TERT Promoter Mutation Detection in Cell-Free Tumor-Derived DNA in Patients with IDH Wild-Type Glioblastomas: A Pilot Prospective Study

Our knowledge about molecular markers in gliomas has undergone a rapid evolution in the last few years. Subsequently, a new generation of biomarkers has become available with the discovery of the genetic alterations that are responsible for the initiation and progression of gliomas, including telomerase reverse transcriptase promoter (TERTp) and isocitrate dehydrogenase (IDH) mutations (1–5).

TERTp mutations are one of the most common recurrent alterations (90%) in glioblastomas and were found to be clonal in both pretreatment and posttreatment autopsy samples in most glioblastoma patients (6). Moreover, the presence of somatic TERTp mutations in glioblastomas confers a dismal prognosis (7, 8). In contrast, only a minority of glioblastomas harbor IDH1 mutations (5%) and were shown to represent a distinct subset with favorable outcomes (5, 8). Consequently, screening for TERTp and IDH1 mutations in glioblastoma patients is of clinical value in survival prediction.

The tDNA contains identical genetic alterations to those detected in the primary tumor itself. Therefore, in numerous cancer types, circulating tDNA detection in the plasma has been shown to be a reliable tool for monitoring, diagnosis, and prognostication (9–15). However, in primary and metastatic brain tumors, the levels of tDNA in the plasma are low and inconsistently detectable (12). Moreover, tDNA assessment has not yet been established in the clinical management of glioma patients and, to date, no study has systematically evaluated the clinical value of TERTp mutation detection in the plasma or cerebrospinal fluid (CSF) in a large cohort of IDH wild-type glioblastomas.

With this in mind, we conducted a pilot prospective study to assess TERTp and IDH hotspot mutations in the CSF-tDNA and plasma-tDNA, collected during surgery. We exclusively evaluated data from patients with TERTp-mutant/IDH wild-type glioblastomas to reduce potential confounders. A further intention of our study is to explore a putative association between the burden of the TERTp mutation in the tDNA and patients’ outcome.

Specimens were collected between July 2015 and September 2017 at the time of surgery from 60 patients who underwent tumor resection of newly diagnosed intracranial tumors that were suspicious for a diffuse glioma. Upon general anesthesia, 8 to 10 mL whole blood was withdrawn from every patient, and 2 to 4 mL CSF was collected during surgery from the tumor site. The CSF collection was performed directly after opening the dura (durotomy), before any surgical manipulation of the tumor, predominantly through dissection of the convexity subarachnoid space.

Blood and CSF samples were collected in specific tubes preserving the tDNA (cell-free DNA-BCT Tubes, Streck). For separation of plasma, blood samples were centrifuged at 300 g for 20 minutes. Without disturbing the buffy coat, the plasma layer (supernatant) was carefully removed and transferred into a new 2 mL low-bind tube. We also collected the pelleted material to assess TERTp mutations in the cellular fraction of the CSF. To completely remove residual cells, both plasma and CSF samples were subsequently centrifuged at 5,000 g for 10 minutes and stored at −20°C until downstream processing.

The present study was conducted according to the guidelines of the Declaration of Helsinki and has been approved by the local Ethics Committee of the Medical Faculty Carl Gustav Carus Dresden. All samples (CSF, blood, and tumor tissue) were collected with written consent from the patients.

Tumor tissues were taken intraoperatively and were snap frozen at −80°C. Frozen tumor tissue was sectioned using a cryotome (Cryostat Jung CM 1800; Leica). To assure a tumor cell content of at least 80% for nucleic acid extraction, control slides stained with hematoxylin and eosin were examined by the local neuropathologist (M. Meinhardt). For DNA isolation, the QIAmp DNA Mini Kit (Qiagen) was used.

