Tissue 2-Hydroxyglutarate as a Biomarker for Isocitrate Dehydrogenase Mutations in Gliomas

Since its initial discovery in 2009, isocitrate dehydrogenase 1/2 (IDH1/2) mutations have been found in a diverse group of cancers such as glioma (1–3), acute myeloid leukemia (AML; refs. 4, 5), intrahepatic cholangiocarcinoma (6, 7), chondrosarcoma (8), angioimmunoblastic T cell lymphoma (9), and others (10–12). IDH1/2 mutations are reported in approximately 70% to 90% of low-grade diffuse gliomas and secondary glioblastomas (GBM), and are associated with younger age at diagnosis and significantly improved overall survival (OS; refs. 1–3, 13, 14). Given the prognostic significance of IDH and its implicated role in gliomagenesis, IDH mutation status and other molecular markers have recently been integrated into the World Health Organization (WHO) classification of gliomas (15).

IDH1 and IDH2 are NADP⁺-dependent enzymes and function in the tricarboxylic acid cycle to catalyze the oxidative decarboxylation of isocitrate to α-ketoglutarate (αKG; ref. 16). IDH1 and IDH2 enzymes share 70% sequence similarity, but are encoded by different genes (IDH1, 2q33; IDH2, 15q26) and localize to different cellular compartments (17). In gliomas, over 90% of IDH mutations involve IDH1 codon 132 with substitution of arginine by histidine (R132H; ref. 18). Accordingly, IHC with an antibody specific for this canonical mutation has greater than 90% sensitivity. The high sensitivity and relatively low cost of IHC makes it the most common method of determining IDH mutation status clinically. Of course, this method will not be sensitive to noncanonical IDH1 or IDH2 mutations (19). Direct sequencing is being increasingly used, but it is associated with higher costs and a longer turnaround time.

Mutations in IDH1/2 confer neomorphic activities that convert αKG to 2-hydroxyglutarate (2-HG; ref. 20), with preferential accumulation of the R-relative to the S-enantiomer of 2-HG (21). Previous studies have investigated the utility of using 2-HG as a biomarker for IDH mutation status in gliomas without conclusive results (21–26). These studies have been limited by measuring levels of total 2-HG, and by measuring 2-HG in blood or urine as opposed to tumor tissue. Only one study investigated the utility of R-2-HG, but in cerebrospinal fluid (CSF; ref. 26). The use of proton magnetic resonance spectroscopy (MRS) has been advocated by several groups, but it is also limited by its ability to detect total 2-HG only (27, 28). In the current study, we evaluated the utility of quantitating the relative abundance of R-2-HG and S-2-HG using a novel high-pressure liquid chromatography tandem mass spectrometry (HPLC-MS/MS) technique as a surrogate for IDH mutation status.

The study was approved by the Research Ethics Board at University Health Network, Toronto, Canada, and conducted in accordance with the Declaration of Helsinki. Fresh-frozen glioma tissues from 87 patients were obtained from the University of Toronto Brain Tumor Biobank. Matched preoperative serum samples were available for 29 patients. Patient and disease characteristics such as age, sex, tumor histology, tumor grade, clinically reported IDH1 mutation status, 1p/19q co-deletion status, treatment details, and survival outcomes were determined by retrospective chart review.

Separately, glioma tissues were obtained from 11 patients at the time of surgical resection. A written informed consent was obtained from each patient.

Tissue and serum samples were analyzed for R-2-HG and S-2-HG utilizing HPLC-MS/MS. Tissue or serum samples were thawed, weighted or measured, and homogenized with 500 μL of double distilled water. Protein was precipitated using acetonitrile, and samples were then centrifuged at room temperature (24°C) for 10 minutes at a speed of 10,000 rpm. The resulting supernatants were transferred to polypropylene tubes and evaporated under vacuum. The residues were reconstituted with 100 to 200 μL of methanol water (90%:10%; v/v) and 10 μL was injected into a Shimadzu CBM-20A System coupled with a triple quadrupole mass spectrometer (API 3200, Applied Biosystems MDS SCIEX). Chromatographic separation was achieved through an Astec CHIROBIOTIC Chiral HPLC analytic column (250 × 4.6 mm, 5-μm particles, Sigma Co). Data collection, peak integration, and processing were performed with Analyst 1.6.2 (Applied Biosystems MDS SCIEX). Assays were performed blinded to clinically reported IDH mutation status. The lower limit of quantitation was 10.0 ng/g of tissue for both R-2-HG and S-2-HG.

