Comparative evaluation of hyperpolarized [¹³C]pyruvate and [¹³C]lactate for imaging neuronal and glioma metabolism

Abstract

Glucose and lactate are primary substrates in cerebral energy metabolism. Hyperpolarized [1-¹³C]pyruvate has become as a powerful imaging agent for metabolic neuroimaging due to its central role in glucose and lactate metabolism, ability to cross the blood-brain barrier, and translational utility in neurological disorders. In particular, [1-¹³C]pyruvate enables assessment of mitochondrial metabolism in the cerebral cortex through its conversion to [¹³C]bicarbonate. While it is not yet confirmed that production of [¹³C]bicarbonate primarily reflects neuronal metabolism, the higher affinity of neuronal transporters for lactate over pyruvate has motivated interest in hyperpolarized lactate as a more physiologic probe of neuronal metabolism. Here, we identify the predominant cellular source of [¹³C]bicarbonate and evaluate [1-¹³C]lactate as an imaging agent for neuronal metabolic imaging. Ex vivo NMR and mass spectrometry imaging of freeze-clamped brain tissue after bolus injection of [U-¹³C₃]pyruvate revealed that pyruvate dehydrogenase dominates pyruvate carboxylase in the cortex, supporting neuronal origin of [¹³C]bicarbonate production. Although the bicarbonate fraction among total ¹³C products in vivo was higher following hyperpolarized [1-¹³C]lactate injection, the signal sensitivity was markedly reduced due to lactate’s shorter T₁ and larger endogenous pool. Isotopomer analysis of brain tissue harvested two minutes after injection of [U-¹³C₃]pyruvate or [U-¹³C₃]lactate showed comparable labeling of mitochondrial intermediates. In glioma-bearing rats, in vivo imaging revealed an elevated pyruvate-to-lactate ratio within the tumor, highlighting altered redox and transport dynamics in malignancy. These findings demonstrate that both hyperpolarized [1-¹³C]pyruvate and [1-¹³C]lactate can effectively probe neuronal and glioma metabolism, although pyruvate outperforms lactate in detecting [¹³C]bicarbonate.

Graphical Abstract

Cerebral metabolism is managed by an intricate network of neurons, astrocytes, microglia, and oligodendrocytes, sustained by a continuous supply of oxygen and glucose. Glucose metabolism via glycolysis and oxidative phosphorylation generates ATP to meet the high energy demands of the brain. While glucose is the canonical substrate, lactate has also emerged as a physiologically important fuel, especially for neurons, supported by the astrocyte-neuron lactate shuttle theory^1–4 and broader cell-cell lactate shuttle theory^5–7. On the contrary, pyruvate, the oxidized metabolite of lactate, has received less attention as a substrate due to its much smaller endogenous pool size.

Due to the pivotal role that connects glycolysis, lactate fermentation, and mitochondrial metabolism, pyruvate is an ideally suited exogenous substrate for probing cerebral metabolism and has been widely used in hyperpolarized (HP) ¹³C magnetic resonance imaging and spectroscopy (MRI/MRS) brain studies^8–14. Intravenous administration of HP [1-¹³C]pyruvate enables detection of its conversion to [1-¹³C]lactate via lactate dehydrogenase (LDH) and to [¹³C]bicarbonate via pyruvate dehydrogenase (PDH), offering insight into cytosolic redox state and mitochondrial metabolism¹⁵. In particular, the detection of [¹³C]bicarbonate has been unambiguously interpreted as a marker of mitochondrial PDH activity since phosphoenolpyruvate carboxykinase (PEPCK)-mediated decarboxylation from [1-¹³C]pyruvate via pyruvate carboxylation (PC) pathway¹⁶ is considered negligible in the brain¹⁷. However, whether the cellular source of HP [¹³C]bicarbonate is neuronal or glial PDH remains unclear, especially given the supraphysiological pyruvate concentrations achieved during bolus injection.

While intracellular pyruvate is the immediate precursor of [¹³C]bicarbonate, the assumption that it derives directly from the injected HP pyruvate overlooks key transport and metabolic dynamics. Pyruvate must traverse the blood-brain barrier and multiple cellular membranes, during which it may undergo rapid interconversion with lactate via LDH. This reversible reaction can occur in the blood, extracellular space, or glial compartments, raising the possibility that pyruvate used for mitochondrial metabolism may in fact arrives in the neuron as lactate. This challenges the assumption that observed bicarbonate signal directly reflects pyruvate uptake and oxidation. Indeed, it was reported that eliminating HP [1-¹³C]lactate signals decreased the HP bicarbonate detection in the brain by more than 40%¹².

