TERT and DNMT1 expression predict sensitivity to decitabine in gliomas

Abstract

Background

Decitabine (DAC) is an FDA-approved DNA methyltransferase (DNMT) inhibitor that is used in the treatment of patients with myelodysplastic syndromes. Previously, we showed that DAC marks antitumor activity against gliomas with isocitrate dehydrogenase 1 (IDH1) mutations. Based on promising preclinical results, a clinical trial has been launched to determine the effect of DAC in IDH-mutant gliomas. The next step is to comprehensively assess the efficacy and potential determinants of response to DAC in malignant gliomas.

Methods

The expression and activity of telomerase reverse transcriptase (TERT) and DNMT1 were manipulated in patient-derived IDH1-mutant and -wildtype glioma lines, followed by assessment of cell proliferation with DAC treatment alone or in combination with telomerase inhibitors. RNA sequencing, Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment, and correlation analysis were performed.

Results

IDH1-mutant glioma tumorspheres with hemizygous codeletion of chromosome arms 1p/19q were particularly sensitive to DAC and showed significant inhibition of DNA replication genes. Our transcriptome analysis revealed that DAC induced expression of cyclin-dependent kinase inhibitor 1A/p21 (CDKN1A), along with downregulation of TERT. These molecular changes were also observed following doxorubicin treatment, supporting the importance of DAC-induced DNA damage in contributing to this effect. We demonstrated that knockdown of p21 led to TERT upregulation. Strikingly, TERT overexpression increased DNMT1 levels and DAC sensitivity via a telomerase-independent mechanism. Furthermore, RNA inhibition (RNAi) targeting of DNMT1 abrogated DAC response in TERT-proficient glioma cells.

Conclusions

DAC downregulates TERT through p21 induction. Our data point to TERT and DNMT1 levels as potential determinants of response to DAC treatment.

Key Points.

1. DAC-induced TERT repression is mediated by p21.

2. TERT and DNMT1 confer DAC sensitivity in gliomas.

Importance of the Study.

Our work shows that epigenetic therapy with low doses of DAC results in significant responses particularly in gliomas expressing high levels of TERT and DNMT1. Given that DAC is currently being tested in a clinical trial setting, molecular characterization of response to DAC may ultimately allow us to stratify glioma patients to maximize the therapeutic efficacy of DNA demethylating agents.

Adult diffuse gliomas are largely divided into 3 major molecular groups: isocitrate dehydrogenase (IDH)-mutant gliomas with hemizygous codeletion of chromosome arms 1p/19q (oligodendrogliomas, IDH-mutant-codel), IDH-mutant gliomas with intact 1p/19q (astrocytomas, IDH-mutant-noncodel), and IDH-wildtype gliomas. In addition to molecular-based classification, glioma DNA methylation subtypes have been classified based on DNA methylation profiles.¹ Because 2-hydroxyglutarate accumulation by mutant IDH induces profound epigenetic alterations, IDH-mutant gliomas are characterized by G-CIMP (the glioma CpG island methylator phenotype).² A consequence of the hypermethylation is the epigenetic silencing of tumor-suppressor genes (TSGs) in many types of cancers, including gliomas.³

To target cancer-specific epigenetic aberrations, DNA methyltransferase inhibitors (DNMTi) decitabine (DAC) and 5-azacytidine emerged as potent anticancer agents for hematological malignancies and solid tumors.^4,5 The mode of action of DNMTi involves incorporation into DNA and triggering subsequent degradation of DNMTs by covalent trapping of the enzymes.⁶ Based on this mechanism, the drugs effectively induce DNA hypomethylation and reactivate TSGs.⁷ Interestingly, DAC upregulates p21 in a DNA damage-dependent manner.⁸

Transient DAC treatment at low doses triggers a sustained antitumor effect in leukemia and breast cancer cells.⁹ In preclinical glioma models, the efficacy of DNMTi has been shown in both IDH-mutant oligodendroglioma and astrocytoma models.^10,11 Intriguingly, JHH-273, the IDH1-mutant astrocytoma model, appeared to require long-term treatment with serial implantation to inhibit tumor growth.¹¹ Furthermore, 5-azacytidine results in reduced expression of the platelet-derived growth factor receptor A (PDGFRA) oncogene in IDH-mutant cells due to the reestablishment of a chromatin structural loop that insulates PDGFRA.¹² Importantly, 5-azacytidine increases the therapeutic effect of temozolomide in preclinical models of IDH-mutant gliomas.¹³ In light of these promising preclinical data, a clinical trial has been initiated to determine the maximum tolerated dose of ASTX727, a combination of decitabine and a cytidine deaminase inhibitor (cedazuridine), at Massachusetts General Hospital (ClinicalTrials.gov NCT03922555). Thus, an improved understanding of the mechanisms of response to DNMTi in gliomas will be important to select patients who will most likely benefit from the drug.

