Background
Temozolomide (TMZ), an imidazotetrazine derivative of the alkylating agent dacarbazine, is a first-line drug for the treatment of patients with glioblastoma. However, the TMZ efficiency is quite modest, with median overall survival ranging 9.4–19.0 months for radiotherapy combined with TMZ versus 7.3–17.1 months for radiotherapy alone [
1]. TMZ is also used in the treatment of brain metastases, melanoma, lymphomas, refractory leukaemia, neuroendocrine tumours, pituitary tumours, Ewing’s sarcoma, primitive neuroectodermal tumours, lung cancer and other tumours [
2]. Most tumour cells are intrinsically resistant or rapidly acquire resistance to TMZ at pharmacotherapeutic concentrations [
3‐
6]. Long-term TMZ treatment of glioblastoma cells induced profound changes in heterochromatin organization and DNA methylation [
7], transcriptome [
8‐
12], proteome [
13,
14], kinome [
15], and metabolome [
8,
10], remodeling of the entire electron transport chain and activation of oxidative stress responses [
16,
17]. These changes impacted morphology, proliferation, adhesion, migration, invasion, and drug cross-resistance in a versatile manner [
7,
8,
14,
18‐
23]. Such a complex phenotype adaptation certainly indicates intricate cellular and molecular defense mechanisms against TMZ. Additionally, the versatile phenotype responses to long-term TMZ treatment (Table
1) may point to the TMZ-promoted genome changes, which affect the organization and functionality of the genetic network (gene content, RNA and protein expression and their interaction). In fact, an acquisition of chemotherapy resistance is generally accompanied by genome evolution and, conversely, chromosomal instability (CIN) correlates with (multi)drug resistance [
24‐
34].
Table 1
Long-term TMZ treatment of tumour cells results in versatile phenotype responses
A172 | No change | | | | | 100 μM/1 mo | |
C6 | No change | ↓ | | ↓ | ↓ in soft agar | 100 μM/1 mo in vitro or 50 mg/kg/10 injections in vivo | this study |
D54 | Changed | ↓ | ↑ G0/G1 ↓ G2/M | ↑ | | up to 0.5 mM/ 5 or 10 mo | |
CSC | | ↓ | | | ↓ in vivo | | |
HEK293 derivatives | No change | ↓ | | | ↓ in soft agar | up to 120 μM/ 3 mo | |
HeLa derivatives | No change | No change or ↑ | | | ↑ in soft agar | up to 120 μM/ 3 mo | |
Hs683 | | ↓ | | ↓ | ↓ in vivo | up to 1 mM/ 10 mo | |
LN-308, LNT-229, LN-18 | No change | ↓ | No change | | | up to EC50/6 mo | |
T98G | | ↓ | | No change | No change in vivo | up to 1 mM | |
T98G | No change | No change | | No change | ↑ in soft agar | 100 μM/1 mo | this study |
U87 | Changed | | | ↑ | | up to IC50 = 150 µM/3 weeks | |
U251 | | | ↑ G2/M | | | | |
U251 | No change | ↓ or ↑ | | ↓ or ↑ | No change in soft agar | up to 100 μM/ 5 or 10 weeks | this study |
U373 | | ↑ | | ↑ | ↑ in vivo | up to 1 mM | |
CIN refers to the rate of gain or loss of whole chromosomes and portions of chromosomes, whereby the rate is defined as cell-to-cell variability or variability between cellular populations [
35]. The dynamic numerical and structural chromosomal aberrations (genome chaos) result in profound alterations in gene expression, reprogramming of metabolic and signaling pathways and the generation of biochemical/phenotype diversification of cancer cells. Long-term drug-treated cells demonstrate transcriptomic and proteomic changes, and differ from parental cells at the molecular and cellular levels [
26,
30]. Despite extensive studies, the role of CIN in the generation of TMZ-driven phenotype diversity and TMZ-based therapeutic failure has been poorly addressed.
Here, we characterized the genome-phenotype evolution of long-term TMZ-treated glioblastoma cell lines. TMZ treatment influenced genomic stability and phenotype diversity in a cell type-dependent manner by selecting resistant genotype-phenotype variants or generating novel versatile phenotypes by promoting CIN. Our data indicate that in addition to the reported TMZ-driven hypermutation phenotype [
36‐
38], TMZ-instigated changes in genome stability and heterogeneity may contribute to the versatile phenotypic responses of tumour cells.
