Genetic alterations associated with acquired temozolomide resistance in SNB-19, a human glioma cell line

Gliomas are the most common brain tumor. They are usually treated by surgical resection combined with irradiation and alkylating agent–based chemotherapy. However, their prognosis remains poor. The therapeutic benefits obtained are transient and recurrence is the rule, together with resistance to chemotherapy. The limited efficacy of chemotherapy is generally attributed to two factors: intrinsic or acquired chemoresistance and the blood-brain barrier impeding the delivery of cytotoxic agents (1).

Temozolomide is a recently introduced, oral, and generally well-tolerated second-generation alkylating agent. Temozolomide has yielded a significant clinical benefit in high-grade gliomas (2). Temozolomide-induced O⁶-G-methylation is reversed by O⁶-methylguanine-DNA methyltransferase (MGMT) in a reaction leading to irreversible inactivation of the protein. Increased levels of MGMT or loss of the mismatch repair capacity confer resistance to temozolomide (3). However, some tumors display resistance to DNA-methylating drugs independently of either their MGMT level or their mismatch repair status, suggesting that other major mechanisms are involved in the resistance to alkylating agents, including the loss of a functional p53 which is able to induce cycle arrest and apoptosis (4), the p53 independent Chk1-mediated G₂-M arrest which protects the cell from mitotic catastrophe (5), or the Akt pathway, frequently activated in glioblastoma, which bypass the temozolomide G₂-M arrest on the one hand, and on the other hand, protects cells by suppressing senescence and mitotic catastrophe (6).

Comparative genomic hybridization (CGH) analysis of cell lines or primary tumors has revealed regions of chromosomal imbalances associated with acquired chemoresistance (7–9). None of these studies, however, concerned glioma cell lines and resistance to temozolomide.

In the present study, we used the temozolomide-sensitive SNB-19 cell line to generate temozolomide-resistant variants by culturing cells in the presence of incremental concentrations of temozolomide. The molecular alterations associated with acquired temozolomide resistance were identified using cytogenetic molecular techniques, including conventional cytogenetics, fluorescent in situ hybridization (FISH), CGH, and CGH arrays. Chromosomal regions having gained or lost DNA during the acquisition of resistance were disclosed. The additional use of microarray analysis allowed us to detect quantitative alterations of mRNA transcripts associated or not with genetic changes. We also evaluated the “kinetics” of these changes during the acquisition of temozolomide resistance.

The glioma-derived cell line, SNB-19 (10), was cultivated in DMEM (Sigma, St. Quentin Fallavier, France) supplemented with 10% FCS (Dutscher, Brumath, France), penicillin G (102 IU/mL), streptomycin (50 μg/mL; Sigma), and l-glutamine (2 mmol/L; Sigma). Cells were maintained at 37°C in a humidified 5% CO₂/95% air incubator, and harvested for passage with 0.5 mg/mL trypsin and EDTA (0.2 mg/mL) when they had reached confluence. For resistant cell selection, temozolomide (Interchim, Montluçon, France) was dissolved in DMSO as a 17 mmol/L stock solution, and diluted extemporaneously in complete medium at the desired concentration. At the first step of selection with 3 μmol/L of temozolomide, 12 clones were isolated from among >50, and were cultured separately or frozen. Two clones were treated with incremental concentrations of temozolomide (3, 5, 10, 20, 30, 60, and 150 μmol/L). At each step of selection, cells were exposed to a higher temozolomide concentration as soon as regrowth was apparent. Briefly, cells were plated in 25 cm² flasks in their usual medium and allowed to attach overnight for drug selection, then the medium was replaced by 0.5 mL of temozolomide-containing fresh medium. After the first 2 hours, temozolomide-free fresh medium was added. 1,3-Bis(2-chloroethyl)-1-nitrosourea (carmustine, Bristol-Myers Squibb) was dissolved in ethanol at 17 mmol/L and carboplatin (Merck, Lyon, France) was diluted in water at 10 mg/mL.

Cell proliferation was determined by using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) test (Sigma) and the methyl blue test. The amount of tetrazolium dye or methyl blue was proportional to the number of viable cells. Briefly, 5 × 10³ cells were plated per well in 24-well polystyrene plates (ATGC, Noisy-le-Grand, France) in culture medium and allowed to adhere overnight. The medium was then replaced with fresh medium containing (or not) increasing concentrations of drug (0.03–150 μmol/L for the parental cells and 10–250 μmol/L for the variants) or the same concentration of the excipient (DMSO or ethanol) as that used for the highest drug dose used for 96 hours. For the MTT assay, the medium was replaced by MTT (2 mg/mL) diluted in medium (80%) for >3 hours and then replaced by DMSO, as previously described (11). For the methyl blue test assay, the surviving adherent cells were fixed with methanol and stained with methylene blue (1% in PBS buffer), and cell-fixed dye was eluted with 0.1 N of HCl (12). Absorbance (A) was measured in an automatic scanning photometer at a wavelength of 550 nm. Each experimental point was done in triplicate. The percentage of live cells was calculated as follows: P = (A in treated cells / A in control cells) × 100. Each assay was done at least thrice. Results are expressed as medians with the range. Statistical significance was determined using the Mann-Whitney U test (P ≤ 0.05).

