All chemicals and reagents were purchased from commercial sources and used without further purification. L2HG disodium salt and (+)-O,O′-diacetyl-l-tartaric anhydride (DATAN) were obtained from Sigma-Aldrich (St. Louis, MO, USA). D2HG disodium salt and dl-2-hydroxyglutaric acid-d₃ (dl-2HG-d₃) disodium salt were obtained from Cayman Chemical (Ann Arbor, MI, USA) and Toronto Research Chemicals (Toronto, ON, Canada), respectively. Guide cannulas and microdialysis probes (Meta-Quant) were purchased from Brainlink B.V. (Groningen, Netherlands). Mass spectrometry-grade chemicals and solvents were obtained from Fisher Scientific (Fair Lawn, NJ, USA) or Sigma-Aldrich (St. Louis, MO, USA). All other chemicals and materials were obtained from commercial suppliers.

Male NCr nude mice (nu/nu, 4 weeks old) were purchased from Taconic Farms (Germantown, NY, USA). All animal experiments were approved by the Institutional Animal Care and Use Committee according to the National Institutes of Health guidelines. U87 cells that had been stably transfected with the R132H IDH1 mutation were kindly provided by Agios Pharmaceuticals (Dr. Michael Su, Cambridge, MA, USA), and used as previously described (4). R132H_mIDH1/U87 cells were grown as a monolayer in Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum, 1× penicillin and streptomycin, and 300 μg/mL of G418 selection antibiotic and maintained in a humidified atmosphere of 5% CO₂ in air at 37°C.

Male NCr nude mice (5 weeks old) were anesthetized using an intraperitoneal dose (0.1 mL/10 g body weight) of 10% ketamine hydrochloride (100 mg/mL) and 8% xylazine (20 mg/mL) in water and placed in a stereotaxic frame. Tumor cells were prepared fresh from culture and kept in suspension (1 × 10⁵ cells/μL in phosphate-buffered saline) on ice until implantation. Each mouse had 3 μL of suspension injected (3 × 10⁵ tumor cells) into the right caudate putamen (0.7 mm anterior and 2.2 mm lateral from the bregma) at a depth of 2.5 mm. The injection site on the skull was sealed with sterile bone wax, and the incision was closed and fixed with wound glue. Mice received a standard mouse diet and water ad libitum and received 3 days of post-operative analgesic care that consisted of a daily 5-mg/kg subcutaneous dose of carprofen. Animals were weighed, assessed for body condition score, and monitored for neurological symptoms daily.

Thirteen days after tumor implantation (corresponding to at least 7 days before the microdialysis studies), guide cannulas were implanted. Mice were again anesthetized and placed in the stereotaxic frame in the same manner as above. Incision sites from the first surgery were opened, and the injection site located on the skull. A guide cannula was implanted at the injection site and secured with two anchor screws and dental cement. Mice were then individually housed for the duration of the study and received a standard mouse diet and water ad libitum and received 3 days of post-operative analgesic care that consisted of a daily 5-mg/kg subcutaneous dose of carprofen. Animals were weighed, assessed for body condition score, and monitored for neurological symptoms daily.

Microdialysis studies were completed 20–23 days after glioma cell implantation. This generally coincided with either a 20% loss in body weight, the first sign of neurological symptoms, or an observation of a body condition score of 2 or lower. For microdialysis, a harness was secured around the mouse’s torso, and the dummy probe from the guide cannula was removed and replaced with a brain microdialysis probe that contains two inlets; one referred to as the make-up flow rate and the other the low flow rate inlet that perfused the dialysis membrane. The probes were perfused at a total flow rate of 1 μL/min (both inlets initially at 0.5 μL/min from each pump) for at least 20 min with sterile water and artificial cerebrospinal fluid (147 mM NaCl, 2.7 mM KCl, 1.2 mM CaCl₂, and 0.85 mM MgCl₂). Flow rates were then adjusted as the probes were continued to be perfused for an additional 40 min at 0.9 μL/min with sterile water from the make-up pump and 0.1 μL/min with artificial cerebrospinal fluid from the low flow rate pump. After this equilibration period, the study commenced with brain microdialysis samples collected in 20-min intervals into a CMA (Harvard Apparatus, Holliston, MA, USA) 470 fraction collector that maintained the samples at 4–6°C for the duration of the study. Samples were transferred to sterile 200 μL polypropylene tubes after collection and stored at −80°C until they were analyzed.