As previously reported in several types of cancers,^14,15 our study demonstrates that DAC led to the substantial reduction of TERT expression in IDH1-mutant-codel lines with TERT promoter (TERTp) mutations. But the underlying mechanism of how DNMTi reduces TERT expression remains incompletely understood. Here, we elucidate a mechanism of DAC-mediated TERT repression. In addition, we demonstrate that DAC response correlates with TERT and DNMT1 levels in preclinical models of diffuse IDH-mutant gliomas and IDH-wildtype glioblastomas (GBMs). Since TERTp mutations are present in almost all IDH-mutant-codel gliomas and up to 80% of IDH-wildtype GBMs, and are absent in IDH-mutant-noncodel gliomas,¹⁶ DNMTi may provide antitumor efficacy in a wide range of gliomas.

Materials and Methods

Cell Lines and Cell Culture

The following patient-derived IDH1-mutant and -wildtype glioma lines were used: TS603 and TS667 from Memorial Sloan Kettering Cancer Center; NCH612 and NCH1681 from Heidelberg University Hospital; SU-AO3 from Stanford University; and L0627, L0615, L0616, L0104, L0125, and L0512 GBM tumorspheres from San Raffaele Scientific Institute (kindly provided by Dr Rossella Galli). Cell lines were regularly subjected to panel sequencing analysis¹⁷ and immunoblotting to confirm the expression of mutant IDH1 (R132H). Glioma lines were grown in NeuroCult (StemCell Technologies) with 2 µg/mL heparin sulfate (Sigma), 20 ng/mL epidermal growth factor (StemCell Technologies), and fibroblast growth factor 2 (StemCell Technologies). Immortalized human astrocytes (IHAs)¹⁸ were maintained in DMEM (Sigma) supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin streptomycin. All cells were cultured in a 5% CO₂ humidified incubator at 37°C.

Plasmids and Reagents

For TERT overexpression and short hairpin (sh)RNA-mediated DNMT1 depletion, pBABE-puro-hTERT-HA, pSuper-retro-puro-shRNA-DNMT1i (CDS), and -shRNA-DNMT1i (3′ untranslated region [UTR]) (Addgene plasmids #1772, #24950, and #24951, respectively) were used. As a control, scramble shRNA (Addgene plasmid #30520) was used. The p21 shRNA plasmid kit (TL305469) was purchased from Origene. Viral transduction was performed as previously described.¹⁹ The following drugs were used for in vitro treatments: BIBR1532 (Selleck Chemicals, S1186), pifithrin-α (Selleck Chemicals, S2929), doxorubicin (Selleck Chemicals, S1208), decitabine (Sigma, A3656), and 6-thio-dG (Cayman Chemicals, 789-61-7).

TERT Promoter Mutation Analysis

For screening of TERT promoter mutations, PCR amplification was performed in a 25 µL reaction mixture containing 12.5 µL 2× KAPA HiFi HotStart ReadyMix (Roche), 1.25 µL DMSO, 10 µM each primer, and 100 ng DNA. Reactions were incubated for 15 minutes at 95°C followed by 45 cycles of 30 seconds at 95°C, 30 seconds at 62°C, and 60 seconds at 72°C. As previously described,²⁰ PCR primers for the proximal TERT promoter are as follows:

Forward: 5′-M13-GTAAAACGACGGCCAGTCACCCGTCC TGCCCCTTCACCTT-3′ Reverse: 5′-GCACCTCGCGGTAGT GG-3′

Immunoblotting

The following primary antibodies were used for the immunoblots: anti-hemagglutinin (HA) (Cell Signaling, 3724), anti-DNMT1 (Cell Signaling, 5032), anti-glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (Cell Signaling, 2118), anti-beta actin (ACTB) (Cell Signaling, 4967), anti-EGF receptor (EGFR; Cell Signaling, 2232), anti-p21 (Cell Signaling, 2947), anti–IDH1 R132H (Dianova, DIA-H09), and anti-ACTB (Sigma, A3854).

