SIRT1 Is Downregulated in Gastric Cancer and Leads to G₁-phase Arrest via NF-κB/Cyclin D1 Signaling

On the basis of incidence and mortality rate, gastric cancer is among the top five leading causes of cancer worldwide. In 2008, there were approximately 12.7 million cancer cases worldwide, and gastric cancer accounted for 8% of all cancers (1). The majority of new gastric cancer cases occur in Eastern Asia, Eastern Europe, and South America. Although gastric cancer rates have decreased globally, it remains high in developing countries, especially China (1, 2). Most of the patients with gastric cancer are not diagnosed until they have reached advanced stages of the disease when, despite the use of conventional therapies such as surgery, chemotherapy, and radiotherapy, patients have a poor prognosis. However, early-stage gastric cancer is a potentially curable disease with a high 5-year overall survival rate (3). Therefore, studies about new diagnostic or prognostic markers and therapeutic targets for gastric cancer are urgently needed.

Sirtuin 1 (SIRT1), a class III NAD⁺-dependent histone deacetylase, is the mammalian homolog of yeast silent information regulator 2. SIRT1 substrates include not only histones but also nonhistone proteins. Through these targets, SIRT1 plays a crucial role in multiple pathways such as cellular metabolism, stress response, calorie restriction, and aging (4, 5). An abundance of recent data has demonstrated that activation of SIRT1 has physiologic benefits, including increased health and longevity, and assists in certain metabolic disorders (4–6). However, the function of SIRT1 in cancer is controversial. SIRT1 was initially regarded as a potential oncogene because of its first nonhistone substrate, p53, a well-known tumor suppressor. SIRT1 deacetylates p53 with specificity for the C-terminal Lys382 residue and silences p53 activity as a transcription factor (7, 8). In addition, two tumor suppressors, hypermethylated in cancer (HIC1) and deleted in breast cancer (DBC1), have been identified as negative regulators of SIRT1 (9, 10). A recently published study has provided new evidence for the role of SIRT1 as a tumor promoter by describing a positive feedback loop between C-Myc and SIRT1 in hepatocellular carcinomas (11). However, studies from other laboratories have indicated that SIRT1 acts as a tumor suppressor. First, acetylation of H3K56, a substrate for SIRT1, is increased in multiple cancers (12). Second, SIRT1 has been reported to play a key role in repairing DNA-breaks and maintaining genome stability (13, 14). Moreover, SIRT1 has been shown to suppress the growth of colon cancer by inhibiting E2F1 or β-catenin (15, 16). The role of SIRT1 in oncogenic progression seems to be specific for tumor type and signaling pathway involved.

In this study, we investigated the expression and function of SIRT1 in gastric cancer. We compared the expression of SIRT1 between normal and malignant gastric tissues. Then we analyzed cell viability, clonogenic potential, cell-cycle distribution, and apoptosis and investigated the mechanism for how SIRT1 suppresses tumorigenesis. The inhibitory activity of SIRT1 on gastric cancer development was also evaluated in vivo.

Forty-four patients with primary gastric cancer were included in this study, which was approved by the local ethics committee (No. 2011008 for Ethics Approval). The patients underwent gastrectomy at Jinan Central Hospital between 2011 and 2012. The tissue specimens were collected and stored as previously described (17). There were 11 female patients and 33 male patients, with a median age of 60 (range, 35–82). Details of patient and disease characteristics are documented in Table 1.

Human gastric cancer cell lines AGS, MGC-803 (Cell Resource Center, Shanghai Institute of Biochemistry and Cell Biology at the Chinese Academy of Sciences, Shanghai, China), BGC-823, and SGC-7901 (China Center for Type Culture Collection, Wuhan, China) were cultured in F12 (AGS) or RPMI 1640 (BGC-823, SGC-7901, and MGC-803) containing 10% FBS, 100 U/mL penicillin and 2 mmol/L l-glutamine at 37°C in a humidified atmosphere containing 5% CO₂. Both of the cell banks routinely perform cell line authentication by short tandem repeat (STR) profiling and all of the cell lines were passaged in our laboratory for no more than 6 months after receipt. High-purity cycloheximide (CHX; 3-[2-(3,5-dimethyl-2-oxocyclohexyl)2-hydroxyethyl-]-glutarimide) was purchased from Beyotime, dissolved in dimethyl sulfoxide (DMSO), and added into the culture medium at a concentration of 20 μg/mL. The vectors expressing wild-type SIRT1 (pECE-SIRT1) and the catalytic-domain mutant form of SIRT1 (pECE-SIRT1-H363Y) were purchased from Addgene and recombined into pcDNA3.1 (Invitrogen). Chemically modified siRNA targeting SIRT1, NF-κB, cyclin D1, and control siRNA were purchased from GenePharma. The sequences of the earlier siRNA are listed in Supplementary Table S1.

