Simple Esterification of [1-¹³C]-Alpha-Ketoglutarate Enhances Membrane Permeability and Allows for Noninvasive Tracing of Glutamate and Glutamine Production

Abstract

Alpha-ketoglutarate (α-KG) is a key metabolite and signaling molecule in cancer cells, but the low permeability of α-KG limits the study of α-KG mediated effects in vivo. Recently, cell-permeable monoester and diester α-KG derivatives have been synthesized for use in vivo, but many of these derivatives are not compatible for use in hyperpolarized carbon-13 nuclear magnetic resonance spectroscopy (HP-¹³C-MRS). HP-¹³C-MRS is a powerful technique that has been used to noninvasively trace labeled metabolites in real time. Here, we show that using diethyl-[1-¹³C]-α-KG as a probe in HP-¹³C-MRS allows for noninvasive tracing of α-KG metabolism in vivo.

Graphical Abstract

Alpha-ketoglutarate (α-KG) is a key intermediate in the citric acid cycle and a precursor for glutamate and glutamine. Atypical metabolism of α-KG has been linked to altered protein synthesis and catabolism, as well as amplified malignant progression.^1,2 In addition, aberrant metabolism of α-KG is implicated in certain cancers, such as those driven by isocitrate dehydrogenase 1 (IDH1) mutations.³ α-KG is quite hydrophilic, however, with its relative impermeability through cell membranes limiting the study of its effects.⁴ In 2007, MacKenzie and colleagues demonstrated that while the addition of extracellular α-KG did not significantly increase intracellular α-KG, the addition of monosubstituted α-KG esters allowed for over 2-fold increase in intracellular α-KG concentrations.⁴ In 2015, Zengeya and colleagues developed a new synthesis of monosubstituted and disubstituted 3-(trifluoromethyl)benzyl (TFMB) α-KG and showed that the addition of these cell-permeable TFMB-α-KG esters increased the activity of α-KG-dependent dioxygenases in vitro.⁵

Hyperpolarized carbon-13 nuclear magnetic resonance spectroscopy (HP-¹³C-MRS) has become an important tool for studying real-time metabolism in vivo. The most common use of this technique has been to track lactate production after injection with [1-¹³C]-pyruvate.⁶ Since cancer cells have a tendency to be highly glycolytic, tracking the flux of pyruvate to lactate using HP-¹³C-MRS has been used, among other applications, to noninvasively predict tumor aggressiveness and track progression following treatment in both pancreatic ductal adenocarcinoma and prostate cancer xenograft tumors.^7,8 [1-¹³C]-fumarate, [1-¹³C]-lactate, and other 13C-labeled metabolites have also been used for noninvasive tracing of real-time metabolic flux in vivo by HP-¹³C-MRS.^9–12

For a molecule to be a successful HP-¹³C-MRS probe, it must typically have a low molecular weight and have the labeled carbon as a carbonyl or carboxylic acid.¹³ For a probe to be useful in vivo, the molecule must be delivered to the desired tissue and cross the cellular membrane rapidly and at high concentrations to allow for detection.¹³ Since high concentrations of the probe are needed, the probe must be nontoxic and maintained near physiologic pH.

One of the reasons [1-¹³C]-pyruvate is an ideal probe for HP-¹³C-MRS is because it can be rapidly transported into cells or across the blood brain barrier by monocarboxylate transporters (MCTs).^14–17 While mitochondrial transporters for α-KG have been characterized,¹⁸ there are no dicarboxylate transporters on cell membranes. As α-KG therefore relies on passive diffusion to cross cell membranes, passive permeability is an important factor in determining its ability to be used as a hyperpolarization probe. The previously developed α-KG esters are not compatible with hyperpolarization, as they are slightly heavier than the desired molecular weight and contain benzyl side groups that are not stable in the radical conditions needed for polarization.^4,5 Here, we propose that a simple esterification to diethyl-α-KG will increase permeability without impeding hyperpolarization, allowing for α-KG metabolism to be tracked in vivo using HP-¹³C-MRS. Accordingly, we synthesized diethyl-[1-¹³C]-α-KG, which allows for the tracing of α-KG metabolism in vivo.

