Helicobacter pylori has been recognized as a definite carcinogen for gastric cancer (GC); however, the pathogenesis of H. pylori infection remains unclear. Runt-related transcription factor 3 (RUNX3) is a candidate tumor suppressor gene whose deficiency is causally related to GC. However, in H. pylori infection-associated GC, the role of RUNX3 has not been studied.
The authors used real-time methylation-specific polymerase chain reaction analysis to determine methylation status of the RUNX3 promoter in a spectrum of gastric lesions, including 220 samples of chronic atrophic gastritis, 196 samples of intestinal metaplasia, 134 samples of gastric adenoma, 102 samples of dysplasia, and 202 samples of GC with paired noncancerous mucosa tissues and corresponding blood specimens. The association of abnormal methylation with precancerous gastric lesions was evaluated along with the association between RUNX3 methylation and H. pylori infection, and the concordance of methylation levels was investigated between serum and tissues.
The results indicated that increasing RUNX3 promoter methylation was correlated with distinct stages of GC progression. GC tissues had the highest methylation proportion (75.2%) compared with precancerous gastric lesions, including chronic atrophic gastritis (15.9%), intestinal metaplasia (36.7%), gastric adenoma (41.8%), and dysplasia (54.9%). H. pylori infection, a major risk factor for GC, contributed to the inactivation of RUNX3 in gastric epithelial cells through promoter hypermethylation. The levels of RUNX3 methylation in serum were in significant concordance with the methylation levels observed in GC tissues (P = .887).
Gastric cancer (GC) is the most common cancer in eastern Asia and the third most frequent cancer across the world.1 In addition, it is the second most common cancer-related cause of death after lung cancer.1, 2 On the whole, from 65% to 70% of incident cases and deaths from GC occur in less developed countries. Despite its decreasing trend over many decades, GC remains a major public health problem in the world.1-3 Gastric carcinogenesis has been relatively well described, and 2 distinct pathways of carcinogenesis have been proposed. It has been demonstrated that gastric adenocarcinomas develop through a stepwise progression of several genetic and epigenetic events, which propel the tumor down the mucosa-adenoma-carcinoma pathway.4 Environmental exposures and genetic factors reportedly to cause GC, but the debate continues regarding whether environment or heritability plays the principal role.5 Knowledge of the factors that influence gastric carcinogenesis is important in the development of effective strategies for prevention and treatment. Helicobacterpylori infection, the preneoplastic lesion chronic atrophic gastritis, intestinal metaplasia, free radicals, and familial GC predisposition have been identified as risk factors for GC.6-11 Recently, studies have demonstrated that H. pylori infection may double the risk of stomach cancer. Infection with a particular type of H. pylori called cytoxin-associated antigen A (cagA)-positive H. pylori can increase the risk even more.4, 8-14 Although many studies have focused on the correlation between H. pylori infection and GC, the mechanism remains unclear. Recent studies have indicated that H. pylori infection plays an important role in the inactivation of tumor suppressor genes and that this inactivation contributes to the pathogenesis of H. pylori.15-20
One of the runt domain family of transcription factors, runt-related transcription factor 3 (RUNX3), reportedly is inactivated in various cancers, including hepatocellular carcinoma,21 lung adenocarcinoma,22 esophageal adenocarcinoma,23 GC,24 oral squamous cell carcinoma,25 colorectal cancer,26 breast cancer,27 and adenoid cystic carcinomas.28 In addition, as an important downstream target of transforming growth factor beta (TGFβ) superfamily signaling, RUNX3 acts as a tumor suppressor by regulating a series of cancer-related genes, such as tumor protein 53 (p53),29 cyclin-dependent kinase inhibitor 1A (P21),30 AT motif-binding factor 1 (ATBF1),31 notch 1 (Notch),32 cyclin-dependent kinase inhibitor 1B (P27), and Caspace3.33 Some studies have indicated that loss of RUNX3 contributes to hyperplasia and intestinal metaplasia of gastric mucosa epithelial cells in an animal model,34 whereas the restoration of RUNX3 activated the apoptotic pathway in GC.31 In recent years, it has been observed that RUNX3 activity is reduced in GC tissues because of hemizygous deletion,35 promoter hypermethylation,24 proteasome degradation,36, 37 and protein mislocalization.38 However, the exact correlation between RUNX3 inactivation and H. pylori infection in gastric mucosa epithelial cells remains unknown, and further study is necessary.
