Functional activation of pyruvate dehydrogenase in human brain using hyperpolarized [1-¹³C]pyruvate

Abstract

Purpose:

Pyruvate, produced from either glucose, glycogen, or lactate, is the dominant precursor of cerebral oxidative metabolism. Pyruvate dehydrogenase (PDH) flux is a direct measure of cerebral mitochondrial function and metabolism. Detection of [¹³C]bicarbonate in the brain from hyperpolarized [1-¹³C]pyruvate using ¹³C MRI provides a unique opportunity for assessing PDH flux in vivo. This study is to assess changes in cerebral PDH flux in response to visual stimuli using in vivo ¹³C MRS with hyperpolarized [1-¹³C]pyruvate.

Methods:

From seven sedentary adults in good general health, time-resolved [¹³C]bicarbonate production was measured in the brain using 90° flip angles with minimal perturbation of its precursors, [1-¹³C]pyruvate and [1-¹³C]lactate, to test the hypothesis that the appearance of [¹³C]bicarbonate signals in the brain reflects the metabolic changes associated with neuronal activation. With a separate group of healthy participants (n = 3), the likelihood of the bolus-injected [1-¹³C]pyruvate being converted to [1-¹³C]lactate prior to decarboxylation was investigated by measuring [¹³C]bicarbonate production with and without [1-¹³C]lactate saturation.

Results:

In the course of visual stimulation, the measured [¹³C]bicarbonate signal normalized to the total ¹³C signal in the visual cortex increased by 17.1 ± 15.9% (P = 0.017) whereas no significant change was detected in [1-¹³C]lactate. ¹H BOLD fMRI confirmed the regional activation in the visual cortex with the stimuli. Lactate saturation decreased bicarbonate-to-pyruvate ratio by 44.4 ± 9.3% (P < 0.01).

Conclusion:

We demonstrated the utility of ¹³C MRS with hyperpolarized [1-¹³C]pyruvate for assessing the activation of cerebral PDH flux via the detection of [¹³C]bicarbonate production.

INTRODUCTION

A series of metabolic processes, including increases in blood flow, oxygen and glucose consumption, and oxidative phosphorylation, are associated with neuronal activation^1,2. These processes are essential in meeting elevated energy needs for the required brain function. Conventional blood oxygen level-dependent (BOLD) fMRI detects the signal perturbation due to deoxygenation of hemoglobin, and has been the dominant methodology for studying brain function and activation³. However, since BOLD fMRI depends on metabolic changes which are several steps away from neuronal activation, it is susceptible to other physiological changes and artifacts, which can result in diluted signal sensitivity, Figure 1.

Figure 1. Functional Activation of Cerebral Metabolism.

In response to stimulation, oxidative metabolism and the activity of one of its key regulating enzymes, PDH, increase preceding concomitant changes in blood oxygen consumption, which results in changes in the BOLD signal. PDH: pyruvate dehydrogenase; BOLD: blood oxygen level dependent.

Pyruvate dehydrogenase (PDH), the enzyme that regulates pyruvate flux into the tricarboxylic acid (TCA) cycle, plays a critical role in regulating the metabolic processes involved in cerebral oxidative phosphorylation⁴. PDH flux, therefore, is a direct measure of cerebral mitochondrial function and metabolism. The advent of dissolution dynamic nuclear polarization (DNP) of ¹³C-labeled pyruvate enabled in vivo assessment of oxidative phosphorylation in the mitochondria as the production of [¹³C]bicarbonate from administered hyperpolarized [1-¹³C]pyruvate is a biomarker of PDH flux⁵. High-resolution ¹³C imaging of cerebral metabolism in rodents using hyperpolarized [1-¹³C]pyruvate⁶ and its recent translation to human studies^7–9 have reported that [¹³C]bicarbonate is primarily produced in the cerebral cortex where the majority of neuron cell bodies are located. The absence or miniscule amounts of labeled metabolic products specific to pyruvate carboxylase in astrocytes supports that the detected [¹³C]bicarbonate is predominantly produced from neurons^9–11. Coupling neuronal metabolism with hyperpolarized bicarbonate has not been actively explored in preclinical models since the pyruvate metabolism is heavily impacted by the anesthetic condition^12–14, therefore, detection of in vivo [¹³C]bicarbonate production in humans presents exciting opportunities to assess the activation of neuronal metabolism directly.