Discussion
In this study we characterized genome-phenotype changes of long-term TMZ-treated glioblastoma cell lines and found that TMZ may either increase or reduce genomic diversity (CCA/NCCAs) and tumour cell aggressiveness. An increase of resistance to TMZ re-challenge seems to be the only fundamental common and predictable trait intrinsic to all long-term TMZ-treated cells; all other phenotype responses were versatile (Table
1). Our data indicate that changes in genome stability and diversity may be responsible for individual and heterogeneous phenotypes of long-term TMZ-treated cells. It is worth emphasizing that U251TMZ1 and U251TMZ2 cell lines, established by parallel selection of the same parental cell line with the same chemotherapy agent under similar treatment conditions with the only difference in the duration of treatment (10 versus 5 weeks), underwent individual genomic and phenotypic evolution. The development of a heterogeneous range of drug-resistant lines with individual genomic and/or phenotypic changes from the same cell line, treated with the same chemotherapy agent (e.g., cisplatin, puromycin) was reported previously [
3,
31].
The therapy-driven glioblastoma genome evolution was scarcely reported previously. An analysis of primary cell cultures established from three surgery glioblastoma specimens of the same patient (primary specimen and two consecutive recurrences after lomustine and TMZ therapy) demonstrated a distinct subclonal architecture, abnormalities in karyotypic pattern, and rates of proliferation and migration [
46]. Extending research on additional matched primary and recurrent glioblastomas, authors revealed that therapy either increased chromosomal aberrations in some cases that correlated with relatively short overall survival or reduced genome diversity in other cases and these patients showed a much longer overall survival [
46]. Recent sequencing of primary and TMZ-treated recurrent gliomas showed the TMZ-driven amplification of mutation heterogeneity (hypermutation phenotype) in IDH1-mutant but not IDH1-wild-type astrocytic gliomas [
36‐
38]. High levels of MGMT methylation and intrinsic or acquired mutations in the key MMR genes and/or MGMT were associated with hypermutation phenotype [
36]. However, these studies were primarily focused on alterations in DNA sequence rather than on CIN.
The resistance acquisition to TMZ was widely attributed to O
6-methylguanine-DNA methyltransferase (MGMT). Despite a relatively low proportion of the TMZ-driven cytotoxic O
6-methylguanine lesion formation (5 %), the methylated promoter of
MGMT was considered one of the most robust predictor of TMZ response with inverse correlation [
47,
48]. However, this generally good correlation between
MGMT methylation and TMZ treatment response was recently challenged. The TCGA Research Network reported that
MGMT promoter methylation could serve as a predictive biomarker only in the glioblastoma classical subtype but not in the other subtypes (mesenchymal, proneural or neural) [
49]. Moreover, clonal analysis of glioblastoma samples demonstrated inter-tumor variability in
MGMT promoter methylation and MGMT protein expression levels, which were inconsistent with TMZ responses [
6]. Similarly, no correlation between the TMZ sensitivity and
MGMT promoter methylation, mRNA or protein expression was revealed for eleven diffuse large B cell lymphoma cell lines [
4]. Here we found that long-term TMZ-treated cells reduced sensitivity to TMZ re-challenge without changing MGMT mRNA or protein expression levels. On the other hand, previous reports based on transcriptome analysis elucidated that complex individual genetic networks rather than a specific common mechanism conferred a different TMZ sensitivity [
4]. Furthermore, the TMZ-resistant variants of Hs683, U87, and LNZ308 cell lines demonstrated individuality in global miRNA expression, and the integrative miRNA/mRNA network analysis revealed obvious differences in the genetic network in comparison to control cells [
12]. A measurement of global kinase activity of five TMZ resistant cell lines revealed no common kinase-driven pathway of TMZ resistance, and two TMZ resistant lines demonstrated extreme kinomic activity differences in comparison to control cells [
15]. Altogether, adaptation of tumour cells to long-term TMZ cytotoxicity and genotoxicity is associated with profound diverse changes in the transcriptome, proteome, kinome and metabolome [
8‐
15], the versatile phenotype responses (Table
1), involvement of many proteins/miRNAs (Additional file
8: Table S8) (see also the recent large synthetic lethal screens for “TMZ-sensitizing genes”) [
9,
50] and DNA repair pathways [
51].