As illustrated in Figure 1, neurons predominantly express MCT2, a high-affinity (K_m = ~0.7 mM), low-capacity monocarboxylate transporter (V_max = 29 nmol/min/mg protein), making them uniquely suited to uptake low concentrations of extracellular lactate¹⁸. In contrast, astrocytes express MCT1 and MCT4, which favor lactate export due to lower affinity (K_m = 3–5 mM for MCT1, 15–30 mM for MCT4) but higher capacity (V_max = 174 nmol/min/mg protein)^{19, 20}. The high affinity of MCT2 enables neurons to efficiently import even low concentrations of lactate, suggesting that HP lactate may enter neurons more readily than HP pyruvate in certain physiological contexts. Moreover, lactate is a physiologically relevant substrate for neuronal oxidation and may better reflect the in vivo metabolic state than pyruvate delivered exogenously. These considerations raise the intriguing question of whether HP lactate, rather than pyruvate, may offer a more specific and physiologically meaningful approach to probing neuronal metabolism in vivo. Indeed, the feasibility of HP lactate was previously demonstrated in healthy mouse brain by Takado, et al.²¹ and also applied to investigate neuroprotective lactate metabolism after stroke^{22, 23}. These studies, however, focused on the rate-limiting role of the blood-brain barrier in the cerebral lactate uptake and post-stroke changes of lactate metabolism.

Figure 1. Pyruvate and lactate metabolism in astrocytes and neurons.

Astrocytes express MCT1, which can mediate the transport of pyruvate and lactate, and MCT4, which primarily exports lactate to the extracellular space during high metabolic activity. Neurons predominantly express the high-affinity transporter MCT2, optimized for lactate uptake rather than pyruvate. PC activity is largely astrocyte-specific, whereas pyruvate oxidation through PDH is the dominant pathway in neurons. In astrocytes, LDH-A preferentially converts pyruvate to lactate, a reaction favored by their more reducing cytosolic redox state (higher NADH/NAD⁺ ratio). In contrast, neurons predominantly express LDH-B, which favors the reverse reaction consistent with their more oxidized redox environment. LDH, lactate dehydrogenase; MCT, monocarboxylate transporter; PC, pyruvate carboxylase; PDH, pyruvate dehydrogenase; TCA, tricarboxylic acid.

The mechanism of lactate production from HP pyruvate is also complex as it involves multiple metabolic and exchange processes, further complicating its interpretation. Neurological disorders, including high-grade gliomas, brain injury, and neuroinflammation, often exhibit high LDH activity and expanded lactate pools²⁴, enabling rapid chemical exchange with HP pyruvate^{25, 26}. This makes it difficult to determine whether elevated HP lactate production reflects true metabolic flux or simply large lactate pools. In this context, HP [1-¹³C]lactate may also offer added value by isolating lactate utilization pathways, clarifying tumor-specific metabolic mechanisms, and potentially improving specificity over conventional ¹H MRS.

In this study, we identify the predominant cellular source of HP [¹³C]bicarbonate in the brain, evaluate HP [1-¹³C]lactate as a direct imaging agent of neuronal oxidative metabolism, and test whether HP lactate provides added value in imaging lactate metabolism in glioma.

METHODS

Animal Preparation and Tumor Implantation

Nineteen male Wistar rats (Charles River Laboratories, Wilmington, MA, USA) were used for this study, as summarized in Table 1. Male rats were selected to reduce the number of animals required by minimizing potential brain metabolic variations associated with menstrual cycle^27–29. A subset of the rats (n = 6) was implanted with C6 rat glioma for in vivo imaging with HP [1-¹³C]lactate. Brains were harvested from two glioma-bearing rats for matrix-assisted laser desorption/ionization (MALDI) mass spectrometry imaging (MSI). The remaining animals were used for comparative studies of HP [1-¹³C]pyruvate and HP [1-¹³C]lactate in vivo (n = 3), NMR isotopomer analysis following injection of [U-¹³C₃]pyruvate (n = 3) or [U-¹³C₃]lactate (n = 3), and MALDI MSI after injection of [U-¹³C₃]pyruvate (n = 4). All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the UT Southwestern Medical Center (Protocol#: 2016–101749).

Table 1.

Summary of Experimental Groups and Measurements.

	In Vivo HP MRS/MRI			Ex Vivo NMR		Ex Vivo MSI
Injectate	[1-13C]pyruvate	[1-13C]lactate	[1-13C]lactate	[U-13C3]pyruvate	[U-13C3]lactate	[U-13C3]pyruvate
#animals	3 healthy rats		6 glioma rats	3 healthy rats	3 healthy rats	4 healthy rats