Cell-free DNA was extracted using the Quick-ctDNA Serum & Plasma Kit (Zymo Research) according to the manufacturer's protocols. Briefly, stored CSF and plasma samples were thawed at room temperature. Samples were incubated with proteinase K in digestion buffer for 30 minutes at 55°C in a water bath. After mixing with DNA binding buffer, samples were loaded into the Zymo-Spin III-S column assembly and centrifuged at 1,000 g for 2 minutes. After washing and centrifugation (10,000 g; 30 seconds), ctDNA was eluted in 35 μL TE-buffer (10 mmol/L Tris, 0.1 mmol/L EDTA, pH 8). The DNA concentration was measured by quantitative real-time PCR on an ABI7500 Real-Time PCR System (Applied Biosystems) using primers and probes for the ß-globin locus.

A nested PCR procedure was used to amplify the proximal TERT promoter covering nucleotide numbers C228T and C250T (chr5:1295228 C>T and 1295250 C>T; hg19) from plasma-tDNA samples as well as genomic cellular DNA from the tumor specimens. Primer sequences for the first PCR were 5′-GGCCGATTCGACCTCTCT-3′ (TERT-p-f) and 5′-AGCACCTCGCGGTAGTGG-3′ (TERT-p-r), amplifying a 489 bp fragment of the TERT promoter region. PCR was carried out using the Qiagen Multiplex PCR Kit (Qiagen) in a 50-μL solution, consisting of a Multiplex PCR Mix (1x), Q-solution (1x), 0.2 μmol/L each TERT-p-f/r, and 5 ng of template DNA. PCR was performed on a GeneAmp PCR System 9700 (Applied Biosystems) with an initial denaturation step at 95°C for 15 minutes, followed by 25 cycles of denaturation at 94°C for 1 minute, annealing/extension at 59°C for 3 minutes, and a final extension at 60°C for 10 minutes.

Second-round (nested) PCR was accomplished using the Q5 High-Fidelity Master Mix (New England Biolabs) and fusion-primers 5′-AGTGGATTCGCGGGCACAGA-3′ (forward) and 5′-CAGCGCTGCCTGAAACTC-3′ (reverse) to attach the Ion A and truncated P1 (trP1) adapter sequences to the amplicons, resulting in a 303 bp PCR product. Final reaction mixture (50 μL) consisted of Q5 Buffer (1x), 1 u Q5 polymerases, 0.5 μmol/L of each fusion primer, 0.2 mmol/L dNTPs, and 1 μL of first-round PCR product as template. The PCR consisted of 30 seconds at 98°C; and 30 cycles of 5 seconds at 98°C, 10 seconds at 68°C, and 20 seconds at 72°C with a final extension at 72°C for 2 minutes.

The PCR reactions were purified using a two‐round purification process with Agencourt AMPure XP Reagent (Beckman Coulter) and eluted in 30 to 50 μL ddH₂O. The barcoded PCR products were quantified with a Qubit 2.0 fluorometer (Life Technologies) using the Qubit dsDNA HS Assay (Life Technologies) and sequenced unidirectionally on an Ion Torrent PGM NGS system (Life Technologies), according to the manufacturer's protocols (16).

Briefly, the pooled library was clonally amplified on Ion Sphere Particles (ISP) in an emulsion PCR using the Ion PGM Hi-Q OT2 Kit (Life Technologies). Enrichment of positive Ion Spheres (ISPs) was achieved using DynaBeadsMyOne streptavidin C1 beads (Life Technologies). Quantification of recovered particles was performed using a Qubit 2.0 fluorometer (Life Technologies) and an Ion Sphere quality control kit (Life Technologies). Ion Torrent chips (316 Chip kit v2; Life Technologies) were prepared and loaded according to the manufacturer's protocol. PGM sequencing (Ion PGM Hi-Q Sequencing Kit; Life Technologies) was done with 500 flows, 18 μs water and argon gas to drive fluidics. Data were analyzed with Torrent Suite 3.0 or higher with alignment against Hg19 and the Variant Caller plugin. The quantitative accuracy was documented using dilutions of TERTp-mutant cell line Calu-1 (DSML) in normal DNA (Supplementary Fig. S1). In addition, we determined the diagnostic sensitivity and specificity of our assay by analyzing DNA extracted from the CSF cellular components in patients with non–tumor-related hydrocephalus during shunt implantation (n = 6).