For 75 of the 87 patients in our cohort, IDH1 status was determined via IHC using an IDH1 antibody as per standard clinical practices in a CLIA-certified clinical pathology laboratory (18). For the remaining 12 patients, DNA was isolated from fresh-frozen tissue and PCR amplified using primer pairs specific for exon 4 of the IDH1 gene. The amplified products were then sequenced by Sanger Sequencing at the Center of Applied Genomics (The Hospital for Sick Children, Toronto, Canada). For patients with discordant IHC and 2-HG results, IDH status was determined through sequencing in the genome diagnostics/cancer cytogenetics laboratory in the Department of Clinical Laboratory Genetics (University Health Network, Toronto, Canada). IDH mutation status was further confirmed by 450K methylation array tumor profiling, using G-CIMP status as a readout (29). The 1p19q co-deletion status, determined either by PCR analysis and/or FISH analysis, was available for 53 patients.

Descriptive statistics were used to report the baseline patient and disease characteristics. Categorical variables between groups were compared using Fisher exact test, and continuous variables were compared using the Mann–Whitney–Wilcoxon test. Univariable and multivariable regression analyses were undertaken to identify potential predictors of 2-HG levels, including age, gender, tumor grade, tumor recurrence, treatment latency (defined as the duration between diagnosis and first surgery), IDH1 status and 1p/19q co-deletion status. ROC analysis evaluated the utility of total 2-HG, R-2-HG alone, and the ratio of R- to S- enantiomers of 2-HG (rRS) as potential biomarkers of IDH status, based on the area under the ROC curve and the Youden index, which identifies the cutoff maximizing the sum of sensitivity and specificity.

The Kaplan–Meier method was used to estimate survival rates. OS was defined as the time from diagnosis to death from any cause. Progression-free survival (PFS) was defined as the time from diagnosis to first recurrence or death from any cause. Patients without documented evidence of an event were censored at the date of last follow-up. The Cox proportional hazards regression analysis was used to explore the relationship between groups and survival. Median study follow-up was calculated using the reverse Kaplan–Meier estimator.

Significance level was set at alpha of 0.05 for inferential analyses. All analyses were performed using R version 3.5.0 (R Foundation for Statistical Computing).

Of the 87 patients included, 61 patients were IDH1 mutant tumors and 26 patients had IDH1 wild-type tumors. Matched glioma tissue and serum samples were available for a subset of 29 patients. There were a total of 55 male and 32 female patients, with a median age at diagnosis of 41 years (range, 20–75 years). Baseline patient and disease characteristics stratified by IDH mutation status are summarized in Table 1. In keeping with previous reports, patients with IDH1 mutations were significantly younger than those with IDH1 wild-type tumors, had lower grade tumors, longer treatment latency, and tumors were more likely to harbor a 1p/19q co-deletion. Median follow-up was 9.8 years (range, 1.3–not reached).

In glioma tissues, total 2-HG was significantly higher in IDH1-mutant tumors (median total 2-HG = 2.92 × 10⁵ ng/g) compared with that in IDH1 wild-type tumors (median total 2-HG = 4.00 × 10³ ng/g, P < 0.001, Fig. 1A). Similarly, the median R-2-HG was 2.91 × 10⁵ ng/g in IDH1-mutant tumors and 2.08 × 10³ ng/g in IDH1 wild-type tumors (P < 0.001). The rRS was significantly higher in IDH1-mutant tumors (median rRS = 567) compared with IDH1 wild-type tumors (median rRS = 1.11, P < 0.001, Fig. 1B). Initially, there were 4 samples that were determined to be IDH1 wild-type based on IHC, but with unexpectedly high tissue rRS. Upon sequencing, 3 of these samples were found to have noncanonical IDH1 mutations. The other sample was confirmed to be IDH1 wild type.