For tumor implantation, one million C6 glioma cells, derived from an N-methyl-N-nitrosourea (MNU)-induced tumor³⁰ into the left striatum under aseptic conditions. Rats (6–8 weeks old, 200–250 g) were anesthetized with Isoflurane (~4% induction, ~2.25% maintenance) in a mix of 70% nitrous oxide and 30% oxygen (V-1, VetEquip) and secured in a stereotaxic frame (Kopf Instruments). Body temperature was maintained at 37 °C (RightTemp, Kent Scientific), respiration rate was visually monitored, corneas were covered with ophthalmic ointment, and the skin was shaved and disinfected (ethanol and betadine). A midline scalp incision exposed skull landmarks (i.e., Bregma and Lambda), the head was leveled, a burr hole was drilled (AP +1.0, ML −2.6), and dura was nicked with a 27g needle to facilitate subsequent needle insertion. Cells were injected at DV −4.9 (relative to dura; equivalent to −5.5 vs. Bregma) via Hamilton syringe (#701, 10 μL) fitted with a 26g needle (flat bevel) at a rate of 1.0 uL/min (10 uL total volume) using a syringe controller (UMP3/Micro4, WPI Instruments); the needle was left in place for 10 min, then slowly removed to minimize fluid backflow. The burr hole was sealed with bone wax, and the incision was sutured (5–0 nylon). Buprenorphine ER (0.6 mg/kg s.q.), Carprofen (5 mg/kg s.q.), Lidocaine (2%, topical at incision), and saline (0.5 mL s.q.) were given for pain management and fluid support. Rats were recovered in a heated (30 °C) chamber until ambulatory, then returned to their home cage. Tumor growth was monitored one and two weeks after the implantation and imaged between day 15 and 22 days depending on the tumor growth and equipment availability.

Hyperpolarization and MRI Protocol

All MRI studies were performed using a clinical 3T MRI scanner (750w Discovery, GE Healthcare, Waukesha, WI, USA) and a ¹³C/¹H dual-tuned birdcage RF coil (inner diameter = 50 mm). Rats were anesthetized with isoflurane at ~3% (1.5 L/min) for tail-vein catheterization, then their heads were positioned at the center of the coil. Animal vital signs were monitored throughout the session, and the respiratory rate was maintained at 40–60 breaths/min by adjusting the isoflurane level (1–3%). From all rats, 2D T₂-weighted ¹H MRI was acquired using a fast spin echo sequence (repetition time (TR) = 5,000 ms, echo times (TEs) = 12.8 ms/ 64.0 ms, field of view (FOV) = 96 mm × 96 mm, matrix size = 256 × 192, slice thickness = 2 mm, echo train length = 8, flip angle = 160°). From healthy rats (n = 3, body weight = 147 ± 20 g), in vivo time-resolved ¹³C spectra were acquired using a dynamic free-induction decay (FID) sequence (10° slice-selective RF excitation, slice thickness = 7.7 mm, spectral width = 5,000 Hz, spectral points = 2,048, TR = 3 s) immediately following bolus injection of 60-mM HP [1-¹³C]pyruvate (0.75 mmol/kg body weight). This procedure was repeated using 60-mM HP [1-¹³C]lactate (0.75 mmol/kg) after a 2–4 hour interval to ensure the second injection was not influenced by the first injection. Animals remained awake with free access to food and water between sessions to avoid metabolic changes associated with prolonged anesthesia^{31, 32}. Glioma-implanted rats (n = 6, body weight = 267 ± 13 g) were imaged with a 2D ¹³C chemical shift imaging (CSI; FOV = 48 mm × 48 mm, matrix size = 16 × 16, slice thickness = 7.7 mm, spectral width = 5,000 Hz, spectral points = 512, TR = 75 ms)¹¹ 20 seconds after bolus injection of 120-mM HP [1-¹³C]lactate (1.5 mmol/kg body weight). To confirm the tumor region, contrast-enhanced T₁-weighted ¹H MRI was acquired using spin-echo sequence (TE = 12 ms, TR = 700 ms, 96 × 96 mm², matrix size = 256 × 192, slice thickness = 2 mm). Dissolution-to-injection time was 19–25 s.

Sample preparation and polarization procedure for [1-¹³C]pyruvate and [1-¹³C]-L-lactate are described previously³³. [1-¹³C]Pyruvic acid (Sigma Millipore, Burlington, MA, USA) and sodium [1-¹³C]-L-lactate (Cambridge Isotope Laboratories, Inc. Tewksbury MA, USA) were purchased. Neat [1-¹³C]pyruvic acid was mixed with 15-mM OX063 (GE Healthcare) and neutralized after dissolution¹¹. 2.1-M [1-¹³C]-L-lactate was prepared in 4:1 w/w water:glycerol with 15-mM OX063. A SPINlab^™ polarizer (GE Healthcare, Waukesha WI, USA) that operates at 5T and ~0.8 K was used for polarizing both samples. 180 or 360 μL of the lactate sample was placed in a sample vial and assembled with a research fluid path (GE Healthcare), which included 16 mL of dissolution media containing 0.1 g/L of disodium ethylenediaminetetraacetate (Na₂EDTA). The sample was dissolved after 4–5 hours of polarization, yielding ~6.0 mL of 60- or 120-mM HP [1-¹³C]-L-lactate.

