Invasion suppressor cystatin E/M (CST6): high-level cell type-specific expression in normal brain and epigenetic silencing in gliomas

Abstract

DNA hypermethylation-mediated gene silencing is a frequent and early contributor to aberrant cell growth and invasion in cancer. Malignant gliomas are the most common primary brain tumors in adults and the second most common tumor in children. Morbidity and mortality are high in glioma patients because tumors are resistant to treatment and are highly invasive into surrounding brain tissue rendering complete surgical resection impossible. Invasiveness is regulated by the interplay between secreted proteases (eg, cathepsins) and their endogenous inhibitors (cystatins). In our previous studies we identified cystatin E/M (CST6) as a frequent target of epigenetic silencing in glioma. Cystatin E/M is a potent inhibitor of cathepsin B, which is frequently overexpressed in glioma. Here, we study the expression of cystatin E/M in normal brain and show that it is highly and moderately expressed in oligodendrocytes and astrocytes, respectively, but not in neurons. Consistent with this, the CST6 promoter is hypomethylated in all normal samples using methylation-specific PCR, bisulfite genomic sequencing, and pyrosequencing. In contrast, 78% of 28 primary brain tumors demonstrated reduced/absent cystatin E/M expression using a tissue microarray and this reduced expression correlated with CST6 promoter hypermethylation. Interestingly, CST6 was expressed in neural stem cells (NSC) and markedly induced upon differentiation, whereas a glioma tumor initiating cell (TIC) line was completely blocked for CST6 expression by promoter methylation. Analysis of primary pediatric brain tumor-derived lines also showed CST6 downregulation and methylation in nearly 100% of 12 cases. Finally, ectopic expression of cystatin E/M in glioma lines reduced cell motility and invasion. These results demonstrate that epigenetic silencing of CST6 is frequent in adult and pediatric brain tumors and occurs in TICs, which are thought to give rise to the tumor. CST6 methylation may therefore represent a novel prognostic marker and therapeutic target specifically altered in TICs.

Fig. 1

Localization of cystatin E/M expressing cells in non-neoplastic control brain tissue. (A, B) Immunohistochemical detection of cystatin E/M in normal cortex (A) and white matter (B) from two representative cases in the TMA. Brown color indicates cystatin E/M expression, which is undetectable in neurons (arrowheads), moderately expressed in many but not all astrocytes (open arrows), and highly expressed in oligodendrocytes (closed arrows). (C) Immunofluorescence staining of cultured normal human astrocytes (NHA) for cystatin E/M expression. Co-immunostaining of NHA for cystatin E/M, GFAP (a marker of astrocytes), and DAPI (to visualize nuclei). (D) Higher power image of a representative NHA. Images in A–C are taken at 20X magnification and in part D at 60X magnification.

Fig. 2

Bisulfite genomic sequencing (BGS) DNA methylation analysis of the CST6 promoter in normal brain and cultured cells. (A) Schematic of the CST6 promoter region with the vertical tick marks indicating CpG sites and the bent arrow (+1) denoting the transcription start site defined using NCBI Map Viewer. Below this, the boundary of the regions analyzed by BGS, MSP, and pyrosequencing throughout the paper are shown with numbering relative to the transcription start site. (B) BGS analysis of normal white matter and cortex tissues, NHA, and human oligodendrocyte precursor cells (HOPC). The total percent methylation for all clones is indicated at the left. The BGS region shown in part A is expanded at the top. Each row of circles (representing CpG sites) corresponds to one sequenced clone. Open and filled circles represent unmethylated and methylated CpGs, respectively.

Fig. 3

Representative immunohistochemical staining for cystatin E/M expression in primary brain tumors from the TMA. (A) Two regions from one TMA core showing predominantly tumor (GBM, left panel) and adjacent reactive-appearing brain tissue (right panel). Only the reactive tissue stains positive for cystatin E/M. Cystatin E/M expression in (B–C) glioblastoma (GBM) and (D–E) two anaplastic oligodendrogliomas (AO). Tumor cells show little or no cystatin E/M reactivity.