In the current study, we investigated RUNX3 methylation in a spectrum of gastric lesions, including chronic atrophic gastritis, intestinal metaplasia, gastric adenoma, and GC with corresponding samples of noncancerous mucosa, and we evaluated the association of abnormal methylation with precancerous gastric lesions. We also evaluated the correlation between RUNX3 methylation and H. pylori infection. Finally, we explored the sensitivity of RUNX3 promoter hypermethylation in circulating tumor DNA as a valuable biomarker for the early detection of GC.
MATERIALS AND METHODS
The Institutional Review Board on Medical Ethics of Zhejiang Province Cancer Hospital approved the method of tissue collection, including informed consent. All patients provided informed consent before the collection of samples according to institutional guidelines. Two hundred two archival samples of surgically resected GC and noncancerous mucosa, 134 archival samples of endoscopically resected gastric adenoma, and 518 archival samples of endoscopically obtained non-neoplastic gastric mucosa (including 220 samples of chronic atrophic gastritis, 196 samples of intestinal metaplasia, and 102 samples dysplasia) were studied. Among these adenoma samples, hyperplastic polyps, serrated adenomas, and mixed-type adenomas were excluded. Tumor tissues were collected at the time of surgery from 202 patients with primary gastric adenocarcinoma at Zhejiang Province Cancer Hospital and Zhejiang Province People Hospital between January 2008 and December 2009. The diagnosis for all patients who had not received preoperative radiotherapy or chemotherapy was confirmed not only with gastrointestinal endoscopy followed by pathologic analysis but also by examining paraffin-embedded tissue sections. Demographic, clinical, and histopathologic parameters were collected. Meanwhile, a spectrum of gastric lesions, including 220 samples of chronic atrophic gastritis, 196 samples of intestinal metaplasia, 134 samples of gastric adenoma, and 102 samples of dysplasia, also were collected at Zhejiang Province People's Hospital, the Fifth People's Hospital at Hangzhou Yuhang District, and the Hospital of Traditional Chinese Medicine at Hangzhou Yuhang District. In the benign gastric disease group, at least 3 biopsies were obtained by endoscopy from each patient's gastric mucosa; the first biopsy was used for a rapid urease test, the second biopsy was immediately frozen and stored, and the third biopsy was embedded in paraffin. Eight hundred fifty-four corresponding serum samples were obtained before surgery or endoscopy. At the same time, 120 age-matched and sex-matched, healthy volunteers were selected randomly in Hangzhou City, Zhejiang Province based not only on the absence of disease and weakness but on the absence of any disease, including inflammation. Among these volunteers, there were 73 men and 47 women, and they ranged in age from 29 years to 74 years (median age, 56.5 years). All volunteers provided informed consent.
Analysis of H. pylori Infection
Biopsies were obtained from all patients who had endoscopic evaluations. H. pylori status had been determined with a rapid urease test and the Giemsa staining method. The urease test was done using a freshly prepared solution of urease test reagent, and the test was scored as positive if the color of the solution turned pink within 24 hours. Biopsy materials embedded in paraffin were stained with Giemsa for histopathologic examination, and the biopsy was considered positive for H. pylori infection when 2 tests were positive, and biopsies that had only a single positive result were excluded from further analysis.
After identifying carcinoma, adenoma, or intestinal metaplasia on hematoxylin and eosin-stained slides, portions of carcinoma, adenoma, or metaplastic mucosa were scraped from 20-mm-thick paraffin sections. The materials collected were dewaxed by washing in xylene and then by rinsing in ethanol. The dried tissues or fresh frozen samples were digested with 40 μL of 200 μg/mL proteinase K (Sigma-Aldrich, St. Louis, Mo) at 42°C for 72 hours and subjected to the traditional method of DNA extraction using phenol/chloroform/isoamyl alcohol and ethanol precipitation.
