Evidence of ¹³C-lactate oxidation in the human brain from hyperpolarized ¹³C-MRI

Abstract

Purpose:

To test the hypothesis that lactate oxidation contributes to the ¹³C-bicarbonate signal observed in the awake human brain using hyperpolarized ¹³C MRI.

Methods:

Healthy human volunteers (N = 6) were scanned twice using hyperpolarized ¹³C-MRI, with increased radiofrequency saturation of ¹³C-lactate on one set of scans. ¹³C-lactate, ¹³C-bicarbonate, and ¹³C-pyruvate signals for 132 brain regions across each set of scans were compared using a clustered Wilcoxon signed-rank test.

Results:

Increased ¹³C-lactate radiofrequency saturation resulted in a significantly lower ¹³C-bicarbonate signal (p = 0.04). These changes were observed across the majority of brain regions.

Conclusion:

Radiofrequency saturation of ¹³C-lactate leads to a decrease in ¹³C-bicarbonate signal, demonstrating that the ¹³C-lactate generated from the injected ¹³C-pyruvate is being converted back to ¹³C-pyruvate and oxidized throughout the human brain.

1 |. INTRODUCTION

It is well known that glucose is the primary source of energy in the brain. Mounting evidence suggests that at least some of this glucose is first converted to lactate and moved between cellular compartments before ultimately being converted back to pyruvate and oxidized in the TCA cycle in a process known as “lactate oxidation.”^1,2 In this study, the hypothesis that lactate oxidation contributes to the ¹³C-bicarbonate signal observed in the awake human brain is tested using hyperpolarized [1-¹³C] pyruvate MRI.

Hyperpolarized [1-¹³C]pyruvate MRI is a methodology for metabolic profiling in vivo. Following the intravenous injection of ¹³C-pyruvate, the downstream metabolites ¹³C-lactate and ¹³C-bicarbonate can be readily imaged in the human brain.^3–6 Signal from ¹³C-lactate shows intra-celluar conversion of ¹³C-pyruvate via the enzyme lactate dehydrogenase (LDH) and is higher with increased local rate of lactate production and/or increased local lactate pool size.

The ¹³C-bicarbonate signal represents bicarbonate created when ¹³C-pyruvate is converted to acetyl-CoA on the mitochondrial membrane, with the acetyl-CoA entering the TCA cycle to drive mitochondrial oxidative phosphylation. Thus ¹³C-bicarbonate signal indirectly reflects mitochondrial oxidation of ¹³C-pyruvate.

However, if lactate oxidation is occurring at a significant level in the human brain, some fraction of the ¹³C-bicarbonate could be formed from ¹³C-lactate that has been created in the cytosol and subsequently converted back to ¹³C-pyruvate and consumed in the TCA cycle in mitochondria (see ¹³C-bicarbonate_lac in Figure 1B). Evidence of oxidative consumption of ¹³C-lactate in the rat brain was previously demonstrated,⁷ with ¹³C-bicarbonate signal in the brain reduced when radiofrequency (RF) saturation was applied to the ¹³C-lactate pool. In this study, the hypothesis that ¹³C-lactate oxidation into ¹³C-bicarbonate is occurring in the human brain was tested using two consecutive hyperpolarized [1-¹³C]pyruvate MRI scans with two different flip angles (46° vs. 80°) applied when imaging ¹³C-lactate.

FIGURE 1.

Schematic representation of the experiment. (A) The two main pathways that ¹³C-pyruvate takes in the brain, with differential RF saturation shown applied to the ¹³C-lactate pool. (B) Two distinct sources of mitochondrial ¹³C-bicarbonate are shown, one resulting from ¹³C-pyruvate directly, and the second resulting from ¹³C-pyruvate_lac that has spent some time as ¹³C-lactate during the timescale of the experiment, producing ¹³C-bicarbonate_lac. This experiment was designed to test for ¹³C-bicarbonate_lac signal.

2 |. METHODS

Written informed consent was obtained from study participants (N = 6) between the ages of 23 and 48 who were imaged under a protocol approved by the Research Ethics Board of Sunnybrook Health Sciences Centre and regulated by Health Canada under a Clinical Trial Application. Participants were screened for cognitive impairment using the Montreal Cognitive Assessment.⁸

2.1 |. Data acquisition

Each dose was compounded from a 1.47 g sample of [1-¹³C]pyruvic acid (Sigma Aldrich) using a SPINLab polarizer system (GE Healthcare). A 20-gauge intravenous catheter was placed in the forearm of each participant prior to scanning for injection of the hyperpolarized [1-¹³C]pyruvate solution.