Metaphases were harvested after a 2.5-hour colchicine block. Chromosome spreads were obtained according to previously described techniques (13). Karyotypes were established after R-banding.

DNA was extracted from frozen cells: samples were digested in extraction buffer (10 mmol/L Tris-Cl, 2 mmol/L EDTA, and 400 mmol/L NaCl) and 1.25% SDS and proteinase K (67 μg/mL) at 37°C overnight. DNA was subsequently treated with RNase (10 mg/mL) for 1 hour at 37°C. NaCl (5 mol/L) was added before precipitation with ethanol. The concentration and molecular size of DNA were estimated using a spectrophotometer (Uvikon 923; Fisher Bioblock Scientific, Illkirch, France) and ethidium bromide-stained agarose gels.

CGH experiments were done according to published protocols (14). DNA was labeled with fluorescein-12-dUTP and normal reference DNA obtained from normal lymphocytes, with Texas red-5-dUTP (Vysis, Rungis, France) using a commercially available nick-translation kit (Vysis). Four hundred nanograms of labeled tumor and reference DNA were coprecipitated with Cot-I DNA, and then denatured. Metaphase cell preparations obtained from normal lymphocytes were denatured and incubated with the probes in a humid chamber at 37°C for 3 days. Slides were washed and mounted in antifade solution. Digital images of the three fluorochromes were acquired using a computer-driven cooled CCD (Sensys, Photometrics, Evry, France) camera mounted on a fluorescence microscope (DMRB, Leica, Germany) equipped with selective single bandpass filters and interfaced to the CGH QUIPS software package (Vysis). Chromosomes stained with 4′,6-diamidino-2-phenylindole were identified by computer-generated reverse DAPI banding. The green to red fluorescence ratio along each chromosome was calculated by the appropriate software. CGH analysis of each tumor was done on 15 metaphases. The threshold values for losses and gains were set at 0.9 and 1.1, respectively. Amplification corresponded to a ratio exceeding 1.5.

A genome-wide resource of 3,342 FISH-mapped, sequenced BAC and PAC clones verified for gene and marker contents were represented as immobilized DNA targets on glass slides for array-based CGH analysis allowing a mean resolution of 1 Mb all along the genome. Each clone was spotted in triplicate on a slide with an Aminosilanne coating (Corning UltraGAPS, NH³⁺) with the Microgrid TAS BioRobotics spotter. These slides were a generous gift from the INSERM Unit U520 (France).

After extraction, 1.5 μg of each test and control DNA sample was digested with DpnII enzyme (Ozyme, Saint Quentin en Yvelines, France) and purified with a QIAquick PCR purification kit (Qiagen, Courtaboeuf, France). They were then labeled by random priming using a Bioprime DNA labeling kit (Invitrogen, Cergy Pontoise, France) with the appropriate cyanine dye (Cy3 or Cy5; Perkin-Elmer, Wellesley, MA). The control and test DNAs were coprecipitated with Cot-1 DNA (Invitrogen), denatured and resuspended in hybridization buffer (50% formamide). Competitive cohybridization was done on CGH array slides preblocked by succinic anhydride/N-methyl-2-pyrrolidinone/borate buffer (Sigma-Aldrich).

After a 24-hour hybridization, slides were washed with SDS and saline citrate, dried, and scanned using a 4000B scan (Axon Instruments, Union City, CA). Image analysis was done with Genepix 5.1 software (Axon Instruments) and processed using software developed at the Curie Institute.

Any BAC with less than two replicates flagged for not fulfilling qualitative spot criteria was excluded. A ratio of <0.8 was considered as a loss, a ratio of >1.2 as a gain, and a ratio of >1.5 as amplification.

FISH was done as previously described (15). Several chromosome-specific biotinylated painting probes (Oncor Appligene, Illkirch, France) were hybridized to metaphase spreads. The biotinylated probes were detected by goat anti-biotin antibody (Vector Laboratories, Burlingame, CA) and fluoresceinated anti-goat rabbit antibody (Biosys, Compiègne, France). Chromosomes were counterstained with propidium iodide. Fluorescent signals were detected using an epifluorescence microscope (DMRB, Leica). Twenty metaphase spreads were analyzed to evaluate signal distribution.