Despite the unrivaled signal amplification of DNP, reliable detection and precise quantification of time-varying hyperpolarized signals are still technically challenging, particularly for mitochondrial products such as [¹³C]bicarbonate. Previous imaging studies reported inconsistent and largely variable measurement of cerebral [¹³C]bicarbonate production relative to the [1-¹³C]pyruvate signal in the human brain^7,8. This variability could be due to the biological variability among individuals such as baseline PDH activity, cerebral transport of hyperpolarized pyruvate, or the momentary difference in activation of the brain regions that were sampled. However, it is also possible that the large variation is attributed to the difficulty of reliably imaging [¹³C]bicarbonate in the human brain. Dynamic MRS in combination with RF receive arrays has demonstrated robust regional detection of time-resolved hyperpolarized [¹³C]bicarbonate in human brains⁹. However, dynamic ¹³C MRS of hyperpolarized signals using a constant flip angle limits the RF sampling of product peaks at each timepoint and carries over the remaining “unsampled” signals to the next timepoint, making the sectional quantification of the products complicated. Moreover, lactate may play a significant role in the production of bicarbonate since the interconversion of pyruvate and lactate, catalyzed by the enzyme lactate dehydrogenase, is rapid and can occur before and after reaching brain tissues.

In this study, we demonstrate changes of bicarbonate production in response to visual stimuli in healthy volunteers and investigate the metabolic pathways to [¹³C]bicarbonate formation from hyperpolarized [1-¹³C]pyruvate in the brain, considering [1-¹³C]lactate as a precursor.

METHODS

Study Participants

Ten healthy volunteers (age: 35.5 ± 14.1; body mass index = 25.3 ± 2.9; five male and five female) were recruited by word-of-mouth, Table 1. The protocol of this study is fully compliant with the HIPAA regulation, was approved by the Institutional Review Board (IRB) of The University of Texas Southwestern Medical Center (IRB#: STU 072017–009, STU-2018–0013). Written informed consent was obtained from the study participants. All participants tolerated the procedure without any incident.

Table 1.

Study Participant Demographics.

Participant ID	Age [years]	Sex	Race	Ethnicity	BMI [kg/m2]	Injection 1		Injection 2
						RF pulse	Stimuli	RF pulse	Stimuli
1	23	M	Asian	Non-Hispanic	24.9	1	Visual	1	None
2	45	F	American Indian /Alaska Native	Non-Hispanic	27.7	1	Visual	1	None
3	22	F	Black/African American	Non-Hispanic	21.7	1	None	1	Visual
4	36	M	White	Non-Hispanic	25.2	1	Visual	1	None
5	21	M	White	Non-Hispanic	26.7	1	Visual	1	None
6	40	F	Black/African American	Non-Hispanic	23.0	1	None	1	Visual
7	26	M	White	Non-Hispanic	20.5	1	None	1	Visual
8	60	F	White	Non-Hispanic	29.4	1	None	2	None
9	32	M	White	Non-Hispanic	25.8	2	None	1	None
10	27	F	White	Hispanic	20.9	1	None	2	None

Integrated ¹³C/¹H MR Protocol

All studies were performed using a clinical 3T wide-bore MRI scanner (750w Discovery; GE Healthcare, Waukesha, Wisconsin, USA). Each participant underwent a ¹H/¹³C-integrated MR protocol (45 – 60 min), which includes ¹H BOLD fMRI, ¹H Magnetization-Prepared Rapid Acquisition Gradient Echo (MPRAGE), and two dynamic ¹³C MRS with separate injections of hyperpolarized [1-¹³C]pyruvate (pyruvate concentration = 250 mM, dose = 0.1 mmol/kg body weight, volume = 0.43 mL/kg, injection rate = 5 mL/s) with a 30-minute interval between the injections (Figure 2a). A ¹³C/¹H dual-frequency head coil (Clinical MR Solutions, LLC, Brookfield, Wisconsin, USA) that consists of ¹H quadrature transmit and receive, ¹³C quadrature transmit and 8-channel ¹³C receive arrays was used for the study⁹, Figure 2b.

Figure 2. Experimental Protocol and Setup.