Imaging was performed using a GE MR750 3.0T MRI scanner (GE Healthcare), and a ¹³C birdcage head coil built in-house. A spectral-spatial excitation and three-dimensional dual-echo echo-planar imaging readout were used to acquire images of ¹³C-lactate, ¹³C-bicarbonate, and ¹³C-pyruvate with a temporal resolution of 5 s and an isotropic spatial resolution of 15 mm.⁹ Each volumetric ¹³C-metabolite image was acquired at 5-s intervals over the course of a 60-s acquisition window.

The hyperpolarized [1-¹³C]pyruvate MRI scan was performed twice for each participant, with a 30-min wait time between the two scans. For one of the scans, an increased flip angle was used for excitation of the ¹³C-lactate resonance, in order to increase the saturation of the ¹³C-lactate pool, while the flip angle for ¹³C-bicarbonate and ¹³C-pyruvate remained the same for both scans.

The net flip angle was 80° for ¹³C-lactate, resulting from a 20° spectrally selective excitation pulse applied 24 times for the 24 phase encodes in the slice direction, 80° for ¹³C-bicarbonate, and 11° for ¹³C-pyruvate in the condition designated the “lactate-80°” scan. The “lactate-46°” scan was acquired with a net flip angle of 46° excitation of lactate (24 applications of a 10° excitation pulse), with the flip angles for the other ¹³C-metabolites held constant. The first three participants (volunteers 1–3) had the lactate-46° scan done as the second scan, while the latter three scans had the lactate-46° done as the first scan.

During the half-hour wait time between the two ¹³C-metabolite scans, the ¹³C head coil was interchanged with an eight-channel ¹H neurovascular array (Invivo Inc) for anatomical imaging. Anatomical images were acquired using axial fast spoiled gradient echo images (field of view 25.6 × 25.6 cm², 1 mm isotropic resolution, pulse repetition time 7.6 ms, echo time 2.9 ms, flip angle 11°). All ¹³C-images were reconstructed and resampled to the resolution of the anatomical images in Matlab¹⁰ before being saved in DICOM format.

2.2 |. Data analysis

For each subject, the metabolite images from the full 60-s acquisition window were summed to produce time-integrated ¹³C images. The T1-weighted anatomical images were parcellated into the 132 brain regions in the BrainCOLOR labelling protocol¹¹ using the Spatially Localized Atlas Network Tiles method.¹² The parcellation maps were then used to compute mean ¹³C-pyruvate, ¹³C-lactate and ¹³C-bicarbonate signal for each region and subject using the mrisegstats software (http://nmr.mgh.harvard.edu). These are referred to below as regional metabolite signal values and are in measured arbitrary units but are comparable between subjects and conditions despite additional sources of variance such as slightly differing ¹³C polarization levels.

In order to test for a significant difference in the regional ¹³C-metabolite signal between the two conditions, a clustered Wilcoxon signed-rank test was used.¹³ The paired ¹³C-bicarbonate signals (each pair for the lactate-46° and lactate-80° conditions, for all of the atlas regions from a each individual participant were designated as a cluster and the clusrank package¹⁴ R version 3–3¹⁵) was used to perform the Wilcoxon signed-rank test combining the data from the six participants. This was also performed for ¹³C-lactate and ¹³C-pyruvate signal.

The regional percentage change between the lactate-46° and lactate-80° conditions was calculated by

$$\mathrm{\Delta }\mathrm{\%}=\frac{\left.\left[S_j\left(\text{lactate}-{46}^{\circ }\right)-S_j\text{(lactate}-{80}^{\circ }\right)\right]}{S_j\left(\text{lactate}-{46}^{\circ }\right)}\times 100\mathrm{\%},$$ (1)

where $S_j\left(\text{lactate}-{46}^{\circ }\right)$ denotes the metabolite signal from the $j_{th}$ region in the lactate-46° condition. To assess the relationship between this regional percentage change and the regional ¹³C-metabolite signal level, Spearman’s rank correlations were computed, using the mean of each brain region across subjects. The false-discovery rate method¹⁶ was used to set the threshold for significance at p = 0.013, accounting for the 15 correlations performed.

To rule out the possibility of differences in ¹³C-pyruvate polarization being the primary cause of differences between the two conditions, a t-test was performed between the polarization levels for the two conditions, as well as for the time from dissolution to injection (N = 6).

3 |. RESULTS

The decrease in ¹³C-bicarbonate signal across regions is apparent in the example images from one participant in Figure 2 and in the boxplots showing all of the regions for the two conditions in Figure 3.

FIGURE 2.