In contrast to tissue measurements, total 2-HG, R-2-HG, and rRS were not elevated in matched serum samples and did not differentiate between IDH1-mutant and IDH1 wild-type status. Median rRS was 1.49 for IDH1-mutant samples versus 1.22 for IDH1 wild-type samples (P = 0.08, Fig. 1C).

The potential predictive value of tissue 2-HG was examined with univariable and multivariable analyses. The sensitivity and specificity of rRS for IDH mutation was determined with ROC analysis. On the basis of univariable analyses, the significant predictors of the rRS from glioma tissues included IDH1 mutation status (higher rRS for IDH1-mutant tumors), age (higher rRS for patients under 40 years), tumor grade (higher rRS for low-grade tumors), treatment latency (higher rRS for longer latency), and 1p/19q codeletion status (higher rRS for codeleted tumors). In multivariable analysis, the impact of IDH1 status persisted, and it was clearly the most important predictor of rRS from glioma tissues (P < 0.001; Table 2). In particular, the rRS from glioma tissues was significantly associated with IDH1 status within both low-grade and high-grade glioma subgroups (Fig. 1D and E). This suggests that elevated rRS values are driven by IDH status, controlled for baseline patient and disease characteristics.

ROC analysis showed that the tissue rRS was highly specific and sensitive for IDH mutation status (Fig. 1F). Using a rRS cut-off of 32.26, where higher values designate IDH1-mutant tumors and lower values designate IDH1 wild-type tumors, the area under the ROC curve was greater than 0.99 and the sensitivity and specificity was 97% and 100%, respectively. This discriminatory effect was greater than that with the use of total 2-HG or R-2HG, consistent with the relative abundance of R-2-HG versus S-2-HG in IDH mutations.

For the group of 11 patients with intraoperative samples, these samples were processed fresh without freezing and the time intervals from sample acquisition to results by HPLC-MS/MS were approximately 60 minutes. There were 4 patients with rRS values above the cutoff of 32.26 (189.5, 223.6, 296.7, and 1,054.2, respectively), and 7 patients with rRS values below the cutoff ranging from 0.39 to 0.80 (Supplementary Materials, Table 2). These results were in complete agreement with those of IHC.

Survival outcomes were compared between IDH-mutant and IDH wild-type patients. The potential prognostic effect of rRS was examined by quartiles. As expected, OS was significantly longer in patients with IDH1-mutant versus IDH1 wild-type tumors [median OS: 131 vs. 18 months, HR = 0.08; 95% confidence interval (CI), 0.04–0.15; Fig. 2A]. Similarly, PFS was significantly longer in patients with IDH1-mutant versus IDH1 wild-type tumors (median PFS: 43 vs. 13 months, HR = 0.22; 95% CI, 0.12–0.39; Fig. 2B). For the subgroup of patients with IDH1-mutant tumors, the 1p/19q co-deletion status did not differentiate survival. OS (HR = 0.80; 95% CI, 0.38–1.71; Fig. 2C) and PFS were both similar (HR = 0.73; 95% CI, 0.38–1.44; Fig. 2D).

Stratifying patients with IDH-mutant tumors by rRS, median OS was 200 months for the lowest quartile of rRS values (Q1) versus 66 months for the highest quartile of rRS values (Q4; HR = 0.33; 95% CI, 0.13–0.86; P = 0.02; Fig. 2E). Similarly, median PFS was 58 months for the lowest quartile of rRS values versus 25 months for the highest quartile of rRS values (HR = 0.56; 95% CI, 0.24–1.32; P = 0.18; Fig. 2F). These results suggest that higher rRS values may be associated with a worse prognosis. There was a similar but less conclusive trend when stratifying by total 2-HG (OS; HR = 0.63; 95% CI, 0.25–1.62; PFS: HR = 0.66; 95% CI, 0.29–1.50).