Total RNA was isolated using an RNA extraction kit (Macherey-Nagel, Hoerdt, France) according to the manufacturer's instructions. RNA quality was assessed using an Agilent 2100 Bioanalyser (Massy, France).

For probe preparation, 20 μg of RNA from sensitive or resistant cell lines were reverse-transcribed and labeled using the CyScribe Post-Labeling Kit procedure (Amersham Biosciences, Orsay, France), without modifications. After purification with the CyScribe GFX Purification Kit (Amersham Biosciences), the cDNAs were combined and applied to the arrays (Agilent human 1A oligonucleotide microarray), exactly as recommended by the manufacturer. Slides were scanned in both Cy3 and Cy5 channels with the GenePix 4000 scanner (Axon Instruments) and analyzed by GenePix Pro Software (Axon Instruments). For dye normalization, the LOWESS (locally weighted linear regression) method was applied, using the VARAN software.⁶

cDNA was prepared from each cell line using a combination of random primers (Promega, Charbonnières, France) and Superscript II (Invitrogen) for validation of array results by quantitative PCR. Oligos were designed around intron-exon boundaries for each gene using primer3 software.⁷

In addition to these genes, MGMT expression was also tested at the RNA level using the following primers: 5′-CCTGGCTGAATGCCTATTTC-3′ (forward) and 5′-GATGAGGATGGGGACAGGATT-3′ (reverse). The methylation status of the MGMT promoter was also evaluated, as previously described (16).

SNB-19 parental cells were very sensitive to temozolomide (Table 2). Two clones of resistant variants (SNB-19A4 and SNB-19C1) were isolated at an early stage of drug treatment (3 μmol/L of temozolomide) and then successively exposed to incremental doses of temozolomide (up to 150 μmol/L). After initial killing of a large majority of cells, surviving cells displayed a normal rate of proliferation. They were then submitted to a new selection step up to 150 μmol/L. The IC₅₀ of temozolomide was evaluated by both MTT and methyl blue test techniques, and similar results were obtained. The SNB-19A4 and SNB-19C1 variants showed a 100-fold (P = 0.02) and 55-fold (P = 0.008) stronger resistance to temozolomide than the parental SNB-19 cell line, respectively. After 20 passages in temozolomide-free medium, the temozolomide resistance factor remained constant.

The resistance of both variants to different alkylating agents was compared with that of the parental cells. The IC₅₀ of 1,3-bis(2-chloroethyl)-1-nitrosourea was increased 2.2-fold (P = 0.038) and 1.6-fold (P = 0.037) for SNB-19A4 and SNB-19C1, respectively. The IC₅₀ of carboplatin was increased 4.1-fold (P = 0.12) for SNB-19A4 and remained stable at 3.1 μmol/L for SNB-19C1 (Table 2). Thus, the resistance to temozolomide acquired by these two variants was stable and could be considered specific to temozolomide.

CGH was used to analyze the alterations in SNB-19A4 and SNB-19C1 variants. SNB-19 showed numerous rearrangements (Fig. 1A), as previously published (10). DNA was extracted from resistant cells at the highest temozolomide concentration (150 μmol/L) and hybridized with normal DNA (data not shown). Most of the numerous imbalances observed were shared by the parental cells and the temozolomide-resistant cells lines. To eliminate these shared imbalances, the DNA from each resistant variant was cohybridized with the DNA from the parental SNB-19 cells. We were therefore able to identify alterations in the temozolomide-resistant clones that were absent in the parental cell line. Such alterations concerned the loss of a 2p region involving the 2p16-2p25 bands, loss of partial amplification of the 4p14-4q21 region and loss of amplification of the 16q12-16q22 region (Fig. 1B and C); these last two regions appeared with a normal ratio during hybridization versus control DNA. During a second step, the same cohybridizations were done with DNA extracted at a lower level of temozolomide resistance (3, 30, and 60 μmol/L). These alterations were present as early as the 30 μmol/L dose of temozolomide but were not detected at the 3 μmol/L dose (data not shown). The IC₅₀ of temozolomide for the SNB-19C1 variant selected at 30 μmol/L of temozolomide was the same as that for the variant selected at 150 μmol/L (IC₅₀ = 45 μmol/L).