NMR Isotopomer Analysis with Bolus Injection of Pyruvate or Lactate

¹³C NMR isotopomer analysis was performed using brain tissues collected from six rats (240 ± 71 g) to compare the metabolic fates of [U-¹³C₃]pyruvate or [U-¹³C₃]lactate following bolus injection (0.75 mmol/kg in 60-mM solution) to mimic HP studies. Approximately two minutes after the injection, the rats were sacrificed, and the brain tissues were harvested and freeze-clamped with liquid nitrogen. The frozen tissue was extracted with 4× sample size 20% perchloric acid and then lyophilized at −85 °C (benchtop SLC, VirTis). Each sample was resuspended in 300 μL of D₂O solution that contained 1-mM DSS and 1-mM ethylenediaminetetraacetate (EDTA). Proton-decoupled ¹³C spectra of perchloric acid extracts were acquired using a 14.1-T NMR spectrometer (Bruker, Billerica MA USA; Avance III HD) with a 10-mm cryoprobe (Bruker) as previously described³⁴.

MALDI MSI

Cryosections (12 μm thickness) were obtained from posterior forebrain to midbrain regions (healthy rats) or glioma-bearing brain regions (tumor-implanted rats) and mounted onto mass spectrometry-compatible slides. Serial adjacent sections were collected for hematoxylin and eosin (H&E) staining. For MALDI MSI, tissue sections were coated with 7 mg/mL N-(1-naphthyl) ethylenediamine dihydrochloride in 70% methanol using an HTX M5 Robotic Sprayer (Htx Technologies LLC, Chapel Hill, NC, USA). The coating was applied over eight passes with the following parameters: flow rate of 0.120 mL/min, track speed of 1,200 mm/min, track spacing of 2 mm, a crisscross spray pattern, and a nozzle temperature of 75 °C. MSI was performed using a timsTOF fleX MALDI QTOF mass spectrometer (Bruker, Billerica, MA, USA) in negative ion mode, acquiring spectra across the m/z 50–600 range at a spatial resolution of 100 μm, with 1000 laser shots accumulated per pixel. Instrument settings included Funnel 1 RF: 75.0 Vpp, Funnel 2 RF: 100.0 Vpp, Multipole RF: 150.0 Vpp, Collision Energy: 5.0 eV, Collision RF: 500.0 Vpp, Quadrupole Ion Energy: 5.0 eV, Transfer Time: 40 μs, and Pre-Pulse Storage: 3.0 μs.

Data Analysis and Statistical Evaluations

All HP data sets were processed using MATLAB (R2024b; Mathworks Inc., Natick, MA, USA) as described previously³⁵. For dynamic ¹³C MRS, the FID data were apodized by a 5-Hz Gaussian filter in the spectral domain. After a zero-filling by a factor of two, a 1D fast Fourier transform was applied. [1-¹³C]Pyruvate, [1-¹³C]lactate, [¹³C]bicarbonate, [1-¹³C]pyruvate-hydrate, and [1-¹³C]alanine peaks were quantified by integrating the corresponding peaks in the absorption mode after 0^th order phase correction. Total HP ¹³C (TC) was calculated by summing all HP ¹³C peaks, and total product (TP) was calculated by summing [1-¹³C]pyruvate and [¹³C]bicarbonate when [1-¹³C]lactate was injected and by summing [1-¹³C]lactate and [¹³C]bicarbonate when [1-¹³C]pyruvate was injected. HP [1-¹³C]alanine was excluded when calculating TC and TP due to its negligible presence in the brain. Differences in TP-to-TC, bicarbonate-to-TC, and bicarbonate-to-TP as well as bicarbonate signal-to-noise ratio (SNR) between lactate and pyruvate injected groups were evaluated by paired t-tests (α = 0.05, two-tailed) using Prism 10 (version 10.5.0; GraphPad Software, San Diego, CA, USA).

For CSI, the raw k-space data were apodized by a 25-Hz Gaussian filter in the spectral domain and by a generalized Hanning filter (α = 0.66) in the spatial domain. After zero-filling by a factor of four in both spectral and spatial dimensions, a 2D inverse fast Fourier transform in spatial domains and a 1D fast Fourier transform in time domain were carried out. All the presented spectra were phase-corrected for both the 0^th and 1^st orders, and the rolling baseline was removed by subtracting a fitted spline function to the signal-free regions from the spectrum. Metabolite maps of [1-¹³C]pyruvate, [1-¹³C]lactate, [¹³C]bicarbonate, and [1-¹³C]alanine were generated by integrating the corresponding peaks in the absorption mode after 0^th order phase correction in each voxel. [1-¹³C]Pyruvate-hydrate map was not generated as the peak was not reliably detected. For comparisons between the tumor and contralateral normal-appearing brain, regions of interest (ROIs) were selected based on the initial hyperintense region in the contrast-enhanced T₁-weighted ¹H MRI and normal-appearing regions in the contralateral hemisphere. Produced [¹³C]bicarbonate and [1-¹³C]pyruvate were quantified by integrating the peaks in the ROI-averaged spectra from the tumor and the contralateral side, then normalized by the total ¹³C signal (sum of [1-¹³C]lactate, [1-¹³C]pyruvate, [¹³C]bicarbonate and [1-¹³C]alanine) within the ROI. In addition, the bicarbonate-to-pyruvate ratio was compared between the ROIs. For display purpose, ¹³C metabolite maps were overlaid on the corresponding ¹H T₁w MRI and a brain mask was applied to the metabolite ratio maps to maintain reasonable window level. Paired t-tests (α = 0.05, two-tailed) were used to assess differences between tumor and contralateral regions.