Mice were randomly separated into the awake–asleep and asleep–awake cohorts. Mice that were awake first had brain microdialysis performed for 2 h, then were placed under anesthesia (1 L/min O₂, 2.5% isoflurane for induction, 1 L/min O₂, 1% isoflurane for maintenance). Sampling continued for an additional 2 h. Mice that were under anesthesia first were placed under anesthesia (same as in the other cohort), but immediately after the equilibration period of the probe. After a 2-h asleep period, the anesthesia was stopped, and a 1-h recovery period commenced prior to a 2-h awake sampling period. All mice regained consciousness by the end of the recovery period. At the conclusion of microdialysis, brain, tumor, and plasma were collected.

At the conclusion of brain microdialysis, with mice under heavy anesthesia (1 L/min O₂, 2.5% isoflurane), cardiac puncture was performed to obtain a whole blood sample using a 1-mL syringe and 23-G needle. Total recovered volume was immediately injected into a K₂EDTA-coated tube. The tube was inverted at least 12 times to ensure that all of the blood was exposed to anticoagulant, and then the sample was stored on ice until processing. The mouse was euthanized by cervical dislocation while under isoflurane anesthesia. Whole blood with anticoagulant was transferred to a 1.5- or 0.5-mL microcentrifuge tube as appropriate and centrifuged at 3000 rpm and 4°C for 15 min. Supernatant (plasma) was carefully removed from tube while leaving pellet and Buffy coat undisturbed and transferred to new 0.5-mL Eppendorf tubes in 100 μL aliquots, labeled, and stored at −80°C until analyzed.

Whole brain and tumor were harvested from each mouse. Indentation of the microdialysis probe was checked to ensure proper sampling location in the tumor. Brain and tumor samples were separated into three sections: tumor, brain (ipsilateral to tumor), and brain (contralateral to tumor). Sections were rinsed with ice cold sterile water and stored on ice until they were weighed. Each section was transferred to a sample tube to be weighed. For every gram of tissue present, 5 mL of ice cold sterile water was added, resulting in a final concentration of 200 mg tissue/mL. Homogenization was performed using a PolyTron PT 2100 tissue homogenizer set to 30,000 rpm for 5 s. The homogenizer was cleaned before and after each tissue sample. Homogenates were transferred to labeled 1.5-mL microcentrifuge tubes and stored at −80°C until ready to be analyzed.

The compartmental model used for D2HG simulations specifies three subcompartments – intracellular interstitial fluid and CSF – for both the glioma and normal cell compartments (see Results). This is consistent with the physiologically based pharmacokinetic models of tissue distribution of intracellular, interstitial fluid, and vascular subcompartments; however, a brain vascular compartment was not considered as an important aspect of the current model for D2HG since plasma concentrations are minimal, about 1 μM, and would not recycle across the blood–brain barrier in significant amounts. The model only considers the two most relevant cell spaces; the glioma cells and those normal cells in close proximity that would most likely be exposed to D2HG in the adjacent interstitial fluid compartment. The model also includes a terminal CSF compartment that represents the subarachnoid spaces that do not exchange material with the brain interstitial fluid compartments (19, 20). The model differential equations and parameters are provided in Table S5 in Supplementary Material. Physiological volumes were calculated from standard relationships based on an average measured tumor weight of 100 mg and assuming a density of 1 g/mL. The CSF bulk flow rate was taken from the literature (21), and the brain interstitial fluid flow (6.0 × 10⁻⁷ L/h) was an intermediate value in the range of 0.15–0.29 μL/min/g estimated in rats (22). Other than the influx rate of D2HG into the intracellular compartments, which was considered to be non-linear, all intercompartmental clearances and the systemic clearance from plasma are considered to be linear or concentration independent. The non-linear intracellular influx was represented as a Michaelis–Menten transport process based on the likely mechanism that NaDC3 is a key dicarboxylic acid transporter (23, 24). The model rate equations and parameters are provided in the Supplemental Material.