Total RNA in tissue specimens or cells was isolated using TRIzol reagent (Invitrogen) and converted into cDNA using the PrimeScript RT reagent kit (Takara). qRT-PCR was performed for genes, including SIRT1, cyclin D1, and β-actin as previously described (18). The sequences of the amplification primers are listed in Supplementary Table S1. The levels of SIRT1 mRNA in clinical specimens were calculated on the basis of the threshold cycle (C_t) values and normalization of β-actin expression using the |$2^{- {\rm \Delta \Delta}C_{\rm t}}$| method (19). The mRNA expression of cyclin D1 was normalized to β-actin relative to the control using the |$2^{ -{\rm \Delta \Delta}C_{\rm t}}$| method. Experiments were performed in triplicate and repeated three times.

Total protein from the tumor specimens or cells was extracted using RIPA lysis buffer (Beyotime) as previously described (18). The protein concentration was determined using a BCA Protein Assay Kit (Pierce). The membrane was probed with antibodies against SIRT1 and β-catenin (Abcam), cyclin D1, cyclin D2, cyclin D3, cyclin E1, cyclin-dependent kinase (CDK)4, bcl-2, bax, caspase-3, NF-κB p65, and c-Jun (Cell Signaling Technology), cyclin D2, p21, and p16 (Santa Cruz Biotechnology). Horseradish peroxidase (HRP)–conjugated anti-rabbit or -mouse antibody was used as the secondary antibody. The protein bands were visualized using an ECL system (Pierce). β-actin (Cell Signaling Technology) served as a loading control.

The tumor specimens from patients with gastric cancer and the nude mice were deparaffinized, rehydrated, and antigen retrieved. The antibodies against SIRT1 (1:200; Santa Cruz Biotechnology) and Ki67 (1:500; Abcam) were added to the sections and then incubated overnight at 4°C. HRP-conjugated anti-rabbit immunoglobulin G (IgG) and 3,3′-diaminobenzidine (DAB) staining were used to visualize the primary antibody. Then, the slides were counterstained with hematoxylin and imaged by an Olympus light microscope using cellSens Dimension software. The percentages of Ki67-positive cells were counted under microscope (magnification, ×400) and five fields were counted for each section.

The CellTiter96 AQueous One Solution Cell Proliferation Assay (Promega) was performed to indicate cell proliferation. Briefly, 2 × 10³ cells were seeded in a 96-well plate and were allowed to grow for 24 hours. Twenty-four hours later, 20 μL MTS was added to each well. After incubation for 3 hours at 37°C, the absorbance at 490 nm was recorded on a Varioskan Flash Multiplate Reader (Thermo Scientific). Percentage of vehicle was calculated by the following formula: [(average absorbance of treated group − average absorbance of blank)/(average absorbance of control group − average absorbance of blank)] × 100%. Assays were performed in triplicate and repeated three times.

Cells were seeded into 6-well plates (300 cells per well) and incubated for 10 days until the colonies were large enough to be clearly discerned. The cells were fixed with methanol and stained with crystal violet and the number of colonies with more than 50 cells was counted manually. Experiments was performed in triplicate and repeated three times.

Cells were harvested, fixed with precooled 70% ethanol at 4°C overnight, and then stained with propidium iodide (PI) containing RNase A at 37°C for 30 minutes in the dark. Cell-cycle distribution was determined using a flow cytometer (BD Biosciences). Each experiment was carried out in triplicate and the data were analyzed with MultiCycle software (Phoenix Flow Systems). Experiments were performed independently three times.

Detection and quantitation of apoptosis were performed by labeling of DNA strand breaks using an In Situ Cell Death Detection Kit, TMR red (Roche Applied Science). After transfection, cells were seeded onto coverslips and labeled with terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) according to the manufacturer's instructions. Treatment with 0.2 mmol/L H₂O₂ for 12 hours served as the positive control. The nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; Beyotime) and the slides were imaged by a fluorescence microscopy (Olympus) using cellSens Dimension software. Experiments were performed independently three times.