In agreement with previously published work, the HP-¹³C-MRS of [1-¹³C]-α-KG phantom showed two major peaks at 172 and 181 ppm, corresponding to [1-¹³C]-α-KG and [1-¹³C]-α-KG hydrate, respectively. The natural abundance C5 and C2 peaks for α-KG were detected as minor peaks at 184 and 208 ppm, respectively (Figure 1A). After a tail vein injection of hyperpolarized-[1-¹³C]-α-KG into HCT116 IDH1 R132H xenograft tumor-bearing mice, the peaks for α-KG were easily identified but broad (Figure 1B). A peak at 175 ppm corresponds to the cyclized confirmation of α-KG. The presence of this peak is consistent with previously published work.^19,20

Figure 1.

[1-¹³C-α-KG] is not permeable enough for glutamine tracing in vivo. (A) Hyperpolarization of commercial [1-¹³C]-α-KG (from ref 20). (B) Representative spectra from HCT116 IDH1 R132H subcutaneous xenograft tumor after injecting with [1-¹³C]-α-KG. (C, D) Post-mortem mass spectrometry analysis of xenograft tumors measuring glutamate (panel (C)) or 2-HG (panel (D)) following injection with either unlabeled α-KG (Ctrl) or [1-¹³C]-α-KG. N = 3.

A small peak at 177 ppm was detected, suggesting that [1-¹³C]-glutamate may be present. Post-mortem mass spectrometry confirmed that [1-¹³C]-α-KG was metabolized to [1-¹³C]-glutamate in the xenograft tumors, as concentrations of [1-¹³C]-glutamate were significantly increased when comparing xenograft tumors injected with [1-¹³C]-α-KG to control (Figure 1C, N = 3 for each group, p = 0.019).

Although a small glutamate peak was present, the low permeability of α-KG resulted in low product formation, making quantification and the resolution of glutamate and glutamine difficult. Mass spectrometry analysis of the xenograft tumors confirmed the insignificant accumulation of labeled [1-¹³C]-α-KG as there was no difference in [1-¹³C]-2-HG concentration in IDH1 R132H xenograft tumors with or without the addition of ¹³C-labeled α-KG (Figure 1D).

Since α-KG is not actively transported across cell membranes,⁴ the negligible degree of product formation is likely a consequence of its reduced intracellular concentration. Since esterification with TFMB alcohol has been shown to increase permeability, we hypothesized that simple esters would also increase the permeability of α-KG without compromising the ability of the α-KG ester to be hyperpolarized. Using a parallel artificial membrane permeability assay to measure the ability of a probe to cross an artificial membrane barrier, we showed that diethyl-α-KG has a 90-fold increase in permeability over α-KG (Figure 2A). Diethyl-α-KG also showed an almost 5-fold increase in permeability over dimethyl-α-KG (DM-α-KG), which is currently used as a cell-permeable derivative of α-KG.^21–23 Diethyl-α-KG also forms ethanol after ester cleavage, which is less toxic than the methanol side product formed by cleavage of DM-α-KG.²⁴

Figure 2.

Signficantly increased permeability of diethyl -α-KG and synthesis and hyperpolarization of diethyl-[1-¹³C]-α-KG. (A) α-KG, dimethyl-α-KG, and diethyl-α-KG permeability measured via a parallel artificial membrane permeability assay. (B) Reaction scheme for the synthesis of diethyl-[1-¹³C]-α-KG. ¹²C enriched carbons are not labeled for clarity purposes. See the Supporting Information (SI) for full synthesis and details. (C) Hyperpolarization and (D) spectra of hyperpolarized diethyl-[1-¹³C]-α-KG.