MTT Assay and Cell Proliferation Measurement

Cells were cultured in proliferation conditions and seeded at 20 000 (IDH1-mutant), 10 000 (IDH1-wildtype), or 1000 (IHA) cells per well in 96-well plates. To determine the effect of DAC treatment on cellular proliferation, the media containing DAC was replaced every other day. The cells were then incubated with 10 µL of MTT reagent (Trevigen) for 2 hours, and solubilized with 100 µL of detergent for 4 hours at 37°C. Absorbance at a wavelength of 570 nm was measured, and the values were converted to percentage using DMSO-treated cells as a control. For measuring cellular proliferation, TS603 cells were seeded at 100 000 cells per well in laminin-coated 6-well plates and treated with 500 nM DAC. The cells were counted using a hemocytometer 10 days later.

Soft Agar Anchorage-Independent Growth Assay

To measure cellular anchorage-independent growth, 100 000 cells were seeded between 0.4% (top) and 0.8% (bottom) agar (Lonza). DAC (500 nM) was added to the media for 7 days. At 1 to 2 months, 6-well plates were stained with 0.005% crystal violet and imaged using the Bio-Rad ChemiDoc MP imaging system.

Quantitative Real-Time PCR

Total RNA was isolated using the RNeasy Plus Mini Kit (Qiagen), and converted to cDNA using the iScript cDNA Synthesis Kit (Biorad). Quantitative real-time (qRT) PCR reactions were performed with SYBR Green PCR Master Mix (Thermo Fisher Scientific) using LightCycler 480 II (Roche). The 2^−ΔΔCt method was used to calculate fold change by using GAPDH as an internal control.

p21 Forward: 5′-GAGGCCGGGATGAGTTGGGAGGAG-3′ p21 Reverse: 5′-CAGCCGGCGTTTGGAGTGGTAGAA-3′ TERT Forward: 5′-GCGTTTGGTGGATGATTTCT-3′ TERT Reverse: 5′-CAGGGCCTCGTCTTCTACAG-3′ DNMT1 Forward: 5′-GAGGAAGCTGCTAAGGACTAGTTC-3′ DNMT1 Reverse: 5′-ACTCCACAATTTGATCACTAAATC-3′ DNMT3B Forward: 5′-TACACAGACGTGTCCAACATGGGC-3′ DNMT3B Reverse: 5′-GGATGCCTTCAGGAATCACACCTC-3′ GAPDH Forward: 5′-TGGGGAAGGTGAAGGTCGG -3′ GAPDH Reverse: 5′-CTGGAAGATGGTGATGGGA-3′

Telomerase Activity Measurement

Telomerase activity was measured using the TRAPeze Telomerase Detection Kit (Millipore). Cells were lysed in 1× Chaps lysis buffer according to the manufacturer’s instructions. The protein (1.5 μg) from each cell lysate was used for the telomeric repeat amplification protocol (TRAP) PCR reaction. The amplified TRAP products were loaded into a high sensitivity DNA chip (Agilent), and the chip was run using Agilent 2100 Bioanalyzer (Agilent). The electropherogram was acquired and analyzed using 2100 Expert software version B.02.10.SI764. The TERT activity of control group was determined as 100% after normalization to an internal control.

RNA Sequencing

Cells were treated with 500 nM DAC for 7 days prior to total RNA isolation. RNA quantity and quality were validated using 4200 TapeStation (Agilent), and samples with an RNA integrity number greater than 9.8 were used for sequencing. RNA-seq libraries were created with TruSeq Stranded mRNA Library Prep Kit (Illumina). Sequencing was performed on the Illumina HiSeq 2000 using single-read 50 bp sequencing (DKFZ Genomics Core Facility). The data are available in the Gene Expression Omnibus (dataset GSE153443).

RNA Sequencing Analysis

Quality control of fastq files was confirmed with FastQC software (http://www.bioinformatics.babraham.ac.uk/projects/fastqc). The STAR aligner was used to map the reads to hg19. FeatureCounts was used to count hg19-annotated genes. Normalization of the raw counts and differential expression analysis was performed using DESeq2 package. Upstream regulator analysis was performed using Qiagen Ingenuity Pathway Analysis (IPA). Overlapping genes were identified using VENNY 2.1 (https://bioinfogp.cnb.csic.es/tools/venny/index.html). Gene set enrichment analysis (GSEA) was carried out with help of WebGestalt²¹ (q < 0.05). Heatmaps, dot, correlation, and volcano plots were generated using the pheatmap, clusterProfiler, ggpubr, and EnhancedVolcano packages, respectively, in R. TERT, DNMT1, and DNMT3B expression data were obtained from the GlioVis data portal.²²

Statistical Analysis

An unpaired Student’s t-test or one-way ANOVA followed by Dunnett’s multiple comparisons was used to compare differences between 2 groups or among multiple experimental groups, respectively. Data are graphed as mean ± SEM or SD. GraphPad Prism software v8.3 was used for all significance calculation. P-values below 0.05 were considered significant.

