Many human medulloblastoma tumors overexpress repressor element-1 silencing transcription (REST)/neuron-restrictive silencer factor, which can be functionally countered by REST-VP16

Medulloblastoma is believed to originate mostly from the undifferentiated neuroectodermal stem cells in the cerebellum (1–5). Although most medulloblastomas represent undifferentiated phenotypes, others express markers representing various levels of differentiation in various cell lineages, such as neuronal and glial cells, and undifferentiated medulloblastoma cells can be further differentiated when exposed to various reagents (1–3, 6, 7). A fraction (∼15%) of patients with sporadic medulloblastoma harbor genetic defects in the patched (PTCH) gene (8). In addition, some mutant ptc^+/− mice produce medulloblastomas (9). The PTCH gene is an integral part of the hedgehog-hip-patched-smoothened pathway (10), and a mutation in any of those genes may give rise to medulloblastoma (11). Mutations in the p53, β-catenin, and Rb pathways are also involved in this disease (3, 12–14). Other mechanisms involving JC virus and various other pathways regulated by platelet-derived growth factor receptor A, insulin-like growth factor I receptor, or Trk receptor tyrosine kinase, etc., also have been correlated with medulloblastoma (4, 14–16). Perhaps these observations reflect the fact that different mechanisms under different genetic backgrounds trigger medulloblastoma tumorigenesis.

One challenge facing medulloblastoma research is understanding the mechanism or mechanisms that regulate most of these tumors. Our previous work indicated that several medulloblastoma cell lines, compared with either neuronal progenitor cells or fully differentiated neurons (17, 18), overexpress repressor element-1 silencing transcription/neuron-restrictive silencer factor (REST/NRSF). REST/NRSF is a global transcriptional repressor (19, 20) responsible for silencing the transcription of most neuronal differentiation genes by binding to a 23-bp consensus DNA sequence, the repressor element-1 binding site/neuron restrictive silencer (RE1/NRSE), which is present in the regulatory regions of these genes. REST/NRSF is expressed in most, if not all, nonneuronal cells in vivo (19, 20), but it is not expressed at high levels in differentiated neurons during embryogenesis. However, later studies showed that it is expressed in certain mature neurons in adults (19, 20), suggesting that REST/NRSF may play a complex role that in turn may depend on its cellular and physiologic environment. REST/NRSF-dependent promoter repression requires interaction with several cellular cofactors, including Co-REST, mSin3A, and histone deacetylase complex, and it requires histone deacetylase activity (19–21).

We constructed a recombinant transcription factor, REST-VP16, by replacing the repressor domains of REST/NRSF with the activation domain of the herpes simplex virus protein VP16 (17, 18, 22, 23), which was designed to counter the effects of REST/NRSF. Our work showed that in transient transfection experiments, REST-VP16 operates through RE1/NRSE, competes with endogenous REST/NRSF for DNA binding, activates plasmid-encoded neuronal promoters in medulloblastoma cells and other mammalian cell types, and activates cellular REST/NRSF target genes even in the absence of factors otherwise required to activate such genes. Furthermore, high-efficiency expression of REST-VP16 mediated by Ad.REST-VP16 in medulloblastoma cells countered endogenous REST/NRSF-mediated repression of neuronal promoters, stimulated the expression of endogenous neuronal genes, and triggered apoptosis through the activation of caspase cascades. In addition, i.t. injection of Ad.REST-VP16 in established s.c. medulloblastoma tumors in nude mice inhibited tumor growth. Using immunohistochemical analysis, here we show that a majority of the 21 human medulloblastoma tumor specimens we examined overexpressed REST/NRSF (6 strongly and 11 weakly). In contrast, adjacent normal cerebellum tissue sections were negative for REST/NRSF expression, indicating that overexpression of REST/NRSF is a new and major marker for human medulloblastoma tumors. Furthermore, we also found that Ad.REST-VP16 blocked the i.c. tumorigenic potential of medulloblastoma cell lines and inhibited growth of established tumors in nude mice, suggesting that expression of neuronal differentiation genes in medulloblastoma cells can interfere with the tumorigenic potential of the cells. Based on these experiments, we suggest that overexpression of REST/NRSF in neuronal cells inhibits the transcription of multiple neuronal differentiation genes, blocks these cells at a predifferentiation stage, and contributes to medulloblastoma tumorigenesis.