Diethyl-[1-¹³C]-α-KG is not commercially available. Accordingly, we modified our previously published synthesis of [1-¹³C-5-¹²C]-α-KG²⁰ to produce diethyl-[1-¹³C-5-¹²C]-α-KG. Because our starting material and intermediates from our previous synthesis were enriched with ¹²C at the fifth position, we synthesized diethyl-α-KG that was also enriched with ¹²C at the fifth position. As the 5-¹²C has no relevance to the remainder of this discussion, we will subsequently refer to diethyl-[1-¹³C-5-¹²C]-α-KG as DE-[1-¹³C]-α-KG. In brief, the morpholino amide intermediate 3 was formed from 1 and 2 via the previously described oxidative-Stetter reaction in 90% yield (Figure 2B). In the final step, DE-[1-¹³C]-α-KG (4) was formed via a transesterification of 3 in 78% yield. The full synthesis of DE-[1-¹³C]-α-KG is presented in the SI. As with our previous synthesis, DE-[1-¹³C]-α-KG was synthesized in high yields and on gram scale.²⁰

DE-[1-¹³C]-α-KG was successfully hyperpolarized (Figure 2C). DE-[1-¹³C]-α-KG was also sufficiently polarized within 2 h, whereas [1-¹³C]-α-KG required a minimum of 5 h for sufficient polarization (data not shown). The in vitro T1 value of DE-[1-¹³C]-α-KG measured at 3 T was 38.8 ± 0.4 s, comparable to the T1 value of [1-¹³C]-α-KG (43.3 ± 0.3 s). The resulting HP-¹³C-MRS showed two major peaks corresponding to DE-[1-¹³C]-α-KG (163 ppm) and DE-[1-¹³C]-α-KG hydrate (174 ppm) (Figure 2D).

Following the successful hyperpolarization of DE-[1-¹³C]-α-KG, we proceeded to test its ability to measure α-KG metabolism in vivo. With DE-[1-¹³C]-α-KG in IDH1 mutant xenograft tumors, the esters were rapidly cleaved to generate [1-¹³C]-α-KG, with the largest peak corresponding to [1-¹³C]-α-KG (172 ppm) and only small peaks remaining for DE-[1-¹³C]-α-KG (174 and 163 ppm for DE-[1-¹³C]-α-KG hydrate and DE-[1-¹³C]-α-KG, respectively; see Figure 3A). The increased permeability of DE-[1-¹³C]-α-KG was corroborated as the downstream metabolites [1-¹³C]-glutamate (178 ppm) and [1-¹³C]-glutamine (177 ppm) were seen in the resulting HP–¹³C-MRS (Figure 3A).

Figure 3.

Diethyl-[1-¹³C]-α-KG allows for tracing of α-KG metabolism in vivo. (A) Representative 1D spectra of HCT116 xenograft tumor after injection with diethyl-[1-¹³C]-α-KG. (B) Time course analysis of HCT116 xenograft tumor after injection with diethyl-[1-¹³C]-α-KG.

Time-course tracing of these peaks showed differential signal decay of glutamate and glutamine, suggesting that they were not contaminants of the hyperpolarized reagent (Figure 3B). With the improved permeability of DE-[1-¹³C]-α-KG, the [1-¹³C]-glutamate and [1-¹³C]-glutamine peaks were able to be resolved. The production of [1-¹³C]-glutamate was seen as a peak at 178 ppm, which appeared 43 s after injection of DE-[1-¹³C]-α-KG. The peak at 178 ppm initially increased but then steadily decreased while a peak at 177 ppm developed, corresponding to [1-¹³C]-glutamine.

This work shows that simple esterification of α-KG to DE-[1-¹³C]-α-KG allows for increased cell permeability. Once injected, the esters were rapidly cleaved to produce α-KG detectable in mouse xenograft tumors, which was subsequently metabolized to glutamate and glutamine. The ability to detect downstream metabolites of α-KG via HP-¹³C-MRS shows that DE-[1-¹³C]-α-KG was able to maintain polarization through membrane diffusion and ester cleavage. While all HP–¹³C-MRS experiments must be designed to be completed in a rapid time frame because of losses in polarization,²⁵ the T1 for DE-[1-¹³C]-α-KG was sufficient to allow for tracing of α-KG metabolism in vivo. The increased permeability of DE-[1-¹³C]-α-KG increased the resolution of the HP–¹³C-MRS peaks. This allowed not only for [1-¹³C]-glutamate to be detected, but also for the [1-¹³C]-glutamate and [1-¹³C]-glutamine peaks to be resolved. With these peaks resolved, a time course analysis allowed for tracking of the multistep metabolism of α-KG to glutamate to glutamine.