Results

DAC Treatment Reduces DNMT1 Levels, Cellular Proliferation, and Anchorage-Independent Growth of IDH1-Mutant-Codel Gliomas

For our study, we tested 4 patient-derived glioma tumorspheres (TS603, NCH612, SU-AO3, and NCH1681) that harbor endogenous IDH1 R132H mutation (Fig. 1A). As previously reported,¹⁰ an epigenetically targeted lower dose (500 nM) of DAC reduced DNMT1 protein levels, cell proliferation, and anchorage-independent growth of the IDH-mutant-codel patient-derived line TS603 (Fig. 1B–D). To explore DAC-mediated growth inhibition in other IDH1-mutant-codel gliomas, we utilized NCH612 (partial deletion of 1p/19q) and SU-AO3 (complete 1p/19q codeletion) tumorsphere lines (Fig. 1A). As expected, DAC treatment led to DNMT1 protein depletion in both lines (Fig. 1E). Furthermore, 7-day exposure to 500 nM DAC was sufficient to inhibit colony formation for 8 weeks (NCH612) or 4 weeks (SU-AO3), indicating significant and durable growth inhibition (Fig. 1D). Next, we measured the cell proliferation rate following DAC treatment to compare DAC sensitivity of IDH1-mutant tumorsphere lines. Intriguingly, the inhibitory effects of DAC on colony formation and cell proliferation were more pronounced in NCH612 and SU-AO3 lines compared with the TS603 line (Fig. 1G–I). Moreover, DAC treatment did not influence proliferation of the NCH1681 (IDH1-mutant-noncodel, grade III astrocytoma) line (Fig. 1F and J), suggesting that DAC-mediated proliferation defect may be more profound in IDH-mutant-codel gliomas.

Fig. 1.

IDH1-mutant-codel glioma cells are sensitive to the DNA methyltransferase inhibitor decitabine (DAC). (A) Table represents the status of chromosome 1p/19q in IDH1-mutant glioma lines. (B) DNMT1 protein expression and (C) cell counting measurements in TS603 treated with DMSO or 500 nM DAC for 10 days. (D) The indicated glioma lines were treated with DMSO or 500 nM DAC for 7 days in soft agar, and anchorage-independent growth was monitored. (E and F) DNMT1 protein expression after 7 days of DMSO or 500 nM DAC in NCH612, SU-AO3, and NCH1681. (G–J) Measurement of MTT assay of NCH612, SU-AO3, TS603, and NCH1681 in the presence of DMSO or the indicated concentrations of DAC for 7 days. Mean ± SEM of 3 independent experiments. **P < 0.01, ***P < 0.001.

DNA Replication, Lysosome, and Cell Cycle Pathways Are Enriched in DAC-Treated IDH1-Mutant-Codel Gliomas

To assess gene expression profiles in response to DAC treatment, the IDH1-mutant-codel lines (SU-AO3, NCH612, and TS603) were treated with a low dose (500 nM) of DAC for 7 days, and differentially expressed genes (DEGs) between DMSO and DAC-treated cells were identified by RNA sequencing (Supplementary Table 1, q < 0.05). The DAC-resistant NCH1681 line was also included for comparison. The majority of DEGs were upregulated, but interestingly a small subset of genes was downregulated (Supplementary Figure 1A–D). As expected, IPA identified DAC as one of the upstream regulators for all 4 IDH1-mutant glioma lines (Supplementary Figure 1E).

To further explore DAC-mediated transcriptional changes, we performed Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis. Consistent with differential sensitivity to DAC in IDH1-mutant glioma lines (Fig. 1), “DNA replication” and “Cell cycle” pathways were enriched in 3 DAC-sensitive IDH1-mutant-codel lines but not in the DAC-resistant NCH1681 line (Fig. 2A). In agreement with this, “DNA replication” was the most negatively enriched gene set in GSEA (Supplementary Figure 2 and Supplementary Table 2). In addition, “Lysosome” pathway was also robustly enriched in 3 IDH1-mutant-codel lines (Fig. 2A), implying that their functional consequences may be similar in response to DAC treatment. Notably, DAC treatment induced enrichment of “p53 signaling pathway” in 3 IDH1-mutant lines (Fig. 2A).