There were two cohorts of nine mice each studied in a crossover fashion with either an awake–asleep cycle or an asleep–awake cycle. Each awake and asleep study period was of a 2-h duration – 6 microdialysis samples – that included an equilibration period of 1-h prior to the start of the study and a 1-h re-equilibration period in the asleep–awake cohort to ensure the mice were fully conscious prior to the 2-h awake period sample collection. The concentrations of both D2HG and L2HG were monitored in BTIF by microdialysis in mice bearing intracerebral mIDH1/U87 brain tumors. Concentrations of L2HG were found to be under the lower limit of quantitation (1.35 μM) in all microdialysis samples collected. D2HG concentrations were measured in all microdialysis samples from each of the 18 mice (Figures 1A,B), wherein it can be seen that BTIF concentrations varied considerably between the mice with a mean ± SD of 122.3 ± 84.2 μM for all mice included in the study.

In order to analyze the D2HG BTIF concentrations, we defined a steady-state period to be the longest consecutive time period (a minimum of three samples or a 1-h period) of measured concentrations with a CV ≤ 25%. All 18 mice contained segments in both the awake and asleep states that fulfilled the steady-state inclusion criteria (Figure 2; Table 1). There was no sequence effect observed between the two cohorts (p = 0.98), indicative that the timing of anesthesia did not have an effect on the D2HG concentrations in the two states. Thus, there were a total of 36 study periods – 18 awake and 18 asleep – that produced mean D2HG BTIF concentrations in the asleep state of 117.6 versus 126.9 μM in the awake state.

Table 1 shows the concentrations of D2HG in total brain tumor homogenates, wherein it can be seen the values were in the millimolar range (assuming a tumor density of 1.00 g/mL), which is consistent with previous studies analyzing D2HG levels in R132H_mIDH1/U87 tumors (4, 26). The mean D2HG brain tumor concentration for all mice was 6.8 ± 3.6 mM, with a range from ~500 μM to >10 mM. No tumors contained quantifiable L2HG (LLOQ = 170 μM). The concentrations of D2HG in BTIF ranged from 0.5 to 7.6% (average = 2.4%) of the total tumor concentrations, and thus, quite variable between mice (Table 1). The tumor D2HG concentrations did not vary significantly between cohorts (p = 0.76), indicative that anesthesia did not impact the overall D2HG concentration in the tumors. The inter-animal variability in D2HG BTIF and tumor concentrations is likely due to the inherent heterogeneity of the tumors caused by regions of necrosis and proliferation that would affect the number of viable cells producing D2HG. In addition, the tumor homogenate concentrations reflect distribution into normal cells and its subpopulations that may be subjected to differential distribution. Finally, the heterogeneity is dependent on the actual collected sample; mIDH1/U87 tumors are well demarcated, so gross dissection to yield a tumor sample is not likely to introduce large errors; however, microdialysis sampling is dependent on the tumor environment and the restrictions imposed by the spatial resolution of microdialysis probe (27). Previous studies utilizing U87MG orthotopic xenograft models have displayed a heterogenous population of tumor cell types (28) as well as variable metabolic profiles among various tumor cells, usually correlated with hypoxic versus normoxic regions of tumors (29, 30).