Digital droplet PCR (ddPCR) was performed on a Bio-Rad QX200 using the Bio-Rad TERT C228T_113 Assay (Assay ID dHsaEXD72405942). The detailed method is provided in the Supporting Information.

IDH1/2 mutation detection was performed in the tumor tissue based on conventional Sanger sequencing, using established primer sets, as published elsewhere by our group (17).

MGMT promoter methylation status was determined by the methylation-specific PCR, as described by Esteller and colleagues (18).

All patients in the study received a pre- and postop. MRI (day one after surgery) with T1, T1 post-contrast, T2, and FLAIR sequences. Radiologic progression was determined using the criteria for Response Assessment in Neuro-Oncology (19). In addition, the preoperative T1-weighted postcontrast MR scans were used to measure the enhancing tumor size and obtain volumetric measurements. A combination of manual segmentation using the ITK-SNAP software (20) and intensity filtering based on individually adjusted thresholds was used to select the enhancing tumor tissue with and without the necrotic tumor portion.

To investigate correlations between the successful detection of TERTp-mutant tDNA and tumor location, we used a previously described MRI-based classification of gliomas (21) to assess the spatial relationship of the tumor with the subventricular zone (SVZ) and the cortex as follows:

The Kaplan–Meier technique was used to estimate progression-free survival (PFS) and overall survival (OS) and tested for significance by the log-rank test. PFS was calculated from the day of surgery until MRI-confirmed tumor progression or end of follow-up. OS was defined as the interval from the day of first surgery until death or the end of follow-up. The Mann–Whitney U and Fisher exact tests were used to test for association of clinical variables and CSF-tDNA, whereas the Spearman test was used to correlate coefficients among outcomes of the contrast-enhancing tumor volume with the abundance of the mutant CSF-tDNA. Multivariate Cox regression models were applied to assess the impact of CSF-tDNA on patient's outcome. All P values are two-sided, and all analyses were conducted using the SPSS software package (Version 21.0 SPSS Inc.).

Out of 60 patients who underwent surgery for an intracranial tumor, we identified 38 patients (24 males and 14 females) with TERTp-mutant/IDH wild-type glioblastomas who were enrolled into our study (Fig. 1). The median patient age at initial diagnosis was 67.6 years (range, 64.3–71.4 years). Patient characteristics and tumor features are shown in Table 1.

Gross total resection of the enhancing tumor—as evaluated by early postoperative MRI—was achieved in 20 patients, whereas 18 patients underwent a subtotal resection. Radiotherapy was applied in all cases, and 31 patients were treated concomitantly with temozolomide. Temozolomide therapy was prematurely discontinued in 7 patients, due to disease progression, and in 2 patients, because of hematologic toxicities.

Our study had a median follow-up of 12.36 months (4.25–33.0 months). The median PFS of patients with a newly-diagnosed GBM was 5.7 months (4.2–7.2 months), and the median OS was 9.9 months (6.7–13.2 months). Twenty-nine patients (78.3%) died during the time of follow-up. The last update of progression status was performed in April 2018.

The median coverage for tumor tissue samples was 48,170 reads (range, 42,716–60,268 reads); for CSF, it was 48,890 reads (range, 8,680–198,576 reads); and for plasma samples, it was 49,700 reads (range, 6,647–107,900 reads). Using this approach, we could detect TERTp mutations in CSF-tDNA at allelic frequencies as low as 0.4% for C228T and 0.2% for C250T. The turnaround time from processing the CSF/plasma sample to final mutation data report was 3 to 5 days.