Fig. 2.

DAC results in down-regulation of TERT expression in IDH1-mutant-codel lines, and its expression inversely correlates with CDKN1A/p21. (A) KEGG enrichment analysis of DAC-treated IDH-mutant glioma lines. DEGs with q < 0.05 were analyzed using the clusterProfiler package in R. The size of the dot indicates the number of genes per pathway, and the color of the dot correspond to the q-value. (B) Venn diagrams demonstrating the overlap of DAC-driven transcriptional changes (fold change > 1.4 and q < 0.05) in IDH1-mutant glioma lines. (C) Heatmap showing significant expression changes of 35 genes found only in SU-AO3, NCH612, and TS603 but not in NCH1681. (D) p21 protein levels in DAC-treated TS603 and SU-AO3 as determined by immunoblot. (E) p21 and TERT mRNA expression after 7 days of 500 nM DAC in TS603 and NCH612 as assessed by qRT-PCR. Mean ± SEM of 3 independent experiments. ***P < 0.001. (F) Inverse correlation between CDKN1A (p21) and TERT expression with or without DAC treatment was analyzed using the Pearson correlation method. Arrows indicate the changes in response to DAC in individual glioma lines.

Induction of p21 by DAC Inversely Correlates with TERT

Next, we asked whether a common set of genes was deregulated in DAC-responsive tumorspheres. Because our data suggested that IDH1-mutant-codel lines are more responsive to DAC, we focused on 35 overlapping (22 up- and 13 downregulated) genes specific to SU-AO3, NCH612, and TS603 (Fig. 2B). In accordance with KEGG analysis from individual glioma lines (Fig. 2A), of these 35 genes, we found cyclin-dependent kinase inhibitor 1A (CDKN1A; p21), CDT1, GADD45A, MCM5, PLK3, and TERT associated with cell cycle of G1/S phase transition (Fig. 2C). We validated the marked p21 induction and TERT downregulation upon DAC treatment in IDH1-mutant-codel lines by immunoblotting and qRT-PCR analysis (Fig. 2D, E). Interestingly, TERT and CDKN1A(p21) expression were inversely correlated (Fig. 2F), implying that DAC-mediated p21 induction is closely associated with TERT repression.

DAC-Mediated TERT Repression Is Dependent on the p53/p21 Pathway

We further monitored p21 and TERT levels after DAC withdrawal. In both DAC-treated SU-AO3 and TS603 lines, increased TERT expression was observed at 25 days after DAC removal (Fig. 3A). In parallel, the expression of p21 was substantially suppressed in DAC-withdrawn cells compared with the initial levels at day 0 following DAC treatment (Fig. 3A), suggesting that DAC-mediated alterations in p21 and TERT levels are reversible. Interestingly, our results also indicate that DAC-withdrawn cells remain sensitive to subsequent DAC treatment (Fig. 3B).

Fig. 3.

DAC-mediated p21 induction requires p53, and TERT expression depends on p21. (A) The changes of p21 and TERT expression of DAC-withdrawn SU-AO3 and TS603 cells as determined by qRT-PCR. The expression levels were normalized to values from DMSO-treated corresponding cells as a control. (B) Measurement of cell proliferation of DAC-withdrawn TS603 cells treated with DMSO or DAC for 7 days. (C) SU-AO3 cells were pretreated with 20 µM pifithrin-α (PFTα) for 24 hours followed by 500 nM DAC. The expression of p21 (C) and cell proliferation (D) were measured by qRT-PCR and MTT assay, respectively. (E) p21 and TERT expression of SU-AO3 and TS603 cells treated with 1 µM doxorubicin for 24 hours. (F) TERT expression in TS603 cells transduced with scramble-shRNA or p21 shRNA. Mean ± SEM or SD of at least 3 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001.

To determine whether p53 is involved in p21 upregulation, SU-AO3 cells were pretreated with 20 µM pifithrin-α (PFTα, p53 inhibitor) for 24 hours before exposure to 500 nM DAC. Induction of p21 expression following DAC treatment was dramatically suppressed in PFTα-pretreated cells, supporting a p53-dependent p21 induction mechanism (Fig. 3C). Moreover, pretreatment of SU-AO3 cells with PFTα abrogated DAC-mediated proliferation defect (Fig. 3D). To further confirm the role of p53 as a mediator of DAC-induced TERT repression, we examined the p21 and TERT levels following doxorubicin treatment. Interestingly, the same effects on p21 and TERT expression were observed in doxorubicin-treated SU-AO3 and TS603 cells (Fig. 3E). To determine whether p21 is required for TERT repression by DAC, we generated TS603 lines expressing scramble shRNA or p21 shRNA (Fig. 3F). As shown in Fig. 3F, p21 knockdown led to a significant 3-fold upregulation of TERT, confirming the essential role of p21 in TERT repression.

