ZIP4 is a novel molecular marker for glioma

Tissue Samples

We included all 555 available samples from 3 independent human glioma cohorts: Chinese Glioma Genome Atlas (CGCA) cohort (n = 186), the US National Cancer Institute Repository for Molecular Brain Neoplasia Data (REMBRANDT) cohort (n = 335), and the University of Texas (UT) cohort (n = 34). For the CGCA cohort, 186 frozen glioma tumor samples were obtained from patients with newly diagnosed glioma treated at the Glioma Center of Beijing Tiantan Hospital from 2006 through 2009. Clinicopathological information (age, sex, preoperational Karnofsky performance scale [KPS] score, and treatment) were obtained from medical records of Tiantan hospital. Tumor histology of all patients was confirmed independently by 2 neuro-pathologists on the basis of the 2007 edition of World Health Organization (WHO) classification of central nervous system tumors. All the patients in the study received conventional therapy consisting of maximal surgical resection, followed by radiotherapy and/or chemotherapy. For low-grade gliomas, the patients first underwent maximal resection and then received radiotherapy of 50–54 Gy in 1.8 Gy fractions. For high-grade gliomas, the patients received operation, followed by 60 Gy radiotherapy for 6 weeks with concurrent temozolomide chemotherapy, and then adjuvant temozolomide-based chemotherapy. Patients who received radiotherapy or chemotherapy before admission were excluded from this study. The study was approved by the Research Ethics Committee of Tiantan Hospital. Informed consent forms were obtained from all patients.

The second independent cohort is the REMBRANDT cohort. REMBRANDT is a joint program of the National Cancer Institute and the National Institute of Neurological Disorders and Stroke, which hosts clinical and genomics data from clinical trials involving patients with gliomas. Clinicopathological information of 335 patients with gliomas is available. In the third cohort, 34 surgically resected glioma tissue samples were obtained from the University of Texas Medical School at Houston (UT cohort). Clinical information of those de-identified samples was obtained through the database of the Neurosciences Research Repository at UT Medical School. This study was approved by the UT Institutional Committee for the Protection of Human Subjects.

Tumor grade is characterized as grade II (astrocytomas) and grade IV (glioblastomas). Because of limited availability of grade III glioma samples from the CGCA and UT cohorts and extreme difficulty to obtain normal brain tissue samples, we only included grade II and grade IV samples in this study. The overall survival time is defined as the interval from diagnosis to either death or the last observation taken. Data were censored at the last follow-up for living patients. Summary statistics of the CGCA and REMBRANDT cohorts are presented in Table 1.

RNA Extraction and Microarray Analysis

For the CGCA cohort, all the tissue samples were immediately snap-frozen in liquid nitrogen after surgery. A hematoxylin and eosin–stained frozen section was prepared for assessment of the percentage of tumor cells before RNA extraction. Only samples with greater than 80% tumor cells were selected. For microarray assay, total RNA from frozen tumor tissues was extracted using the Total RNA Isolation Kit (Ambion, Austin, TX) according to the manufacturer's protocol. RNA concentrations were measured using the NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Houston, TX). The cDNA and biotinylated cRNA were synthesized and hybridized to the Agilent Whole Human Genome Array (Agilent, Santa Clara, CA) according to the manufacturer's instructions. Data were acquired using Agilent G2565BA Microarray Scanner System and Agilent Feature Extraction Software (version 9.1), and probe intensities were normalized using GeneSpring GX 11.0. A similar procedure was performed in the UT cohort to collect and process the tissues samples. All the samples were immediately snap-frozen in liquid nitrogen after surgery. Total RNA from frozen tumor tissue samples was extracted using the Total RNA Isolation Kit (Ambion, Austin, TX), and the concentration was measured using the NanoDrop ND-1000 spectrophotometer as above.