The cancer stem cell hypothesis postulates a significant role of glioblastoma cancer stem cells (GSC) in therapy resistance and tumour recurrence. However, a recent study showed that clones of GSC had distinct tumourigenic potential that was determined by their genetic diversity rather than expression levels of different GSC-associated markers (CD133, CD15, A2B5 or CD44) [
52]. Moreover, although TMZ treatment induced conversion of non-GSC into GSC both in vitro and in vivo [
53], the majority of patient-matched GSC and non-GSC cultures (25 tested) had a similar TMZ responsiveness and in some cases GSC were even more sensitive [
54]. These studies highlight the primary importance of genetic heterogeneity in tumorigenic potential of CSC-associated populations, and furthermore point to dynamic plasticity of tumor cells under TMZ therapy and no superiority of GSC over bulk tumor population in TMZ resistance.
There are approximately twenty current clinical studies using mTOR inhibitors for the treatment of gliomas [
55]. However, phase II studies with recurrent glioblastoma reported no efficacy of TEM in the combination with TMZ, sorafenib, bevacizumab, or erlotinib [
55]. Although we used a clinically relevant TEM concentration [
56], the discrepancy between cell culture responses (Fig.
7b) and patient responses is obvious. It is worth noting that TEM is also able to induce/promote CIN in tumor and normal cells [
57]. A targeted therapy failure in patients with recurrent glioblastoma after radiotherapy with TMZ [
58] highlights the necessity to lower the evolutionary potential of a tumour and constrain its dynamics by directing efforts at reducing tumour population diversity, at potentiating the immune system and homeostasis of the individual.
In conclusion, our current data improve the knowledge on the TMZ-instigated genome evolution and highlight the primary importance of genetic instability in chemotherapy failure as the more different combinations of molecular mechanisms exist within a cancer cell population, the more likely a population adapts to drug cytotoxicity/genotoxicity. TMZ treatment-associated changes of the genetic network (gene content, RNA and protein expression and their interaction), which are governed by changes of the genome context (number and structure of chromosomes and their nuclear topology) may offer an explanation for why the versatile and opposite phenotype responses of long-term TMZ treated tumor cells were observed in different studies (Table
1). Although our study is limited to the use of established glioblastoma cell lines, our results are consistent with a recent report on evolution of low-grade gliomas to aggressive high-grade glioblastoma in 6 of 10 patient cases due to an increased mutation load upon TMZ therapy [
38]. Our results and the latter study suggest that the therapeutic promotion of excessive genetic instability/heterogeneity is a double-edged sword: while the primary response in the form of increased overall survival will be positive, the price for moderate inhibition of tumour growth will be changes in the genomic landscape, tumour subclonal architecture, and, eventually, promotion of cancer evolution, which ultimately impacts the therapeutic management of recurrence.
Methods
Cell cultures
Human U251 (Bank of Cell Lines from Human and Animal Tissues, R.E. Kavetsky Institute of Experimental Pathology, Oncology and Radiobiology, Kyiv, Ukraine), T98G (ATCC) and rat C6 (Pirogov Russian State Medical University, Moscow, Russia) glioma cell lines were grown in DMEM (HyClone, Thermo Scientific, UK) supplemented with 10 % fetal bovine serum (FBS, HyClone) and 100 µg/ml penicillin/100 u/ml streptomycin (Sigma, USA) in an environment of 95 % air/5 % CO2. U251, T98G, and C6 cell lines are isocitrate dehydrogenase 1 (IDH1)-wild-type. U251 is MGMT-negative; T98G is MGMT-positive.
Pharmacological agents
Temozolomide (TMZ, Sigma), Temsirolimus (TEM, Abcam Biochemicals, USA) and U0126 (Abcam Biochemicals) were dissolved in DMSO to a concentration of 100 mM. The final DMSO concentration in the culture medium did not exceed 0.4 %. Stock solutions of all drugs were stored at −20 °C.
TMZ treatment of glioblastoma cells in vitro
U251, T98G, and C6 glioblastoma cell lines were treated with DMSO or TMZ (Sigma) twice with 25 μM, twice with 50 μM and then with 100 μM TMZ twice per week during 5 weeks (U251TMZ2, T98GTMZ and C6TMZ) or 10 weeks (U251TMZ1), followed by several weeks of washout (in the TMZ-free medium) before in vitro tests. DMSO did not exceed 0.1 % of the culture medium.
