Introduction

(d)-2-Hydroxyglutarate (D2HG) levels are increased in patients with the rare autosomal disorder D2HG aciduria [10] and mildly elevated in some other metabolic disorders such as multiple acyl-CoA dehydrogenase deficiency [12], dihydrolipoyl dehydrogenase deficiency [16], pyruvate decarboxylase deficiency [9] and pyruvate carboxylase deficiency [28]. Some of the patients with D2HG aciduria carry mutations in the (d)-2-hydroxyglutarate dehydrogenase gene (D2HGDH) [27].

The determination of D2HG levels has gained considerable interest upon the observation of mutations in the isocitrate dehydrogenase 1 (IDH1) and isocitrate dehydrogenase 2 (IDH2) genes conferring a novel enzymatic activity to the respective mutant proteins: instead of decarboxylating isocitrate to αKG, the mutant proteins convert αKG to D2HG [11]. Furthermore, mutated IDH1 protein loses the reverse conversion from αKG to isocitrate which was reported for IDH1 wild type [17].

IDH1 and IDH2 mutations occur in a heterozygous and somatic manner in diffuse astrocytomas and oligodendroglial tumors of WHO grades II and III [2, 13, 29], secondary glioblastoma [2, 20], acute myeloid leukemia [19, 24], angioimmunoblastic T cell lymphoma [6], chondrosarcoma [1], and cholangiocarcinoma [3]. IDH2 mutations have also been observed in patients with D2HG aciduria exhibiting no D2HGDH mutations [15].

IDH1 and IDH2 share high amino acid sequence identity. Mutations occur in the substrate binding region [25] and in IDH1 most frequently affect R132 and in IDH2 R140 and R172. The mutations in tumors are all somatic, always heterozygous and have in common that the novel enzymatic activity is conferred by the mutated IDH1 or IDH2 proteins. Thus, elevated D2HG levels might be a suitable surrogate marker for IDH1 and IDH2 mutations in selected human neoplasms.

D2HG in human tissues or body fluids is detected either by liquid-chromatography–mass spectrometry (LC/MS) or gas-chromatography–mass spectrometry (GC/MS). Both methods are labor intensive and time consuming and are not suitable for high-throughput analysis or screening. Not be known widely, D2HG previously has been analyzed in must, wine and ardent spirits [26] based on utilization of the enzyme (d)-2-hydroxyglutarate dehydrogenase (HGDH) from the anaerobic bacterium Acidaminococcus fermentans [4]. Thus inspired, we set out to develop a sensitive enzymatic assay allowing for cost-effective screening of human tissues and body fluids for D2HG.

Materials and methods

Overview on detection of D2HG

The two-step coupled enzyme reaction is shown in Fig. 1. In a first step, HGDH bidirectionally converts D2HG and αKG. Coenzyme in this reaction is the hydride acceptor NAD+. Buffer conditions, substrate and coenzyme concentrations are set to favor the conversion of D2HG to αKG thus resulting in the accumulation of NADH [4, 26]. In a second step, hydrogen from NADH is transferred by diaphorase to non-fluorescent resazurin which results in the production of fluorescent resorufin [7]. Fluorometric detection is carried out with excitation at 540 ± 10 nm and emission of 610 ± 10 nm.

Fig. 1
figure 1

Scheme for enzymatic assay. a (d)-2-Hydroxyglutarate dehydrogenase (HGDH) catalyzes the reduction of NAD+ to NADH by oxidation of (d)-2-hydroxyglutarate to α-ketoglutarate. b NADH-detection by diaphorase/resazurin system. Fluorescent product resorufin is exited at 540 nm and detected at 610 nm

Full size image

Expression and purification of HGDH

The gene encoding 2-hydroxyglutarate dehydrogenase from A. fermentans was cloned into pASK-IBA7plus expression plasmid (IBA GmbH, Göttingen, Germany) by one of us [30]. pASK-IBA7plus carries carbenicillin resistance and N-terminal to the cloning site a Strep-tag®. Expression is induced by 0.4 μM anhydrotetracycline (IBA GmbH). E. coli (BL21) cells were transfected with the plasmid and plated on LB agar (1.5 % agar, 1 % tryptone, 1 % NaCl, 0.5 % yeast extract) containing carbenicillin (100 μg/l) at 37 °C.

