For the Chinese TMA cohort, 285 glioma patients of different WHO grades were recruited between January 2000 and December 2010 from the Department of Neurosurgery at Changzheng Hospital, Second Military Medical University (SMMU) (Shanghai, China). Sixteen normal brain tissue samples were taken from trauma outpatients during decompression treatment. Clinical information was obtained from medical records of the patients. Tumor histology was confirmed by two neuropathologists independently. All participants provided written informed consent. The present study was approved by the Ethics Review Committee of the Second Military Medical University and Fudan University.

The tissue microarray construction and immunohistochemistry analyses were performed by Shanghai Biochip Co., Ltd. (Shanghai, China), as described previously (21,22). The immunohistochemical staining was read and scored by two independent pathologists in a blinded fashion. The total immunoreactivity score was the product of the staining scores for density and intensity (23). High CapG expression, indicating strongly positive cases (+), had a score of >3, while low expression indicated weakly positive cases (−) with a score of ≤3.

Total RNA was extracted from surgically resected tissues using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) following the manufacturer's protocol. Reverse transcription of total RNA was performed using ReverTra Ace (Toyobo, Tokyo, Japan). The PCR primers are avaiable upon request. Quantitative real-time PCR was run using the ABI Prism 7900HT instruments (Applied Biosystems, Foster City, CA, USA) in the presence of Thunderbird SYBR qPCR Mix (Toyobo). All samples were tested in triplicate and repeated 3 times and melting curve analysis was used to check the specificity of amplification. The 2^−ΔΔCt method (24) was used to calculate relative gene expression.

The Cancer Genome Atlas (TCGA) (13) cohort (N=528; the Affymetrix platform), an independent dataset of GBM, was included in the present study. For the TCGA cohort, multidimensional data of clinical information, molecular subtypes, copy number alterations, common mutations, methylation levels and gene expression were obtained from the TCGA data portal (http://tcga-data.nci.nih.gov/tcga/) and cBioPortal (http://www.cbioportal.org/public-portal/), to analyze the expression profile and prognostic significance of the CapG gene, and their correlations with several molecular alterations. In the analyses of the TCGA cohort, CapG gene levels were dichotomized to low and high according to the ROC curve.

U87 and U251 human glioma cell lines were purchased from the Japanese Cancer Research Resources Bank. The glioma cells were cultured in Dulbecco's modified Eagle's medium (DMEM) (Life Technologies, Foster City, CA, USA) supplemented with 10% fetal bovine serum (FBS) and penicillin (100 U/ml)/streptomycin (100 U/ml). To generate cell lines stably overexpressing CapG, we cloned CapG from cDNA of the U251 cell line by PCR and cloned it into the CD513B plasmid (System Biosciences, Mountain View, CA, USA) at the EcoRI and BamHI restriction sites. To stably interfere with CapG expression, we selected sequences targeting CapG into the HpaI-XhoI sites plasmid LentiLox 3.7 (pLL3.7) (data available upon request). However, the shLacZ (25) and CD513B empty vector (EV) were used as controls for pLL3.7-shCapG and CD513B-CapG, respectively. Lentiviral production and infection were performed as previously described (26).

The tissue and cell samples were lysed in radioimmunoprecipitation assay (RIPA) buffer containing protease inhibitor cocktail (Sigma, St. Louis, MO, USA). The lysates were separated on 10 or 12% SDS-PAGE gel and electrotransferred to polyvinylidene difluoride membranes (Millipore, Billerica, MA, USA). After blocking with 5% non-fat milk for 1 h at room temperature and incubating with primary antibodies. CapG (Abcam, Cambridge, MA, USA), CDK2, CDK6, Myc, cyclin A1, cyclin D1, cyclin E, CDK4, E2F1, p15, RB and phospho-RB (ProteinTech, Group, Inc., Chicago, IL, USA), the membranes were incubated with the secondary antibody (Epitomics, Burlingame, CA, USA). Anti-GAPDH antibody (Vazyme, Nanjing, China) was used as a loading control.