Lentiviral vectors containing SIRT1 cDNA, SIRT1 short hairpin (sh)RNAs, or their controls were constructed by GenePharma and used to transfect BGC-823 cells. For shRNAs, the same targeting sequences were used as the siRNAs. For stable infection, BGC-823 cells infected with different lentiviral vectors were cultured in complete medium supplied with 2 μg/mL puromycin (Acros Organics) for 4 weeks.

Female athymic BALB/c nude mice (6–8 weeks) were purchased from Peking University (Beijing, China) and maintained under specific pathogen-free conditions at the Key Laboratory of Cardiovascular Remodeling and Function Research, Qilu Hospital, Shandong University. The study was approved by the local ethics committee (No. 001 in 2011 for Animal Ethics Approval). The mice were randomly divided into six groups and BGC-823 cells infected with or without different lentiviral vectors (1 × 10⁶ cells in 0.1 mL PBS) were subcutaneously injected into the flank region of each group. In detail, group I, BGC-823 cells without infection were injected (regarded as control; n = 8); group II, BGC-823 cells infected with the control-lentiviral expression vector were injected (regarded as LV-C; n = 8); group III, BGC-823 cells infected with the SIRT1-lentiviral expression vector were injected (regarded as LV-S; n = 8); group IV, BGC-823 cells infected with the control-lentiviral shRNA were injected (regarded as LV-Ci; n = 14), group V and VI, BGC-823 cells infected with one of the two different SIRT1-lentiviral shRNAs were injected respectively (regarded as LV-sh-1 and LV-sh-2; 8 mice for each group). The subcutaneous tumor size was measured with a caliper and the tumor volume was calculated by the formula (length) × (width²)/2. Mice were sacrificed 4 weeks after implantation. Tumors were harvested and processed for Western blot analysis and immunohistochemistry studies.

The difference in SIRT1 mRNA levels between cancerous and matched normal gastric tissues was analyzed using Student t test. The comparison of cell proliferation, foci number, and cell-cycle distribution was performed using one-way ANOVA with a Tukey post-hoc test or a Student t test. The data were expressed as mean ± SD. The statistical analyses were performed using Statistical Package for the Social Sciences (SPSS, version 16.0). P values less than 0.05 were considered statistically significant.

To explore the role of SIRT1 in gastric cancer, we first determined the protein levels of SIRT1 using immunohistochemistry in gastric tissues from patients with gastric cancer. The cancerous and matched normal gastric specimens from 44 patients were analyzed. The SIRT1 protein was detected in all the normal gastric tissues with positive cell fractions ranging from 78% to 100% (Fig. 1A and Table 1). The SIRT1 protein was present in both cytoplasmic and nuclear compartments of the epithelial cells, and most of the positive cells exhibited intermediate to strong signals (Fig. 1A and Table 1). However, the expression of SIRT1 was detected in 59% (24 of 44) of the cancerous samples. The SIRT1-positive gastric cancer samples exhibited weak to intermediate staining with similar distribution patterns (Fig. 1B and C and Table 1).

We further analyzed the SIRT1 mRNA levels in 20 available gastric cancer and the corresponding normal tissue samples by qRT-PCR, including two specimens that lacked SIRT1 protein detection in the cancerous region. Consistent with the protein profile, the presence of SIRT1 transcripts was not observed in these two cancer specimens. Among the other 18 pairs of samples, the cancerous samples had a significantly lower level of SIRT1 mRNA expression compared with the matched normal tissues (Fig. 1D and Table 1). These data indicate that SIRT1 is downregulated in gastric cancer, which inspired us to further investigate the details of SIRT1 in gastric cancer progression.

To determine the function of SIRT1 in gastric cancer cells, we separately transfected the mock vector (pcDNA), the vectors expressing wild-type SIRT1 (pcDNA-SIRT1), and catalytic-domain mutant form of SIRT1 (pcDNA-SIRT1-H363Y) into the gastric cancer cell lines AGS, BGC-823, and SGC-7901. Furthermore, we knocked down SIRT1 expression using SIRT1-specific siRNAs in the gastric cancer cell lines AGS, BGC-823, SGC-7901, and MGC-803. The efficient transfection was verified by Western blot analysis (Fig. 2A and Supplementary Fig. S1). The MTS assay showed that overexpression of SIRT1 significantly inhibited cell proliferation, which was absent when the catalytic-domain mutant form of SIRT1 was transfected (Fig. 2B). Moreover, there was a clear increase in cell proliferation in SIRT1-depleted cells (Fig. 2B). Thus, SIRT1 exerts an inhibitory activity on cell proliferation in gastric cancer dependent on its deacetylase activity.