DAC-Mediated Proliferation Defects Depend on TERT

We next assessed the effect of IDH1 mutation on DAC sensitivity on TERT IHA cell lines overexpressing IDH1 R132H.²³ In addition to IDH-mutant IHAs, parental IHAs also exhibited remarkable DAC response (Fig. 4A), implying that molecular determinants besides mutant IDH may contribute to the pronounced DAC sensitivity.

Fig. 4.

Ectopic TERT expression augments DAC sensitivity of TS603 and TS667 lines, independent of telomerase activity. (A) Cell proliferation rate of parental (control) and mutant IDH1-expressing IHA lines in response to 500 nM DAC for 5 days. (B) TERT expression of TS603 and TS667 cells transduced with Ctrl (tdTomato) or TERT are shown using qRT-PCR. Cell proliferation rate of TS603 (C), TS667 (D) expressing Ctrl or TERT, and NCH612 (E) treated with 500 nM DAC, 20 µM BIBR1532, or a combination of both drugs for 7 days as assessed by MTT assay. (F) Measurement of MTT assay of SU-AO3 in the presence of 250 nM DAC, 40 µM BIBR1532, or a combination of both drugs for 7 days. (G and H) TS603 and TS667 expressing TERT were treated with 500 nM DAC, 1 µM 6-thio-dG, or a combination of both drugs for 7 days, and cell proliferation was determined by MTT assay. Mean ± SEM of at least 3 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001.

TERTp mutations frequently occur in oligodendrogliomas and GBM.²⁰ Using Sanger sequencing, we identified two hotspot mutations (either C228T or C250T) in the TERTp in IDH1-mutant-codel lines but not in NCH1681 and TS667 (IDH-wildtype, GBM) lines (Supplementary Figure 3A). To determine whether DAC-mediated proliferation defects are dependent on TERT expression, we transduced TS603 and TS667 lines with a retroviral construct expressing TERT-HA for overexpression of TERT (Fig. 4B). Strikingly, TERT overexpression significantly enhanced DAC sensitivity in both glioma lines (Fig. 4C, D). Further supporting these findings, we used BIBR1532, a small-molecule inhibitor of TERT, which attenuated sensitivity toward DAC in TS603 or TS667 stably expressing TERT, suggesting that TERT is required for DAC-induced impaired proliferation (Fig. 4C, D). Although BIBR1532 or DAC alone exhibited significant inhibitory effects on cell proliferation in NCH612 and SU-AO3 lines, BIBR1532 in combination with DAC failed to elicit the additive or synergistic effect of the 2 drugs, implying that TERT inhibitor interferes with DAC or vice versa (Fig. 4E, F).

Consistent with its role as a regulator of telomerase, TERT overexpression increased telomerase activity of the TS603 line, while TERT inhibition by BIBR1532 treatment reduced its activity (Supplementary Figure 3B). We further sought to determine whether altered telomerase activity would influence DAC sensitivity. In contrast to the findings with the TERT inhibitor, the proliferation defect was not restored by 6-thio-dG, a telomerase inhibitor (Fig. 4G, H). In addition, no association was found between selective DAC response patterns and endogenous telomerase activity assessed by TRAP assay (Supplementary Figure 3C). Taken together, these results demonstrate that TERT confers DAC sensitivity to glioma cells most likely through telomerase-independent mechanisms.

TERT and DNMT1 Expression Correlate with DAC Response in IDH-Wildtype GBMs

Given that TERTp mutations are enriched in both IDH-mutant oligodendrogliomas and 80% of IDH-wildtype GBMs, we asked whether TERT expression could predict response to DAC in IDH-wildtype gliomas. We queried The Cancer Genome Atlas (TCGA) expression data and identified high TERT expression in the classic-like GBM subtype that harbors EGFR amplifications and mutations (Supplementary Figure 4A). We assessed DAC response in 6 additional GBM lines with or without EGFR overexpression.²⁴ Similar to IDH1-mutant gliomas, we observed variable response to DAC in GBMs (Fig. 5A and Supplementary Figure 5A–F). DAC resulted in the most effective growth inhibition in L0627 where TERT was highly expressed. Notably, TERTp hotspot mutations were found in all GBM lines (Fig. 5A).