Example images from the lactate-46° (left) and lactate-80° (right) conditions. All images are from the same participant. The ¹³C-lactate images (upper row) show that ¹³C-lactate signal (colour overlay) is increased in the lactate-80° condition, as expected. Conversely, the ¹³C-bicarbonate signal (middle row) is decreased in the lactate-80° condition. The ¹³C-pyruvate signal (bottom row) shows decreased signal in the lactate-80° condition.

FIGURE 3.

Boxplots of mean regional ¹³C-lactate, ¹³C-bicarbonate, and ¹³C-pyruvate signal for each of the six participants. The boxes in each panel shows the regional ¹³C-metabolite signal from the lactate-46° (left) and lactate-80° (right) conditions. Lines connect the same brain region between the two different conditions for each participant.

The clustered Wilcoxon signed-rank tests showed a significant difference between conditions for ¹³C-bicarbonate (p = 0.04, Z = −1.783), supporting the main hypothesis that increased RF saturation of the ¹³C-lactate pool decreases the ¹³C-bicarbonate signal. The ¹³C-pyruvate signal also showed a decrease although this did not reach significance (p = 0.08, Z = −1.4). For the ¹³C-lactate signal itself, the higher flip angle gave the largest and most significant change (increase) in signal (p = 0.01, Z = +2.3), as was expected.

The mean percentage change between the lower and higher flip angle was +25% ±7% for ¹³C-lactate, −21% ± 5% for ¹³C-bicarbonate and −6% ± 6% for ¹³C-pyruvate. A negative percentage change indicates reduced ¹³C-metabolite signal for the lactate-80° condition relative to the lactate-46° condition. The ratio of ¹³C-bicarbonate signal from the lactate-80° scan to ¹³C-bicarbonate signal from the lactate-46° scan is plotted for each of the 132 brain regions and plotted in Figure 4A. Each box is composed of the N = 6 ratios for each brain region.

FIGURE 4.

¹³C-metabolite signal measured in the lactate-80° condition divided by ¹³C-metabolite signal measured in the lactate-46° condition for each of the 132 brain regions. The ¹³C-bicarbonate signal ratio (A) is <1.0 in most regions. The ¹³C-pyruvate signal ratio (B) is also <1.0 in most regions, but less so than in (A). Each box comprises the N = 6 ratios for the region listed on the horizontal axis. The red line shows unity, which denotes equal signal in the two conditions.

The majority of brain regions have mean values below 1.0 (redline), which indicates reduced ¹³C-bicarbonate signal with increased ¹³C-lactate RF saturation. A similar effect was observed for ¹³C-pyruvate signal (Figure 4B), with the ratio for most regions below 1.0, indicating reduced ¹³C-pyruvate signal with increased RF saturation of ¹³C-lactate. Both of these results are consistent with the red pathway steps in Figure 1B being a significant source of ¹³C-bicarbonate signal in the human brain.

Figure 5 the percentage change (denoted Δ %) in regional ¹³C-bicarbonate signal between the lactate-46° and lactate-80° conditions are plotted vs. the mean regional signal (from the lactate-80° condition) for each of the three ¹³C-metabolites in Figure 6. There was a weak but significant negative correlation between the percentage change in ¹³C-bicarbonate signal and the mean regional ¹³C-lactate signal, suggesting that brain regions that produced more ¹³C-lactate were associated with increased oxidative consumption of ¹³C-lactate in the TCA cycle (after being converted back to ¹³C-pyruvate).

FIGURE 5.

Correlation matrix of the percent changes in each ¹³C-metabolite signal between conditions as a function of each ¹³C-metabolite (*p ≤ 0.05, **p ≤ 0.01, ***p < 0.001).

FIGURE 6.

The percentage change (Δ %) in regional ¹³C-pyruvate signal between the two conditions plotted versus mean regional ¹³C-metabolite signal. A significant negative correlation is seen in (A). Each point is a particular brain region averaged across the six healthy volunteers. The Spearman correlation coefficient and corresponding p-value are listed in each plot.

Conversely, the percentage change in ¹³C-pyruvate signal between the lactate-46° and lactate-80° conditions showed a weak but significant positive correlation with the regional ¹³C-metabolites signal (see Figure 7). The opposing correlations observed for Δ % ¹³C-pyruvate and Δ % ¹³C-bicarbonate may be explained by the hypothesis that regions with lower net ¹³C-pyruvate uptake tend to contain a higher fraction of ¹³C-lactate-derived ¹³C-pyruvate (¹³C-pyruvate_lac in Figure 1) simply because the local ¹³C-pyruvate pool size is smaller and thus more easily enriched by the ¹³C-lactate-to-¹³C-pyruvate conversion occurring at (or approaching) equilibrium. In addition, the overall smaller percentage change in ¹³C-pyruvate signal between ¹³C-lactate saturation conditions may be explained by the the presence of extracellular ¹³C-pyruvate that would be expected to be less affected by ¹³C-lactate saturation.

