of most crucial indicators of gastric cancer prognosis and treatment planning. Current imaging methods have limited accuracy in predicting lymph node metastasis. We sought to identify protein markers in primary gastric cancer and to define a risk model to predict lymph node metastasis. The Protein Pathway Array (PPA) (initial selection) and Western blot (confirmation) were used to assess the protein expression in a total of 190 freshly frozen gastric cancer samples. The protein expression levels were compared between samples with lymph node metastasis (n = 73) and those without lymph node metastasis (n = 57) using PPA. There were 27 proteins differentially expressed between lymph node positive samples and lymph node negative samples. Five proteins (Factor XIII B, TFIIH p89, ADAM8, COX-2 and CUL-1) were identified as independent predictors of lymph node metastasis. Together with vascular/lymphatic invasion status, a risk score model was established to determine the risk of lymph node metastasis for each individual gastric cancer patient. The ability of this model to predict lymph node metastasis was further confirmed in a second cohort of gastric cancer patients (33 with and 27 without lymph node metastasis) using Western blot. This study indicated that some proteins differentially expressed in gastric cancer can be selected as clinically useful biomarkers. The risk score model is useful for determining patients' risk of lymph node metastasis and prognosis.

Gastric cancer is the fourth most common malignancy and the second leading cause of cancer-related death, after lung cancer, in the world. More than 80% of patients are diagnosed at an advanced stage or experienced tumor recurrence after surgical resection.1 Despite of some decline in the incidence and prolonged overall survival of gastric cancer patients in the past decade thanks to the advancement of treatments (e.g., surgery, chemotherapy and radiotherapy), gastric cancer continues to be a major health problem with a high mortality rate.2

Most gastric cancer patients are diagnosed at Stages III or IV, with 50–75% of patients presenting with lymph node metastasis.3 The preoperative determination of lymph node status is critical in tumor staging and in planning optimal management of gastric cancer patients. For early gastric cancer without lymph node involvement, less invasive treatment (e.g., endoscopic mucosal resection) can be performed. For localized gastric cancer without lymph node involvement, surgical resection with limited lymph node dissection is recommended to reduce postoperative morbidity and mortality. For advanced gastric cancer with lymph node involvement, surgical resection with extensive lymphadenectomy is necessary to achieve better outcome.4 Currently, preoperative assessment of lymph node status is mainly based on imaging studies such as computerized tomography, magnetic resonance imaging and positron emission tomography/computed tomography. However, many studies show that lymph node size determined by imaging techniques is not a reliable indicator of lymph node metastasis.5, 6 Therefore, preoperative markers that can reliably predict lymph node metastasis in gastric cancer patients are urgently needed.

Metastasis is a multistep process involving tumor cell detachment from a primary site, invasion into surrounding connective tissue, vascular or lymphatic channels and colonization at target organs. However, the exact mechanism of lymph node metastasis remains unclear. A number of genes and proteins are involved in the metastasis process, such as cell adhesion molecules (e.g., cadherin, selectin and intergrin), proteolytic enzymes (e.g., matrix metalloproteinase and urokinase-type plasminogen activator) and cell motility factors (e.g., autocrine motility factors and high mobility group proteins).7, 8 These proteins play critical roles in lymph node metastasis and thus may be used as biomarkers to predict lymph node metastasis of gastric cancer.

In this study, we investigated the expression of 193 cancer-related proteins and phosphoproteins in primary gastric cancer tissues with lymph node metastasis and tissues without lymph node metastasis using the Protein Pathway Array (PPA) method.9–13 PPA is a recently developed high-throughput protein assay, which combines multiplex immunoblot with computational analysis. We were able to identify several differentially expressed proteins which are associated with vascular/lymphatic invasion and lymph node metastasis.