(a) The experimental MRI protocol that consists of ¹H BOLD fMRI, ¹H MPRAGE, and two dynamic ¹³C MRS with hyperpolarized [1-¹³C]pyruvate. (b) ¹³C/¹H dual-frequency RF coil. Signals detected in the two posterior ¹³C receive channels (colored in blue) of the ¹³C/¹H dual-frequency RF coil were used to assess the metabolic changes due to visual stimulation. The activation map shown in the center of the coil is the averaged BOLD response to the visual stimuli. (c) Spectral-spatial excitation profiles of two RF pulses used to excite the [1-¹³C]lactate resonance with 10° (RF1) and 90° (RF2) and their spectral profiles (RF1 as red line; RF2 as blue line) at the central position (z = 0). (d) ¹H BOLD fMRI and one of the ¹³C MRS scans were acquired with block-designed checkerboard stimuli while the other ¹³C MRS was without stimulation. AUC: area under the curve; BOLD: blood oxygenation level dependent.

Following a ¹H fast GRE localizer, B₀ inhomogeneity was inspected using ¹H 2D GRE (FOV = 240 mm × 240 mm, spatial resolution = 6.25 mm × 6.25 mm, slice thickness = 2 cm, #slices = 2, TEs = 2.80 ms, 6.10 ms) and the shim currents were retained for ¹³C scans. Simultaneously with each pyruvate injection, time-resolved ¹³C spectra were acquired from a brain slab in axial or oblique plane that included the visual cortex using a dynamic ¹³C MRS sequence (TR = 4 s, spectral width = 10,000 Hz, spectral points = 4,096, slab thickness = 3 cm). The ¹³C MRS acquisitions used a custom-designed spectral-spatial RF pulse to excite hyperpolarized metabolites with distinct flip angles. The ¹³C acquisition center frequency was set to the in vivo [1-¹³C]pyruvate resonance, which was calculated from the resonance frequency of water protons as described previously¹⁵. Between the two ¹³C scans, a 2D ¹H BOLD fMRI (FOV = 240 mm × 240 mm, matrix size = 64 × 64, slice thickness = 5 mm, #slice = 24, TE = 28 ms, TR = 2 s, flip angle = 80°) with the visual stimuli and 3D ¹H MPRAGE images (FOV = 256 mm × 256 mm × 180 mm, matrix size = 256 × 256 × 90, TE = 3.8 ms, TR = 8.7 ms, preparation time = 38 ms, flip angle = 8°) were acquired. The BOLD fMRI and MPRAGE scans were repeated using a 24-channel ¹H head coil (GE Healthcare) for improved signal sensitivity.

Two spectral-spatial RF pulses were designed using the Spectral-Spatial RF Pulse Design Package¹⁶ with the gradient constraints for the wide-bore system (amplitude ≤ 25 mT/m and slew rate ≤ 100 T/m/s). One RF pulse (RF1, pulse duration = 21.9 ms) excited 90° on bicarbonate, 5° on pyruvate and 10° on lactate and the other (RF2, pulse duration = 21.9 ms) was designed to excite 90° on bicarbonate, 5° on pyruvate, and 90° on lactate, Figure 2c.

Experiments with Visual Stimuli

A subgroup of the participants (#1 - #7) was exposed to visual stimulation paradigm using a block design of circular blue/yellow checkerboards (4 Hz, stimulation ON; duration = 20 s) that alternated with a black screen with a cross in the center (stimulation OFF; duration = 20 s) during one of the ¹³C MRS scans, Figure 2d. This paradigm was followed to maximize activation and the power for detection of the response to stimulation¹⁷. The ¹³C brain slab was prescribed to include the visual cortex. As the scan started, the block-designed screen started playing. The second pyruvate injection was performed without the visual stimulation. RF1 was used for both the ¹³C MRS. The same stimuli (4 min) were applied to ¹H BOLD fMRI and ¹³C MRS for consistent stimulation. The dual-frequency ¹H/¹³C RF head coil with 8-channel ¹³C receive array was used during the first 45 minutes of the scan session. The BOLD fMRI and MPRAGE scans were repeated using a 24-channel ¹H head RF coil (GE Healthcare) for improved signal sensitivity.

Experiments with Lactate Saturation

¹³C MRS from the other subgroup (participants #8 - #10) were acquired using RF1 and RF2 to assess the lactate saturation effect without the stimuli. The dual-frequency ¹H/¹³C RF head coil was used throughout the experiment.