In most patients with suspected gliomas, the diagnosis is based on preoperative imaging and tissue acquired at the time of surgical resection. IDH status is a key part of the integrated genotype-phenotype diagnosis of gliomas and directs treatment decisions postoperatively. However, it is usually determined postoperatively using IHC for the canonical IDH1 R132 mutation. An important limitation of this method is that the small number of other noncanonical IDH1 mutations or mutations in IDH2 will not be detected. This is highlighted by the 3 patients in this study who were deemed to be IDH wild type by IHC, but were found to have high rRS ratios and subsequently confirmed to have noncanonical IDH1 mutations by sequencing. In addition to IHC, many centers rely on sequencing to determine IDH mutation status; however, the resources required and turnaround times can be deterrents for routine clinical use. In addition, sequencing does not provide information on the activity of mutant IDH enzymes. An evolving method for IDH determination is MRS, an imaging modality that can estimate the tumor 2-HG levels noninvasively and in real time based on the spectral pattern (27, 28). MRS accuracy varies widely depending on the tumor volume (sensitivity, 8%–91%), with very low sensitivity in smaller tumors. Another technology, desorption electrospray ionization mass spectrometry (DESI-MS), was recently used to detect 2-HG from tissue sections rapidly at the time of surgery (30, 31). Furthermore, DESI-MS could be used to determine surgical margins, thereby assisting in the decision-making during surgery. Similarly, matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF) was shown to be able to detect 2HG in tissue sections rapidly (32). However, none of these methods can distinguish between R-2-HG and S-2-HG. In contrast, the technique we present leverages the biochemical consequence of all mutant IDH mutations and uses the ratio of R/S as a biomarker.

To date, several direct and quantitative 2-HG detection methods have been reported. Gross and colleagues (5) were among the first to report higher 2-HG levels in the serum of patients with IDH1 R132H-mutant AML compared with wild-type AML. Similarly, in a cohort of 82 patients with de novo AML, median 2-HG levels were significantly higher in serum samples from IDH1/2-mutant patients, in comparison with serum samples (21.2 vs. 1.2 μmol/L) from IDH wild-type patients (33). From these studies, serum 2-HG levels appear to be a reliable surrogate of IDH1/2 mutations in patients with AML. Significantly higher 2-HG has been reported in the serum of patients with IDH-mutant cholangiocarcinoma, breast cancer, and colon cancer (11, 12, 34, 35). However, current evidence does not support the use of serum 2-HG for IDH mutation detection in gliomas. Capper and colleagues (22) reported no differences in preoperative serum 2-HG levels of WHO grade II and III IDH-mutant versus wild-type glioma (1.6 vs. 1.3 mmol/L). Lombardi and colleagues (24) reported similar 2-HG levels in the serum of patients with IDH1-mutant versus wild-type glioma (97.2 vs. 97.0 ng/mL). Interestingly, when 2-HG testing was extended to urine samples, patients with IDH wild-type glioma had a significantly higher concentration of 2-HG than those with IDH-mutant glioma (7.3 vs. 4.6 μg/mg), contrary to expectation. Results from our study do not support the use of total 2-HG, R-2-HG, or rRS as biomarkers for IDH mutation status from patient serum samples. This may be attributed to the very low diffusion rate of 2-HG from glioma cells due to the blood–brain barrier, in comparison with AML and cholangiocarcinoma where 2-HG is produced outside of the central nervous system (36). Kalinina and colleagues used a HPLC-MS method and found that R-2-HG in CSF of IDH-mutant patients increased by >17 fold compared with IDH wild-type patients (26). However, sensitivity and specificity were 84% and 90%, respectively.

Taken together, previous methods of 2-HG detection have reported inconsistent results and have mainly focused on detection of total 2-HG or R-2-HG, as opposed to the abnormal accumulation of R-2-HG relative to S-2-HG as a result of the neomorphic activity of IDH mutations. Our results revealed that assessment of tissue rRS reliably determined IDH mutation status with very high sensitivity and specificity. Outlier values were predominantly due to noncanonical IDH1 mutations, which were not detectable using conventional IHC but identified with sequencing. IDH mutation status was the most important predictor of tissue rRS, after accounting for patient age, gender, tumor grade, tumor recurrence, treatment latency, and 1p/19q co-deletion status.