In the tumor tissue, the detected TERTp hotspot mutations were located at 146 bp (referred to as C250T, n = 7) and in 31 cases at 124 bp (referred to as C228T) upstream of the translation start of the TERT promoter. In the matched CSF-tDNA, the TERTp mutation was successfully detected in 35 of 38 patients (92.1%; Table 1).

Remarkably, in four IDH-mutant/TERTp-mutant gliomas (oligodendrogliomas), we detected only one TERTp mutation in the CSF-tDNA (Supplementary Table S1). In contrast, no TERTp mutations were detectable in the CSF-tDNA of any of the patients with an IDH-mutant/TERTp wild-type glioma (n = 7) or an IDH wild-type/TERTp wild-type glioma (n = 5) or in patients with brain metastases (n = 6). Additional sequencing of CSF cell-free DNA from 6 patients with a non–tumor-related hydrocephalus revealed no TERTp mutations. Taken together, these results show a specificity of 100% (95% CI, 87.6–100%) and a sensitivity of 92.1% (95% CI, 78.6–98.3%) in determining the TERTp mutations in the CSF of TERTp-mutant glioblastoma patients. Repeated analyses of a subgroup of samples (n = 15) from the CSF-tDNA indicated a high reproducibility of our results, with a median SD from mean of 7.6% (range, 1.3%–34.4%).

Moreover, to validate our results, ddPCR was performed in duplicates of five TERTp-mutant and four TERTp wild-type samples. The ddPCR showed concordant results with those determined by the Ion Torrent system in all 9 cases. However, our study has a limited power to sufficiently address the question which method is superior in terms of TERTp mutation detection in the CSF-tDNA.

The sensitivity of TERTp mutation assessment in the plasma-tDNA was far lower [7.9% (95% CI, 1.6–21.4%) (n = 3)] than that in the CSF (Fisher exact test, P < 0.0001). All 3 patients with detectable TERTp mutation in plasma-tDNA had large glioblastomas with tumor infiltration of the choroid plexus and encasement of one major cerebral artery.

To investigate whether the tumor size might affect the TERTp mutation detection in the CSF, we performed volumetric tumor measurements. The median gross tumor volume was 59.2 cm³ (range, 1.8–177.2 cm³), and the median contrast-enhancing tumor volume was 45.5 cm³ (range, 1.5–105.33 cm³).

Among the 29 newly-diagnosed GBM cases with detectable TERTp mutations in the CSF and an available follow-up, we observed a statistically significant linear correlation between the variant allele frequencies (VAF) and either the gross tumor volume (Spearman-Rho = 0.55; P = 0.002) or the (Gadolinium-) contrast-enhanced tumor volume (Spearman-Rho = 0.54; P = 0.003; Fig. 2).

In terms of anatomical location, all but one tumor was adjacent to the CSF space through infiltration of the subventricular zone and/or the cortical subarachnoidal space (groups 1–3; Fig. 3 and Table 1). Interestingly, the only patient with a tumor that was entirely encapsulated by the brain parenchyma (Group 4, Table 1) was 1 of the 3 patients with nondetectable TERTp mutations in the CSF (Supplementary Fig. S2).

Notably, we did not observe a significant association between the TERTp mutation VAF in the CSF-tDNA and patients’ age, sex, or performance status.

To explore a potential role of the TERTp mutation VAF in the CSF-tDNA in predicting patients’ survival, Kaplan–Meier curves were generated and analyzed based on the TERTp mutation VAFs in the CSF-tDNA. It was concordantly observed that patients with low VAF in the CSF-tDNA had a longer OS when compared with patients with high VAF, irrespective of using the lower quartile VAF of 11.45%, the lower third VAF of 13%, or the median VAF of 20.3% to dichotomize the patients (P = 0.035, P = 0.008, and P = 0.062, respectively; Fig. 4).