Next, we determined the effect of SIRT1 on colony formation of gastric cancer cells. Upregulation of SIRT1 caused a significant reduction in foci number as well as sizes in gastric cancer cells. This inhibitory effect of SIRT1 was dependent on its deacetylase activity because transfection of the mutant SIRT1 did not have an observable effect on colony formation (Fig. 2C and E). However, knockdown of SIRT1 significantly increased the foci number compared with the corresponding control group (Fig. 2D and E). These results indicate that SIRT1 inhibits colony formation of gastric cancer cells.

To examine the inhibitory effects of SIRT1 on gastric cancer cells, we performed cell-cycle analysis. We observed that overexpression of SIRT1 increased the proportion of cells in G₁ phase. This increase of cell fractions in G₁ phase was accompanied with a concomitant decrease of cell populations in S and G₂–M phases (Fig. 3A). The induction of G₁ phase arrest by SIRT1 was also dependent on its deacetylase activity because transfection of the mutant SIRT1 did not have an effect on the cell-cycle distribution (Fig. 3A). Further evidence was obtained from the RNA interference experiments. Depletion of SIRT1 significantly decreased the fractions of cells in G₁ phase, whereas the number of cells in S and G₂–M phases increased obviously (Fig. 3A). To validate these results, we examined the expression of cell-cycle regulators operative in G₁ phase, including cyclin D1, cyclin D2, cyclin D3, cyclin E1, CDK4, p21, and p16. As shown in Fig. 3B, when wild-type SIRT1 was upregulated, expression of cyclin D1 was obviously inhibited, which was absent when the mutant SIRT1 was transfected. In contrast, when SIRT1 was efficiently decreased by a specific siRNA, expression of cyclin D1 was clearly increased (Fig. 3B and Supplementary Fig. S1). However, no significant changes were observed in the expression of other regulators, including the activators of G₁–S transition (cyclin D3, cyclin E1, and CDK4) and the inhibitors of CDKs (CDKI; p21 and p16). However, we failed to detect expression of cyclin D2 in AGS and BGC-823 cells. These data indicate that SIRT1 induces G₁ phase arrest in gastric cancer cells by inhibiting cyclin D1.

Previous studies have shown that deacetylation by SIRTs can decrease the stability of proteins by ubiquitination and subsequent proteasomal degradation (20, 21). Here, we detected the stability of cyclin D1 using CHX to inhibit protein synthesis. Overexpression of SIRT1 in BGC-823 cells did not affect the stability of cyclin D1 protein (Fig. 4A). Furthermore, depletion of SIRT1 did not exert significant effects on the stability of cyclin D1 protein either (data not shown). The qRT-PCR results indicated that expression of cyclin D1 mRNA was negatively regulated by SIRT1 (Fig. 4B), which suggests that SIRT1 transcriptionally inhibits the cyclin D1 gene. Among the three common transcription factors that regulate the mRNA levels of cyclin D1, NF-κB, β-catenin, and AP-1, expression of NF-κB p65 was inhibited by SIRT1, whereas expression of the other two transcription factors did not change with SIRT1 (Fig. 4C). These data suggest that SIRT1 may inhibit the transcription of cyclin D1 through downregulation of NF-κB.

Next, RNA interference experiments were performed. Inhibition of NF-κB in SIRT1-depleted cells rescued cyclin D1 mRNA to a level comparable with the controls (Fig. 4D and E). Moreover, The changes in cell proliferation, colony formation, and cell-cycle distribution were also reversed by downregulation of NF-κB or cyclin D1 in SIRT1-depleted cells (Fig. 4F–H; the percentages of cells in G₁ phase: SIRT1 + control siRNA vs. SIRT1 + NF-κB siRNA, 47.5 ± 0.89 vs. 54.53 ± 2.75, P = 0.007; SIRT1 + control siRNA vs. SIRT1 + cyclin D1 siRNA, 47.5 ± 0.89 vs. 59.33 ± 1.1, P < 0.001; SIRT1 + NF-κB siRNA vs. control siRNA, 54.53 ± 2.75 vs. 56.3 ± 2.1, P = 0.683; SIRT1 + cyclin D1 siRNA vs. control siRNA, 59.33 ± 1.1 vs. 56.3 ± 2.1, P = 0.405). Taken together, our results indicate that SIRT1 induces G₁ phase arrest in gastric cancer cells via NF-κB/cyclin D1 signaling.