Following overexpression or knockdown of CapG, the glioma cells were seeded in 96-well plates (2,000 cells/well) in sextuple, and the cell proliferation assay was performed using Cell Counting Kit-8 (CCK-8; Dojindo Laboratories, Kumamoto, Japan) according to the manufacturer's instructions. Cell cycle analysis was performed as described previously (27). Glioma cells were washed with 1Χ PBS 3 times and fixed with 75% ice-cold ethanol. Fixed cells were incubated in 1Χ PBS with propidium iodide (PI; 25 µg/ml) and 10 mg/ml RNase for 30 min in the dark before being analyzed by flow cytometry. All assays were performed in 3 independent experiments.

Cells were cultured to monolayers in 6-well plates and a reference line was drawn at the back of each well. Three scratches were made from one end to the other end of the well with a sterile p200 pipette tip. The width of the wound distance was measured at baseline. The cells were cultured in serum-free medium for 24 and 48 h, respectively. Wound closure rate was expressed as the distance migrated after 24 or 48 h relative to the basic width at 4 h. Migration experiments were performed in triplicates and repeated at least 3 times.

Cells were suspended in FBS-free DMEM at a concentration of 1×10⁵ (U87 cells) and 2.5×10⁵ (U251 cells) for Transwell invasion assays using BD BioCoat Matrigel invasion chambers (BD Biosciences, Franklin Lakes, NJ, USA) according to the manufacturer's instructions. The suspended cells (500 µl) were pipetted into the upper chamber, while the lower chamber was filled with 750 µl DMEM with EGF (15 ng/ml). After 24 h of incubation at 37°C, the filter side of the upper chamber was scrubbed with cotton swabs to remove non-invading cells and Matrigel matrix. The migrated cells were fixed in methanol and stained with crystal violet. Then, the membranes were cut from the chamber and placed on microscope slides for observation. Each invasion condition was performed in triplicates and repeated at least 3 times.

All experiments were conducted in triplicate and are expressed as the means and standard deviation by two-tailed Student's t-test. Analysis of patient survival outcome was conducted by Kaplan-Meier and Cox regression using GraphPad Prism 5 (GraphPad Software Inc., La Jolla, CA, USA) and SPSS statistics 17.0, respectively (SPSS, Inc., Chicago, IL, USA) (*P<0.05, **P<0.01, ***P<0.001 and ‘ns’, not significant).

In the present study, a cohort of 285 glioma specimens and 16 normal controls from a Chinese population was enrolled in the TMA. It included 12 grade I (pilocytic astrocytomas), 108 grade II (diffuse astrocytomas), 46 grade III (anaplastic astrocytomas) and 119 grade IV (glioblastomas, GBM) specimens according to the WHO grading schedule. The clinicopathological parameters of the TMA are presented in Table I. The patient median overall survival (OS) was 30 months for all glioma cases, with a 5-year survival rate of 4.3%. Next, we examined the protein level of CapG by immunohistochemical assay performed on the TMA and observed its cytoplasmic expression in most samples. Furthermore, the immunoreactivity of CapG was significantly increased with WHO grade in the human glioma tissues when compared with normal controls: high grade gliomas (HGG; grade III/grade IV) showed stronger immunoreactivity than the low grade gliomas (LGG, grade I/II) (Fig. 1A and B).

We then measured the transcriptional level of CapG in a subset of 36 human gliomas (16 LGG and 20 HGG specimens) and 16 normal controls via quantitative RT-PCR assay. These samples were randomly selected from the cohort. Consistent with the TMA data, CapG was considerably increased in human glioma specimens according to WHO grade (Fig. 1C). These results suggest that CapG is frequently elevated in human glioma tissues.