CGH array analyses were done with the same cell lines, first cohybridized with normal cell DNA (Fig. 2A), then again cohybridized between the DNA of each resistant variant (DNA extracted at 150 μmol/L temozolomide) and the DNA of the parental SNB-19 cells (Fig. 2B and C). These results confirmed those of the conventional metaphase-based CGH analysis. However, the altered regions were more accurately defined (2p16.1-2p25.3, 4p14.4-4q21.22, and 16q12.1-16q22.1) and further loss of amplification of the 1p13.2-1q21.1 region shared by SNB-19A4 and SNB-19C1 was detected.

SNB-19 exhibited a hyperdiploid karyotype with 59 to 61 chromosomes. SNB-19A4 was also hyperdiploid with 54 to 55 chromosomes and SNB-19C1 was hypopentaploid with 102 to 107 chromosomes (data not shown).

FISH was done with chromosomal painting probes for chromosomes 2, 4, and 16 on metaphases because CGH showed shared rearrangements involving these three chromosomes. FISH revealed a derivative chromosome in the parental SNB-19 cell line, that involved chromosomes 2, 4, and 16 with at least five breakpoints (Fig. 3). This derivative chromosome was lost in both SNB-19A4-resistant and SNB-19C1-resistant variants. However, it was difficult to assign the lost chromosomal material found by CGH to the loss of this marker alone because several chromosomes were hybridized to these probes.

We did a comparative transcriptome analysis between parental SNB-19 cells and the two temozolomide-resistant variants. RNA isolated from SNB-19A4 and SNB-19C1 was cohybridized with RNA isolated from SNB-19 parental cells in a dye-swap experiment using Agilent arrays. The search for differential expression evidenced quantitative alterations of 245 genes between SNB-19A4 and SNB-19, and 284 genes between SNB-19C1 and SNB-19. By cross-checking the two lists of genes, 78 genes were found to exhibit similar alterations in the two temozolomide-resistant variants. Differential expression was confirmed by quantitative PCR analysis for 18 out of these 78 genes (Table 2).

Quantitative alterations of gene expression and genomic alterations were confronted. The location of the gene was known in 72 out of the 78 differentially expressed genes (Table 3; Fig. 4). Among them, 11 genes (15%) were located in the rearranged region (Table 3).

In order to test whether MGMT expression might be responsible for part of the resistance mechanism in A4 and C1 variants, we tested MGMT expression at the RNA level, but it was not detectable. Analysis of the methylation status of the MGMT promoter showed that it was fully methylated in all cell lines. Consequently, the MGMT level did not allow us to discriminate SNB-19 from the resistant variants.

Malignant glial tumors are well known to be refractory to chemotherapy. Significant clinical benefits have been obtained with temozolomide, a recently introduced alkylating agent (2). The main purpose of the present study was to generate temozolomide-resistant variants from SNB-19, a malignant glioma cell line, in an attempt to identify the genetic alterations associated with acquired resistance to temozolomide. Several temozolomide-resistant variants were obtained by stepwise selection with incremental doses of temozolomide in the culture medium. Genomic alterations appeared as early as the 30 μmol/L concentration, which roughly corresponds with the mean peak concentration obtained after conventional treatment (i.e., 150 mg/m²; ref. 17).

The resistance of these two variants can be considered specific to temozolomide given the difference in the P value between temozolomide and 1,3-bis(2-chloroethyl)-1-nitrosourea and no cross-resistance was noted with carboplatin, a third alkylating agent also used to treat glioblastoma-bearing patients. This was surprising, because resistance to a given alkylating agent often extends to other alkylating agents and even to other drugs (18, 19). Ma et al. (18) found that tumor cell–acquired resistance to temozolomide was associated with increased MGMT activity and alterations of apoptosis-controlling gene, i.e., increased Bcl2 and BclXL and decreased Bad, Bax, and BclXs. However, none of these genes were identified in our microarray analysis.

Hypermethylation of the MGMT CpG island can cause transcriptional silencing in cell lines unable to repair O⁶-methyl (20). We found no changes on MGMT expression which stay undetectable. The promoter methylation was also unchanged, with both parental cell lines and resistant variants giving 100% methylation (data not shown). A mismatch repair deficiency has not been addressed specifically but such an abnormality is infrequent in gliomas. A differential expression was not found on MGMT genes, mismatch repair genes, or other genes involved in drug resistance mechanisms including base excision repair (21), drug-metabolism, and detoxification (mainly through higher levels of glutathione or glutathione-S-transferase; ref. 22).