NMR data were analyzed using TopSpin (Bruker) to calculate the fractional enrichments of lactate, glutamate, and succinate. Phase-correction of ¹³C NMR spectra and peak integration were processed using TopSpin (Bruker; version 4.0.9). Fractional enrichments for doubly ¹³C-labeled metabolites were calculated relative to the corresponding natural abundance (1.1%) ¹³C singlets by calculating from the integrated doublets (D) and singlets (S): $\int D/\int S\times 1.1/100$. For lactate C2, the fractional ¹³C enrichment was calculated by ($\int D_{12}+\int D_{23}+\int Q)/\int S\times 1.1/100$, where Q indicates quartets. For succinate, due to its symmetric structure, C1 and C2 resonances are identical to C3 and C4 resonances, respectively, and each singlet in succinate resonance is accounted for 2.2 %. Fractional enrichment of succinate, therefore, was calculated by $\int D/\int S\times 2.2/100$. Unpaired t-tests (α = 0.05, two-tailed) were used to compare the fractional ¹³C abundance between lactate-injected and pyruvate-injected groups.

For MALDI MSI data, a list of targeted metabolites and their predicted ¹³C isotopologues was imported into SCiLS Lab (Bruker), and ion images were generated for each m/z. Theoretical isotope distributions were calculated using Bruker’s Isotope Pattern tool. Signal intensities for each isotopologue were exported as a simple text format (csv). In addition to intensities of unlabeled metabolites, PDH- and PC-mediated ¹³C-labeled citrate and glutamate were identified by M+2 and M+3 isotopologues, respectively. The abundance of each ¹³C isotope as a percentage of the monoisotopic peak was calculated and subtracted by the predicted natural abundance. Paired t-tests (α = 0.05, two-tailed) were performed to compare the abundance of M+2 and M+3 isotopologues. For spatial correlation between the PDH-driven products and neurons, pixel-wise linear regression was performed in MATLAB using polyfit function to evaluate correlations between M+2 glutamate and M+0 N-acetylaspartate (NAA), a neuronal biomarker.

RESULTS

MALDI MSI detected M+2 isotopologues of citrate (2.26 ± 0.19% of M+0 after correcting predicted natural abundance) and glutamate (1.97 ± 0.53% of M+0) from bolus injected [U-¹³C₃]pyruvate in healthy rat brains while M+3 isotopologues were hardly detectable (citrate: 0.65 ± 0.07%, P = 0.027; glutamate: 0.04 ± 0.01%, P = 0.018), indicating PDH is the primary enzymatic path towards the TCA cycle rather than PC, Figure 2. The M+2 glutamate distribution was positively correlated with NAA (ρ = 0.233, P < 10⁻¹⁰), supporting that the source of M+2 glutamate is neuronal metabolism.

Figure 2. Metabolic fate of a pyruvate bolus in the brain.

(A) Astrocytes contain astrocyte-specific PC, which is considerably more active than PDH. (B, C) Following decarboxylation via PDH, M+2 isotopologues of citrate and glutamate were detected in rat brain tissue, collected two minutes after bolus injection of [U-¹³C₃]pyruvate, whereas M+3 isotopologues were negligible. (D) The distribution of M+2 glutamate correlated positively with NAA, supporting that the neuronal origin of the observed M+2 glutamate. NAA, N-acetylaspartate; PC, pyruvate carboxylase; PDH, pyruvate dehydrogenase.

In vivo HP study from rat brains showed that the total products in the brain were smaller when HP [1-¹³C]lactate was injected (3.6 ± 0.4% of TC) than with HP [1-¹³C]pyruvate injection (31.3 ± 4.49% of TC, P = 0.011), Figure 3. [1-¹³C]Alanine was excluded from the calculation of both TC and TP because its predominant source is peripheral tissues outside the brain. [¹³C]Bicarbonate/TC was approximately half the signal when [1-¹³C]lactate was injected (1.4 ± 0.2%) as compared to the [1-¹³C]pyruvate injection (2.9 ± 0.3%, P = 0.013), consistent with a previous mouse brain study reporting roughly two-fold higher bicarbonate-to-substrate ratio when pyruvate, rather than lactate, was used as the substrate²³. The portion of [¹³C]bicarbonate in TP was, however, higher with [1-¹³C]lactate injection (39.4 ± 3.9%) than [1-¹³C]pyruvate injection (9.4 ± 1.2%, P = 0.006). The peak SNR of HP [¹³C]bicarbonate in the time-accumulated ¹³C spectra was lower with HP lactate (14.3 ± 4.3) as compared with HP pyruvate injection (20.9 ± 4.5, P = 0.021). Significantly larger total HP products relative to TC are likely due to substrate competition between the large endogenous (unlabeled) lactate pool and the injected (labeled) lactate pool in exchange with pyruvate before converting to bicarbonate. As the tissue-level concentrations of endogenous lactate and injected lactate are largely comparable, the labeled bicarbonate becomes much smaller when HP lactate was injected. This does not apply to ¹³C-pyruvate injection since the endogenous pyruvate pool is much smaller.