Hyperpolarization of [1-¹³C]Pyruvate

Preparation and injection of hyperpolarized [1-¹³C]pyruvate were performed as documented in the Investigational New Drug approval by the Food and Drug Administration (IND#: 133229). Up to three pyruvate samples were simultaneously polarized for each subject using a clinical SPINlab^™ polarizer (GE Healthcare). 1.47g of GMP-grade [1-¹³C]pyruvic acid (>99%; Sigma Aldrich, St Louis, Missouri, USA), mixed with 27.7 mg of AH111501 EPA radical (Syncom, Groningen, Netherlands), was inserted into a clinical fluid path in a sterile environment and polarized for 3 – 4 hours until dissolved by 38 mL of sterile water at 130 °C. The dissolved pyruvate solution was immediately mixed with 36.5 mL of room temperature TRIS/NaOH media (333 mM/600 mM) and tested using a quality control (QC) device (GE Healthcare) prior to the injection. Pyruvate concentration (238.9 ± 5.6 mM), pH (7.7 ± 0.5), and residual radical (1.4 ± 0.7 μM) of the injectate were measured. In parallel to the QC analysis, the hyperpolarized pyruvate solution was transferred to the study participant for an intravenous injection (transfer time = 56.4 ± 6.6 s).

¹³C Data Reconstruction and Analysis

Dynamic ¹³C MRS were reconstructed from the GE MR raw data using MATLAB R2021b (Mathworks, Natick, Massachusetts, USA) as described previously¹⁸. For assessing metabolic responses to visual stimuli (participants #1 – #7), ¹³C data from channels #2 and #3 of the ¹³C receive arrays were selectively combined and analyzed due to the proximity to the visual cortex. For comparing the RF pulses (participants #8 – #10), data from all ¹³C channels were combined. Metabolite peaks, [¹³C]bicarbonate, [1-¹³C]lactate, and [1-¹³C]pyruvate, were quantified from phase-corrected ¹³C spectra by integrating the corresponding peak in absorption mode. The total ¹³C signal, which was used to normalize [¹³C]bicarbonate and [1-¹³C]lactate peaks, was calculated by summing [¹³C]bicarbonate, [1-¹³C]lactate, and [1-¹³C]pyruvate signals. For display purposes, the baseline was subtracted from the time-averaged spectra (0 – 160 s) by fitting a spline to the signal-free regions of the smoothed spectrum and normalized by the [1-¹³C]pyruvate peak.

BOLD fMRI Analysis

All the BOLD fMRI data were analyzed using the Statistical Parametric Mapping Software version 12 (SPM12; Functional Imaging Laboratory, University College London, London, UK). In order to eliminate the possibility of potential motion artifacts during acquisition, each dataset in the NIfTI-1 format was first realigned by registering the 144 images to the first image in the series using a least-squares approach and a six-parameter spatial transformation. The realigned images were then normalized to the Montreal Neurological Institute (MNI) template space and convolved with a Gaussian kernel of a full width at half maximum of 8 mm isotropically.

Statistical Analysis

Data are presented as mean ± standard deviation. Statistical significance of changes in hyperpolarized ¹³C metabolite signals between injections was evaluated by paired t-tests (α = 0.05, two-tailed), assuming the normality of hyperpolarized ¹³C metabolite signals and metabolite ratios^19,20. Indeed, Miller, et al. showed that distribution of metabolite ratios is approximately normally distributed in the high signal-to-noise ratio regime (SNR > 10)²⁰. The normality of ¹³C metabolic ratios was also confirmed in a brain tumor patient study¹⁹. For statistical analysis of the fMRI data, model specification was performed on each individual subject’s images using a 1^st-level mass-univariate method based on general linear models with the masking threshold of 0.8. A brain activation map was generated using the contrast comparison between checkerboard stimulation and rest. Then, to acquire the mean activation of all subjects, a one-sample t-test in 2^nd-level analysis was performed on the smoothed images. The mean activation clusters were generated for P < 0.05 and superimposed on the SPM12 canonical brain atlas (avg152T2).