Resistance to temozolomide also involves p53 status, Chk1- and Chk2-mediated G₂ checkpoint pathways, and Akt activation. In response to temozolomide, p53 induces cell cycle arrest and apoptosis (4). On the other hand, MGMT inactivation may facilitate AT-GC mutations on p53 (23, 24), explaining the correlation between MGMT inactivation and p53 dysfunction (25, 26). In response to temozolomide, Chk1 induces G₂ arrest preventing from cell death independently of p53 status (5). Akt pathway activation, which is a frequent feature in glioblastoma, bypasses these mechanisms, preventing both G₂ arrest and cell death (6). Again, none of the genes differentially expressed in our microarray analysis involved these pathways.

Gliomas are known to display very rearranged karyotypes with numerous chromosomal imbalances, due to sequential chromosomal endoreduplications and losses of chromosomal material (13). Among the imbalances found in SNB-19, some are frequent in gliomas: gain of 1q32 and gain of the chromosome 4 region bearing the PDGFRA gene (15, 27).

SNB-19 and the two resistant variants were aneuploid and their karyotypes were very complex, but strikingly, there were fewer rearrangements in the resistant variants. The two temozolomide-resistant variants harbored genomic differences with the parental SNB-19 cell line: loss of a 2p region and loss of amplification of the chromosome 4 and 16q regions. This observation does not fit with the commonly accepted hypothesis assuming that the acquisition of drug resistance, evolving through accidental chromosome recombinations, harbors more chromosomal rearrangements (28).

A likely explanation is that temozolomide-resistant variants with less rearranged genomes preexisted in the parental cell line and were selected through exposure to temozolomide. This could be related to cell passaging, but these clones were cultured totally independently after the first 3 μmol/L step. Moreover, two additional variants displaying the same alterations were further isolated (data not shown). The less complex rearrangement pattern of the temozolomide-resistant variants indicates that they may have more in common with an ancestral genotype than the temozolomide-sensitive parental cells, which is compatible with the idea that chemoresistance resulted from the enrichment of preexistent cells.

The expression of 17,000 genes was quantified by microarray analysis. Seventy-eight genes seemed to be differentially expressed in both variants: 15% were located in regions that had sustained losses (11). Among the 78 genes, only two, CLK1 and RAI, were previously associated with a response to cisplatin treatment (29, 30). However, none were associated with resistance to temozolomide. Thirty (19 underexpressed and 11 overexpressed) of these 78 genes seemed to be of specific interest because of their function.

Among the underexpressed genes, some have been described as tumor suppressor genes in solid tumor, IGFBP7 (31), FEZ1/LZTS1 (32), and GJA1 (33), or as inductors of apoptosis, such as CTGF (34). Many are involved in differentiation, notably in the nervous system, or in embryonic pathways such as GJA1 (35), GPM6A (36), FBLN-5 (37), NEDD9 (38), BMPR1B (39), AFP (40), EPHA3 (41), CYP4F2 (42), SEC3L1 (43), HELLS (44, 45), EDNRB (46), and PLEKHB1 (47). PLEKHB1 has recently been found to be down-regulated in grade 4 (glioblastomas) versus grade 1 gliomas (pilocytic astrocytoma). LEF1 is a transcription factor inhibiting the Wnt pathway, a key regulator of cell development and of the stem cell pool (48).

Several overexpressed genes in our temozolomide-resistant variants are involved in developmental pathways such as FZD2 (implicated in the Wnt pathway; ref. 48), or in stem cell self-renewal and proliferation such as LIF (49, 50), ELL (51), and CNN1 (52).

Thus, a group of differentially expressed genes in the SNB-19 temozolomide-resistant variants points to a shift towards a cancer stem cell–like genotype. The concept of a cancer stem cell was recently explored in human brain tumors and cell lines (53, 54). Tumor stem cells retain the ability to self-renew and might explain why many cancers relapse under conditions in which most tumor cells seem to be killed following therapy (54). An emerging model proposes that tumor stem cells are naturally resistant to chemotherapy through their quiescence, their capacity for DNA repair and increased ABC transporter expression (55). In conclusion, our cytogenetic analyses suggest that acquired temozolomide resistance in this malignant glioma model involves the selection of preexistent resistant cell variants. The gene expression profile of temozolomide-resistant variants displayed a complex picture with decreased expression of genes involved in differentiation and increased expression of genes known to be expressed in stem cells. Altogether, these data suggest that acquired temozolomide resistance could have resulted from the selection of less differentiated preexistent resistant cells with stem cell–like characteristics in the parental tumor.

We thank Olivier Delattre for his generous gift of the CGH array slides and for granting access to his analysis platform; Dr. Jean-Gabriel Judde for critical reading of the manuscript; Dr. François Apiou, Anne-Marie-Dutrillaux, and Michèle Gerbeault for their technical assistance in karyotype, FISH, and CGH analyses; and Lorna Saint Ange (Institut Gustave-Roussy) for editing.