Sodium Bisulfite Modification and Real-Time Methylation-Specific Polymerase Chain Reaction
DNA samples were treated with bisulfite to convert all unmethylated cytosine to uracils while leaving methylated cytosines unaffected. Using EpiTect Bisulfite Kits (Qiagen, Hilden, Germany), 1 or 2 μg of genomic DNA were denatured by treatment with NaOH, modified by sodium bisulfite, then purified by using the phenol/chloroform method. Primers for the methylated and unmethylated promoter regions of RUNX3 were obtained from a previous report28 (methylated: forward, 5′-TAT TCG TTA GGG TTC GTT CGT TGC-3′; reverse, 5′-ACG ACC GCG AAC GAA CTT CGA AAC-3′ [201 base pairs]; unmethylated: forward, 5′-TAT TTG TTA GGG TTT GTT TGT TGT-3′; reverse, 5′-ACA ACC ACA AAC AAA CTT CAA AAC-3′ [201 base pairs]). Bisulfite-modified DNA was used for real-time methylation-specific polymerase chain reaction (real-time MSP) using FastStart DNA Master SYBR Green I (Takara Biotechnology Inc., Shiga, Japan) on an ABI 7500 PCR instrument (Applied Biosystems, Foster City, Calif). All samples were analyzed with primer sets for both methylated and unmethylated DNA. The relative amount of methylation in each unknown sample was calculated as the percentage methylation = 100 × (number of copies of methylated DNA/[number of copies of methylated + unmethylated DNA]).39 The sum of unmethylated plus methylated DNA (U + M) was used as an approximation of the total number of target gene copies. Methylated DNA was scored according to the methylated percentage (0, <20%; 1, 20%-40%; 2, 40%-60%; 3, 60%-80%; and 4, >80%; scores of 0, 1-3, and 4 were considered unmethylated, partially methylated, and fully methylated, respectively). PCR products were confirmed by 2% agarose gel electrophoresis, observed by ethidium bromide staining, then observed and photographed under ultraviolet illumination (BioSpectrumAC BioImaging Systems; Ultra-Violet Products Inc., Upland, Calif). Human genomic DNA (New England Biolabs, Beverly, Mass) treated with SssI methyltransferase in vitro was used as a positive control. Peripheral blood DNA from healthy untreated individuals was used as a negative control.
The correlations between RUNX3 methylation and clinicopathologic factors were analyzed using Pearson chi-square tests and continuity correction tests. Multinomial logistic regression analysis was used to analyze the risk factors for RUNX3 methylation. Because the H. pylori infection status of noncancerous mucosa was the same as that in GC, chronic atrophic gastritis was used as the reference group instead.
Correspondence analyses and Z tests were used to calculate kappa statistics and corresponding P values that evaluated the concordance of RUNX3 methylation status between serum and tissue samples from patients with GC. To evaluate the ability of serum RUNX3 methylation levels to predict GC in serum, a receiver operating characteristic curve analysis was performed, and the area under the ROC curve (AUC), accuracy, sensitivity, specificity, false-positive rate, and false-negative rate were calculated. The Kaplan-Meier method was used to analyze correlations between RUNX3 methylation and patient survival. All statistical analyses were carried out using Predictive Analytics Software Statistics (PASW Statistics, version 18; Polar Engineering and Consulting, Nikiski, Alaska). All statistical tests were 2-tailed, and the significance level was set at P < .05.
Methylation Frequency of the RUNX3 Promoter in Different Gastric Lesions
In total, samples from 854 patients (1056 samples), including 220 samples of chronic atrophic gastritis, 198 samples of intestinal metaplasia, 134 samples of gastric adenoma, 102 samples of dysplasia, and 202 samples of GC with corresponding noncancerous mucosa, which were used as normal controls, were assessed by real-time MSP analysis of the RUNX3 promoter; and the different methylation levels were scored from 0 to 4, representing negative methylation and methylation levels 1 through 4 (Fig. 1). Samples of chronic atrophic gastritis, intestinal metaplasia, and dysplasia were classified further into 3 subgroups with mild, moderate, and severe. In addition, GC samples were classified into 4 subgroups according to disease stage (stages I, II, III, and IV) based on the seventh edition of the American Joint Committee on Cancer TNM staging system. Table 1 indicates the distribution of clinicopathologic characteristics.