RESULTS

Participants in the first subgroup showed increased bicarbonate productions in the posterior ¹³C coil components (channel #2 and #3) while viewing the checkerboard stimulus, as compared to those measured without the checkerboard stimuli, Figure 3a. The averaged BOLD response to the visual stimuli is shown in Figure 2b. Bicarbonate signal intensity (bicarbonate normalized to the total ¹³C signal) in time-averaged (0 – 160 s) ¹³C spectra increased by 17.1 ± 15.9 % from 0.154 ± 0.068 at rest to 0.178 ± 0.081 with the stimuli (P = 0.017), Figure 3b. Lactate normalized to the total ¹³C signal also mildly increased from 0.310 ± 0.072 to 0.321 ± 0.075 (P = 0.064). The peak SNR of each metabolite in the time-averaged ¹³C spectrum was 138 ± 52 for [¹³C]bicarbonate (n = 14), 186 ± 68 for [1-¹³C]lactate, and 438 ± 223 for [1-¹³C]pyruvate. The pyruvate time-courses were comparable between the two injections, but time-advanced [1-¹³C]pyruvate appearance was detected from some of the participants (#1, #4, #5) at the onset of the first stimulus beginning at 20 s, Figure 3c. During the first stimulus, the overall [¹³C]bicarbonate signal was elevated (Figure 3d) but [¹³C]lactate signal was maintained or increased by smaller amount from selected participants (#1, #3, #4, #5). The largest signal increase was detected at 4.6 ± 2.8 s from the start of the stimulus for [1-¹³C]lactate and 5.1 ± 2.0 s for [¹³C]bicarbonate. Signal changes during the subsequent stimuli presentation were under the detection limit.

Figure 3. Activation of Pyruvate Dehydrogenase Flux in Response to Visual Stimulation.

(a) Time-averaged ¹³C spectra acquired from a prescribed 3-cm slab that includes the visual cortex using the RF coil and the non-lactate-saturating RF pulse (RF1) with and without visual stimulation. (b) Bicarbonate AUC, normalized to the total ¹³C signal, increased with the stimuli as compared to the baseline in all participants. The presented time-courses of hyperpolarized (c) [1-¹³C]pyruvate, (d) [1-¹³C]lactate, and (e) [¹³C]bicarbonate signals with and without visual stimulation are from a representative participant (#1, 23 y.o. male). A coefficient was applied to the time-courses to match the maximum pyruvate signals between the injections. *: P < 0.05

Time-resolved ¹³C spectra were acquired from the second subgroup with two consecutive injections of hyperpolarized [1-¹³C]pyruvate using two different RF pulses. A larger [¹³C]bicarbonate signal was detected using RF1 without lactate saturation as shown in the time-averaged ¹³C (Figure 4a) and the time-courses (Figure 4c,d). The measured bicarbonate-to-pyruvate ratio from time-averaged ¹³C spectra was 0.549 ± 0.045 with RF1 and 0.307 ± 0.066 with RF2 (P = 0.0098), which is a decrease by 44.4 ± 9.3 %. The lactate-to-pyruvate ratio increased by 56.7 ± 5.7 % from 0.838 ± 0.281 with RF1 to 1.304 ± 0.400 with RF2 (P = 0.021), Figure 4b.

Figure 4. Effect of Saturating [1-¹³C]Lactate on [¹³C]Bicarbonate Formation.

(a) Time-averaged ¹³C spectra from a representative healthy participant (#9, 32 y.o. male) using the RF pulses. (b) Decreased [¹³C]bicarbonate and increased [1-¹³C]lactate were measured with the lactate-saturating RF pulse (RF2) as compared to the measurements without lactate saturation (RF1). Time-courses of (c) [¹³C]bicarbonate and (d) [1-¹³C]lactate acquired from the participant. *: P < 0.05; **: P < 0.01

DISCUSSION

Pathway Analysis of Cerebral [¹³C]Bicarbonate Production from Hyperpolarized [1-¹³C]Pyruvate

Metabolic profiles of neurons and astrocytes are distinct in terms of utilizing energy substrates to maintain metabolic homeostasis and to adaptively support high energetic demands of neurons. Pyruvate, the end-product of glycolysis and the redox pair, lactate, play a pivotal role in oxidative metabolism in the brain. Any form of its primary substrates, including glucose, glycogen, and lactate, is converted to pyruvate prior to entering the mitochondria for oxidative metabolism.