To substantiate our findings, a multivariate analysis was performed including all first-diagnosed glioblastoma patients with detectable TERTp-mutant CSF-tDNA and available follow-up (n = 29). The TERTp mutation VAF with a cutoff of 13% was found to be an independent prognostic factor for OS [P = 0.001, HR = 16.02; 95% confidence interval (CI), 3.15–28.43; Table 2]. By exploring an interaction between the tumor volume and a TERTp mutation VAF cutoff of 13% in the CSF-tDNA as one factor in the multivariate analysis, the HR was 2.07 (95% CI, 1.33–3.22, P = 0.001). All other included clinical parameters such as age, extent of resection, and MGMT promoter status were not independent prognostic factors in our cohort, although it is possible that there are other clinical variables which may account for the differences in outcome.

To examine a potential application of CSF-tDNA as a marker for postoperative residual disease, CSF was collected in 5 patients who had intraoperatively detectable TERTp mutation in the CSF-tDNA and underwent a lumbar drainage placement to treat a postoperative CSF leakage. Interestingly, we saw a dramatic decrease of VAF levels (P = 0.043, Wilcoxon-signed rank test; Supplementary Fig. S3). However, although GTR was achieved in all 5 patients, TERTp mutation in the CSF-tDNA was postoperatively detectable in only two cases. Remarkably, persistent postoperative TERTp mutation detection in the lumbar spine CSF was associated with a decreased PFS of both patients (5.1 months), compared with the 3 patients with nondetectable mutation who had a longer PFS (median, 11.9 months; 95% CI, 8.3–15.5 months). However, given the small group of patients, our study was likely underpowered to make any significant conclusions about differences in survival outcomes.

Detection of circulating tumor material has been shown to be of clinical value for diagnosis, prediction of response to treatment, and prognosis in cancer patients (10, 11, 14, 15, 22). In this pilot study, we demonstrate that TERTp mutation detection in intraoperatively collected CSF is feasible with a very high sensitivity and specificity in glioblastoma with spatial relationship to a CSF reservoir in the brain, consistent with the findings from previous studies (23, 24). Remarkably, we observed a significant association between the VAF of the TERTp mutation in the CSF-tDNA and the outcome of glioblastoma patients in our cohort. Using different cutoffs, our data consistently showed that an increasing burden of TERTp-mutant alleles in the CSF-tDNA was associated with unfavorable outcomes, suggesting a high reliability of the TERTp mutation VAF levels in the CSF-tDNA as an outcome prediction tool in glioblastoma patients.

Furthermore, we found a lower sensitivity of TERTp mutation detection in a distinct subset of TERTp-mutant/IDH-mutant gliomas (oligodendrogliomas), confirming the findings of a recent study that reported on the inability to detect molecular alterations in the CSF-tDNA in some low-grade gliomas (24). This may be explained by the higher cellularity of glioblastomas compared with oligodendrogliomas, which is presumably associated with larger quantities of tDNA shedding into the CSF.

A further potential application for CSF-tDNA is its use as a marker for postoperative residual disease. Presently, predicting the disease-free status of patients is predominantly based on clinical and radiologic criteria. Interestingly, fluctuations in circulating tDNA levels have been shown to occur prior to changes seen in imaging studies (10). Moreover, it has been demonstrated that levels of circulating tDNA increase with disease progression and correspondingly decrease with response to therapy in solid tumors and glioblastomas (10, 11, 25, 26). In the current study, we demonstrate an association between patients’ outcome and intra-/postoperative TERTp mutation CSF-tDNA levels in five cases. A significant postoperative reduction of the TERTp mutation VAF in the CSF-DNA when compared with the “baseline” intraoperative VAF level was associated with favorable survival outcomes. Although preliminary, this finding suggests that clearance of TERTp mutations in the CSF may indicate residual disease in glioblastoma patients and may be used for disease monitoring. However, this observation needs to be verified in a larger cohort of patients before any definitive conclusions can be made.

