Gliomas, accounting for 30% of adult primary brain tumors, are the most common primary tumors of the central nervous system and are classified into four clinical grades, with the most aggressive and lethal tumors being grade IV glioblastoma multiforme (GBM) 12. The median survival of GBM patients is less than one year from initial diagnosis, and many of the commonly used chemotherapeutic agents have limited effects on these malignant tumors 3. Recently, there has been increasing hope that temozolomide, a novel alkylating agent, will prove efficacious in the treatment of human glioma 45. However, a number of studies have suggested that, in tumors, O⁶-methylguanine-DNA methyltransferase (MGMT) provides resistance to treatment with temozolomide, unless expression is lost by promoter methylation or there is direct inhibition of MGMT activity 6. Considering the attractive efficacy of temozolomide, one of the greatest challenges facing the field may be to identify therapeutic agents that suppress MGMT expression, as such drugs may sensitize resistant glioma cells to temozolomide. Thus, establishment of more sophisticated systems to understand the functions and regulation of MGMT are highly desired.

Normal human cells, genetically modified by retroviral-mediated gene transfer, have proven important, because such systems are useful for identifying factors directly contributing to tumorigenesis, in contrast to tumor-derived cell lines which harbor numerous unknown genetic abnormalities 78. GBMs are believed to arise from astrocyte cells by means of stepwise accumulation of genetic abnormalities 49, and immortalized normal human astrocyte (NHA) cells had been established by introducing the telomerase catalytic subunit (hTERT) in combination with human papillomavirus E6/E7 to inactivate both p53 and pRb pathways 10. It had been systematically demonstrated that the immortalized NHA cells, expressing activated Ras (H-RasV12) or expressing both H-RasV12 and an active form of Akt (myrAKT), formed tumors consistent with human anaplastic astrocytoma or GBM in intracranial- and flank-xenografts models 1011. These studies indicate that such systems are useful for glioma research to understand direct functions and regulation of genetic elements during transformation and gliomagenesis.

Here we created similar experimental models using NHA cells by introducing the simian virus 40 early region (SV40ER) instead of human papillomavirus E6/E7. The SV40ER encodes both small-t antigen, a suppressor of protein phosphatase 2A (which is downregulated in half of human glioma) 12 and large-T antigen, which directly binds to and inactivates p53, as well as pRB and the closely related proteins p107 and p130 13. Using such genetically modified NHA cells, we demonstrate that MGMT is downregulated during oncogene-mediated transformation of astrocyte cells. Since our results indicate that downregulation of MGMT expression was not primarily dependent upon promoter hypermethylation, inhibitors of DNA methylation or histone deacetylases (HDACs) may be used in combination with alkylating agents for improved treatment of some GBM cases. We propose that the NHA cell systems are useful for investigating the mechanisms underlying MGMT expression and for improving glioma therapies.

To create an experimental model of human gliomas, NHA cells were introduced with genes for hTERT (T), SV40ER (S), H-RasV12 (R), and myrAKT (A) (Figure 1A; NHA/TS, NHA/TSR, NHA/TSRA cells). The NHA/TS cells grew much faster than parental NHA cells, exhibiting continuous growth beyond population-doubling 50 (data not shown). However, the NHA/TS cells did not form colonies in soft agar or tumors in xenografts into recipient nude mice (Figure 1B; Table 1), indicating that the NHA/TS cells were immortalized but not transformed. In contrast, the NHA/TSR and NHA/TSRA cells formed a considerable number of colonies in soft-agar, as well as tumors in flank-xenografts into nude mice (Table 1). NHA/TSRA cells displayed much more refractile morphology and loose attachment to the culture dish relative to NHA/TSR cells (Figure 1B). Although the difference in the anchorage-independent growth property was quite subtle between NHA/TSR and NHA/TSRA cells (Table 1), the histopathological features of two types of flank xenografts were clearly different. In tumor tissues formed by NHA/TSR cell injection, atypical astrocytic cells with higher cellularity exhibits histopathological features of Grade III glioma as anaplastic astrocytoma of WHO classification. On the other hand, the TSRA-xenografts composed of similar atypical astrocytes displayed numerous areas of necrosis with nuclear pseudo-palisading pattern. Consistent with the histological feature as GBM, the immunohistochemical index of Ki-67 positivity is higher in NHA/TSRA tumors than those of NHA/TSR derived tumors (Figure 2).