TMZ treatment of C6 cells in vivo
The animals were kept in accordance with the Guidelines on Laboratory Practices adopted by the Ministry of Health of the Russian Federation (Order 267, 19 June 2003). The protocol stipulating animal treatment was approved by the Ethics Committee of N. I. Pirogov Russian State Medical University, and all rules and regulations were followed during experimentation on animals. Glioma modeling was performed by the intracerebral stereotactic implantation (Leica stereotactic device, USA) of C6 cells (5 × 10
5) into the striatum region of ketamine-anesthetized adult female Wistar rats as described previously [
59]. Rats with C6 glioma received 20 % DMSO (n = 1, C6R1) or TMZ (n = 1, C6R2TMZ) injected intraperitoneally (i.p.) three times per week at a dose of 50 mg/kg. Rats were sacrificed after 10 injections. Gliomas were aseptically harvested, mechanically disaggregated, and a cell suspension was seeded into adherent dishes. Cells were used at the passages 3–10 for analysis.
Conventional cytogenetics
Chromosome samples were prepared as described previously [
23]. 200 metaphase plates were calculated for distribution of chromosome across cells. 20 metaphases (U251 and T98G derivatives) or 10 metaphases (C6 derivatives) were described for chromosome abnormalities, according to the International System for Human Cytogenetic Nomenclature (ISCN 2013). Clonal chromosome aberrations (CCAs) were defined as aberrations found at least in two cells among examined metaphases, whereas non-CCAs (NCCAs) as aberrations detected in only one cell. The frequency of NCCAs in a cell line was calculated by dividing the number of metaphases displaying NCCAs to the total number of examined metaphases (×100 %). Only structural NCCAs were considered.
Array comparative genome hybridization (aCGH)
A total DNA was isolated using NucleoSpin Blood DNA extraction kit (Macherey–Nagel, Germany) according to the manufacturer’s instructions. To analyze copy number alterations (CNAs), aCGH was performed as detailed previously [
23]. Human and rat cell lines were analyzed on the CytoSure Aneuploidy Array 15 k (Oxford Gene Technologies, UK) and 180 K microarrays (Agilent Technologies, USA), respectively. Image analysis of human and rat samples was carried out with CytoSure Analysis Software (Oxford Gene Technologies) and Agilent CytoGenomics Edition 2.9.2.4, respectively.
Cell proliferation in a high and low-glucose medium
Cells were seeded onto 6 cm dishes at a density of 5 × 104 (U251 and T98G derivatives) or 1 × 104 (C6 derivatives) and grown in high-glucose (4.5 g/L) or low-glucose (1 g/L) DMEM with 10 % FBS. On the 7th day of seeding, cells were harvested, incubated with trypan blue, and calculated using hemocytometer. Experiments were repeated at least three times.
Cell viability test
U251, T98G (5 × 10
4 cells) and C6 (1 × 10
4 cells) derivatives were seeded onto 6 cm dishes and incubated overnight. The cells were treated for 7 days with a single dose of TMZ (20 and 100 µM), TEM (2 µM), U0126 (5 µM) or DMSO. Experiments were repeated at least three times. Cell viability was evaluated by trypan blue exclusion assay instead of metabolically-based MTT or ATP assays, which are prone to over/underestimate cell viability under cytotoxic stress [
45].
5 × 10
3 cells were placed in 1.5 ml of 0.35 % low gelling temperature agarose (Gibco, Life Technologies, USA) with DMEM supplemented with 10 % FBS. 0.35 % top agarose was poured on 1.5 ml of solidified 0.5 % base agarose/10 % FBS/DMEM. Cells were seeded in triplicates in a 35-mm dish and grown at 37 °C for 21 days to allow colony formation. Colonies were visualized by staining with 0.005 % crystal violet, photographed, counted using OpenCFU software [
60], and expressed as the means of triplicates of four independent experiments.
Scratch wound healing assay
Using a P200 pipette tip, the scratches were made by scraping across the confluent cell monolayer. Pictures were taken at 0 and 16 h (C6 derivatives) or 24 h (U251 and T98G derivatives) and automated image analysis was carried out using TScratch software [
61] to avoid any potential bias in quantifying an extent of migration. At least twelve wound healing areas for each cell line were photographed and analyzed to take into account the differences in cell density and widths of scratches. The per cent of wound area closure was calculated taking open wound area at 0 h for 100 %.