Fig. 5.

TERT and DNMT1 expression levels correlate DAC response in IDH1-wildtype GBM lines. (A) The relative percentage of cell proliferation rate of 6 different GBM lines in the presence of 500 nM DAC for 5 days as assessed by MTT assay. The absorbance value of the DAC-treated line is normalized to value obtained from the DMSO-treated corresponding line as a control (upper). Heatmap illustrating TERT, DNMT1, and DNMT3B expression of GBM lines as assessed by qRT-PCR (bottom). (B and C) Correlation between endogenous DNMT1/DNMT3B expression and DAC response was analyzed by qRT-PCR and MTT assay, respectively. Color of dots indicates each GBM line.

To further identify molecular determinants of response to DAC in gliomas, we focused on overlapping genes (n = 111) correlated with TERT expression in IDH-mutant-codel and classic-like glioma subtypes from TCGA dataset (Supplementary Figure 4B). Intriguingly, we found that DNMT1 and DNMT3B expression correlated with TERT in both glioma subtypes (Supplementary Table 3). In addition, DNMT1 and DNMT3B were found to be highly expressed in the classic-like subgroup among IDH-wildtype GBMs (Supplementary Figure 4C, D). Given that DAC plays a role as a suicide substrate for DNMT enzymes, we next measured endogenous DNMT1 and DNMT3B levels in the GBM lines. Our results indicate that expression of DNMT1, but not DNMT3B, correlated with DAC response (Fig. 5A–C). These data suggest that TERT or DNMT1 expression may be useful in predicting response to DAC.

DNMT1 Is Required for DAC Response in TERT-Proficient Gliomas

Previous studies demonstrated that the protein levels of DNMTs are strongly associated with DAC response in embryonic stem cells and breast cancer organoids.^25,26 Because protein and gene expression levels of DNMT3A and DNMT3B were relatively low in 4 different IDH1-mutant lines, and DNMT3B expression showed only a weak association with DAC treatment in GBM lines, we focused on DNMT1 protein levels to determine whether its levels could be a potential biomarker for DAC response in gliomas. We confirmed higher levels of DNMT1 in the DAC-sensitive lines (SU-AO3, NCH612, L0627, and L0616), illustrating that DNMT1 protein levels are closely linked to DAC response in gliomas (Fig. 6A, B).

Fig. 6.

Silencing of DNMT1 abrogates DAC-induced decreased cell proliferation in TERT-proficient glioma lines. (A and B) Levels of endogenous DNMT1 protein in IDH1-mutant and -wildtype glioma lines as determined by immunoblot. (C) Immunoblots with the indicated antibodies in TS603 expressing Ctrl or TERT. (D–F) TS603 TERT, TS667 TERT, and L0627 transduced with scramble-shRNA or DNMT1-shRNA were treated with DMSO or DAC for 7 days, and cell proliferation rate was assessed by MTT. DNMT1 protein levels as determined by immunoblot. Mean ± SEM of at least 3 independent experiments. *P < 0.05, ***P < 0.001.

Given that TERT stimulates DNMT3B expression in hepatocellular carcinoma,²⁷ we asked whether TERT overexpression increases DNMT1 levels. We confirmed that in TS603 TERT overexpressing cells, DNMT1 was increased at the protein level (Fig. 6C). Thus, we hypothesized that DNMT1 induction by TERT may result in increased DAC sensitivity. To examine whether DNMT1 knockdown abolished TERT-mediated DAC efficacy, retroviral expression vectors encoding shRNA targeting DNMT1 were transduced in the L0627, TS603, and TS667 lines that constitutively express TERT. Intriguingly, cell proliferation was significantly decreased in DNMT1 knockdown lines but not in control cells expressing scramble shRNA, indicating that silencing of DNMT1 recapitulates the effect of DAC, a DNMT inhibitor (Fig. 6D–F). Consistent with our above observations that cells with low DNMT1 levels were exquisitely resistant to DAC, DNMT1 knockdown by shRNA protected against the DAC-mediated proliferation defect (Fig. 6D–F). Overall, DNMT1 is required for the anti-proliferative efficacy of DAC in TERT-overexpressing gliomas.