For histologic examination, surgically excised human brain tumor tissue and adjacent normal tissue samples fixed in 10% buffered formalin and embedded in paraffin were obtained from the Brain Tumor Center tissue bank at M.D. Anderson Cancer Center, stained with H&E, and examined under light microscope as previously described (17). For immunohistochemical assays, the brain sections were stained with an anti-REST antibody (a gift of Gail Mandel) essentially as previously described (17). In brief, the routinely processed, formalin-fixed, paraffin-embedded tissue was sectioned at a thickness of 4 to 5 μm. The tissue sections were deparaffinized, incubated in 0.3% hydrogen peroxide in methanol for 10 minutes at room temperature, washed, and treated with 1% acetic acid in PBS for 5 minutes. Nonspecific background binding was blocked with normal goat serum. The slides were incubated with primary anti-REST antibody overnight at 4°C. The secondary polyclonal antibody (biotinylated goat anti-rabbit antibody, diluted 1:100, Oncogene Science, Cambridge, MA) was applied for 1 hour at 37°C. Detection was accomplished after incubation with avidin alkaline phosphatase (DAKO Corporation, Carpinteria, CA), diluted 1:200, for 1 hour at 37°C, followed by application of an alkaline phosphatase substrate (Life Technologies, Gaithersburg, MD) and incubation in the dark for 20 to 30 minutes. The staining intensity was graded as negative (−), weakly positive (+), or strongly positive (++), and the number of tumor cells that were positively stained was expressed as a percentage of the total number of tumor cells, estimated on the basis of 15 to 20 representative fields from each section.

D283 (or Daoy) medulloblastoma cells were purchased from American Type Culture Collection (Rockville, MD) and grown as previously described (17). Replication-incompetent adenoviral vectors encoding the vector DNA (Ad), green fluorescent protein cDNA (Ad.GFP), or the REST-VP16 cDNA (Ad.REST-VP16) under the control of cytomegalovirus promoter/enhancer elements were generated and used to infect the Daoy cells as previously described (17). Deparaffinization and immunohistochemical staining were done on a Ventana BenchMark XT Autostainer (Ventana, Tucson, AZ). Initially, antigen retrieval was incorporated using citrate buffer (pH 6.0) in a rice steamer for 45 minutes at 92°C. After rinsing in PBS, the sections were incubated with the primary antibody, synaptophysin (Abcam, Cambridge, MA; monoclonal mouse clone sy38), at the dilution of 1:50. The avidin-biotin-peroxidase conjugation method was used by the detection kit, LSAB (DAKO). The immunoreaction was visualized using 3-amino-9-ethylcarbazole as the chromogen. Sections then were counterstained with Mayer's hematoxylin for visualization of nuclear detail.