Material and Methods

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Patients and tissue samples

This study included 190 primary gastric cancer samples (130 in the test cohort and 60 in the validation cohort) (Table 1), which were obtained from patients who underwent D2 gastrectomy after informed consent between August 2008 and January 2011 at The First Hospital of Jilin University, Jilin, China. This study was reviewed and approved by the Institution Ethical Review Boards of The First Hospital of Jilin University. The representative tumor tissues were dissected and frozen in a liquid nitrogen tank within 30 min of removal after immediate pathological examination. Tumor samples of 3 × 3 × 5 mm3 were taken from areas without gross necrosis and washed three times with iced-cold normal saline. Samples were kept at −80°C until analysis. The histological diagnosis of resected specimens was confirmed by at least two well-trained pathologists.

Table 1. Patients' demographics and clinicopathological characteristics
inline image

All 190 patients had gastric cancer at pT3 stage (invasion to serosa) and none had dissemination to distant organs (M0) (Table 1). In the test cohort, 57 patients were lymph node metastasis negative (LNNs) and 73 were lymph node metastasis positive (LNPs) at the time of surgery. These samples were randomly divided into a training set (40 LNNs and 50 LNPs) and a validation set (17 LNNs and 23 LNPs). In the validation cohort, 27 patients had no lymph node metastasis and 33 had lymph node metastasis at the time of surgery. The mean follow-up period was 446 days (ranging from 22 days to 1,162 days) for the test cohort and 399 days (ranging from 77 days to 1,171 days) for validation cohort. The last day of follow-up was December 1, 2011. The survival time was counted from the date of surgery to the last day of follow-up or date of death. The cause of death of majority patients was cancer recurrence or distant metastasis. However, patients who died of other causes or were lost to follow up were also included in the analysis (i.e., overall survival).

Protein pathway array

Total protein from each sample was extracted as previously described.10 Briefly, 1 ml of 1× lysis buffer (Cell Signaling Technology, Danvers, MA) with 1× protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN) and 1× phosphatase inhibitor cocktail (Roche Applied Science, Indianapolis, IN) was added to each tissue sample and the lysate was sonicated twice for 15 sec each time on ice, and then centrifuged at 14,000 rpm for 30 min at 4°C. The protein concentration was determined using BCA Protein Assay kit (PIERCE, Rockford, IL). Three hundred μg of protein lysate was loaded in one well across the entire width of a 10% SDS polyacrylamide gel and separated by electrophoresis as described previously.10, 13 After electrophoresis, the proteins were transferred electrophoretically to a nitrocellulose membrane (Bio-Rad, Hercules, CA), which was then blocked for 1 hr with blocking buffer consisting of 3% BSA in 1× TBST containing 20 mM Tris-HCl (pH 7.5), 100 mM NaCl and 0.1% Tween-20. Next, the membrane was clamped on a Western blotting manifold (Mini-PROTEAN II Multiscreen apparatus, Bio-Rad, Hercules, CA) which isolates 20 channels across the membrane. The multiplex immunoblot was performed using a total of 193 protein-specific or phosphorylation site-specific antibodies (Supporting Information Table 1 for list and sources of the antibodies). Seven sets of antibodies (a total of 30–36 protein-specific or phosphorylation site-specific antibodies in the first five sets and 15–18 protein-specific antibodies in the last two sets) were individually used for each membrane (each sample was transferred electrophoretically to 2 nitrocellulose membranes and each membrane performed multiplex immunoblot at most four sets of antibodies) and all antibodies (from various companies) were validated independently before inclusion in PPA. For the first set of 36 primary antibodies, a mixture of two antibodies in the blocking buffer were added into each channel and then incubated at 4°C overnight. The membrane was then washed with 1×TBS and 1×TBST, and was further incubated with secondary anti-rabbit (Bio-Rad, Hercules, CA), anti-mouse(Bio-Rad, Hercules, CA) or anti-goat(Santa Cruz Biotechnology, Santa Cruz, CA) antibody conjugated with horseradish peroxidase for 1 hr at room temperature. The membrane was developed with chemiluminescence substrate (Immun-Star HRP Peroxide Buffer/Immun-StarHRP Luminol Enhancer) (Bio-Rad, Hercules, CA), and chemiluminescent signals were captured using the ChemiDoc XRS System (Bio-Rad, Hercules, CA). The same membrane was then stripped off using stripping buffer (Restore Western blot stripping buffer, Thermo Scientific, Rockford, IL) and then used to detect a second set of primary antibodies as described above.