The improved passive diffusion of DE-[1-¹³C]-α-KG has the added benefit of not requiring a transporter, thereby reflecting a clear readout of α-KG metabolism. Although tracking of the conversion of hyperpolarized [1-¹³C]-pyruvate to lactate is commonly used as a measure of glycolysis, Rao and colleagues recently showed that the conversion of pyruvate to lactate, as measured by HP-¹³C-MRS, can be rate-limited by MCT1.²⁶

Since glutaminolysis has been shown to be a major energetic pathway for cancer cells, aminotransferase inhibitors, which prevent cancer cells from metabolizing glutamine, are being developed for clinical use.^27,28 However, not all cancers are dependent on glutaminolysis equally: Yuneva and colleagues showed, for example, that liver cancers with Myc mutations rely heavily on glutaminolysis, while MET-initiated liver cancers increase glutamine synthetase, the opposing pathway to glutaminolysis.²⁹ Breast cancer subtypes also show differential dependence on glutaminolysis: basal breast cancers rely on glutamine input while luminal-type breast cancers are less reliant on glutaminolysis for growth.³⁰

Therefore, the ability to establish glutamine dependence in cancers will be vital to determine the susceptibility to glutaminase inhibitors. Kung and colleagues showed that glutamine synthetase expression can predict glutamine independence and promote resistance to the glutaminolysis inhibitor BPTES.³⁰ As tracking glutamine production from α-KG can help determine glutamine synthetase activity, we propose that HP-¹³C-MRS using DE-[1-¹³C]-α-KG could be a useful tool to noninvasively determine tumor dependence on glutaminolysis. Further study is needed to show the difference in HP-¹³C-MRS results between glutamine-dependent and independent tumor types, but we propose that using DE-[1-¹³C]-α-KG as a probe in HP-¹³C-MRS may allow for noninvasive classification of susceptibility to glutaminolysis inhibition.

While we did not detect peaks for [1-¹³C]-succinate or [1-¹³C]-citrate in the HP–¹³C-MRS in HCT116 R132H xenograft tumors, tracing the production of these metabolites may be possible using DE-[1-¹³C]-α-KG in cell lines with higher levels of oxidative phosphorylation. With its ability to be hyperpolarized, permeate cell membranes, and be rapidly converted to downstream metabolites in mouse xenograft tumors, we establish in this work the promise of DE-[1-¹³C]-α-KG as an HP–¹³C-MRS probe to study α-KG metabolism in vivo.

METHODS

Chemicals and Reagents.

[1-¹³C]-α-KG, dimethyl-α-KG, diethyl-α-KG, α-KG, D-2-HG, l-glutamate, and N-acetyl-glutamine (NAG) were purchased from Sigma-Aldrich (St. Louis, MO). UPLC/MS grade acetonitrile was obtained from Sigma–Aldrich. Water was purified through a Milli-Q Integral 5 system supplied by EMD Millipore (Billerica, MA). Ammonium acetate and ammonium hydroxide were purchased from Sigma–Aldrich. See the SI for details regarding the synthesis of labeled α-KG derivatives.

Reagent Preparation.

Labeled α-KG derivatives were synthesized as described in the SI. Sigma [1-¹³C]-α-KG and DE-[1-¹³C]-α-KG were dissolved in a mixture of D₂O:d-glycerol = 1:1 containing 17.3 mM of Ox063 at a concentration of 5.9 M. After three freeze-and-thaw cycles using liquid nitrogen, followed by vortexing for 1 min, samples were stored at 4 °C until use.