To further investigate the correlation between CapG expression level and clinical prognosis, we performed Kaplan-Meier analyses and found that glioma patients with higher CapG expression had significantly shorter overall survival (OS, log-rank P=1.142E-06, HR=2.171, 95% CI=1.572–2.999, Fig. 1D) and progression-free survival (PFS, log-rank, P=1.963E-05, HR=1.913, 95% CI=1.408–2.600, Fig. 1E) than those with low expression. Moreover, we observed that HGG patients with higher CapG expression also had an unfavorable OS (log-rank P=0.018, HR=1.501, 95% CI=1.067–2.225, Fig. 1F) and PFS (log-rank P=0.027, HR=1.476, 95% CI=1.034–2.107, Fig. 1G) than those with low CapG expression.

Next, our univariate and multivariate Cox regression analyses revealed that high CapG expression was an independent risk predictor of both OS (P=0.003, HR=1.674, 95% CI=1.198–2.339) and PFS (P=0.008, HR=0.008, 95% CI=1.123–2.12) in our cohort (Table II). In addition, we further identified CapG as an independent prognostic factor for OS (P=0.021, HR=1.541, 95% CI=1.067–2.225) and PFS (P=0.032, HR=1.476, 95% CI=1.034–2.107) of HGG patients; higher levels of CapG predicted poorer survival (Table II).

We next investigated the expression profile of the CapG gene in an independent cohort from TCGA (13). Consequently, mRNA expression of CapG increased in 97.7% (516/528) of the GBM specimens compared to the normal controls (Fig. 2A). Since TCGA network has introduced a novel molecular classification of GBM according to gene expression, which are proneural, neural, classical and mesenchymal (28), we then assessed CapG expression in each of these subtypes. As shown in Fig. 2B, we observed that CapG expression was drastically decreased in the proneural subtype in comparison with the others. Nevertheless, CapG expression was still elevated in GBM samples.

To further explore the expression profile of the CapG gene, we scanned associations between its expression and these common genetic alterations in GBM (13,29) including amplifications of CDK4, CDK6, EGFR, AKT3, PDGFRA, MDM2, MET, MDM4, MYCN, CCND2, PIK3CA, mutations in PTEN, TP53, IDH1, EGFR, NF1, RB1, PIK3R1, ATRX, PIK3CA, CTNND2, SPTA1, GABRA6, LZTR1, and deletions of PTEN, CDKN2A, CDKN2B, CDKN2C, NF1, PARK2, RB1. As a result, we found that CapG expression exhibited a correlation with the amplification of CDK6, EGFR, MET, SYK, ANXA1, MYCN, PDGFRA, and mutations of PTEN, IDH1, RB1.

Since we ascertained that CapG expression is correlated with the GBM subtype, we then explored whether the differential CapG expression with the above-mentioned common genetic alterations was dependent on the specific subtype. Consequently, cases with CDK6 amplification or EGFR amplification or MET amplification had higher levels of CapG expression than those with no copy number alterations in the proneural subtype, while there were no dramatic changes in non-proneural subtypes (Fig. 2C-E). Moreover, cases with SYK amplification had lower CapG expression level than those with no copy number alterations in the proneural subtype, and there was also no obvious change in the non-proneural subtype (Fig. 2F). In contrast, cases with PTEN mutation or RB1 mutation showed a higher level of CapG expression than cases with wild-type only in the non-proneural subtype (Fig. 2G and H). However, cases with ANXA1 amplification, MYCN amplification, PDGFRA amplification, or IDH1 mutation exhibited no significant CapG expression change in the proneural or non-proneural subtypes when compared with those without these genetic alterations (data available on request).

To further explore the relationship between CapG expression and GBM patient clinical outcome in the TCGA cohort, we compared the OS and PFS of all GBM patients with differential CapG expression and observed statistically significant disadvantage in prognosis for GBM patients with higher CapG expression (OS, log-rank P=2.018E-04, HR=1.498, 95% CI=1.209–1.857, Fig. 3A; PFS, log-rank P=0.005, HR=1.398, 95% CI=1.107–1.767, Fig. 3E). Multivariate Cox regression further validated the CapG expression as an independent predictor for the GBM patient OS (P=0.002, HR=1.546, 95% CI=1.170–2.042) and PFS (P=0.029, HR=1.400, 95% CI=1.035–1.893) in the TCGA cohort (Table III).