Using the NHA cell system described above, we found that the mRNA level of MGMT was decreased in immortalized (NHA/TS) and transformed (NHA/TSR and NHA/TSRA) cells and that MGMT expression was also downregulated in the flank xenografts derived from NHA/TSR and NHA/TSRA cells (Figure 3A). Immunoblot analyses also demonstrated that introduction of SV40ER decreased MGMT level in NHA cells and that the levels were further decreased in transformed NHA cells (Figure 3B). Conversely, in human lung fibroblasts TIG3/T cells (TIG3 cells expressing hTERT) 14, neither SV40ER nor H-RasV12 suppressed MGMT expression (Figure 3B), suggesting the astrocyte-specific mechanisms of MGMT regulation. This is an interesting observation as TIG3/TS cells are refractory to Ras-induced genetic alteration 1516, morphological changes 17 and thereby cellular transformation 14. Although such observation prompted us to further test the tumor suppressive functions of MGMT, the additional expression of MGMT in NHA/TSR and NIH3T3/R cells (NIH3T3 cells transformed by H-RasV12) failed to suppress anchorage-independent growth (Table 1), suggesting that MGMT may be unable to restore transformed phenotypes. In fact, tumors have long been noted to be heterogeneous in MGMT expression 56, and approximately half of malignant gliomas expressed MGMT 18. However, it is possible that the mechanisms underlying resistance to downregulation of MGMT may render TIG3 fibroblasts refractory with respect to oncogene-mediated transformation, because MGMT functions to protect normal cells from exogenous carcinogens 6. In support of this idea is the observation that MGMT mRNA was downregulated in v-fos-transformed 1920 and Ras-transformed (TA unpublished data) rodent fibroblasts, showing highly malignant phenotypes.

To understand the mechanism underlying the downregulation of MGMT in transformed NHA cells, we first analyzed the MGMT expression following treatment with 5-aza-2'-deoxycytosine (5-aza-dC) and/or Valproic acid (VPA), because it has been widely acknowledged that aberrant methylation of MGMT-promoter contributes to the gene expression changes 6. Inhibitors of HDACs, such as VPA and trichostatin A, act synergistically with 5-aza-dC (a DNA methyltransferase inhibitor and demethylating agent in dividing cells) to further increase the expression of genes silenced in association with promoter hypermethylation 21. Reverse transcriptase (RT)-PCR analysis revealed that treatments with 5-aza-dC alone or combined treatment with 5-aza-dC/VPA slightly increased MGMT mRNA levels, when the PCR was performed in a condition where the positive control reaction (mRNA from the NHA/TS cells) was saturated (data not shown), suggesting that the downregulation may have been partially mediated by promoter hypermethylation. However, combined treatment with 5-aza-dC/VPA was insufficient to restore MGMT protein levels in NHA/TSR and NHA/TSRA cells whereas the treatment clearly induced acetyltation and demethylation of histone H3 (Figure 4A), suggesting that additional factors, which are expressed in the parental NHA cells but inactivated in the transformed NHA cells, may be required to increase MGMT protein levels.

In an attempt to identify such missing factors contributing to gene regulation of MGMT, we tested if the expression of wild type p53 restores MGMT levels, as MGMT was slightly decreased in the NHA/TS cells where p53 is inactivated by large T antigen (Figure 3). Although several glioma studies have demonstrated correlations between inactivation of the p53 pathway and lower levels and/or activity of MGMT 222324, the introduction of wild type p53 failed to restore MGMT protein levels in transformed NHA cells (Figure 4B). We also confirmed previous reports that both U87MG (harboring wild type p53) and U251MG (harboring p53 mutations) glioma cells expressed low or no levels of MGMT 252627. We further found that treatment with 5-aza-dC/VPA failed to induce MGMT protein expression in both cell lines (Figure 4C), whereas the treatment slightly increased the mRNA levels (by RT-PCR; data not shown). Since gliomas harboring p53 mutations and immuno-positive for p53 (G2, G7, and G8) as well as gliomas harboring wild type p53 and immuno-negative for p53 (G4 and G10) expressed MGMT mRNA (Figure 5, Table 2), MGMT might be regulated primarily by the factors other than p53 in gliomas. Conversely, MGMT is clearly downregulated in normal astrocyte cultures from p53 deficient mice 28, suggesting the differences in the regulation of MGMT between normal and transformed cells.

Since MGMT seems to be regulated in several ways, it is possible that inhibitors of HDACs may be used for therapies for some glioma patients in combination with alkylating agents, such as temozolomide. Since VPA, which inhibits both class I and class II HDACs 29, has displayed potent in vitro and in vivo antitumor activities against brain tumor cells 30, it is one attractive candidate agent. The treatment with VPA appreciably inhibited growth of the NHA/TSRA cells in a dose dependent manner (Figure 6A) and two-week-treatment with 0.5 mM VPA (within the range of serum levels achieved in pediatric high-grade glioma patients) 31 suppressed anchorage-independent growth (Figure 6B). Although hyperacetylation of histone H3 was noted in the NHA/TSRA cells treated with VPA for two weeks, the treatment failed to restore protein (Figure 6C) and mRNA levels of MGMT (data not shown). Since protein levels of p21^WAF1 and p27^KIP1 were elevated in VPA-treated NHA/TSRA cells (Figure 6C), the antiproliferative effect might be mediated, at least in part, by cell cycle arrest, as widely acknowledged. These findings suggest that HDAC inhibitors, which suppress tumor cell growth without affecting MGMT expression, may be used in combination with alkylating agents to some glioma patients.