Figure 3. ¹³C MRS with HP [1-¹³C]lactate and HP [1-¹³C]pyruvate.

Time-averaged ¹³C spectra from in vivo rat brains following bolus injection of (A) HP [1-¹³C]lactate and (B) HP [1-¹³C]pyruvate. (C,D) Dynamics of HP metabolites when [1-¹³C]lactate and [1-¹³C]pyruvate was injected. HP injectate signals were scaled by (C) 1/5 and (D) 1/10 for better visualization. (E) The fraction of summed products relative to total HP ¹³C signal and (F) the fraction of bicarbonate relative to total HP ¹³C signal were both greater with HP pyruvate injection than with HP lactate injection. (G) The proportion of bicarbonate within the total product signal was higher after HP lactate injection, although (H) the peak SNR of bicarbonate was lower compared with HP pyruvate injection. [1-¹³C]Alanine was excluded from the calculation of total HP signal and summed products, as it is predominantly produced in peripheral tissue outside of the brain.

¹³C NMR isotopomer analysis of healthy rat brains harvested at two minutes after bolus injection of either [U-¹³C₃]pyruvate or [U-¹³C₃]lactate was performed to identify differences in cerebral metabolic profiles from each agent. The analysis showed that the ¹³C spectral patterns and fractional enrichments of lactate, glutamate, and succinate were nearly identical between the two groups, Figure 4. Specifically, the fractional enrichment of [4,5-¹³C₂]glutamate at the C5 position was 2.3 ± 1.4 % in pyruvate-injected rats and 2.2 ± 1.2 % in lactate-injected rats (P = 0.942), indicating that similar amounts of ¹³C entered the TCA cycle through PDH following either substrate. The fractional enrichment of [3,4-¹³C₂]succinate was also comparable between groups when measured at the C3 position (3.9 ± 1.5 % for pyruvate and 2.6 ± 1.4 % for lactate; P = 0.343) and at the C4 position (0.68 ± 0.24 % for pyruvate and 1.02 ± 0.29 % for lactate; P = 0.186). Lactate labeling tended to be higher in the lactate-injected group (0.51 ± 0.06 % at C1) than in the pyruvate-injected group (0.33 ± 0.08 %) although the difference was not statistically significant (P = 0.119). While doublets (D34) were observed at the C3 positions of glutamate and glutamine, reflecting PDH-mediated ¹³C labeling at C4, PC-mediated labeling such as glutamate D23 at C2 was negligible in both the pyruvate- and lactate-injected groups, confirming that the contribution from PC was minimal.

Figure 4. ¹³C NMR of brain tissue extracts.

Brain tissues were collected 2 minutes after bolus injection of [U-¹³C₃]lactate or [U-¹³C₃]pyruvate. (A) PDH-mediated downstream products, including [5,6-¹³C₂]citrate, [4,5-¹³C₂]glutamate, [4,5-¹³C₂]glutamine, [3,4-¹³C₂]succinate, [1,2-¹³C₂]GABA, and [3,4-¹³C₂]aspartate, were detected in (B) ¹³C NMR spectra from both lactate-injected (red) and pyruvate-injected (blue) groups. As shown in the expanded spectral regions highlighting (C) carboxyl, (D) carbonyl, and (E) methylene carbons of selected metabolites, the brain spectra showed comparable fractional enrichments and isotopomer labeling patterns for lactate, succinate, and glutamate between lactate- and pyruvate-injected groups. (F) The absence of D23 doublets at the C3 positions of glutamate and glutamine indicates negligible contribution of PC pathway to their formation. Filled and open circles represent ¹³C-labeled and unlabeled (¹²C) carbons, respectively.

Consistently with previous C6 glioma studies^{36, 37}, elevated HP injectate signals were detected from the tumor region, Figure 5. Interestingly, despite the enlarged lactate pool size, more pyruvate was produced in the tumor (pyruvate/TC = 0.0256 ± 0.0113) than the contralateral normal-appearing brain region (0.0149 ± 0.0091, P = 0.001), suggesting elevated LDH activity. Bicarbonate production decreased in the tumor (bicarbonate/TC = 0.0058 ± 0.0015) compared to normal brain (0.0104 ± 0.0038, P = 0.024), reflecting impaired PDH flux. Bicarbonate-to-pyruvate ratio decreased in the tumor accordingly (0.244 ± 0.067 vs 0.815 ± 0.336, P = 0.008).

Figure 5. Metabolite maps of C6 glioma with HP [1-¹³C]lactate.