In concordance with earlier studies, our assay to detect TERTp-mutant plasma-tDNA had a low sensitivity (12). Diaz and colleagues reported that mutant DNA fragments were found in high concentrations in the circulation of most patients with metastatic cancer—e.g., melanoma, breast, or colon cancers—and at lower concentrations in patients with localized cancers, as in gliomas or prostate cancer (13, 27). Similarly, a recent study showed that levels of CSF-tDNA were considered significantly enriched when compared with the corresponding plasma in 12 individuals with primary or metastatic brain tumors (28). In order to improve this shortcoming, new technologies are needed. For instance, an innovative technique employed focused ultrasound to increase the release of brain tumor material into the bloodstream, essentially making blood a viable surrogate for tumor detection (29).

Although we suggest that TERTp mutation detection in the CSF-tDNA has the potential to be developed as a diagnostic instrument in the management of glioblastoma patients, we acknowledge that our approach to collect CSF at the time of surgery was not intended as a preoperative diagnostic tool. From a clinical standpoint, the preoperative detection of TERTp mutation in the CSF-tDNA by lumbar puncture would be more readily accessible than by intraoperative surgery, notwithstanding the consideration of lumbar punctures as an invasive procedure. Consequently, the assessment of genomic alterations at several longitudinal time points as a “liquid biopsy” is a consequential approach to monitor tumor progression and to correlate with radiographic findings. Ultimately, this serial approach may avoid unnecessary surgical procedures and lead to better management of glioblastoma patients. Lastly, a recent study showed a very low congruence for same patient-paired prostate cancer samples between different laboratories (30), raising the question whether the results generated by our assay could be generalized to other laboratories or attained by using different assays. Therefore, to complement our study, confirmatory observational cohorts with more patients and further techniques are certainly needed to determine precise cutoffs of the TERTp mutation VAF that may be used to define patients at high risk for rapid progression.

In conclusion, our pilot study highlights the value of CSF-tDNA for an accurate and reliable detection of TERTp mutations in glioblastomas. Furthermore, our findings suggest that high TERTp mutation VAF levels in the CSF-tDNA may be a suitable predictor of poor survival in glioblastoma patients. This would allow for a more precise risk stratification of patients for clinical trials. Further studies are needed to validate the findings of this exploratory analysis.

C. Thiede is an employee of AgenDx GmbH, reports receiving speakers bureau honoraria from Illumina, Novartis, and Roche, is a consultant/advisory board member for Novartis and Roche, and reports receiving commercial research grants from Bayer and Novartis. No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of Harvard Catalyst, Harvard University and its affiliated academic healthcare centers, or the NIH.

Conception and design: T.A. Juratli, D. Krex, C. Thiede

Development of methodology: T.A. Juratli, A. Zolal, C. Schuster, D. Krex, C. Thiede

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.A. Juratli, S. Stasik, S. Richter, D. Daubner, S. Hennig, M. Makina, T. Lautenschlaeger, G. Schackert, D. Krex, C. Thiede

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.A. Juratli, S. Stasik, A. Zolal, D. Daubner, M.A. Juratli, S. Hennig, M. Meinhardt, T. Lautenschlaeger, D. Krex

Writing, review, and/or revision of the manuscript: T.A. Juratli, A. Zolal, M.A. Juratli, S. Hennig, M. Meinhardt, T. Lautenschlaeger, G. Schackert, D. Krex, C. Thiede

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.A. Juratli, D. Daubner, R. Thowe, D. Krex

Study supervision: T.A. Juratli, D. Krex

The authors thank Drs. Dino Podlesek, Stephan Sobottka, Matthias Kirsch, and Thomas Pinzer for helping collect CSF.

This work was conducted with support from Harvard Catalyst|The Harvard Clinical and Translational Science Center (National Center for Research Resources and the National Center for Advancing Translational Sciences, NIH Award UL1 TR001102) and financial contributions from Harvard University and its affiliated academic healthcare centers.

This work is supported by the MeDDrive (grant number 60.378) to Dr. T.A. Juratli and the BMBF (grant 01GS0872) to Dr. C. Thiede. Dr. T.A. Juratli was awarded the Aesculap EANS research Prize 2018 for this study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.