Much of our understanding of the molecular basis of gliomagenesis derives from the study of established cell lines that are explanted from human tumors. Such cell lines are often assumed to be representative of the original diseases, and they have been extensively employed for the identification and preclinical testing of potential therapeutic compounds. However, they harbor an unknown number of genetic lesions, which complicates the study of specific molecules and pathways. Indeed, with respect to p53-mediated MGMT regulation, there have been contradictory observations, where transient knockdown of p53 caused MGMT downregulation in SF767 glioma cells without affecting promoter methylation 28 but increased MGMT expression in T98 cells 27. Since such discrepancy may depend on unidentified mutations specific for individual cell lines, it is difficult to provide constant conclusions using tumor-derived cell lines alone. Therefore, normal human cells as well as genetically engineered mouse models will prove very useful, which allow us to generate experimental models of human cancers by direct manipulation with defined genetic elements 78.

Here we established glioma models from NHA cells and demonstrated that MGMT is downregulated in the transformed astrocyte cells. Although numerous studies have proposed the strong linkage between MGMT expression and promoter hypermethylation of the gene, treatment with epigenetic compounds (5-aza-dC/VPA) was unable to increase MGMT protein levels in the transformed NHA cells or in cultured tumor cells. Some previous reports demonstrated the epigenetic regulation of MGMT expression by treating tumor cell lines with 5-aza-dC and testing the MGMT levels by RT-PCR 3233. However, based on the data presented here, such limited evaluation might not be sufficient, as RT-PCR assay overestimates the MGMT restoration. Since inappropriate evaluation would mislead the development of advanced therapies, protein levels or enzyme activity of MGMT in the cells treated with such compounds should be examined and compared with those in normal cells. In fact, the use of alkylating agents in combination with HDAC inhibitors has been hampered, probably because it has been suspected that such agents would increase MGMT expression. However, we showed that two-week-treatment with VPA inhibited tumor cell growth without increasing MGMT expression, suggesting potential clinical use although further preclinical studies are required.

As the treatment with 5-aza-dC/VPA slightly increased mRNA levels of MGMT in cultured glioma cells (including U87MG, U251MG and transformed NHA cells), methylation status might be involved in MGMT regulation to some extent. However, such treatment was insufficient to restore protein expression of MGMT, suggesting that the regulation mechanisms may largely depend on other cellular factors. Although p53 does not seem to contribute directly to MGMT expression, it is possible that other transcription factors, whose expressions have been downregulated or silenced during transformation and gliomagenesis, may collaborate with HDAC inhibitors to increase MGMT protein. Interestingly, the sonic hedgehog pathway has been shown to regulate the self-renewal of CD133-positive glioblastoma cells, which were resistant to temozolomide treatment 34. Furthermore, some of CD133-positive glioma cultures highly expressed MGMT as well as target genes of the sonic hedgehog pathway 35. Thus, it is possible that the regulation of MGMT expression may be partially mediated by the sonic hedgehog pathway in some cases.

As discussed above, MGMT seems to be regulated by a number of ways in gliomas. The detailed mechanisms should be further analyzed as MGMT level is a critical determinant for efficacy of therapies with alkylating agents. Thus, identification of molecules and compounds that increase MGMT expression by screening NHA/TSR or NHA/TSRA cells with cDNA- and chemical-libraries would be very useful for the development of rational drug combination. We propose that the NHA cell system creates refined human glioma models for the systematic dissection of genetic alterations and elucidation of the complexities of the signaling pathways important for gliomagenesis. These systems also provide powerful means to find and authenticate molecules of particular promise for therapeutic targeting, and the present study provides an important proof-of-principle test for such systems.

MGMT: O⁶-methylguanine-DNA methyltransferase

GBM: glioblastoma multiforme

NHA: normal human astrocyte

hTERT (T): human telomerase catalytic subunit

H-RasV12 (R): activated H-Ras

myrAKT (A): myristoylated form (active form) of AKT

SV40ER (S): simian virus 40 early region

HDAC: histone deacetylase

VPA: Valproic acid

5-aza-dC: 5-aza-2'-deoxycytosine

RT-PCR: reverse transcriptase polymerase chain reaction

GAPDH: Glyceraldehydes-3-phosphate dehydrogenase

The author(s) declare that they have no competing interests.

KS designed the research, carried out all experiments except for immunohistochemistry, and drafted the manuscript. TA participated in the design of the study and contributed to new reagents. EA carried out H&E staining and immunohistochemistry. KT participated in epigenetic studies. SK collected and analyzed clinical samples. ST carried out histological analyses, conceived of the study, participated in its design and coordination, and helped to draft the manuscript. All authors read and approved the final manuscript.