Real-Time Polymerase Chain Reaction (PCR)

The expression levels of ZIP4, matrix metallopeptidase 9 (MMP-9), vascular endothelial growth factor A (VEGF-A), platelet-derived growth factor subunit A (PDGF-A), interleukin 6 (IL-6), interleukin 8 (IL-8), and insulin-like growth factor binding protein 2 (IGFBP-2) in 34 surgical samples from UT were analyzed with real-time PCR using the SYBR Supermix Kit (Bio-Rad, Hercules, CA). PCR included the following components: 100 nM each primer, diluted cDNA templates, and iQ SYBR Green supermix and running for 40 cycles at 95°C for 20 s and 60°C for 1 min. PCR efficiency was examined by serially diluting the template cDNA, and the melting curve data were collected to check PCR specificity. Each cDNA sample was run as triplicates, and the corresponding no–reverse-transcriptase (RT) mRNA sample was included as a negative control. The β-actin primer was included in every plate to avoid sample variations. The relative mRNA level was presented as unit values of 2^[Ct(β-actin) – Ct(gene of interest)]. The primer sequences for ZIP4, MMP-9, VEGF-A, PDGF-A, IL-6, IL-8, and IGFBP-2 are listed in Supplementary Table S1.

Statistical Analysis

Statistical analysis was performed using the statistical package SPSS (version 16.0; SPSS Inc., Chicago, IL). For the microarray data, if multiple probes were used for a single candidate gene, the reference sequence of each probe was searched in GenBank and the probes detecting exons were chosen for further analysis. Of the 24 zinc transporters, the probe for ZnT4 was missing in the CGCA database and the probe for ZIP11 was missing in the REMBRANDT database. Microarray expression data of genes were further normalized by subtracting the mean and dividing by the standard deviation for the probe set. The differences of the transcription levels of the 14 ZIPs and 10 ZnTs between grade II and grade IV tumors were compared using Student's t test. Both REMBRANDT and CGCA samples were dichotomized into low and high gene-expression groups, with cutoffs set at the median value of the ZIPs and ZnTs expression levels. Overall survival times were compared using the Kaplan-Meier method and formally compared using the log-rank test. The strength of the associations was measured using hazard ratio, which is the ratio of the instantaneous death rate comparing surviving patients with high zinc transporter level with patients having low zinc transporter level. The prognostic value of the 14 ZIPs and 10 ZnTs were evaluated using multivariate Cox proportional hazard regression models, adjusting for patients' tumor grade, age, and preoperational KPS score. Associations between ZIP4 and MMP-9, VEGF-A, PDGF-A, IL-8, IL-6, and IGFBP-2 transcription levels were assessed using simple linear regression models. All statistical tests were 2-sided, and P < .05 were considered to be statistically significant. When multiple comparisons were involved, Bonferroni correction was used and the significance cutoff was .05/24 = .002.

Gene Profiling Identifies Dysregulated Zinc Transporters in Human Glioma

We analyzed 186 tumor samples from patients with glioma in the CGCA cohort. Of these patients with glioma, 98 (52.7%) received a diagnosis at Glioma Center of Beijing Tiantan Hospital of grade II gliomas and 88 (47.3%) received a diagnosis of grade IV gliomas. The expression levels of the 24 zinc transporters were compared between the grade II gliomas and the grade IV gliomas. The results are shown in the heatmap diagram in Fig. 1. Specifically, after Bonferroni correction for multiple comparisons, ZIP3, ZIP4, ZIP8, ZIP14, ZnT5, ZnT6, and ZnT7 were found to be significantly upregulated in grade IV patients, compared with grade II patients (P< .002). On the other hand, ZIP11, ZnT3, ZnT9, and ZnT10 were significantly downregulated in grade IV patients, compared with grade II patients (P< .002).

To verify these findings, we also conducted similar analysis for the 335 patients with gliomas in the REMBRANDT cohort, in which 139 (39.2%) and 196 (60.8%) patients received a diagnosis of grade II and grade IV gliomas, respectively. The results on the comparison of the expression levels between the 2 groups are presented in Supplementary Fig. S1. These results are consistent with the findings from the CGCA cohort. Specifically, ZIP1, ZIP3, ZIP4, ZIP7, ZIP8, ZIP13, ZIP14, ZnT5, ZnT6, and ZnT7 were upregulated and ZnT10 was downregulated in grade IV, compared with grade II gliomas (P< .002). The consistent findings across the CGCA and REMBRANDT cohorts suggest that specific zinc transporters may play important roles in glioma pathogenesis and progression. A summary of mean differences, 95% confidence intervals (CIs) of differences, and P values based on 2 sample t test for the CGCA and REMBRANDT cohorts is provided in the Supplementary Table S2.