Moreover, we next performed survival analysis for each molecular subtype and detected that only the proneural (OS, log-rank P=0.001; HR=1.961, 95% CI=1.313–2.928, Fig. 3B; PFS, log-rank P=0.044; HR=1.570, 95% CI=1.012–2.435, Fig. 3F) subtype retained statistical significance. In addition, we carried out survival analyses stratified by the status of these common genetic alterations in GBM and found the effect of CapG on GBM patient prognosis was associated with EGFR amplification or CDK6 amplification. It is noteworthy that the prognostic value of CapG was highly significant in individuals with wild-type CDK6 (OS, log-rank P=0.001, HR=2.261, 95% CI=1.409–3.628, Fig. 3C; PFS, log-rank P=0.017, HR=1.966, 95% CI=1.119–3.111, Fig. 3G) or EGFR (OS, log-rank P=0.011, HR=2.154, 95% CI=1.191–3.897, Fig. 3D; PFS, log-rank P=0.033, HR=2.113, 95% CI=1.064–4.197, Fig. 3H), but not significant in individuals with amplification of CDK6 or EGFR. Furthermore, patients with low CapG expression and wild-type EGFR or CDK6 had the most favorable prognosis, whereas patients with high CapG expression and amplification of EGFR or CDK6 had the poorest prognosis (Fig. 3C, D, G and H).

Given that CapG is significantly upregulated in human glioma, we inferred that CapG may promote glioma cell proliferation. To confirm our hypothesis, we performed lentiviral infection to stably overexpress or knockdown CapG in U87 and U251 glioma cells. The efficiency of ectopic expression of CapG was verified by real-time PCR (data not shown) and western blotting (Fig. 4A). Thus shCAPG-1 and shCAPG-3 were chosen for the following study according to their silencing efficiency.

We then studied the effect of CapG on cellular proliferation in U87 and U251 cells using CCK-8 assay. The results showed that overexpression of CapG significantly promoted the proliferation in comparison with the control groups, whereas knockdown of endogenous CapG markedly inhibited cellular growth in both cell lines (Fig. 4B). Colony formation assay consistently showed that overexpression of CapG in U251 cells markedly increased the number of colonies, whereas knockdown of CapG significantly decreased the colony numbers compared with the controls (Fig. 4C).

To elucidate the mechanism by which depletion of CapG inhibits cell proliferation in U87 and U251 cells, we performed flow cytometry to analyze whether the inhibition is correlated with specific cell cycle changes. Compared with the shLacZ groups, U87 and U251 cells with CapG depletion showed marked increases in the percentage of cell in the G0/G1 phase and substantial decrease in the S phase (Fig. 5A). To further investigate the molecules involved in the changes in the cell cycle, we determined the expression of several important regulators of the G1 phase by western blot assays (30–32). As shown in Fig. 5B, we observed that depletion of CapG significantly suppressed CDK2, Myc, cyclin A1 and D1, and phospho-RB expression in the U87 and U251 cells. However, the expression levels of cyclin E, CDK4, CDK6, E2F1, p15 and RB had no significant changes after CapG depletion. Collectively, these results suggest that knockdown of endogenous CapG induces G0/G1 cell cycle arrest in glioma cells.

Recent research has shown that suppression of CapG expression influences the metastasis of prostate cancer cells (20,29). We then examined the impact of CapG on glioma cell migration and invasiveness in vitro. In wound healing assays, CapG overexpression increased cell migration markedly within 24 or 48 h compared to the control cells, while silencing of CapG resulted in slower wound healing (Fig. 6A-D). However, the effect on cell invasiveness was next assessed by Transwell invasion assays. Consistent with the abovementioned migration analyses, CapG upregulation substantially increased the invasive ability of the glioma cells while CapG depletion resulted in a significant decrease in invasiveness of the cells (Fig. 6E and F). Therefore, glioma cell migration and invasiveness were positively correlated with CapG expression levels.