For production of HGDH, cells were grown aerobically in 100 ml of LB medium until the culture reached an OD of 0.6. Expression in the bacteria was induced with anhydrotetracycline and bacteria were grown overnight on a shaker (Certomat R, B. Braun AG, Melsungen, Germany) at 220 rpm and 37 °C. Bacteria were harvested, sedimented and the resulting pellet stored at −80 °C.

The bacterial pellet was suspended in 3 ml 20 mM potassium phosphate buffer, pH 7.4 and sonicated for 15 min on ice (Branson Cell Disruptor B15, Danbury, CT, USA) at 20 % power. The fragmented bacterial extract was centrifuged at 4,000g for 30 min at 4 °C (Heraeus Megafuge 1.0R, Thermo Scientific, Osterode, Germany). In a cold room set to 4 °C, the supernatant was applied in 500-μl steps to a prewashed Strep-tag® column (IBA GmbH). Washing and elution was performed according to the supplier’s instruction in six steps with 0.1 ml elution buffer each. 5 μl of each of the 6 elutions were separated by 4–12 % SDS-PAGE (Invitrogen, Carlsbad, CA, USA). Coomassie staining was employed to display induction of HGDH expression. HGDH-positive eluents were pooled and protein concentration was determined by Pierce BCA protein assay kit (Thermo Scientific, Houston, TX, USA). The protein concentration was adjusted to 0.1 μg/μl by adding ddH2O and the HGDH was stored in 100 μl aliquots at −80 °C.

Preparation of fluids

The preparation of the medium supernatant or urine differed from that of serum. Cell culture supernatant or urine was deproteinized using deproteinization kit (Biovision, Mountain View, CA, USA). Briefly, to 100 μl of sample 20 μl of perchloric acid solution was added, following incubation on ice for 5 min. Samples were centrifuged for 2 min at 13,000g (Heraeus Fresco 21, Thermo Scientific). This procedure was performed in triplicate (n = 3). From each supernatant 96 μl were transferred to a new tube and 4 μl of neutralization solution were added. After incubation on ice for 5 min, samples were centrifuged for 2 min at 13,000g. Supernatants were collected and stored at −20 °C.

Serum was obtained by centrifugation of 5 ml blood collected in Sarstedt S-Monovette Serum Gel tubes (Sarstedt, Nümbrecht, Germany) at 2,500g for 10 min at 20 °C. Serum was stored at −80 °C. To 100 μl serum 1 μl Proteinase K solution (40 μg/μl) (Promega, Madison, WI, USA) was added and incubated over night at 37 °C. Deproteinization was performed as described in triplicate (n = 3) but due to the higher protein concentration in serum 25 μl of perchloric acid solution was added, and 95 μl of the supernatant was mixed with 5 μl neutralization buffer.

Preparation of cells and fresh frozen tissues

Confluent cells were collected from 6-well plates (BD Bioscience, San Jose, CA), dissolved in 220 μl cell lysis buffer (Cell signaling, Danvers, MA, USA) and freeze-thawed three times. To remove cell debris samples were centrifuged for 5 min at 13,000g.

Fresh frozen tissue (mean 72 mg) was dissolved in 340 μl cell lysis buffer (Cell signaling) and homogenized by sonication for 1 min at 20 % power (Branson). To this suspension 3 μl of Proteinase K (40 μg/μl) was added followed by incubation for 2 h at 37 °C.

Protein concentration was determined by Pierce BCA protein assay kit (Thermo Scientific). 100 μl of the lysate was deproteinized as described for cell culture medium or urine in triplicate (n = 3). Supernatants were collected and stored at −20 °C.