Although metabolic processes of steady-state infused ¹³C-labeled thermal (non-hyperpolarized) glucose or lactate have been studied using ¹³C MRS over timeframes ranging from minutes to hours, the metabolism of bolus injected pyruvate or lactate in the hyperpolarization timescale has not been explored. While intracellular pyruvate is undoubtedly the direct precursor of bicarbonate, it has often been assumed that this bicarbonate is converted straight from an administered hyperpolarized pyruvate bolus. However, [1-¹³C]lactate, another major product of hyperpolarized [1-¹³C]pyruvate in the brain, can complicate this procedure. Unlike [¹³C]bicarbonate whose production is mitochondrion-specific, bolus-injected pyruvate can be rapidly converted to lactate in red blood cells and other organs before reaching brain tissues, Figure 5. The significantly reduced [¹³C]bicarbonate signals in the dynamic ¹³C MRS obtained with lactate saturation confirms that [1-¹³C]lactate generated from a [1-¹³C]pyruvate bolus is a substantial contribution to [¹³C]bicarbonate formation in the brain. Therefore, preserving hyperpolarization in [1-¹³C]lactate in addition to [1-¹³C]pyruvate is required for optimal detection of total [¹³C]bicarbonate production. This observation is consistent with previous studies that saturated lactate prior to bicarbonate imaging^21–23.

Figure 5. Pathway to Hyperpolarized [¹³C]Bicarbonate Formation in The Brain.

Intravenously injected [1-¹³C]pyruvate bolus can be converted to [1-¹³C]lactate by red blood cells or other organs before being delivered to the brain. ¹³C-labeled pyruvate and subsequent metabolic products are in blue font. GLUT: glucose transporter; LDH; lactate dehydrogenase; MCT: monocarboxylate transporter; MPC; mitochondrial pyruvate carrier; PC: pyruvate carboxylase; PDH: pyruvate dehydrogenase; RBC: red blood cell

Contribution of Neuronal Metabolism to [¹³C]Bicarbonate Production

It has been proposed that glutamate stimulates astrocytes to increase aerobic glycolysis, followed by lactate shuttling to neurons²⁴. According to this astrocyte-neuron lactate shuttle (ANLS) theory²⁵, lactate is produced from glucose or glycogen in astrocytes, released via monocarboxylate transporter 4 (MCT4), and transferred to neurons²⁶. However, the ANLS hypothesis remains controversial as it does not fulfill critical stoichiometric requirements between glucose uptake, lactate synthesis and release, and neuronal oxidation²⁷. An infusion study with [3-¹³C]lactate reported that the relative lactate transport capacity of neurons and astroglia can exceed glucose oxidation rates and, therefore, creates a shared lactate pool²⁸. Along with the lactate shuttle theory which describes the roles of lactate in delivery of substrates²⁹, the direct lactate diffusion concept supports that pyruvate and lactate are adequate substrates for monitoring neural metabolism and neuronal activity. In particular, at supraphysiological plasma lactate concentrations as seen with the injection of hyperpolarized pyruvate or lactate, it was estimated that the contribution of plasma lactate to brain metabolism via direct lactate transport can increase up to 60%²⁸. Additionally, significantly smaller in vivo lactate permeability was reported in astrocytes than neurons when lactate was injected intravenously³⁰, supporting our observation that neurons are a significant source of [¹³C]bicarbonate production during stimulation.

The association between neuronal activity and cellular metabolism was recently investigated using hyperpolarized [1-¹³C]pyruvate ex vivo by Grieb, et al.³¹. In this study, hyperpolarized [¹³C]bicarbonate and [1-¹³C]lactate in rat cerebrum slices were monitored after blocking action potential generation and inhibiting ATP-synthase using tetrodotoxin and oligomycin, respectively, and reported that PDH activity, measured by [¹³C]bicarbonate, in the cerebrum increased by 4.4-fold with tetrodotoxin and decreased by 7.4-fold with oligomycin. This study supports that neuronal activity affects PDH activity and, thus, production of [¹³C]bicarbonate.

In Vivo Detection of Cerebral PDH Activation in Humans

Assessing PDH flux has several advantages over the conventional imaging approaches. As illustrated in Figure 1, a cascade of physiological, metabolic processes is involved in brain activation. When the brain activation triggers oxidative phosphorylation in the mitochondria, PDH activity and subsequent oxidative metabolism in neurons increase during stimulation^32,33. The increased oxidative metabolism demands a continuous supply of oxygen and glucose to neurons and astrocytes, leading to an increase in cerebral blood flow (CBF), cerebral blood volume, and cerebral metabolic rate of oxygen consumption.