Cell invasion assay
A 24-well tissue culture plate-based Chemicon cell invasion assay (QCM ECMatrix 550, Millipore, USA) was performed according to the manufacturer’s protocol. 2 × 105 cells were seeded to the inserts. After 24 h, five fields of invaded cells in each well were randomly photographed and counted manually. Test was performed two times.
Real time quantitative PCR
Total RNA was extracted from cell lines using TRI Reagent (Sigma, #T9424) according to the manufacturer’s recommendations. Equal amounts of total RNA (5 μg for 20 μl reaction mixture) were transcribed into cDNA with random hexamer primers and RevertAid Reverse Transcriptase (Thermo Scientific, #EP0441). Twofold diluted cDNA and gene specific primers were mixed with Maxima SYBR Green qPCR Master Mix (2X) (Fermentas, #K0251) according to the manufacturer’s recommendations. qRT-PCR was run in triplicates on CFX96 RT-PCR Detection System (Bio-Rad). The amplification procedure of target genes was as follows: initial denaturing step at 95 °C for 10 min, followed by 40 cycles of denaturation at 95 °C for 15 s, annealing at 59 °C for 30 s and extension at 72 °C for 30 s. Melting curve analysis was performed to confirm amplification specificity. To calculate the relative gene expression ratios (fold-change), CT method (also known as the 2−ΔΔCT method, expressed as ratios relative to control values after normalization to the internal control TBP—TATA-binding protein) was applied. CT values were derived using Bio-Rad CFX Manager 3.1.
Primers
TBP Forward: TGCACAGGAGCCAAGAGTGAA; Reverse: CACATCACAGCTCCCCACCA; CD133 Forward: CGTGGATGCAGAACTTGACAACGT; Reverse: ATACCTGCTACGACAGTCGTGGT; SOX2 Forward: GCCGAGTGGAAACTTTTGTCGGA; Reverse: CGTGTACTTATCCTTCTTCATGAGCGTC; OCT4 Forward: GGAGAAGGAGAAGCTGGAGCA; Reverse: GGCAGATGGTCGTTTGGCTGAATA; NANOG Forward: GTCTGGACACTGGCTGAATCCT; Reverse: CTCGCTGATTAGGCTCCAACCAT; MGMT Forward: CCTGGCTGAATGCCTATTTCCACCA; Reverse: GGATGAGGATGGGGACAGGATTGC.
Western blot analysis
Total cell lysates were analyzed as described earlier [
23]. The following antibodies were used: mouse anti-MGMT (Novus Biologicals; #NB100-168), rabbit anti-PTCH2 (Cell Signaling; #2464), rabbit anti-ASK1 (Cell Signaling; #8662), rabbit anti-MDM2 (Thermo Fisher Scientific Pierce; #PA5-11353), rabbit anti-p53 (Millipore; #04-1083), rabbit anti-phospho-p53 (Ser6) (Millipore; #04-540), rabbit anti-PARP (Cell Signaling; #9542), rabbit anti-ERK1/2 (Millipore; #06-182), rabbit anti-phospho-ERK1/2 (Thermo Scientific Pierce; #MA5-1574), rabbit anti-AKT1 (Millipore; #07-416), mouse anti-phospho-AKT1 (Santa Cruz; sc-52940), mouse anti-β-actin (Sigma-Aldrich; A1978), the Epithelial-Mesenchymal Transition antibody sampler kit (Cell Signaling; #9782), anti- rabbit (Cell Signaling, #7074) and anti-mouse (Cell Signaling, #44209).
Statistical analyses
A two-sided t test was used to calculate the significance values (Statistica 10 Software, USA). Data showing p values of *P < 0.05, **P < 0.01, and ***P < 0.001 were considered significant. All experimental data are reported as mean and the error bars represent the experimental standard error (±standard deviation, SD).
Authors’ contributions
Conception and design: AAS, VPB; Collection and assembly of data: AAS, SVA, KVK, DOM, VPB, NLH, OAK; Data analysis and interpretation: AAS, SVA, DOM; Contribution of reagents, materials, and analysis tools: KVK, YSV, VPC, SSA, VVD. AAS wrote the paper. All authors read and approved the final manuscript.