Preparation of FFPE tissue

Prior to extraction all tissues used were confirmed by light microscopy to have a tumor content exceeding 80 %. Ten 10-μm sections were cut from paraffin blocks and collected in a tube. The cut surface of the tissues approximated 400 mm2 but was estimated in detail for every tissue preparation to calculate used tissue volume. D2HG content was calculated in pmol per mm3 tissue volume. Deparaffinization was performed by adding 1 ml xylene (Roth, Karlsruhe, Germany), followed by vortexing and centrifugation for 5 min at 13,000g. Pellets were re-suspended in 1 ml xylene by vigorous mixing followed again by centrifugation for 5 min at 13,000g. The deparaffinized tissue pellets were dried in the open tube under a fume hood for 4 h. Dried pellets were solved in 340 μl cell lysis buffer and homogenized by sonication for 1 min at 20 % power (Branson). To this suspension 3 μl of Proteinase K (40 μg/μl) was added followed by incubation for 2 h at 37 °C. Deproteinization of 100 μl sample in triplicate (n = 3) was performed according to the procedure described for human serum.

Standards

Different tissues required individual standards. Standard for D2HG detection in medium supernatant, urine, or serum was prepared with concentrations of 0.5, 1, 2.5, 5, 7.5, 10, 25 and 50 μM D2HG in fresh medium, urine or serum. Standard for D2HG detection in cells, frozen tissue or FFPE tissue was prepared with concentrations of 0.5, 1, 2.5, 5, 7.5, 10, 25 and 50 μM D2HG in ddH2O. All standards were prepared with the identical procedure of the corresponding samples, i.e. medium supernatant, urine, serum or cells, fresh frozen tissue or FFPE tissue preparations.

D2HG assay

Total reaction volume was 100 μl. 10 ml of assay solution was prepared freshly for each 96-well plate subjected to D2HG assay. Assay solution contained 100 mM HEPES pH 8.0, 100 μM NAD+ (Applichem, Darmstadt, Germany), 0.1 μg HGDH, 5 μM resazurin (Applichem) and 0.01 U/ml diaphorase (0.01 U/ml, MP Biomedical, Irvine, CA, USA). Just before use 75 μl of assay solution were added to 25 μl sample volume and incubated at RT in the dark for 30 min in black 96-well plates (Thermo Scientific). Fluorometric detection was carried out in triplicate with 25 μl deproteinized sample analyzed in each reaction with excitation at 540 ± 10 nm and emission of 610 ± 10 nm (FLUOstar Omega, BMG Labtech, Offenburg, Germany).

Alternative method to detection with diaphorase/resazurin

Although in this paper all detections rely on fluorometric detection via diaphorase/resazurin, we also successfully detected NADH generation by a colorimetric method employing the electron acceptor PMS (N-Methylphenaziniummethylsulfat, Applichem) and the tetrazolium salt XTT [2,3-bis(2-methoxy-4-nitro-5-sulfo phenyl)-2H-tetrazolium-5-carboxyanilite disodium salt, Applichem]. The assay solution contained 100 mM HEPES pH 8.0, 100 μM NAD+, 0.1 μg HGDH, 8.25 μM PMS and 125 μM XTT. XTT is reduced with the help of the electron transfer agent PMS to the highly colored soluble formazan. The assay reaction was incubated at RT in the dark for 30 min in 96-well plates. Colorimetric detection was carried out with the FLUOstar Omega at 490 or 450 nm.

Patient material, GC/MS and statistics

Fresh frozen tumor tissue from 15 glioma patients with predetermined IDH status was selected from the archive of the Department of Neuropathology, Heidelberg. From 11 of these patients, formalin-fixed paraffin-embedded tissue was also examined (Supplementary table 2). Serum from 12 AML patients was obtained from the University of Dresden. Of these, five patients (P1–P5) have been tested with IDH1/IDH2 mutations while eight patients (P6–P13) were wild type (Supplementary table 2).

Sample preparation for the measurement of D2HG with GC/MS was performed as previously described [23, 24].

Results

Selection of buffer and optimization of pH, HGDH, diaphorase and resazurin concentrations

Different buffers were tested including 0.1 M Tris–HCl pH 8.0, 0.1 M K+-phosphate pH 8.0, 0.1 M HEPES pH 8.0 and 0.1 M MOPS pH 7.9. Best results were obtained with 0.1 M HEPES pH 8.0 (Supplementary figure 1a). Therefore, all further analyses were carried out with 0.1 M HEPES pH 8.0. Next, the effect of pH on the performance of our assay was assessed. D2HG concentrations of 5 and 50 μM were used. Optimal activity was achieved for pH range from 7.4 to 8.0 (Supplementary figure 1b). Resorufin has previously been shown to yield best fluorescence signal in neutral and basic solutions [5]. Therefore, we chose pH 8.0 to perform all further assays.