Table 1. The Distribution of Different Clinicopathologic Characteristics (Sex, Age, Lesion Site, and Helicobacter pylori Infection) in Different Types of Gastric Lesions
Stepwise Cumulation of RUNX3 Promoter Methylation During Disease Progression
Patients who had methylation scores of 1 to 4 were considered the methylation group, and the frequency of methylation for this group is illustrated in Figure 2. In these patients, as the tissue type progressed from noncancerous mucosa to chronic atrophic gastritis, intestinal metaplasia, gastric adenoma, dysplasia, and GC, the proportion of RUNX3 promoter methylation increased from 7.4% (n = 15) 15.9% (n = 35), 36.7% (n = 72), 41.8% (n = 56), and 54.9% (n = 56) to 75.2% (n = 152), respectively (Fig. 2A). The methylation proportion also was correlated positively with the severity of lesions (mild, moderate, and severe) and with GC stage (stages I-IV) (Fig. 2B). In addition, the frequency of RUNX3 methylation was correlated with depth of tumor invasion (P < .0001), lymph node metastasis (P < .0001), distant metastasis (P = .024), lymphatic vessel invasion (P = .024), and venous invasion (P = .0002) but not with growth pattern (P = .274), differentiation (P = .178), age (P = .243), or tumor location (P = .960) (Fig. 2C).
Risk Factors for RUNX3 Promoter Methylation
To further analyze the risk factors for RUNX3 methylation, all patients were reclassified into 3 groups according to their methylation scores (0, 1-3, and 4). Multinomial logistic regression analysis indicated that the risk of a score of 4 increased significantly when the lesion type progressed from intestinal metaplasia to gastric adenoma, dysplasia, and GC compared with progression to chronic atrophic gastritis. The odds ratio (OR) was 19.66 (95% confidence interval [CI], 4.13-93.57) for intestinal metaplasia, 25.75 (95% CI, 5.12-129.41) for gastric adenoma, 53.39 (95% CI, 10.40-273.99) for dysplasia, and 498.22 (95% CI, 103.90-2389.17) for GC. The correlation also existed between the risk of a score of 4 and H. pylori infection. The OR for H. pylori infection was 68.71 (95% CI, 36.64-104.37), which was higher than the OR for all the benign lesion types but lower than the OR for GC (Table 2). In addition, there was no correlation between the risk of RUNX3 promoter methylation and other clinicopathologic variables, such as sex, age, and tissue location (data not shown).
Table 2. The Risk of Runt-Related Transcription Factor 3 Promoter Methylation According to Clinicopathologic Variables
Multinomial logistic regression analysis was adjusted for age, sex, and lesion site.
With a methylation score of 0 as reference, a significant 498.22-fold increased risk of having a methylation score of 4 (95% CI, 103.90-2389.17) was observed among patients who had GC compared with patients who had CAG. In addition, a significant 68.71-fold increased risk of having a methylation score of 4 (95% CI, 33.64-104.37) was observed among patients who were positive for HP infection compared with HP-negative patients.
The methylation status of all patients with and without H. pylori infection is illustrated in Figure 3A and Figure 3B, respectively. The frequency of methylation for all patients who were positive for H. pylori infection increased when the lesions progressed from chronic atrophic gastritis to intestinal metaplasia, gastric adenoma, dysplasia, and GC (Fig. 3B). However, among patients who were negative for H. pylori infection, RUNX3 methylation was not observed in chronic atrophic gastritis, intestinal metaplasia, gastric adenoma, or mild dysplasia but increased when the lesions progressed from moderate dysplasia to severe dysplasia and from stage I to stage IV in patients with GC (Fig. 3A).
Circulating Hypermethylated RUNX3 as a Marker of Early Diagnosis and Prognosis for Gastric Cancer
Real-time MSP was performed on serum samples from 120 normal individuals without any gastric lesions or H. pylori infection, from 652 patients with benign lesions, and from 202 patients with GC. Regardless of the frequency of RUNX3 methylation in tissues, the frequency of RUNX3 methylation detected in serum was negative in almost all benign and normal samples, except for 2 patients who had severe dysplasia. However, circulating methylated RUNX3 was detected in almost all patients with GC who had RUNX3 methylation detected in tissue samples (Fig. 4). Correspondence analyses and Z tests revealed significant accordance (kappa statistic, 0.887; P < .001) of methylation levels in GC serum and tissue samples, with 95.5% accuracy, 94.1% sensitivity, 100% specificity, a 0% false-positive rate, and a 5.9% false-negative rate (Table 3). For the diagnostic value of RUNX3 methylation levels in serum samples to predict GC, the results indicated 94.2% accuracy, 70.8% sensitivity, 99.8% specificity, a 0.2% false-positive rate, and a 29.2% false-negative rate (Table 4). The receiver operating characteristic analysis of methylated RUNX3 detection in serum and tissues revealed significant discriminative capacity. The AUC was 0.854 (95% CI, 0.815-0.892; P < .0001) for serum and 0.773 (95% CI, 0.734-0.812; P < .0001) for tissue (Fig. 5).