These experiments were done following M.D. Anderson Institutional Animal Care and Use guidelines. Daoy cells infected with Ad or Ad.REST-VP16 (2 × 10⁵ cells present in 5 μL cell growth medium) were inoculated in a group of six mice for each experimental condition using an implantable guide-screw system we described elsewhere (24). This method, which uses a guide screw, a stylet, a modified Hamilton syringe with a 26-gauge needle, and an infusion pump, allows delivery of cells to a very specific location in the brain. Several days before the inoculation, the guide screws were placed in the dorsal cranium of mice. The animals were anesthetized by i.p. injection of 1.2% avertin solution made in sterile PBS per gram of body weight (100% avertin solution: 10 g tribromoethylalcohol + 10 mL tertiary amyl alcohol) at a dosage of 15 to 17 μL per gram of body weight. The guide screw entry site was marked at a point 2.5 mm lateral and 1 mm anterior to the bregma, which is located directly above the caudate nucleus. A small, hand-controlled twist drill was used to make the hole, and a specially devised screwdriver was used to thread the screw into the hole and secure it. After a recovery period of 5 days after the guide screws were placed, the cells were inoculated. On the day of implantation, the cells were harvested and resuspended in Ca²⁺-free and Mg²⁺-free PBS. The cells were kept on ice until implantation. Mice with the guide screws in place were reanesthetized as described above. The tumor cell suspension was drawn into a Hamilton syringe, which was fitted with a cuff to control the depth of injection. The needle of the Hamilton syringe was slowly lowered into the center of the guide screw until the cuff rested on the screw surface. Up to 10 mice attached to 10 syringes were positioned in a row in a Harvard Apparatus (Holliston, MA) infusion pump that provided continuously controlled slow injection of the cells into the caudate nucleus at a rate of 1 μL/min. After the entire volume of the cell suspension was injected, the needles were manually removed and with a fine forceps, the stylets were positioned in the screw holes to close the system and prevent the tumor cells from leaking after inoculation. The mice were sacrificed after 4 weeks by CO₂ inhalation and their brains fixed with formalin and embedded in paraffin; 4 to 5 μm brain sections were examined as described above for the human medulloblastoma brain specimens. For i.t. adenovirus injection experiments, i.c. implanted medulloblastoma cells were allowed to grow for 2 weeks and then each mouse was injected thrice at the interval of 3 days with 2 × 10⁵ plaque-forming units of Ad.GFP or Ad.REST-VP16 present in 3 μL volume using the guide-screw system. The mice brains were then examined by magnetic resonance imaging after an additional 4 weeks, sacrificed, and their brains were analyzed by histologic methods as described above. The diameter of each tumor was determined by measuring the average of the largest and the smallest diameters and this value was used to calculate the average volume of each tumor. Because we have small number of independent samples, the P value was calculated by the Mann-Whitney Test using SPSS statistical package (SPSS Inc., Chicago, IL).

The magnetic resonance images were acquired on a clinical 1.5 T GE Signa (GE Medical Systems, Waukesha, WI) in the Small Animal Cancer Imaging Research Facility at University of Texas M.D. Anderson Cancer Center (Houston, TX). Mice were anesthetized for imaging using 1.5% to 5% isoflurane inhalation anesthesia. The i.c. screws were removed from each of the mice before imaging. The images of the brains of the mice were acquired using a 3 cm quadrature volume resonator. Axial T₁-, T₂-, and postcontrast T₁-weighted images using gadolinium-diethylenetriaminepentaacetic acid (Magnebist from Berlix Imaging, Wayne, NJ) as a contrast agent were acquired of the brain. The axial head images covered a 4.0 cm square field-of-view. Data was acquired at 256 × 192 data matrices, resulting in an isotropic in-plane resolution of ∼150 × 200 μm. All slices were 1.5 mm thick and separated by 0.5 mm gaps. The T₁-weighted images were acquired with a spin-echo pulse sequence at TE/TR = 14/450 ms, whereas the T₂-weighted images were acquired with fast spin-echo pulse sequence at TE/TR = 90/5,000 ms with an echo train length of 12.

Apoptosis in tumor sections was determined by using the TACS in situ apoptosis detection kit (R&D Systems, Minneapolis, MN) essentially according to the manufacturer's instructions. Formalin-fixed, paraffin-embedded mouse brain sections were deparaffinized and then digested with proteinase K for permeabilization. The apoptotic DNA fragments were labeled with biotinylated nucleotides using terminal deoxynucleotidyl transferase and the biotinylated nucleotides were then detected by streptavidin-horseradish peroxidase conjugate and diaminobenzidine.