IDH-mutant gliomas are associated with improved survivals compared with IDH wild-type gliomas. Recent data suggest that even IDH-mutant gliomas are a heterogeneous group, and the extent of tumor DNA methylation is associated with survival (37). The presumed mechanism of 2-HG–associated tumorigenesis is through inhibiting αKG-dependent dioxygenases including histone and DNA demethylases. It is reasonable to postulate that higher 2-HG levels are associated with increasing genomic abnormalities, hence poorer outcomes. Our results suggest that tissue rRS appeared to stratify prognosis further for patients with IDH-mutant tumors. Specifically, higher rRS values were associated with a worse prognosis. The median OS was 200 months for the lowest rRS quartile versus 66 months for the highest rRS quartile., These findings further argue for measuring both R-2-HG and S-2-HG, rather than the total tissue 2-HG.

The method reported here has advantages over other methods of determining IDH mutation. An elevated rRS is the biochemical consequence of all known IDH mutations, thereby avoiding sequencing each known mutation. Tissue S-2-HG serves as an inherent internal control, and the use of tissue rRS can overcome limitations associated with measuring total 2-HG only, such as cellularity and variable allele frequencies in the tissue tested and differences in mutant IDH enzyme activities. Most importantly, rRS could potentially provide prognostic information in addition to IDH status, and influence treatment decisions in these patients. For example, it is possible that patients with low rRS can be monitored closely postoperatively to minimize treatment associated toxicities and maintain quality of life for these patients, while those with high rRS may be considered for adjuvant therapies or enrolled into clinical trials. However, our method is based on the availability of glioma tissues and is not suitable for dynamic monitoring of changes in 2-HG as results of treatment or recurrence.

Another potential application of our HPLC-MS/MS–based method is the real-time determination of IDH status at the time of surgical resection. Current evidence suggests that aggressive resection may improve survival outcomes in patients with gliomas; however, the extent of resection must be balanced against surgical risk and long-term neurologic compromise. In a study of 335 glioma patients by Beiko and colleagues (38), 93% of patients with IDH1-mutant tumors underwent complete resection of enhancing disease compared to 67% of patients with IDH1 wild-type tumors, suggesting that IDH1-mutant tumors were more amenable to resection. Moreover, complete resection of IDH1-mutant tumors was associated with a significant OS improvement compared with incomplete resection (19.6 vs. 10.7 months). Kanamori and colleagues recently reported that tissue total 2-HG levels could be determined within 10 minutes using HPLC-MS/MS (39). The current run-time for our method is approximately 60 minutes. We demonstrated that this technique could be applied intraoperatively to determine IDH status and the result made available to the surgeon in real time to enable a personalized surgical approach for patients based on IDH mutation status. This will not be achievable with the current sequencing technology.

In this study, we did not have IDH1 or IDH2 sequencing information for the majority of patients. This limitation was mitigated by performing confirmatory 450K methylation analysis and by directly sequencing any tumor tissue with discordant values. In follow-up to this study, the prognostic significance of rRS needs to be confirmed in a larger cohort of patients.

In conclusion, we report a novel technique that can reliably detect both R-2-HG and S-2-HG in human glioma tissues. The rRS from glioma tissues differentiated IDH1-mutant versus IDH1 wild-type tumors with high sensitivity and specificity, whereas serum samples were unreliable. We demonstrated that this technique could be used for the real-time determination of IDH status intraoperatively, and thus may potentially guide clinical practice such as the extent of surgical resection. Most importantly, tissue rRS appears to stratify survival among patients with IDH-mutant tumors, suggesting that treatment decisions can be further individualized in these patients. This technique is advantageous over other methods of IDH mutation determination, such as IHC, sequencing, or various MRS methods, in that it can provide all this information simultaneously.

No potential conflicts of interest were disclosed.

Conception and design: R. Nejad, K.D. Aldape, G. Zadeh, E.X. Chen

Development of methodology: R. Nejad, G. Zadeh, E.X. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.-W. Sim, R. Nejad, F. Nassiri, K.D. Aldape, G. Zadeh, E.X. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.-W. Sim, R. Nejad, F. Nassiri, W. Mason, G. Zadeh, E.X. Chen

Writing, review, and/or revision of the manuscript: H.-W. Sim, R. Nejad, F. Nassiri, W. Mason, K.D. Aldape, G. Zadeh, E.X. Chen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Nejad, W. Zhang, G. Zadeh, E.X. Chen

Study supervision: K.D. Aldape, G. Zadeh, E.X. Chen

Financial support for this study was provided by the Princess Margaret Cancer Center Foundation.

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.