(A) T₁-weighted (with contrast enhancement) and T₂-weighted ¹H MRI of the tumor-bearing rat brain. (B) HP ¹³C metabolite maps of lactate, pyruvate, bicarbonate, and alanine from the corresponding slice following HP [1-¹³C]lactate injection. (C) Pyruvate production was higher and (D) bicarbonate production was lower in the tumor compared with the contralateral normal-appearing brain. (E) The bicarbonate-to-pyruvate ratio was correspondingly decreased in the tumor. (F) Elevated lactate and pyruvate pools in the tumor region were confirmed by MADL MSI. (G) H&E staining verified tumor location.

DISCUSSION

In this study, we aimed to identify the cellular source of [¹³C]bicarbonate from HP [1-¹³C]pyruvate in the brain, to assess the potential of HP [1-¹³C]lactate as alternative probe for imaging neuronal metabolism, and to evaluate the performance of HP [1-¹³C]lactate in glioma where the lactate pool is enlarged and the redox is shifted. Ex vivo NMR isotopomer and MSI isotopologue analyses were used to identify the primary cellular origin of [¹³C]bicarbonate from HP [1-¹³C]pyruvate or [1-¹³C]lactate. HP [1-¹³C]pyruvate provided greater sensitivity for detecting PDH activity. In contrast, HP [1-¹³C]lactate yielded elevated pyruvate/lactate ratio in gliomas despite high endogenous lactate and NADH/NAD⁺, implying its capability of sensing true LDH-mediated flux.

It is important to consider the enzymatic context for bicarbonate production in the brain. Our data indicate that [¹³C]bicarbonate derived from HP [1-¹³C]pyruvate or HP [1-¹³C]lactate within two minutes after bolus injection primarily reflects neuronal PDH flux. PC is expressed almost exclusively in astrocytes, where its activity greatly exceeds that of PDH³⁸. The absence of detectable PC-derived ¹³C products in our MALDI-MSI and NMR analyses suggests that HP [1-¹³C]pyruvate has limited sensitivity for probing astrocytic mitochondrial metabolism, and that the observed [¹³C]bicarbonate originates mainly from neuronal mitochondria. This interpretation aligns with prior reports noting the difficulty of attributing HP [¹³C]bicarbonate to PC activity^{12, 14, 39}. Studies using HP [2-¹³C]pyruvate in healthy rat³⁵ and human⁴⁰ brain further support PDH-mediated TCA entry, as evidenced by downstream labeling of glutamate and citrate. Moreover, the positive correlation between HP bicarbonate production and neuronal activation provides further support for a neuronal origin of this signal^{12, 41, 42}. It should be noted, however, that glial pyruvate carboxylation is expected to increase markedly under awake conditions, as reported by Oz, et al.⁴³. Future studies conducted in awake animals or using cell-type-specific approaches⁴⁴ will be valuable for clarifying the relative contributions of neuronal and glial pathways.

Several metabolic routes can lead to HP [¹³C]bicarbonate production in neurons. The conventional view assumes that administered HP [1-¹³C]pyruvate is the direct precursor, with bicarbonate production quantified using metrics such as the [¹³C]bicarbonate-to-[1-¹³C]pyruvate ratio or pyruvate-to-bicarbonate conversion rate constants. However, intravenously injected pyruvate can be converted to lactate in red blood cells or in peripheral organs such as the liver before reaching the brain, and this lactate can significantly contribute to neuronal bicarbonate production^{12, 45}. The possibility that pyruvate is first reduced to lactate before oxidative decarboxylation, together with evidence that HP [1-¹³C]lactate itself is a substantial substrate for neuronal PDH, suggests that lactate-derived bicarbonate should be explicitly accounted for in HP data interpretation. Astrocytes also play an important role. Gandhi et al.⁴⁶ demonstrated that astrocytes in the adult rat brain take up extracellular lactate 4.3-fold faster and with 2.3-fold higher capacity than neurons and can rapidly transfer lactate to both astrocytes and neurons. This astrocyte-to-neuron lactate gradient, further corroborated by intravenous L-lactate administration studies from Mächler et al.⁴⁷ provides a plausible route by which lactate generated from HP pyruvate or administered directly enters neurons and contributes to bicarbonate production.

Determining whether lactate is a more effective probe than pyruvate for assessing neuronal PDH activity requires consideration of both metabolic and physical factors. Metabolically, supraphysiological plasma lactate levels such as those achieved after HP pyruvate or lactate injection can contribute up to 60% of brain metabolism via direct transport⁴⁸. Our NMR isotopomer analysis showed that bolus injections of pyruvate or lactate produced nearly identical labeling of mitochondrial intermediates downstream of PDH within two minutes, indicating that both substrates access oxidative metabolism efficiently. However, pyruvate has an advantage in terms of redox balance by generating NAD⁺ during its LDH-mediated conversion to lactate, promoting oxidative decarboxylation via PDH³³. Physical MR properties also favor pyruvate; its longitudinal relaxation time (T₁) is longer (75.4 ± 2.9 s, n = 3) than that of lactate (67.8 ± 0.56 s)³³ at 3 T under identical in vitro conditions (15 mM OX063, no gadolinium), improving signal persistence. Endogenous pool sizes further influence label transfer^{26, 49}. Under normal conditions, lactate concentrations are ~0.1–1 mM in neurons and 0.5–2 mM in astrocytes, whereas pyruvate pools are typically less than 10% of lactate levels, reflecting cytosolic redox balance. As a result, an HP [¹³C]lactate bolus competes with a larger endogenous lactate pool, reducing labeled carbon transfer to pyruvate and bicarbonate, whereas pyruvate injection faces less competition.