Fig. 4

MSP and pyrosequencing DNA methylation analysis of normal and tumor tissues from the TMA. (A) Representative MSP results for two normal and five primary brain tumor samples from the TMA. ‘U’ and ‘M’ represent PCR amplification with primers specific for unmethylated and methylated CST6 promoter, respectively (see also Fig. 2A). (B) Bisulfite pyrosequencing DNA methylation analysis for two normal and two tumor tissues from the TMA. Representative pyrograms are shown for each sample with the percent methylation at each of five CpG sites given above its location in the sequence (see Fig. 2A for the region analyzed by pyrosequencing). (C) Average percent methylation for the five CpG sites shown in part B for four non-neoplastic tissues and 18 tumor tissues derived from the TMA. These samples were also analyzed for cystatin E/M expression in the TMA by IHC. There was an excellent correlation between the MSP and pyrosequencing data as well as the presence of DNA hypermethylation (>20%) and reduced/absent cystatin E/M expression by IHC (Table 1).

Fig. 5

CST6 expression and methylation status in neural stem cells and glioma tumor initiating (tumor stem) cells. (A) Semi-quantitative RT-PCR analysis showing that CST6 is expressed at moderate to high levels in two neural stem cell lines (SCP-23 and SCP-27) cultured under non-differentiating conditions. In contrast, the glioma initiating line H1228 does not express CST6 but can be induced to re-express CST6 following treatment with 5 μM 5-azadC for 10 days (‘+’ lane). Untreated and 5-azadC-treated T98G cells serve as negative and positive controls for CST6 expression, respectively. Expression of Nestin and SOX2 demonstrate the multipotent phenotype of normal stem cells and tumor initiating cells. Amplification of beta-actin serves as a loading control. (B) CST6 BGS methylation analysis of neural stem cells and the tumor initiating line H1228. High-level methylation of the CST6 promoter in H1228 is consistent with the lack of CST6 expression in the absence of 5-azadC treatment. (C) RT-PCR analysis showing that CST6 expression is induced in neural stem cells following differentiation, while this upregulation is blocked in H1228 cells. The graph shows representative gel quantitations and is presented as the fold change in CST6 expression upon differentiation (relative to the undifferentiated cells). ‘NC’ – not calculated since no CST6 expression was detectable. ‘D’ differentiated, ‘U’ undifferentiated.

Fig. 6

CST6 silencing and promoter hypermethylation is common in low and high grade pediatric primary brain tumor cultures. (A) Representative RT-PCR analysis of CST6 expression in primary cell lines derived from different types of pediatric brain tumors cultured in the absence (‘−’) or presence (‘+’) of 5-azadC (5 μM for 3 days) and TSA (100 nM for final 24 hrs, ‘A+T’ lanes). Amplification of beta-actin serves as a loading control. (B) MSP DNA methylation analysis of the same group of tumors from part A showing that CST6 promoter hypermethylation is common and correlates with transcript inducibility by 5-azadC and TSA. (C) CST6 expression by RT-PCR in cells treated singly with 5-azadC (‘A’) or TSA (‘T’). (D) BGS methylation analysis of select pediatric brain tumor cultures. G - glioma, PA - pilocytic astrocytoma.

Fig. 7

Ectopic expression of cystatin E/M suppresses glioma cell motility and invasion. (A) Scratch wound assay. T98G cells were transfected with empty vector (pcDNA3) or cystatin E/M expression vector (pcDNA3-CST6) and allowed to reach ~90% confluence over 72 hours. The monolayer was then wounded with a plastic pipette tip, washed, and photographed over a period of 24 hrs. U-87MG cells were not examined because they do not form a tightly packed monolayer. (B) Cell invasion assay. U-87MG cells were transfected with empty vector or cystatin E/M expression vector. After 24 hrs, cells were replated into the Cultrex cell invasion assay kit (Trevigen), allowed to grow for an additional 24 hrs then cell invasion was measured and set relative to empty vector transfected cells. Similar results were obtained with T98G cells (data not shown). (C) Western blot demonstrating ectopic expression of cystatin E/M in T98G and U-87MG cells. The breast cancer cell line MDA-MB-231 served as a positive control for cystatin E/M expression. Upper and lower bands (gray and black arrowheads) correspond to the glycosylated and non-glycosylated forms of cystatin E/M, respectively.