To determine whether human medulloblastoma tumors overexpressed REST/NRSF, we examined 21 surgically excised, formalin-fixed, paraffin-embedded tissue specimens organized in a tissue array, first by H&E staining (Fig. 1A and B) and then by immunohistochemical analysis with an anti-REST antibody (Fig. 1C; Table 1). All specimens were obtained from the M.D. Anderson Brain Tumor Center tissue bank and were previously diagnosed by histologic examination as medulloblastoma. Six normal cerebellar tissue samples adjacent to tumors, which were histopathologically distinct from the tumor tissue, were included in the array as negative controls. Medulloblastoma patient age ranged from 2 to 34 years (Table 1). The morphology shown in Fig. 1A and B shows that normal cerebellar cortex tissue was comprised of granular-cell neurons with monotonously uniform, round nuclei and interspersed pale neuropil islands (glomeruli), whereas the medulloblastoma displayed disheveled, densely packed tumor cells with irregular, hyperchromatic, pleomorphic nuclei, as expected for these tumors (25). The immunohistochemical analysis (Fig. 1C) done using anti-REST/NRSF antibodies revealed a strong positive signal with HeLa cells (HeLa cytospin), which express REST/NRSF, and they served as a positive control (17). These cells were spun down, fixed with formalin, and paraffin-embedded and processed the same way as the brain tissue. As described above, normal cerebellar tissue adjacent to the tumor was used as a negative control. Whereas no immunoreactivity was observed in normal cerebellar tissue, REST/NRSF positivity was observed in 17 of 21 medulloblastoma tumor samples, with strong signals in 6 and weak signals in 11. The four specimens that did not show immunoreactivity came from patients ages 2, 8, 12, and 25 years, indicating that immunopositivity is not universally present in younger patients. These results indicated that REST/NRSF is a major determinant for human medulloblastoma tumors.

Several medulloblastoma cell lines (D283, Daoy, and D341) overexpress REST/NRSF, and they do not express neuronal differentiation genes (17). These cells also form i.c. tumors in nude mice (6). Previously, we showed that the recombinant molecule REST-VP16, which can compete with REST/NRSF for binding to RE1/NRSE and can activate neuronal differentiation genes (REST/NRSF target genes) instead of repressing them, could also block tumorigenic potential of various medulloblastoma cells at s.c. locations in mice (17). Because the i.c. environment may be critical for medulloblastoma tumorigenesis, we wanted to determine whether REST-VP16 could also block the tumorigenic potential of medulloblastoma cells in orthotopic mouse models. For this purpose, we inoculated D283 (or Daoy cells) cells infected with either Ad or Ad.REST-VP16 into the brain of six 8-week-old nude mice using the implantable guide-screw system described in Materials and Methods. We sacrificed the mice after 5 weeks and examined their paraffin-embedded brain sections by H&E staining (Fig. 2). Whereas no tumor was visible in any of the Ad.REST-VP16–inoculated mice tumors, large tumors appeared in five of six Ad-inoculated mice. Similar results were also obtained with Daoy medulloblastoma cells. These experiments suggest that countering REST/NRSF function and the forced expression of neuronal differentiation genes in medulloblastoma cells can interfere with tumorigenic potential of the cells in orthotopic mouse models.

To examine whether the i.t. injection of adenoviral vectors carrying REST-VP16 would inhibit the growth of preformed medulloblastoma i.c. tumors, uninfected D283 cells (or Daoy cells) were injected i.c. in nude mice as described above. After 2 weeks, one group of mice was injected with Ad.GFP and the other with Ad.REST-VP16. Three injections were carried out with an interval of 3 days each, for a total of three injections. At the end of an additional 4-week period, all mice were analyzed by magnetic resonance imaging, sacrificed, and their paraffin-embedded brain sections were analyzed by H&E to detect the tumor volume (Fig. 3). The Ad-REST-VP16-injected mice showed smaller tumors with a volume of 2.10 mm³ (SE = 0.54) compared with mice injected with Ad.GFP that showed an average tumor volume of 13.57 mm³ (SE = 3.07). The P value for comparing both sets of data is 0.002 using Mann-Whitney test. Again, similar results were obtained with Daoy cells. It is interesting that in the tumorigenicity experiment described in the previous section, in which cells were directly infected with adenovirus in vitro and then inoculated into nude mice, tumors grew in none of the mice receiving Ad.REST-VP16–infected cells. In contrast, the i.t. injection of Ad.REST-VP16 caused inhibition of tumor growth rather than regression of the tumor. This suggests that i.t. injection delivers the virus only to a limited number of cells in the tumor, presumably causing them to undergo apoptosis (17). However, the spread of the virus among neighboring tumor cells was restricted because of the fact that these are replication-defective viruses. Thus, tumor cells that did not receive the virus would grow at a normal rate for the tumor.