Table 3. The Detection of Runt-Related Transcription Factor 3 Promoter Methylation in Gastric Cancer Tissues and Serum Samples
Kappa (P) values, correspondence analyses, and Z tests represent the accordance of runt-related transcription factor 3 (RUNX3) promoter methylation scores between serum and tissue samples from patients with gastric cancer.
Tissue methylation score
Table 4. The Diagnostic Value of Runt-Related Transcription Factor 3 Methylation in Serum DNA
Kappa (P) values, correspondence analyses, and Z tests indicate that runt-related transcription factor 3 (RUNX3) promoter methylation scores are in accordance between serum and tissue samples from patients with gastric cancer.
Gastric cancer tissue
Noncancerous gastric tissue
Kaplan-Meier survival analysis indicated that the cumulative disease-free survival rates for patients with GC who had higher circulating methylated RUNX3 scores were significantly lower for those with lower circulating methylated RUNX3 scores (P < .0001) (Fig. 6B). The same results also were observed in tissue samples (P < .0001) (Fig. 6A).
The RUNX3 gene functions as a tumor suppressor in variety of cancers, eg, hepatocellular carcinoma,21 colorectal cancer26 and esophageal cancer,23 and there is cumulative evidence indicating that the down-regulation of RUNX3 expression correlates significantly with human GC24 and precancerous gastric lesions, such as chronic atrophy gastritis, intestinal metaplasia, gastric adenoma, and dysplasia.24RUNX3 suppressed GC or gastric epithelial cell growth,24, 31 suppressed tumor growth and metastasis in an animal model,40 and was associated inversely with vascular endothelial growth factor expression and elevated microvessel formation.40 In addition, loss of RUNX3 induced epithelial cells to differentiate into intestinal type cells, generating a precancerous state of the stomach in an animal model.34 The inactivation of RUNX3 reportedly is caused by hemizygous deletion,41 hypermethylation,41 and protein mislocalization,38 and RUNX3 promoter hypermethylation has been widely investigated.
RUNX3 promoter methylation has been observed frequently in GC and precancerous gastric lesions, and it was associated negatively with RUNX3 expression.24, 41 Human GC cell line investigations indicate that RUNX3 expression is down-regulated by promoter methylation, and the restoration of RUNX3 suppresses gastric epithelial cell growth by inducing P27 and Caspase 3.33 In the current study, we demonstrated that RUNX3 promoter methylation correlates positively with precancerous gastric lesions and GC progression. The methylation proportion increased from 2.2% to 97.4% and 100% as the lesion type progressed from mild chronic atrophic gastritis, to severe intestinal metaplasia, and severe dysplasia, respectively. The same correlation also was observed during GC progression. The methylation proportion increased from 23.8% to 62.8%, 84.8%, and 100% as GC progressed from stage I to stages II, III, and IV, respectively. GC tissues had the highest methylation proportion (75.2%) compared with precancerous gastric lesions, such as chronic atrophic gastritis (15.9%), intestinal metaplasia (36.7%), gastric adenoma (41.8%), and dysplasia (54.9%). Our results support previous studies through the whole process of disease progression from chronic atrophic gastritis to intestinal metaplasia, gastric adenoma, dysplasia, and GC in a large number of patients.