¹³C MRI of Hyperpolarized 13C-Labeled α-KG.

For all hyperpolarization experiments, 35 μL of Sigma [1-¹³C]-α-KG or DE-[1-¹³C]-α-KG solution containing 2.5 mM of the gadolium chelate (ProHance, Bracco Diagnostics, Milano, Italy) were polarized at 3.35 T and 1.45–1.5 K in the Hypersense DNP polarizer (Oxford Instruments, Abingdon, U.K.) for 3–5 h according to manufacturer’s instructions. The polarized samples were rapidly dissolved in 4 mL of alkaline buffer containing 25 mM Tris(hydroxymethyl)-aminomethane, 50 mg/L ethylendiaminetetraacetic acid, and 37.5 mM NaOH, for the final dissolution buffer to be pH 7.4 after mixture with α-KG.

For in vivo hyperpolarized studies, the sample was rapidly injected after dissolution and spectra were acquired every second for 240 s using a single pulse acquire sequence with a sweep width of 3.3 kHz and 512 FID points, extrapolated to 1024 points using the “forward-backward” linear prediction method of Zhu and Bax.³¹ Spectra were acquired using a 17 mm custom-build ¹³C solenoid leg coil and 3T scanner (MR Solutions, Guildford, U.K.). The baseline was estimated by a modification of the Dietrich first derivative method to generate a binary mask of baseline points,³² followed by spline interpolation using the Whittaker smoother³³ to generate a smooth baseline curve.³⁴ To improve signal-to-noise, the signal matrix was truncated to a rank of 5 after singular value decomposition.³⁵

Animal Experiments.

2.0 × 10⁶ HCT116 IDH1 R132H cells were subcutaneously injected into the right hind legs of nude mice (Nu/Nu) to create xenograft tumors. In vivo experiments commenced when the tumors reached ~1250 mm³ in size, as measured by calipers. For injection of hyperpolarized ¹³C compounds, reagents were injected into the tail veins of mice positioned in the 3T scanner. For extraction of metabolites after HP-¹³C-MRS experiments, tumors were processed as previously published^36–38 with slight modifications. After harvesting, tumors were immediately cut into pieces and flash frozen in liquid nitrogen and stored at −80 °C until use.

Preparation of Tumor for Mass Spectrometry.

Tumors were extracted as previously published.^39,40 Briefly, 50 mg of DE-[1-¹³C]-α-KG was injected intravenously into the tail vein of a nude mouse bearing a HCT116 IDH1 R132H xenograft tumor. Mice were sacrificed by cervical dislocation 10 min after injection, following an intravenous saline flush. The tumor was then excised, immediately flash frozen in liquid nitrogen, and stored at −80 °C until the extraction procedure.

The polar fractions were isolated from the frozen tumor sections using a modification of a previously published procedure for cell extracts.^39,40 Briefly, a section of the frozen tumor was cut and then pulverized in liquid nitrogen using a cryogenic grinder (Freezer/Mill 6875, Spex SamplePrep, Metuchen, NJ). Approximately 50 mg aliquots of the ground tissue powder were weighed and then immediately quenched with 2 mL of acetonitrile at −20 °C. The solution was allowed to thaw on ice and 1.5 mL of ice cold doubly distilled (dd) H₂O was added to the thawed extract. Lipids and nonpolar metabolites were extracted by the addition of 1 mL of −20 °C chloroform with vigorous mixing. Addition of chloroform created a three-phase system consisting of the polar and nonpolar fractions and an interphase layer consisting primarily of proteins. Following centrifugation at 6400 g for 30 min, 90% of the aqueous phase was transferred to a pretared microcentrifuge tube. After removal of the nonpolar chloroform phase, the interphase layer was washed with ice cold 2:1 choloroform/methanol containing 1 mM BHT and recentrifuged. The aqueous phases were then combined and lyophilized. To remove residual proteins, the lyophilized powder was reconstituted in 100 μL of ice cold dd H₂O, followed by 400 μL of ice-cold acetone. Samples were then incubated at −80 °C for 30 min to facilitate protein precipitation. The protein precipitate was isolated by centrifugation for 30 min at 14 000 rpm. The pellet was then washed with 100 μL of 60% acetonitrile and the remaining supernatant after centrifugation relyophilized before analysis by MS.