To determine whether the delivery of REST-VP16 in preformed medulloblastoma tumors indeed caused expression of terminal neuronal differentiation genes and apoptosis, we subjected the tumor sections to immunohistochemical analysis with antisynaptophysin antibody (Fig. 4) and an in situ apoptosis detection assay that measures apoptotic cells by detecting DNA fragmentation (Fig. 5). As shown, Ad.REST-VP16, and not Ad.GFP, injected tumors showed expression of sypatophysin and distinct signs of apoptosis.

Our observations described here indicated that REST/NRSF overexpression is involved in a majority of the human medulloblastoma tumors. REST/NRSF has several differentially spliced isoforms (26, 27). One such isoform, REST4, functions as a dominant-negative regulator by interfering with REST/NRSF activity (27, 28). Because the antibody used in this study was raised against the full-length REST/NRSF protein, it is not clear whether the medulloblastoma samples express the full-length REST/NRSF or any of its isoforms or both. However, we previously found that several human medulloblastoma cell lines expressed the repressor activity of REST/NRSF and subsequently blocked its multiple target neuronal differentiation genes. Furthermore, countering REST/NRSF activity in those cells through REST-VP16 caused activation of the target genes. These results suggested that the human medulloblastoma cells expressed the repressor activity of REST/NRSF and most likely not the REST4 isoform, which would have inhibited REST/NRSF activity and activated neuronal differentiation genes. Similarly, the data shown in Fig. 4 also indicated that the i.c. injected tumor cells did not express the neuronal differentiation genes, but they did so only when infected with REST-VP16. However, experiments done with antibodies specific for defined domains of the REST/NRSF protein would be needed to definitively answer this question.

Here, we also found that expression of REST-VP16 in human medulloblastoma cells caused apoptosis and blocked i.c. tumorigenic potential of the cells in mice. Previously, we found that expression of REST-VP16 in human medulloblastoma cells did not convert them to neurons but rather caused apoptosis (17). The exact mechanism for this process is not known. One possibility is that although REST-VP16 does activate neuronal differentiation genes, it does not block the expression of genes encoding stem cell properties and the simultaneous expression of genes regulating both differentiation and undifferentiation properties caused apoptosis. The second possibility is that the medulloblastoma cells, like most other cancer cells, harbor genetic instability and might not have contained all the cellular machinery required to proceed through normal neuronal differentiation pathways. This seems to be a reasonable possibility because REST-VP16 was found to convert neural stem cells and myoblasts to a physiologically active neuronal phenotype (22, 23). It will be interesting to determine the effect of REST-VP16 expression in medulloblastoma tumors that do not overexpress REST/NRSF. If the latter tumors are blocked in neuronal differentiation by some other mechanisms and they do not contain the functional machinery for normal differentiation, those tumors might behave very similar to the tumors studied here. In contrast, if those tumors are blocked in neuronal differentiation but also contain the functional differentiation machinery, they would likely undergo terminal differentiation as was observed with the neural stem cells (23).

On the basis of our observations, we suggest a model (Fig. 6) in which, when normal neuronal stem cells that may express REST/NRSF undergo differentiation, the expression of REST/NRSF is blocked and the cells differentiate into mature neurons in response to environmental cues leading to cascades of activators that are transcribed in a stage-specific manner. However, under abnormal conditions, the expression of REST/NRSF is either maintained or reinitiated in response to abnormal environmental cues along the differentiation pathway, repressing the transcription of most terminal differentiation genes (REST/NRSF target genes), capturing these still-dividing cells at a stage before full differentiation, and eventually initiating medulloblastoma tumor formation. Our hypothesis is not inconsistent with the possibility that REST/NRSF cooperates with other abnormally expressed regulators and that, after tumors are initiated by REST/NRSF, tumor maintenance and progression are then regulated by other genes. Furthermore, REST/NRSF may have yet-unknown functions in this process that give rise to additional mechanisms.

We thank Belinda Rivera for her assistance with the anesthesia and positioning of the animals during the magnetic resonance procedures, Murlidhar Tekchandani for acquiring the magnetic resonance images, Joy Gumin for help with the terminal deoxynucleotidyl transferase–mediated nick end labeling assay, and Bunmi Owolabi for help with the statistical analysis of the data.