Lack of sub-G₁ and results of the TUNEL labeling experiments indicated that SIRT1 did not trigger apoptosis in gastric cancer cells (Fig. 4H and Supplementary Fig. S2). Moreover, the protein levels of apoptosis-related molecules (bcl-2 and bax) from each group were comparable and no cleaved caspase-3 was detected in the gastric cancer cells (Supplementary Fig. S2).

Next, the effects of SIRT1 on gastric cancer development were verified using a nude mouse xenograft model. For in vivo study, stable lentivirus–infected BGC-823 cells were used. Efficient infection was verified using a fluorescence microscope (data not shown). All of the nude mice survived to the end of our experiments and there were no apparent differences in their body weights (data not shown). Four weeks after implantation, there were no observable differences in the tumor volumes among the three groups of control, LV-C, and LV-Ci. The tumor volumes for the earlier three groups were 1.4297 ± 0.1821, 1.369 ± 0.1443, and 1.2898 ± 0.1947 cm³, respectively (P > 0.05; Fig. 5A and B). Overexpression of SIRT1 significantly inhibited tumor development (LV-C vs. LV-S: 1.369 ± 0.1443 vs. 0.5988 ± 0.1123 cm³; P = 0.001; Fig. 5A and B). Interestingly, tumor had not developed in one of the nude mice in the LV-S group at the end of the xenograft experiments. However, stable depletion of SIRT1 obviously promoted tumor progression (LV-Ci vs. LV-sh-1: 1.2898 ± 0.1947 vs. 2.2344 ± 0.1935 cm³, P < 0.001; LV-Ci vs. LV-sh-2: 1.2898 ± 0.1947 vs. 2.0456 ± 0.18 cm³, P < 0.001; Fig. 5A and B). The inhibitory effects of SIRT1 on BGC-823 xenografts were then corroborated in the tumor samples by detecting the protein levels of Ki67 using immunohistochemistry. The percentages of Ki67-positive cells in groups of control, LV-C, LV-S, LV-Ci, LV-sh-1, and LV-sh-2 were 46.1 ± 1.44, 45.7 ± 0.95, 6 ± 1.13, 45.56 ± 1.56, 87.59 ± 1.93, and 82.23 ± 3.21, respectively (LV-C vs. LV-S, P < 0.001; LV-Ci vs. LV-sh-1, P < 0.001; LV-Ci vs. LV-sh-2, P < 0.001; Fig. 5B). The changes in the expression of cyclin D1 and NF-κB were similar to what we observed in vitro (Fig. 5C). Our results suggest that SIRT1 inhibits the gastric cancer development in vivo, which involves NF-κB/cyclin D1 signaling.

SIRT1 was initially concerned because its homolog in the budding yeast, silent information regulator 2, has been shown to extend the life span of the yeast (22). Subsequent studies have revealed that SIRT1 plays an essential role in metabolic pathways and is linked to health, longevity, and diseases such as cancer (5, 6). The role of SIRT1 in tumorigenesis is an area of considerable debate. First, the expression levels of SIRT1 in different types of cancers are conflicting. The expression of SIRT1 is relatively higher in hepatocellular carcinoma, breast cancer, and thyroid cancer (10, 23, 24) but lower in colon and lung cancer (15, 16, 25) compared with their corresponding normal tissues. Even in the same types of cancer, such as prostate cancer, contrary results have been reported (26, 27). Here, we detected the expression of SIRT1 in gastric cancer and found that both the mRNA and the protein levels of SIRT1 decreased compared with the matched normal tissues. This result conflicts with previous results reported by Cha and colleagues (28). Because the same antibody was used to detect the expression of SIRT1, the conflicting data may be due to the different race from which the tumor samples were obtained.