The concentrations of D2HG were quantifiable (≥1.35 μM) in all non-brain tumor samples. The concentrations (assuming a density of brain tissue of 1.00 g/mL) of D2HG in ipsilateral and contralateral brain ranged from 4.2 μM to 2 mM and 3.0 to 48.6 μM, respectively (Table S4 in Supplementary Material). In contrast, the L2HG levels in ipsilateral and contralateral brain ranged from 1.35 to 10.63 μM and 1.35 to 14.65 μM, respectively; however, concentrations in four ipsilateral and three contralateral brain samples were less than the lower limit of quantitation (1.35 μM).

Plasma samples collected from each mouse (excluding mouse 10, for which a usable sample volume could not be obtained) contained both isomers of 2HG in quantifiable concentrations (≥0.27 μM). The plasma concentrations of D2HG (Figure 2; Table S4 in Supplementary Material) and L2HG ranged from 0.36 to 3.48 μM and 0.41 to 1.44 μM, respectively. Although plasma D2HG concentrations were very low compared with tumor concentrations, there was a strong positive correlation observed between them (r = 0.78, p < 0.001), suggesting that tumor D2HG distributes from the CNS into the systemic circulation.

The high D2HG concentrations in total tumor homogenate bated the question of the cellular localization of D2HG. It was surmised that the high concentrations were not likely confined to glioma cells and might accumulate in normal cells, particularly those in the adjacent microenvironment, using interstitial fluid as a conduit for cell-to-cell transfer. This general phenomenon has been considered previously (22), and within the guise of D2HG being tumorigenic, it is fair to assume its uptake into normal glial or precursor cells and cells needed for an immune response could be a contributing factor. A semi-mechanistic model of D2HG was developed (Figure 3A, see Materials and Methods) that consisted of an eight-compartment model with intracellular, interstitial fluid, and CSF compartments for both the glioma and normal cell compartments. The model parameters included physiological volumes and brain interstitial and CSF fluid flow rates with the remaining transport parameters calibrated to the observed D2HG concentrations in the brain tumor, BTIF, and plasma; being about 6000, 125, and 1.5 μM, respectively. The total tumor homogenate D2HG concentration represents an intracellular glioma cell compartment concentration of 10,000 μM (see Materials and Methods). The model assumed linear intercompartmental transfer of D2HG except for the influx from interstitial fluid – both in the glioma and normal cell compartments – to the intracellular compartment where Michaelis–Menten transport was assumed based on the likelihood D2HG was a substrate for NaDC3 (SLC13A3) (23, 24). The simulations show (Figure 3B) that appreciable D2HG concentrations – 9400 μM – may be attained at steady state in the adjacent normal cell intracellular compartment or 93% of those attained in the glioma intracellular compartment. It can be seen that there is a time lag for each compartment to achieve steady-state D2HG concentrations consistent with a concentration gradient from the glioma interstitial fluid compartment to the normal cell interstitial fluid compartment due to the convective flux of D2HG once it exits glioma cells into the interstitial fluid space. As a simulation exercise there is uncertainty in the parameter values and other parameter sets provide analogous simulated results; nonetheless, the results support the feasibility that high D2HG concentrations in glioma cells permit flux into the interstitial fluid compartment for transfer into normal cells. This possibility prompted an investigation of D2HG action on immune cells, both CD4 and CD8 T-cells.

To test the possible action of D2HG on immune cells, we examined the effect of D2HG on the proliferation and IFN-γ by human T cells that were activated by anti-CD3 antibody (Ab) and anti-CD28 Ab. Addition of D2HG caused a significant decrease (twofold to threefold) in expansion of these activated CD4⁺ and CD8⁺ T cells in a dose-dependent manner (Figure 4A). Interestingly, treatment with 500 μM D2HG enhanced the production of IFN-γ-producing T cells in cultures compared with untreated controls (Figure 4B). These results suggest that glioma cell-derived D2HG may interfere with T cell immune function.