For PPA data analysis, the signals of each protein were determined by densitometric scanning (Quantity One software package, Bio-Rad) and the background was locally subtracted from raw protein signal. The background-subtracted intensity was normalized by “global median subtraction” method to reduce variation among different experiments, that is, the intensity of each protein from each sample divided by total intensities of all proteins from the same sample and then multiplied by average intensities of all proteins in all samples.10

Western blot analysis

Total proteins were extracted from 60 fresh gastric cancer samples as described above. Proteins from each sample were loaded into each well and resolved on 10% SDS-PAGE. After transferring the proteins onto nitrocellulose membranes, the membranes were incubated with primary antibodies and then secondary antibodies labeled with the horseradish peroxidase as described above. The signals of each protein were determined by densitometric scanning (Quantity One software package, Bio-Rad). The primary antibodies included rabbit anti-TFIIH p89, ADAM8, COX-2 and CUL-1 (Santa Cruz Biotechnology, Santa Cruz, CA), goat anti-Factor XIII B (Santa Cruz Biotechnology, Santa Cruz, CA), mouse anti-β-actin (Sigma, St. Louis, MO) and mouse anti-GAPDH (Santa Cruz Biotechnology, Santa Cruz, CA).

Statistical analysis

Significant Analysis of Microarray (SAM) tool (http://www-stat.stanford.edu/∼tibs/SAM/) was used to select the differentially expressed proteins between LNNs and LNPs. K-fold cross-validation (K = 10) was used to select those proteins with a great discriminating power to distinguish LNPs from LNNs using BRB Array Tools v.3.3.0 (htt://linus.nci.nih.gov/BRB-ArrayTools.html).15

For the calculation of the lymph node metastasis risk scores, univariate and multivariate logistic regression analyses were performed.16 The analyses generated a set of independent predictors with corresponding regression coefficient, p value, odds ratio (OR) and 95% confidence interval of the OR. The calibration of the model, that describes agreement between expected and actual observed outcomes, was performed using the Hosmer-Lemeshow goodness-of-fit test. The Cox proportional hazard regression analysis, Kaplan-Meier method and log-rank test were used to determine the overall survival. The logistic regression, the Receiver Operating Characteristic (ROC) curve and the survival analysis were performed using SPSS 17.0 software (SPSS, Chicago, IL). The area under the curve (AUC) was calculated using the MedCalc statistical software (Mariakerke, Belgium). A p value < 0.05 was considered to be statistically significant.

Results

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Identification of differentially expressed proteins between LNPs and LNNs and correlation with clinical parameters of gastric cancer in the test cohort

Totally, 40 LNNs and 50 LNPs in the training set were screened initially for differentially expressed proteins using the PPA method. Among the 193 proteins and phosphoproteins tested, 131 were detected in either LNNs or LNPs (Supporting Information Table 1) and 27 were found to be differentially expressed between LNPs and LNNs using SAM analysis (q < 0.05 and False Discovery Rate <5%). Of these 27 proteins and phosphoproteins, 17 were up-regulated in the LNPs, including TFIIH p89, Factor XIII B, ADAM8, uPA, Rab 7, PSCA, VCAM-1, p-p70 S6 Kinase (Thr389), MMP-9, nm23-H1/2/3, Endoglin, Rap 1, Pax-2, COX-2, DRG1, Cdc25B and CUL-1, while 10 were down-regulated in the LNPs, including PSM, fusin, VSV-G, Stat1, PCNA, FKHR, p-Rb (Ser780), WT1, Wnt-1 and connexin 43. To identify a robust set of proteins for classification, a K-fold cross validation (K = 10) using support vector machine (SVM) was performed using these 27 proteins. For this training set, the rate of correct classification was 83% and the sensitivity and specificity were 86 and 80%, respectively. Twelve proteins with p values <0.01 (Factor XIII B, PSM, ADAM8, VCAM-1, uPA, Rab 7, TFIIH p89, fusin, nm23-H1/2/3, Pax-2, PSCA, Endoglin) were selected as the best classifiers by SVM. These 12 proteins were tested on a separate validation set (17 LNNs and 23 LNPs) using SVM as described above. The rate of correct classification was 86% and the sensitivity and specificity were 88 and 82.5%, respectively.