Cell Permeability Assay.

The permeability of unlabeled α-KG, dimethyl-α-KG, and diethyl-α-KG was measured using a parallel artificial membrane permeability assay (R&D Systems, Minneapolis, MN). α-KG and derivatives were added to donor 96-well plates and allowed to pass through a dodecane membrane supplemented with 2% lectin for 24 h at 37 °C. Concentrations of α-KG and derivatives from the filtrate solution were assayed using an ultraviolet–visible (UV-vis) spectrophotometer (BioTek Synergy H1 Hybrid Multi-Mode Monochromator Fluorescence Microplate Reader, BioTek, Winooski, VT) at 270 μm and compared to a respective standard curve for each α-KG derivative.

Instrumentation for LC-MS/MS Analysis.

LC-MS/MS analysis was performed on an Agilent 6460C Triple Quadrupole mass spectrometer with an ESI source (Agilent Technologies, Santa Clara, CA). The LC inlet was an Agilent 1200 series chromatographic system equipped with 1260 binary pump, 1290 thermostated column compartment and 1260 high performance autosampler. Instrument control and data processing was performed using Agilent’s MassHunter software.

Chromatography.

Metabolites were measured as previously published with minor modifications.^41–43 Chromatographic separations were performed with gradient elution in hydrophilic interaction chromatography (HILIC) mode on a Phenomenex Luna-NH2 (2.0 × 100 mm, 3.0 μm; Phenomenex, Torrance, CA). The mobile phase was composed of aqueous buffer (Solvent A) and organic solvent (Solvent B) each containing 10 mM ammonium acetate with ammonium hydroxide added to adjust pH to 9.0. Solvent A contained a 95:5 mixture of water:acetonitrile, whereas solvent B contained a 95:5 mixture of acetonitrile:water. The analytes were eluted from the column by a linear gradient which started at 60% B, held under initial conditions for 1 min, then decreased from 60% B to 5% B within 8 min and held at 5% B for 5.0 min then returned to the initial conditions. A 10 min equilibrium time between injections was used to ensure reproducible retention times. The flow rate was set at 0.5 mL/min. The column oven was kept at 40 °C throughout the analysis. The injection volume was 5 μL and the autosampler rack temperature was 8 °C. The needle wash solvent was a mixture of 50:50 acetonitrile:water. Tumor samples were mixed with an equal volume of NAG (500 ng/mL in 50:50 acetonitrile:water) as an internal standard for monitoring system performance. A pooled tumor sample, prepared by combining 5 μL from each tumor sample, was used to condition the column before analysis of the actual samples. A QC sample consisting of 1 μg/mL of unlabeled standards (2-HG, α-KG) was used to determine retention times of the labeled compounds.

Mass Spectrometry.

Mass spectrometric data were acquired in positive/negative ion switching mode with the following ESI-MS parameters: gas temperature, 350 °C; gas flow, 13 L/min; nebulizer, 45 psi; capillary voltage, 4000 V. Nitrogen was used as desolvation gas and collision gas and dwell time was set at 80 ms for each transition. Cell accelerator voltage was set to 7 and quantification was done in multiple reaction monitoring (MRM) mode. Precursor and product ion selection was determined experimentally using authentic samples when available. When authentic samples were not available, transitions were based on the unlabeled precursor and adjusted to incorporate the labeled atom.

Data Analysis.

For LC-MS/MS data analysis, peak integration was performed in Agilent’s MassHunter software. Statistical analysis (fold-change and t-test) was performed in MetaboAnalyst 3.0 (www.metaboanalyst.ca).^41,42 No filtering or data normalization was performed before analysis.

Statistics.

Results are presented as means ± SD. Significance was determined by Student’s t-tests. A threshold of p ≤ 0.05 was used to determine significance.