To determine the optimal concentration of HGDH, amounts ranging from 0.01 to 0.25 μg of recombinant enzyme were tested. For the low concentration of 5 μM D2HG, amounts higher than 0.05 μg HGDH did not increase fluorescent readout, and, for the high concentration of 50 μM D2HG, amounts higher than 0.1 μg HGDH did not increase fluorescent readout. Hence, we used 0.1 μg recombinant HGDH per microtiter plate well in the assay, which turned out to be sensitive and economical (Supplementary figure 1c).

To test for the optimal diaphorase concentration, the assay was performed with 0.1 μg recombinant HGDH and D2HG concentrations of 5 and 50 μM. Best results were obtained for a diaphorase concentration of 0.1 U per well (Supplementary figure 1d).

Resazurin accounts for weak background fluorescence. We determined the resazurin concentration with the best signal-to-noise ratio in dependency of D2HG as follows: employing 0.1 μg recombinant HGDH and 0.1 U diaphorase per well, we tested fluorescent readouts from 0.5 and 5 μM (Supplementary figure 2a). The best signal-to-noise ratio was determined by dividing values of control readout containing no D2HG by values obtained in readings with D2HG. In this range, best signal-to-noise ratio was seen for concentrations of 1 μM resazurin (Supplementary figure 2b). Higher D2HG concentrations required higher resazurin levels. For 50 μM D2HG (Supplementary figure 2c) best signal-to-noise ratio was seen with 5 μM resazurin (Supplementary figure 2d). These findings suggest adjusting resazurin concentration in the assay to the expected D2HG values in the samples to obtain best results.

Determination of the range for D2HG concentrations suitable for assay

To estimate the limit of detection (LOD) and the limit of quantification (LOQ) [18], the fluorescence intensity of nine negative control samples (enzyme assay mixture without D2HG) and a standard curve of D2HG dissolved in ddH2O covering the concentration range from 0.5 to 10 μM were measured. LOD and LOQ are defined as LOD = 3σ/S and LOQ = 10σ/S, where σ is the standard deviation (SD) of the background (blank) and S is the slope of the standard curve. The SD (σ) of the measured background fluorescence was calculated with 359.4 fluorescence intensity units. We determined a linear fitted standard curve for D2HG with S = 8,180 fluorescence intensity units/μM D2HG. According to this data LOD is 0.13 μM and LOQ is 0.44 μM. LOD and LOQ were also estimated for D2HG dissolved in cell medium with and without phenol red, human serum and urine (Supplementary table 1).

Another possibility to assess the quality of an assay is the calculation of the Z′-factor. This factor is defined as Z′ = 1 – ((3σc+ + 3σc−))/|μc+ − μc−|) [31]. μc+ and μc− are the means of positive control and negative control signals; σc+ and σc− are SDs of the readings. These values were measured from 12 separate wells with enzymatic assay mixture alone and 50 μM D2HG. We calculated Z′ as 0.95. Per definition an excellent assay has a Z′-factor of 0.5 to <1 [31]. This indicates that our assay is suitable for the measurement of D2HG in the range of 0.44–50 μM (Supplementary table 1).

Assay validation, stereoselectivity of HGDH and proof of principle for assay

The reaction specificity was determined by omitting components from the reaction mixture. Only the complete reaction mixture was able to generate a fluorescence signal corresponding to the D2HG applied (Fig. 2a). We also examined the variation of test results on the same samples over time. D2HG was added to final concentrations of 0.5, 5 and 50 μM to the cell culture medium and to human serum. Intra-run and inter-run variations were satisfactory (Supplementary table 3).

Fig. 2
figure 2

a Omission of either HGDH, diaphorase or NAD+ results in background fluorescence. b and c Readout of D2HG in cell culture medium by enzymatic assay plotted against readout by GC/MS. d and e Readout of D2HG in human blood serum by enzymatic assay plotted against readout by GC/MS

Full size image

Acidaminococcus fermentans HGDH turned out to be highly stereoselective under the assay conditions. D2HG was metabolized effectively at 5 and 50 μM, whereas L2HG was not (Supplementary figure 3c).