As intracellular pyruvate can be formed from both glucose and extracellular monocarboxylates, the contributions of these precursors to cerebral pyruvate pool, and the distinction between these two kinetically different pools, have been described as compartmentalization in neurons⁵⁰ and astrocytes⁵¹. Intracellular redox dependance plays a dominant role in the pyruvate compartmentalization⁵². Astrocytes have a higher cytosolic NADH/NAD⁺ ratio than neurons as they are more glycolytic and export lactate⁵³. This compartmentalization may contribute to the performance difference between HP pyruvate and lactate. Indeed, this pathway-dependent pyruvate compartmentalization effect was well-established in literature^{54, 55} and was reported in a rat heart study using HP pyruvate⁵⁶.

Application of HP [1-¹³C]lactate to glioma imaging showed greater lactate delivery, pyruvate production, and pyruvate-to-lactate ratio in tumors compared with the contralateral normal-appearing brain. The increased lactate and pyruvate levels likely reflect disruption of the blood-brain barrier, consistent with previous HP pyruvate and lactate studies^{21, 36, 57}. The elevated pyruvate-to-lactate ratio in tumors provides unique metabolic insight, reflecting increased LDH activity and reduced redox in C6 glioma,⁵² and overcoming the disadvantage of competition with a larger endogenous lactate pool. In contrast, increased [1-¹³C]lactate production from HP [1-¹³C]pyruvate in glioma has been attributed to a combination of pyruvate-to-lactate flux and isotope exchange with an enlarged lactate pool. Although SNR was compromised, HP [1-¹³C]lactate could detect decreased [¹³C]bicarbonate in the tumor, reflecting impaired PDH activity.

While in vivo HP pyruvate and HP lactate studies showed markedly different metabolic profiles, the ex vivo NMR analyses presented nearly identical profiles following pyruvate and lactate bolus injections. This discrepancy may arise from multiple factors such as T₁ relaxation and differences in measurement timing. In the ex vivo study, brain tissue was collected 2 mins after bolus injection of lactate or pyruvate, whereas most HP signals including bicarbonate appeared in vivo within 2 mins from the HP injection. Metabolic profiles observed by MALDI MSI may reflect a longer post-injection window, as the tissue was frozen whole using liquid nitrogen rather than rapidly freeze-clamped. Future studies should aim to better align measurement timing across modalities, and the use of high-power focused microwaves could further improve metabolic preservation.

L-[1-¹³C]lactate has been proposed as an alternative HP probe and explored in other applications. HP lactate has been used to probe cardiac metabolism in rodents^58–60 and pigs⁶¹, with reliable detection of lactate oxidation to pyruvate and subsequent conversion to alanine and bicarbonate. Lactate metabolism in rat liver for exploring PC pathways has also been investigated³³, as has skeletal muscle metabolism for investigating lactate utilization^{62, 63}. A technical advantage of lactate is the absence of large pyruvate-hydrate peaks, which can obscure accurate quantification of metabolites of interest. However, highly concentrated lactate can form oligomers that obscure certain spectral regions⁶⁰, although they do not interfere with [¹³C]bicarbonate detection at 161 ppm. Lactate plays a central role in cancer metabolism^{64, 65} and supports cancer growth^{66, 67}, yet HP lactate has not been explored in cancer applications. Our glioma findings highlight the potential of HP lactate to assess LDH activity and altered redox balance in cancer.

CONCLUSION

In this study, we present evidence that neuronal PDH is the primary source of HP [¹³C]bicarbonate generated from HP [1-¹³C]pyruvate or HP [1-¹³C]lactate. While lactate offers certain physiological advantages, our findings indicate that HP pyruvate remains the more effective probe for interrogating neuronal and glioma metabolism due to its direct role in mitochondrial oxidation and robust bicarbonate production. Nonetheless, HP [1-¹³C]lactate positions as a compelling alternative agent, particularly suited for capturing physiological lactate uptake and oxidation in neurons, and for revealing metabolic fluxes obscured by pyruvate-lactate cycling. Such properties may provide unique insights into enzyme activities and redox changes in neurological disorders, including gliomas. Collectively, these results support the view that HP lactate is not only a physiologically relevant substrate but also a potential alternative, or superior in some contexts, probe to pyruvate for investigating brain and glioma metabolism.

ABBREVIATIONS

CSI

chemical shift imaging

FID

free induction decay

FOV

field of view

HP

hyperpolarized

LDH

lactate dehydrogenase

MALDI

matrix-assisted laser desorption/ionization

MSI

mass spectrometry imaging

PC

pyruvate carboxylase

PDH

pyruvate dehydrogenase

TE

echo time

TP

total product

TR

repetition time