Association between Zinc Transporters and the Overall Survival Time among Patients with Glioma

To account for the multiple comparison at 24 transporters, Bonferroni correction was used to adjust the significance level, and P value <.002 is declared as statistically significant. The hazard ratio of each zinc transporter was estimated from multivariate Cox proportional hazard regression models, and the results are summarized in Table 2. Among the 24 zinc transporters, ZIP4 showed moderate association with patient survival in the CGCA and moderate association in the REMBRANDT cohorts (hazard ratio = 1.61, 95% CI = 1.02–2.53, P = .040 based on CGCA cohort and hazard ratio = 1.32, 95% CI = 1.08–1.61, P = .007 based on REMBRANDT cohort), and ZIP3 showed significant association in REMBRANDT cohort (hazard ratio = 1.41, 95% CI = 1.16–1.70, P< .001), but not significant in CGCA cohort (hazard ratio = 1.22, 95% CI = 0.91–1.63, P = .181). ZIP6, ZIP7, ZIP9, ZIP10, ZIP11, ZnT8 and ZnT9 showed significant association with survival in CGCA cohort, but not in REMBRANDT cohort. The minor statistical difference in certain ZIPs (such as ZIP3) between the CGCA and REMBRANDT cohort might be attributable to the different age distribution and ethnicity of the patients from those 2 cohorts.

We further evaluated the prognostic values of ZIP4 by Kaplan-Meier plotting method and compared with log-rank test. ZIP4 showed a significant association with survival both in CGCA cohort (low expression vs high expression, 40.3 months [95% CI, 35.7–45.0 months] vs 29.4 months [95% CI, 25.8–33.0 months]; P = .009) and in REMBRANDT cohort (42.5 months [95% CI, 35.3–50.0 months] vs 24.7 months [95% CI, 20.6–28.8 months]; P< .001) (Fig. 2). In grade II glioma group, ZIP4 is significantly correlated with patients' survival in REMBRANDT cohort (67.8 months vs 33.2 months; P< .001), but not in CGCA cohort because of limited number of deaths. ZIP4 does not correlate with prognosis in grade IV glioma subgroup. Prognostic values of other potential zinc transporters (ZIP3, ZIP6, ZIP7, ZIP9, ZIP10, ZIP11, ZnT8, and ZnT9) were also evaluated (Supplementary Figs. S3–S10).

Validation of the Microarray Data by Real-Time PCR

On the basis of the expression profile and the survival analyses described above, we chose ZIP4 for further validation. ZIP4 mRNA level was examined in 34 surgically resected samples from an independent UT cohort. As shown in Fig. 3, the mean ZIP4 expression level in grade IV glioma is 5.9-fold higher, compared with that of grade II gliomas (Student's t-test, P = .040). These results, supported by 3 independent cohorts (CGCA, REMBRANDT, and UT), strongly suggest that ZIP4 may be a novel diagnostic and prognostic marker for glioma and may serve as a new target for molecular targeted therapy of glioma.

Correlation between ZIP4 and Other Key Genes for Cell Growth and Angiogenesis in Glioma

To further investigate the underlying mechanism of ZIP4-mediated glioma pathogenesis, we investigated the associations between the expression level of ZIP4 and other key genes for tumor growth. We found that ZIP4 level was significantly associated with several genes that play important roles in cell growth and angiogenesis, such as MMP-9, VEGF-A, PDGF-A, IL-6, IL-8, and IGFBP-2 (Wald test P< .001), based on the data from the CGCA cohort (Supplementary Fig. S2). The selection of those 6 genes for validation is based on their expression correlation with ZIP4 levels in the 2 cohorts, and their functional involvement in ZIP4-mediated tumor growth as we previously described.^29,30 To validate this finding, we examined the expression of those 6 genes in 34 surgically resected samples from an independent cohort (UT cohort). As shown in Fig. 4, we plotted the expression level of ZIP4 against each of the 6 genes based on the brain tumor samples from the UT cohort with real-time PCR. The associations were statistically significant (Wald test P< .001), suggesting that those genes might play important roles in ZIP4-mediated tumor growth and metastasis of glioma.