The combinational changes of these factors form the basis of BOLD fMRI. The BOLD signal originates from deoxygenated hemoglobin, which is highly paramagnetic, resulting in gradients in the magnetic field and appearing as changes in T₂* relaxation in MRI³. As an indirect detection of neuronal activities, BOLD fMRI has several outstanding limitations. For instance, the concentration of deoxygenated blood can fluctuate due to variances in blood pressure or blood volume, affecting the BOLD signal^34,35. In addition, the vascular delay that occurs during the signal acquisition can underestimate the level of neural activity³⁶. For proper interpretation of brain activity, uncoupling of the neurovascular response is necessary to adjust the effects of the hemodynamic response on BOLD fMRI³⁷. The hemodynamic response (CBF) can be measured by positron emission tomography (PET) techniques but the complex kinetics and large background signals complicate the analysis³⁸.

PET techniques can be also used to measure metabolic changes associated with brain activation such as variances in the cerebral metabolic rates of oxygen and glucose^39,40. In particular, glucose utilization during the focal physiologic neural activity elicited by visual stimulation can be detected using ¹⁸F-fluorodeoxyglucose PET⁴⁰. However, cerebral glycogen metabolism needs to be uncoupled to accurately measure glucose oxidation resulting from stimulation. The glycogen shunt is a metabolic pathway that occurs in cells, in which glucose is converted into glycogen instead of being metabolized through the usual process of glucose oxidation⁴¹. In the brain, a fraction of glucose can be cycled through the cerebral glycogen pool and this cycling can increase with the degree of brain activation⁴². To measure glucose oxidation in the brain, it is necessary to uncouple the glycogen shunt by different methods such as inhibiting the enzymes involved in glycogen synthesis or depleting the glycogen stores.

In contrast, PDH activation is a direct measure of neuronal activity⁴³ that is not affected by the blood oxygen level. As a surrogate biomarker of PDH activity, the stimulus-triggered elevation of [¹³C]bicarbonate production in the visual cortex suggests that the change reflects increased PDH flux due to neuronal activation. Although immediate increases in [¹³C]bicarbonate production to the stimuli were detected in our study, further studies with faster sampling schemes would be necessary to investigate whether the metabolic change occurs prior to the hemodynamic response.

Several studies reported limited cellular transport of pyruvate. Rao, et al. reported that hyperpolarized [1-¹³C]pyruvate-to-[1-¹³C]lactate conversion rates are primarily a functional readout of [1-¹³C]pyruvate transmembrane influx mediated by MCT1⁴⁴, an isoform that is mainly detected on endothelial cell of the blood-brain barrier (BBB). Indeed, opening the BBB facilitated cellular transport of hyperpolarized pyruvate and lactate to the brain^45–47. In particular, Hackett, et al. demonstrated increased production of [¹³C]bicarbonate and [1-¹³C]lactate in the brain from hyperpolarized [1-¹³C]pyruvate after opening the BBB with focused ultrasound⁴⁷. Thus, it is possible that [¹³C]bicarbonate production is already saturated without additional activation. However, MCT expression is regulated in relation to brain activity⁴⁸. In a neural-derived cell line, both MCT1 and MCT4 are upregulated in hypoxia⁴⁹. When the brain activation upregulates PDH activity and subsequent oxidative metabolism in the TCA cycle, the elevated PDH activity increases the uptake of lactate (and pyruvate) whose cellular transport is via MCT1 to astrocytes and via MCT2 to neurons^50–52.

Cerebral Lactate Production

Our study with the second subgroup focused on identifying the source of [¹³C]bicarbonate in the brain. The result showed that [¹³C]bicarbonate decreased by 44 % when lactate was saturated, suggesting that [1-¹³C]lactate produced outside or inside the brain contributes approximately 56 % of [¹³C]bicarbonate. Whether the [1-¹³C]lactate produced outside the brain is the dominant source of [1-¹³C]lactate in the brain is uncertain. An experimental setup such as having a lactate saturation band right outside of the brain (e.g., neck) prior to assessing brain metabolism would estimate the amount of [1-¹³C]lactate production from [1-¹³C]pyruvate in the brain.