To identify the proteins associated with other important clinical characteristics (Table 1) in addition to lymph node metastasis, SAM analysis was performed. The increased expression of MMP-9 was significantly correlated with tumors greater than 5 cm in diameter (q < 0.001). The decreased expression of p-Rb (Ser780) and Rab 7 were significantly correlated with poor histologic differentiation (q < 0.001). The increased expression of ADAM8, Rab 7, nm23-H1/2/3, uPA and TFIIH p89, and decreased expression of PSM were significantly correlated with the presence of vascular/lymphatic invasion (q < 0.001).

Establishment of risk model for predicting lymph node metastasis

To identify the proteins and the clinicopathological variables that are associated with lymph node metastasis, univariate logistic regression analysis was performed. Among the clinicopathological variables (Table 1), including age, gender, family history, tumor location, tumor size, histologic differentiation and vascular/lymphatic invasion, only vascular/lymphatic invasion was found to be correlated with lymph node metastasis. Among 27 differentially expressed proteins and phosphoproteins between LNPs and LNNs, 10 were found to be correlated with lymph node metastasis, including p-p70 S6 Kinase (Thr389), Factor XIII B, PSM, Endoglin, TFIIH p89, nm23-H1/2/3, COX-2, CUL-1, connexin 43 and ADAM8 (Table 2). Therefore, the vascular/lymphatic invasion and the above 10 proteins were included in a stepwise multivariate logistic regression analysis and the vascular/lymphatic invasion, Factor XIII B, TFIIH p89, ADAM8, COX-2 and CUL-1 still stood as independent predictors of lymph node metastasis with p value = 0.903 by Hosmer-Lemeshow goodness-of-fit test (Table 3).

Table 2. Factors correlated with lymph node metastasis based on univariate analysis
inline image
Table 3. Independent predictors of lymph node metastasis based on the multivariate analysis
inline image

On the basis of the multivariate logistic regression analysis, the above six variables (Table 3) were included in the risk score calculation using the following formula: Risk Score = Probability×100, where Probability = eZ/(eZ + 1), where “e” was the base of natural logarithm and “Z” was the result of the logistic regression.16 Z was calculated as: Z = B0 + B1X1 + B2X2 +…Bp Xp, where “B0” was the constant of regression coefficients, “X1Xp” were the independent variables, and “B1Bp” were their corresponding coefficients. In this study, Z = −116.193 + 3.380 (vascular/lymphatic invasion) + 2.882 (Factor XIII B) + 1.310 (TFIIH p89) + 3.415 (ADAM8) + 1.339 (COX-2) + 1.326 (CUL-1). For a given patient, the risk score can be calculated based on the values of 6 variables using the above formula. The risk score ranged from 0 to 100.

The distribution of the risk score (mean = 56.154, ranged 0.002–99.953) of 130 patients was presented in Figure 1. Most of the cases were distributed in the extremes of the range, that is, 31.5% (41 out of 130) of cases distributed between 0 and 20 and 43.8% (57 out of 130) of cases distributed scores between 80 and 100 (Fig. 1a), suggesting good separation of the low-risk group from the high-risk group with this model. Furthermore, the risk of lymph node metastasis was increased with increase of the risk scores (Fig. 1b): 2.4% (1 out of 41) of patients with scores between 0 and 20 had lymph node metastasis, in contrast, 96.5% (55 out of 57) of patients with scores between 80 and 100 had lymph node metastasis.

thumbnail image

Figure 1. Correlation of risk scores with lymph node metastasis in 190 patients (the test cohort = 130 and the validation cohort = 60): (a) Most patients were distributed in the extremes of the risk score ranges (0–20 and 80–100), whereas a few patients were distributed in the middle of the range. (b) The incidence of lymph node metastasis correlated with score levels. The results showed that the higher risk scores, the higher percentage of lymph node metastasis. Solid bars represent the test cohort and open bars represent the validation cohort.