Second, the role of SIRT1 in different types of cancers has also been disputed. Results from the hepatocellular carcinoma indicate that inhibition of SIRT1 in hepatocellular carcinoma cells impairs their proliferation, increases the expression of differentiation markers in vitro, and has been shown to be beneficial in blocking tumor formation in vivo (11, 29). However, the opposite is true in colon cancer. Two different groups have reported that SIRT1 inhibited proliferation and intestinal tumorigenesis both in vitro and in vivo (15, 16). These contrary data suggest that the role of SIRT1 in oncogenic progression is cancer-type specific. In our experiments, SIRT1 exerted inhibitory effects on cell proliferation and colony formation in gastric cancer cells. This inhibitory activity was verified using a nude mouse xenograft model. Our results indicate that SIRT1 is an inhibitor of gastric cancer and that downregulation of SIRT1 promotes gastric tumorigenesis.

The regulation of cell cycle is an important antiproliferation mechanism in cancer (30, 31). Among the seven members of the mammalian sirtuin family, SIRT2, a tubulin deacetylase required for normal mitotic progression, has been frequently linked to cell-cycle transition (4). Few studies have focused on SIRT1 and cell-cycle regulation. A study by Wang and colleagues indicates that SIRT1 binds to E2F1, a cell-cycle and apoptosis regulator, and inhibits its activity (32), which suggests that SIRT1 may participate in the regulation of cell cycle. A subsequent study reveals that in addition to E2F1, SIRT1 interacts with another transcription factor, c-Myc. SIRT1 deacetylates c-Myc and activates its target genes, including cyclin D2, thus plays an essential role in cell proliferation and cell-cycle regulation (33). In our study, overexpression of SIRT1 induced G₁ phase arrest in gastric cancer cells. Detection of regulators of the G₁ phase, including the activators (cyclin D1, cyclin D2, cyclin D3, cyclin E1, and CDK4) and the inhibitors (p21 and p16), indicated that SIRT1 negatively regulated the expression of cyclin D1. Unlike FOXO3 (21), the stability of cyclin D1 protein was not affected by SIRT1. Detection of the mRNA levels of cyclin D1 revealed that SIRT1 inhibited the expression of cyclin D1 at the transcriptional level. Several transcription factors that bind the promoter of cyclin D1 gene have been proven to be inhibited by SIRT1, including NF-κB, β-catenin, and AP-1 (16, 34–37). Among these three transcription factors, only NF-κB was inhibited by SIRT1 at the protein level. Although our results do not eliminate the possibilities that other transcription factors may participate in the regulation of cyclin D1 by SIRT1, inhibition of NF-κB in SIRT1-depleted cells rescued the mRNA levels of cyclin D1. These results suggest that NF-κB is responsible for SIRT1 inhibiting the transcription of cyclin D1. Moreover, the changes in cell proliferation, colony formation, and cell-cycle distribution caused by SIRT1 depletion were reversed by downregulation of NF-κB or cyclin D1, indicating that NF-κB/cyclin D1 signaling accounts for the inhibitory effects on gastric cancer cells induced by SIRT1. The inhibitory effect of SIRT1 on NF-κB/cyclin D1 pathway was also verified by Western blot analysis using xenograft tumor tissues.

Taken together, using clinical gastric cancer specimens, gastric cancer cells, and nude mouse xenograft models, our study shows that SIRT1, a histone/protein deacetylase, exerts tumor-suppressive function in gastric cancer. Through the NF-κB/cyclin D1 pathway, SIRT1 induces G₁ phase arrest in gastric cancer and leads to the inhibition of gastric cancer both in vitro and in vivo. Results from our work demonstrate the inhibitory function of SIRT1 in gastric cancer development and suggest a potential therapeutic effect for SIRT1 activators in the prevention and therapy of gastric cancer.

No potential conflicts of interest were disclosed.

Conception and design: Q. Yang, J. Jia

Development of methodology: Q. Yang, B. Wang, W. Gao, S. Huang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Wang, W. Gao

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Yang, Z. Liu, W. Li

Writing, review, and/or revision of the manuscript: Q. Yang, J. Jia

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Li

Study supervision: B. Wang, J. Jia

This work was supported by the National Basic Research Program of China (973 Program; No. 2012CB911202), the National Natural Science Foundation of China (No. 81101869, 81100103, 81171536, 81000868, and 81371781), and the Independent Innovation Foundation of Shandong University (2012TS108).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.