H. pylori infection, the major risk factor for GC, contributes to the inactivation of RUNX3 in gastric epithelial cells in various ways, including promoter hypermethylation with mediation of nitric oxide (NO) produced by macrophages after the induction of H. pylori infection42; inflammation and oxidative stress triggered by H. pylori infection43; CagA-dependent, proteasome-mediated degradation38; and CagA inhibition of RUNX3 expression through the v-src sarcoma viral oncogene/mitogen-activated protein kinase kinase/extracellular signal-regulated kinase (Src/MEK/ERK) and p38 mitogen-activated protein kinase (MAPK) pathways in gastric epithelial cells.44 In all of these, aberrant methylation has been widely investigated. H. pylori infection-induced, aberrant DNA promoter methylation has been detected in a series of tumor suppressor genes, including cyclin-dependent kinase inhibitor 2A (CDKN2A),45 cytochrome c oxidase subunit II (Cox2),18 trefoil factor 2 (TFF2),46 cadherin 1 type 1 (CDH1),47 GATA binding proteins 4 and 5(GATA-4 and GATA-5),45 and upstream transcription factors 1 and 1 (USF1 and USF2).45 The correlation between H. pylori infection and RUNX3 promoter methylation in GC initially was reported in 2008 by Kitajima et al.48 Furthermore, there is a significant correlation between H. pylori infection and RUNX3 methylation not only in GC cells but also in precancerous gastric lesions like chronic atrophy gastritis, intestinal metaplasia, and dysplasia.24
Compared with previous studies, we report a much more significant correlation between H. pylori infection and RUNX3 promoter methylation. The methylation proportion was correlated positively with precancerous gastric lesions and GC in patients who were positive for H. pylori infection. However, among patients who were negative for H. pylori infection, no methylation was detected in any precancerous gastric lesions with only rare exceptions of moderate and severe dysplasia, which may be considered early stage GC. Our multinomial logistic regression analysis results indicated that the risk of RUNX3 methylation was higher for H. pylori infection (OR, 68.71) than for precancerous lesions (OR: intestinal metaplasia, 19.66; gastric adenoma, 25.75; dysplasia, 53.39). Thus, our current results indicate that the cumulation of RUNX3 promoter methylation may be induced by H. pylori infection in a time-dependent manner. This also may explain a possible molecular mechanism by which H. pylori infection triggers GC. However, among our patients who had H. pylori-negative GC lesions, the proportion of RUNX3 methylation had a positive correlation with disease stage (partially methylated: stage I, 6%; stage II, 10%; stage III,: 21%; and stage IV, 0%; methylated: stage I, 0%; stage II, 14%; stage III, 63%; and stage IV, 100%), and the risk of RUNX3 methylation induced by GC(OR, 498.22) was much higher than that by the H. pylori infection (OR, 68.71). Thus, as GC progressed, RUNX3 methylation levels cumulated in both H. pylori-negative and H. pylori-positive patients, indicating that RUNX3 promoter methylation was induced not only by H. pylori infection but also by other, much stronger risk factors in GC lesions. Thus, we conclude that RUNX3 promoter methylation may interact with GC in a positive-feedback manner; however, more investigation among cell lines and/or animal models will be required to reveal the precise molecular mechanism.
It has been reported that real-time MSP-based quantification of serum RUNX3 methylation has diagnostic value for colorectal cancer.49 Sakakura et al reported that 19 of 65 serum samples had methylated RUNX3 sequences (sensitivity, 95.5%; specificity, 62.5%; AUC, 0.8651); and the RUNX3 methylation index was correlated with cancer stage, histology, lymphatic vessel invasion, and venous invasion.50 Tissue factor pathway inhibitor 2 (TFPI2) methylation (7 of 73 samples; 10%) in serum from patients with GC was correlated significantly with lymph node metastasis.51 In the current study, the level of RUNX3 methylation in serum, possibly caused by circulating nucleic acid released by GC cells, was correlated significantly with methylation in GC tissues (kappa statistic, 0.887). Among all GC tissues, including unmethylated samples, there was still a positive proportion of 70%; however, no RUNX3 methylation was detected in serum samples from patients who had benign lesions except for 2 patients who had severe dysplasia, which was considered early GC. Our analysis of the ability of serum RUNX3 methylation levels to predict GC revealed 94.2% accuracy, 70.8% sensitivity, 99.8% specificity, a 0.2% false-positive rate, and a 29.2% false-negative rate. The specificity of the current study but the sensitivity was lower compared with previous studies. Receiver operating characteristic curve analysis (AUC, 0.854) revealed a high diagnostic value of detecting RUNX3 methylation in serum samples. Kaplan-Meier survival analysis indicated that cumulative disease-free survival rates had a negative correlation with RUNX3 methylation scores in both serum and tissue samples. Thus, we conclude that the quantification of serum RUNX3 methylation has great potential value for detecting and diagnosing GC and even in the postoperative evaluation of patients with GC.
This research was supported by a grant from the Science and Technology General Project of Zhejiang Province (no. 2009C33143).