Download figure to PowerPoint

To further determine the optimal model to predict lymph node metastasis, the receiver operating characteristic (ROC) curve analysis was applied to either single or combined predictors (Fig. 2) in the test cohort. To control over-fitting of the ROC curves, an internal validation was performed using K-fold cross validation (K = 5). The area under the curves (AUC) were calculated based on the above ROC curves for risk score model, five protein combination, vascular/lymphatic invasion, Factor XIII B, TFIIH p89, ADAM8, COX-2 and CUL-1 with AUC 0.932, 0.895, 0.671, 0.791, 0.763, 0.689, 0.580 and 0.659, respectively (Fig. 2). These results suggest that the risk score model which combined vascular/lymphatic invasion and five proteins was the best among all predictors (p < 0.05).

thumbnail image

Figure 2. Receiver operating characteristic (ROC) curves of different predictor models for lymph node metastasis. Eight predictor models i.e., risk score model, five protein combination, vascular/lymphatic invasion, Factor XIII B, TFIIH p89, ADAM8, COX-2 and CUL-1 were included in the analysis. Both risk score model and five protein combination were better than individual predictors (p < 0.05).

Download figure to PowerPoint

Correlation of risk score with overall survival

Since lymph node metastasis is an important predictor of overall survival, we examined the possibility of metastasis risk score to predict overall survival of the patients in the test cohort. On the basis of the risk score distribution, the patients were separated into three groups (Fig. 3a): low-risk group (risk scores ≤10), intermediate-risk group (risk scores >10 and <90) and high-risk group (risk scores ≥90). Among these three groups, lymph node metastasis was present in 2.8, 52.3, and 98% of patients in the low-risk score group, intermediate-risk score group and high-risk score group, respectively. Kaplan-Meier analysis showed that overall survival of the patients was worst in the high-risk group and best in the low-risk group (log-rank test: p = 0.000) (Fig. 3b). To further determine whether the risk score model can be an independent predictor, a multivariate Cox proportional hazard regression analysis was performed taking into consideration other clinicopathology variables, including age, gender, tumor location, tumor size, vascular or lymphatic invasion, histologic differentiation and family history. The data showed that the risk score model still stood as an independent predictor with a hazard ratio of 4.735 (p = 0.005, 95% CI: 1.596–14.047). In addition, age at surgery was also an independent predictor of survival with a hazard ratio of 2.891 (p = 0.009, 95% CI: 1.310–6.380), consistent with our previous report.10

thumbnail image

Figure 3. Kaplan-Miere survival analysis of patients with gastric cancer. (a, c) The patients in the test cohort (a) and the validation cohort (c) were ranked according to their metastasis risk scores. The two lines divided the cases into low-risk group (risk scores ≤10), intermediate-risk group (risk scores >10 and <90) and high-risk group (risk scores ≥90). (b, d) The overall survivals of the patients in the test cohort (b) and validation cohort (d) were determined based on risk score groups. p values were determined by log-rank test.

Download figure to PowerPoint

Validation of the risk score model using second cohort of patients

To further confirm the ability of the risk score model to predict lymph node metastasis and survival, an independent validation cohort of gastric cancer patients was analyzed (Table 1). The expression levels of Factor XIII B, TFIIH p89, ADAM8, COX-2 and CUL-1 in cancer tissues were assessed using conventional Western blot. The patients' risk score were calculated as previously described after adjusting the expression levels against β-actin and GAPDH. The distribution of the scores in the validation cohort is shown in Figure 1a. Similar to the test cohort, most patients were distributed in the extremes of the range: 38.3% (23 out of 60) cases were below 20 and 43.3% (26 out of 60) cases were over 80. The risk of lymph node metastasis was also increased with increase of the risk scores, that is only 8.7% (2 out of 23) cases with scores below 20 had lymph node metastasis, in contrast, 96.2% (25 out of 26) cases with scores over 80 had lymph node metastasis (Fig. 1b).