Tests were performed with D2HG dissolved in cell medium and in human serum. D2HG concentrations in both were 0, 1, 5, 10, 25, 100, 250 and 500 μM. The serial dilutions were examined by both, the enzyme assay and GC/MS. Readings exhibited a very high degree of accordance in medium (Fig. 2b; for low concentrations Fig. 2c) and serum (Fig. 2d; for low concentrations Fig. 2e).

D2HG in urine samples generally produced a higher background than medium and serum-based measurements (Supplementary figure 3a).

We also tested a colorimetric detection method employing PMS and XTT which can be measured by absorption at 450 or 490 nm (Supplementary figure 3b). In direct comparison fluorometric detection is more sensitive than colorimetric detection.

Applications

We applied this assay to in vitro systems and to material from patients with conditions resulting in elevated D2HG concentrations.

We analyzed IDH wild-type glioma cell line LN229, a set of transfected LN229 cells with stable but different expression levels of mutant IDH1 R132H protein and two chondrosarcoma lines, one of them with the IDH1 R132C mutation. Wild-type LN229 and the chondrosarcoma line SW872 without IDH mutation exhibited basal levels of D2HG. The transfected LN229 lines H3, H49, H77, H104, H114 and C67 as well as the chondrosarcoma line HT1080 with the IDH1 R132C mutation exhibited strongly elevated levels of D2HG (Fig. 3a).

Fig. 3
figure 3

D2HG detection in cell lysates (a) and the corresponding cell culture supernatant (b). IDH1/IDH2 wild-type glioma line LN229 and LN229 clones transfected with IDH1 R132H (H3, H49, H77, H97, H104, H114) or R132C (C67); chondrosarcoma line SW872 is IDH1/IDH2 wild type and chondrosarcoma line HT1080 with IDH1 R132C mutation. c D2HG in fresh frozen tumor. Patients 1–6 have glioma with IDH1 mutation. P7–P15 have glioma with wild type of IDH1/IDH2. d Analyses on formalin-fixed paraffin-embedded (FFPE) tissue from these patients

Full size image

D2HG was determined in the medium of the same set of cells. No D2HG was detected in medium of normal LN229 and SW872, both not harboring IDH mutations. Increased levels of D2HG were detected in all cell lines with IDH mutations and the D2HG concentrations correlated with intracellular levels (Fig. 3b).

We analyzed fresh frozen tumor tissue from 15 patients with glioma. Patients P1–P6 were determined by Sanger sequencing to carry IDH mutations (Supplementary table 2). A control set of gliomas P7–P15 were wild type for IDH. Basal levels of D2HG were traceable in 8 of 9 control gliomas with a mean value of 6.1 ± 2.5 pmol/mg tissue. All glioma tissues with mutations exhibited strongly increased D2HG levels ranging from 48.1 to 249 pmol/mg tissue (Fig. 3c).

Next we analyzed FFPE tissue from patients with glioma. Numbering of tumors corresponds to the fresh frozen glioma set. Increased D2HG levels ranging from 51 to 189 pmol/mm3 were determined in gliomas P1–P5 carrying IDH mutations while only basal D2HG averaging 4.1 ± 1.6 pmol/mm3 levels were measured in gliomas P7–P12 wild type for IDH (Fig. 3d).

We determined D2HG in serum of AML patients with IDH1 or IDH2 mutations by both, enzymatic assay and GC/MS. Both approaches detected a strong increase in D2HG in AML patients carrying IDH1/IDH2 mutations. In AML patients without IDH1/IDH2 mutations only basal levels were detected averaging 4.8 μM by GC/MS and 3.7 μM by the enzyme assay (Fig. 4).

Fig. 4
figure 4

D2HG in serum of acute myeloid leukemia (AML) patients. GC/MS data (black bars) and enzymatic assay data (gray bars) demonstrate high concordance. AML with IDH1/IDH2 mutations (P1-AML to P5-AML) exhibit high D2HG levels while AML wild type for IDH1/IDH2 exhibit basal D2HG levels only

Full size image

Discussion

An enzymatic assay for the detection of D2HG in tumor tissue, serum, urine, cultured cells and culture supernatants was developed and applied based on a previously established test for glutamate fermentation [4].