Discussion

Current standard of care options for glioma patients include surgery, radiation, and chemotherapy. However, despite multimodal treatment approaches, the outcomes for glioma patients remain poor. Previous work by us and others has highlighted the efficacy of DNMT1 inhibitors in IDH-mutant gliomas.^10–13 Our results indicate that DAC is effective in a broader range of gliomas, including IDH-wildtype GBMs, and this sensitivity is determined by TERT and DNMT1 expression. Because IDH1 mutation partly enhances DAC sensitivity of IHA cells, whether IDH mutation or 1p/19q codeletion status directly affects DAC response remains to be investigated using larger patient-derived tumorsphere cohorts.

Interestingly, we show that TERT overexpression-mediated DAC efficacy is more pronounced in TS667 (P < 0.001) harboring wildtype TERTp relative to TS603 (P < 0.05) harboring mutant TERTp. This discrepancy is likely due to differential susceptibilities to ectopic TERT expression in individual lines, highlighting the influence of TERT on DAC efficacy. Indeed, differential response to DAC is also observed in myeloid leukemia cell lines such that low levels of TERT mRNA are associated with resistance to DAC-induced cell death.¹⁵

Several studies aimed to demonstrate that DAC sensitivity is mechanistically linked to DNMT levels.^25,26,28 DNMT is considered to be the primary mediator of DAC response directly through the enzyme trapping mechanism, namely, the formation of DNMT-genomic DNA adducts.²⁹ Consistent with this, our study demonstrates that silencing of DNMT1 abrogates DAC-mediated proliferation defects. A cytotoxic effect is also observed in DAC-treated TS603 (data not shown). Thus, DAC efficacy is at least in part a result of increased cell death as well as impaired proliferation.

DAC induces profound transcriptional changes in glioma tumorspheres, and several pathways including DNA replication, cell cycle, and lysosome were significantly enriched in DAC-sensitive IDH1-mutant glioma lines. As reported previously,³⁰ DAC-associated DNA replication downregulation along with a substantial inhibition of cell proliferation indicates that this expression pattern corresponds to DAC response. To this effect, we show that DAC triggers p21 in glioma tumorspheres, suggesting that DNA damage response by DNMTi-dependent p21 induction may lead to inhibition of proliferation.^8,31

Transcriptional repression of TERT is dependent on p21,³² and our results also indicate DAC-mediated p21 induction acts as a causal contributor to TERT downregulation in gliomas. Furthermore, TERT expression correlates with DNMT1 levels in glioma tumorspheres, and this pattern is confirmed using publicly available databases (Supplementary Table 3). Although the acquisition of DAC sensitivity by TERT overexpression is accompanied by markedly increased DNMT1 levels in TS603, the underlying mechanism by which DNMT1 protein levels are elevated by TERT overexpression requires further investigation.

A paradoxical outcome of TERT inhibition as a mediator of DAC sensitivity in TERTp mutated glioma tumorspheres is the question whether TERT repression may ultimately lead to the development of DAC resistance. As a possible treatment regimen for myelodysplastic syndromes, DAC is administered for 5 consecutive days every 28 days.³³ In our in vitro studies mimicking this clinical situation (DAC treatment followed by 25 days without drug exposure), we observed an increase in TERT expression at the end of the withdrawal course. Importantly, at this time point, pretreated cells responded to DAC. Moving forward, long-term treatment courses and optimal DAC scheduling need to be assessed to determine DAC efficacy and potential mechanisms of resistance. Given that the plasma half-life of DAC is about 30 minutes, another important aspect in clinical application is delivery to the brain. Although DAC is known to penetrate the blood-brain barrier, an effective permeability of the blood-brain barrier and optimal drug concentration need to be assessed in future studies.

Here, we show that TERT and DNMT1 are key actors of DAC effects in gliomas. Taken together, based on our encouraging preclinical results, glioma patients with high TERT and DNMT1 levels can potentially benefit from DNA demethylating agents.

Funding

This research was supported by the German Cancer Aid, Max Eder Program grant number 70111964 (S.T.), and Oligo Nation and Operation Oligo Cure (S.T., D.C., I.A-R).

Conflict of interest statement. The authors declare that they have no conflict of interest.

Authorship statement. J-W.P. and S.T. conceived the study. J-W.P., F.S., and B.S. performed the experiments. J-W.P., F.S., and S.T. analyzed and interpreted the results. M.M. and C.H-M. provided materials. I.A-R., D.C., and W.W. provided conceptual input and expertise. S.T. supervised the project. J-W.P. and S.T. wrote the manuscript. All authors contributed to the writing or editing of the manuscript.