As shown in the test cohort (Fig. 3a), the patients were separated into three groups based on risk-score distribution (Fig. 3c). Lymph node metastasis was present in 4.5, 68.8 and 95.5% of patients in the low-risk score group, intermediate-risk score group and high-risk score group, respectively. Kaplan-Meier analysis showed that overall survival of the patients was worst in the high-risk group and best in the low-risk group (log-rank test: p = 0.014) (Fig. 3d). The results further confirmed the ability of metastasis risk score to predict prognosis of patients with gastric cancer.

Discussion

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Lymph node metastasis of gastric cancer is an important factor for prognosis of survival and for planning surgical and chemotherapeutic treatment. We evaluated the expression of 193 proteins and phosphoproteins in primary gastric cancer tissue using PPA and found 27 (14%) were differentially expressed between LNPs and LNNs. Most of these differentially expressed proteins are important in diverse cellular processes including cell cycle regulation, cell proliferation, cell death, cell-to-cell signaling and interaction, cell adhesion, migration and invasion. Some of the dysregulated proteins were reported in gastric cancer by other investigators, consistent with our findings. For example, VCAM-1, MMP-9 and uPA were up-regulated in gastric cancer and closely related to the metastasis,17–19 while FKHR and connexin43 were down-regulated in gastric cancer.20, 21 Some of the dysregulated proteins were reported in other cancers but not in gastric cancer. For example, Pax-2 was up-regulated in renal cell cancer22 and WT1 was down-regulated in lung cancer.23 In addition, our study showed that six proteins (ADAM8, Rab 7, nm23-H1/2/3, uPA, TFIIH p89 and PSM) were significantly correlated with vascular/lymphatic invasion, suggesting their roles in tumor cell migration, invasion and metastasis.

Although various models have been proposed to predict lymph node metastasis, the clinical utility of these models is limited. For example, the Maruyama computer model based on several clinical variables (age, sex, tumor type, depth of invasion, location, diameter of the primary tumor and histological type) was initially proposed by Kampschoer in 1989.24 Subsequent studies showed that although the sensitivity for lymph-node detection was high (97–100%), the specificity was very low (20%).25 Recently, the gene expression profile of primary gastric cancer was applied to identify mRNAs associated with lymph node metastasis.26 Marchet et al. evaluated 5,541 genes in frozen tumor samples obtained from 32 patients with primary gastric adenocarcinomas (21 with lymph node metastasis and 11 without lymph node metastasis). Of 136 differentially expressed mRNAs, three (Bik, aurora kinase B, eIF5A2) could correctly predict lymph node status in 30 of 32 cases. However, limited sample size precludes its clinical application. Our study of a cohort of 130 patients showed that up-regulation of five proteins (Factor XIII B, TFIIH p89. ADAM8, COX-2 and CUL-1) as well as vascular/lymphatic invasion status were independent predictors of lymph node metastasis (Table 3). The metastasis risk model created by combining the expression levels of the above five proteins and vascular/lymphatic invasion status was able to predict lymph node metastasis of each individual patient (Fig. 3a). This model can predict lymph node metastasis in patients with T3 gastric cancer with 84.6% accuracy, 86.3% sensitivity and 91.2% specificity based on ROC analysis. This risk model was further confirmed using an independent cohort of patients (Fig. 3c).

This metastasis risk model is clinically relevant and can be used to guide preoperative surgery planning and medical treatment. For example, a radical gastrectomy with extensive lymphadenectomy should be carried out with subsequent aggressive chemotherapy for the patients with high-risk scores to improve patients' survival. In contrast, for the patients with low-risk scores, a radical gastrectomy without extensive lymophadenectomy may be performed to reduce surgery-related complications and mortality.