Mixed results have been reported in terms of the change in lactate content by visual stimulation^53–56. Our study showed an increasing trend of hyperpolarized [1-¹³C]lactate during visual stimuli but the change was not as obvious as that seen in [¹³C]bicarbonate. This is possibly due to the small flip angle excitation used for lactate and the increased [1-¹³C]lactate delivery to the visual cortex by the elevated hemodynamic response. Further investigation is needed for quantitative investigation of functional changes in lactate.

Limitation and Future Improvement

This proof-of-concept study was performed with a limited number of subjects, and a larger cohort study may characterize functional metabolic changes better, depicted by hyperpolarized ¹³C metabolites, with enhanced statistical power. The demonstrated method has several technical limitations that should be considered for future studies. First, a short observation window is an innate methodological limitation of the functional ¹³C MRS with hyperpolarized [1-¹³C]pyruvate due to the use of a contrast agent that generates transient hyperpolarized signals. Despite the high hyperpolarization level of [1-¹³C]pyruvate and the optimized acquisition protocol, detecting task-induced [¹³C]bicarbonate change was not achievable approximately a minute after a [1-¹³C]pyruvate injection, which may underestimate the true extent of neuronal activation.

Second, data acquisition scheme can be further optimized to improve detection. The acquisition scheme of fully exciting hyperpolarized products at each dynamic time point was introduced for estimating metabolic exchange rates⁵⁷ and metabolic reaction rates⁵⁸. In our study, the MR acquisition parameters were to reliably detect the relevant ¹³C signals in the brain: [1-¹³C]pyruvate, [1-¹³C]lactate, and [¹³C]bicarbonate. While bicarbonate was fully RF-sampled at each timepoint with a 90° excitation to measure produced bicarbonate signal during the time interval between the timepoint and the previous measurement, pyruvate and lactate were sampled with small flip angles to prevent their signals from being destroyed prior to the bicarbonate conversion. The flip angles of 5° and 10° were used for pyruvate and lactate, respectively, considering that hyperpolarized probes in injectate are typically larger than their products. Although the acquisition scheme with the parameters worked successfully, identical flip angles for both precursors would simplify the analysis. With the demonstrated signal sensitivities, even smaller flip angles for the precursors may be plausible for reliable detection of their time-courses and would further improve the bicarbonate sensitivity.

Third, identifying the response time of PDH flux to stimuli was limited, primarily due to the low temporal resolution. Shortening the repetition time, however, would decrease the amount of produced bicarbonate between timepoints, resulting in a reduced sensitivity of [¹³C]bicarbonate detection, and would require more frequent RF excitations, which would shorten the observation time window. These dynamic acquisition and RF sampling schemes need to be investigated sufficiently prior to establishing appropriate imaging strategies.

Another parameter that was not thoroughly explored in this study is the stimulus design. A 20 s stimulus duration and interval were to avoid saturation and fatigue of the visual system and were chosen based on previous BOLD fMRI study design^17,59. Moreover, a 30-minute interval between the ¹³C scans may not be long enough for a complete return of stimulated brain metabolism to baseline, requiring further investigations.

Finally, quantitative techniques may need to be improved for future studies. This study focused on [¹³C]bicarbonate and [1-¹³C]lactate, normalized to the total ¹³C signal. Due to lacking polarization measurements, absolute signal changes were not included in the analysis, and therefore, any change in blood flow could not be identified. For this, it is necessary to normalize the raw signal intensities by compensating experimental parameters such as polarization level and transport time¹⁸. In addition, the RF sampling factors should be considered to improve precision. However, quantifying multi-compartmental distribution of partially sampled, time-varying pyruvate and lactate signals with unknown in vivo characteristics such as T₁ relaxation times remains challenging^60,61. Kinetic analysis can be considered as a potential quantification approach, considering that changes in metabolic reaction rates in the brain were previously demonstrated for data acquired with the product-selective saturating scheme⁵⁸.

In conclusion, we demonstrated the utility of ¹³C MRI with hyperpolarized [1-¹³C]pyruvate for assessing the activation of cerebral PDH flux. Our observation of elevated [¹³C]bicarbonate formation in the visual cortex directly from hyperpolarized [1-¹³C]pyruvate or from its reduced form, [1-¹³C]lactate, in response to visual stimulation establishes the metabolic basis of existing modalities used to assess neuronal metabolism such as BOLD fMRI and PET.