GC/MS has been established as gold standard for detection of D2HG in tissues [23]. Therefore, a set of identical probes consisting of serial dilutions of D2HG in cell culture medium and serum were tested. Comparative analyses revealed high accordance of readings by GC/MS and the enzymatic assay (Fig. 2b–e). Of importance, this accounts also for the low levels of D2HG corresponding to those encountered in serum and tissues from patients inconspicuous in respect to D2HG. Consequently, the comparison between the enzymatic test and GC/MS was extended to human serum samples from AML patients with and without IDH1/IDH2 mutations. Sera from five patients with IDH1/IDH2 mutations were found to contain increased D2HG levels exhibiting comparable readings with both, GC/MS and the enzymatic assay, whereas only basal levels were detected in eight patients without mutations (Fig. 4). In contrast to AML, we did not detect elevated D2HG in serum or cerebrospinal fluid from patients with IDH1 mutated glioma (data not shown). Feasibility of the enzymatic assay for analysis of cell lysates, corresponding culture medium and fresh frozen tumor tissue was demonstrated (Fig. 3a–c). We previously reported detection of D2HG by GC/MS in paraffin-embedded tumor tissue from patients with IDH1/IDH2 mutated glioma [23]. Here, we demonstrate feasibility of the enzymatic test for this tissue preparation (Fig. 3d). Thus, the enzymatic assay proved to be an alternative to GC/MS offering comparable sensitivity.

While readouts of GC/MS and the enzymatic test did not differ, both approaches carry inherent advantages and shortcomings. GC/MS has the considerable advantage of multiple metabolites being tested in a single round of analysis. Screening for glutaric aciduria type 1 and D2HG aciduria include the detection of hydroxyl organic acids by GC/MS, i.e. glutaric and 3-hydroxyglutaric or D2HG acid, respectively [14]. Thus, the enzymatic approach may not be of major significance for screening metabolic disease in the newborn. On the other hand, GC/MS is quite time consuming and parallel analyses depend on employment on multiple sets of machinery. The enzymatic test can effectively be performed on 96-well microtiter plates. In acquired diseases such as tumor lesions, the detection of D2HG by an enzymatic test is of interest. All different tumor-associated mutations in IDH1 and IDH2 have a common denominator: The mutant proteins catalyze very effectively the reduction of αKG to D2HG by NADPH [11]. This might allow taking elevated D2HG levels as a surrogate marker for tumor-relevant IDH1 and IDH2 mutations. Approximately 90 % of the IDH1/IDH2 mutations in primary brain tumors are of the IDH1 R132H mutation type [13] and, therefore, can be efficiently detected by a monoclonal antibody recognizing this most frequent mutation [8]. However, the remaining 10 % disperse among the rare IDH1 and the few IDH2 mutations which currently can only be detected by sequencing methods. These rare IDH1 and IDH2 mutations are readily picked up by D2HG enzymatic assay because the rare mutations generally are even more effective in the catalysis of D2HG (unpublished data).The alterations in IDH1 and IDH2 in AML, chondrosarcoma and biliary carcinoma are much more evenly distributed among the different mutations types [1, 3, 19, 21]. In these tumor entities, a universal surrogate marker might be even more appreciated. Further, D2HG detection might be of great interest upon the development of selective inhibitors of mutant IDH1 and IDH2 proteins. Given the uniform response of different mutations, it might be quite possible that inhibitors will be developed targeting several of the tumorigenic mutations in IDH1 and IDH2, thereby potentially complementing a universal surrogate marker with a universal IDH1/IDH2 mutation-selective drug. In some of the neoplasms such as AML, monitoring of D2HG might be an important parameter for remission or relapse of disease [22]. In these examples the simple and cost-efficient detection of D2HG by a specific enzymatic test may be very effective. Given the high investment required for GC/MS machinery, the enzymatic test is considerably less expensive to perform. It needs to be awaited which will be the favored applications for the tests.

In conclusion, we developed a sensitive, rapid and inexpensive enzymatic test for the detection of D2HG in fluids or solid tissues. This test has special strengths in simultaneous analyses of large sample sets and for general screening.