Our study showed that several proteins were associated with vascular/lymphatic invasion and/or lymph node metastasis. Factor XIII is a protransglutaminase which is involved in the final stages of the clotting cascade. In blood plasma, Factor XIII exists as a heterotetramer composed of two catalytic subunits (Factor XIII A) and two inhibitory subunits (Factor XIII B).27 This study showed that the expression of Factor XIII B was increased in gastric cancer with lymph node metastasis and its level correlated with lymph node metastasis. The increased expression of Factor XIII B was also observed in several cancer cell lines, including RT4, UICC3, T24, T24t, MDA-MD 231, HeLa, H1975 and K562 (data not show), suggesting that these tumor cells produce this protein. ADAM (a disintegrin and metalloprotease) is a group of glycoproteins with a multidomain consisting of pro-metalloprotease, disintegrin-like, cysteine rich, EGF-like and transmembrane domains.28 By acting on a large panel of membrane-associated and extracellular substrates, they control several cell functions such as adhesion, fusion, migration and proliferation. An overexpression of ADAM8 (also known as CD156) was observed in various cancers and was correlated with metastasis and unfavorable prognosis, including lung cancer,29 renal cell cancer,30 pancreatic cancer31 and prostate cancer.32 Our results demonstrated that increased expression of ADAM8 in gastric cancer correlated with vascular/lymphatic invasion as well as lymph node metastasis, consistent with previous reports.

The general transcription factor II H (TFIIH) is a multisubunit, multifunctional protein complex which consists of two subcomplexes: a seven-subunit core complex [xeroderma pigmentosum B (XPB; p89), xeroderma pigmentosum D (XPD; p80), p62, p52, p44, p34, and trichothiodystrophy A] and a three-subunit cycle-dependent kinase (CDK)-activating kinase (cdk7/cyclin H/MAT1). TFIIH participates in regulating transcription of RNA polymerase I and II and repairing damaged DNA via nucleotide excision repair pathway.33 TFIIH p89 (XPB) encoded by the ERCC3 gene contains seven helicase motifs.34 Although increased expression of ERCC 1 or 2 was observed in various cancers including gastric cancer,35 there is no report on the expression level of ERCC3 (TFIIH p89) in gastric cancer. CUL-1 (Cullin 1) is an important component of ubiquitin E3 ligases and ubiquitinates a broad range of proteins involved in cell-cycle progression, signal transduction and transcription.36 Recently, Bai et al. reported that Cullin 1 overexpression in gastric cancer was significantly correlated with gastric cancer TNM stage, depth of invasion, lymph node metastasis as well as worse overall survival rates.37 Furthermore, CUL-1 knockdown inhibits cell growth by up-regulating p27 expression and decreases cell adhesion ability by suppressing the expression of SRC family kinases and focal adhesion kinase.37 Cyclooxygenase (COX) is a key enzyme that catalyzes the formation of prostaglandin (PG), thromboxanes and other eicosanoids from arachidonic acid. COX-2 plays an important role in inflammation and carcinogenesis.38 Recent studies demonstrated that COX-2 is overexpressed in many malignant tumors and may associate with metastasis, such as breast cancer,39 lung cancer,40 colorectal cancer41 and gastric cancer.42 In this study, increased expressions of TFIIH p89, CUL-1 and COX-2 in gastric cancer correlated with vascular/lymphatic invasion as well as lymph node metastasis, consistent with previous reports.

Although our study demonstrated a strong association of these proteins with vascular/lymphatic invasion as well as lymph node metastasis, several limitations do exist. This study included only 190 patients from a single institution, thus the findings may not generalize to other gastric cancer patients. Also, only 193 antibodies were included in this study, thus, other potential protein markers may have been missed. Last, this study only included patients with pT3 gastric cancer to avoid potential confounding factors, therefore, a further study needs to determine the association of these proteins with other tumor stages (T1 and T2).

In conclusion, our data showed a significant difference in protein expression in primary gastric cancer with or without lymph node metastasis. The expression of proteins strongly correlated with vascular/lymphatic invasion, lymph node metastasis and overall survival, suggesting the potential roles of these proteins in gastric cancer metastasis. Finally, the metastasis risk score can be used for planning personalized treatment (surgery and chemotherapy) of each individual patient with gastric cancer. Future studies will focus on understanding the roles of these proteins in lymph node metastasis of gastric cancer and confirming the ability of the risk score model to predict the status of lymph node metastasis in a different cohort of patients.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

The authors wish to sincerely thank Mr. Yang Li for his excellent technical support in protein preparation, Dr. Xiufen Liu and Dr. Miao Cui for PPA training, Dr. Suyan Tian and Dr. Fei Yin for statistical support.

References

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
IJC_27864_sm_SuppTab1.doc38KSupporting Information Table 1

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.