FIGURE 7.

The percentage change (Δ%) in regional ¹³C-pyruvate signal between the two conditions plotted versus mean regional ¹³C-metabolite signal. All three plots show a significant positive correlation. Each point is a particular brain region averaged across the six healthy volunteers. The Spearman correlation coefficient and corresponding p-value are listed in each plot.

Table S1 shows the liquid-state ¹³C polarization readouts from the quality-control module on the SPINLab polarizer immediately after dissolution, as well as time-to-injection after dissolution. The t-tests comparing the polarization and time-to-injection for the lactate-80° vs. the lactate-40° scans were both insignificant (p = 0.95 and p = 0.77, respectively).

4 |. DISCUSSION

Previous studies have shown evidence of the oxidative metabolism of lactate in the brain.^17,18 In this study, it was shown that increased RF saturation of the ¹³C-lactate pool resulted in reduced ¹³C-bicarbonate signal. The is in agreement with a prior rat study that assessed the effects of ¹³C-lactate saturation in multiple organs, with the strongest increase in ¹³C-bicarbonate signal observed in the brain,⁷ though supraphysiological ¹³C-pyruvate concentrations being constrained within vasculature by the blood-brain barrier may be why the effect of ¹³C-lactate saturation is more profound in the brain than in other organs.

The results demonstrate that a measurable fraction of ¹³C-lactate in the brain is being converted back to ¹³C-pyruvate and consumed in mitochondria, resulting in ¹³C-bicarbonate. An increase in ¹³C-pyruvate signal in the lactate-46° condition compared to the lactate-80° condition was also observed. This is consistent with the model shown in Figure 1 wherein ¹³C-lactate is converted back to ¹³C-pyruvate prior to conversion to acetyl-CoA via pyruvate dehydrogenase. The increase in both ¹³C-pyruvate and ¹³C-bicarbonate under the lactate-46° condition was observed across most brain regions assessed. Differences between brain regions were not subjected to statistical testing due to the small sample size. Such differences might be expected though, as differential expression patterns of the enzymes LDH and pyruvate dehydrogenase between brain regions, as well as the expression of regulating factors like ur were observed in maps of single-cell RNA-sequencing from the brain.¹⁹

A hypothetical mechanism for the observed decrease in ¹³C-bicarbonate signal is that there is a separate compartment with shifted equilibrium, which could be an intracellular compartment such as the mitochondria, that promotes the conversion of ¹³C-lactate back to ¹³C-pyruvate and subsequent conversion to acetyl-coA on the mitochondrial membrane (lower pathway in Figure 1B). This is supported by the presence of LDH isoforms in mitochondria that favor the conversion of ¹³C-lactate into ¹³C-pyruvate.²⁰ Different equilibria for reactions dependent on NADH/NAD+ in the cytosol versus the mitochondria have been documented, with shuttling mechanisms such as malate-aspartate shuttle facilitating the transfer of reducing equivalents between the two compartments.²¹

It is well documented that lactate and pyruvate pools within the cytosol are in exchange and that this can confound the interpretation of ¹³C-metabolite signals.²² This principle of “label exchange near equilibrium” is used in the enzyme kinetics literature to analyze the dynamics of an enzyme-substrate system after introduction of a trace quantity of labeled substrate to a mixture of enzyme, substrate and product near equilibrium. Measurement of the resulting dynamics of the labeled substrate can be used to accurately measure the enzyme activity under physiological conditions.^23–25 Short-lived ternary complexes of LDH-NADH-pyruvate or LDH-NAD+-lactate and long-lived binary LDH-NAD(H) could result in rapid label exchange without dissociation of the LDH-NAD(H) complex and resulting in zero net flux. The two-way interconversion between the ¹³C-pyruvate and ¹³C-lactate pools would enable RF saturation of the ¹³C-lactate pool to affect the ¹³C-pyruvate pool, and therefore, the ¹³C-bicarbonate pool in the mitochondria. This exchange could account for the observed effect without any net flux between ¹³C-lactate and ¹³C-pyruvate. For example, if ¹³C-lactate with RF saturation is converted to saturated ¹³C-pyruvate at the same time as an unsaturated ¹³C-pyruvate is converted to ¹³C-lactate, no net flux has occurred, yet a saturated ¹³C-pyruvate is available for decarboxylation and reduced ¹³C-bicarbonate signal. This also confounds estimation of the percent lactate that is turned over per unit time.

However, it is important to note that decarboxylation of ¹³C-pyruvate into acetyl-CoA, producing ¹³C-bicarbonate, is a nonreversible reaction.²⁶ Each ¹³C label stays attached to the same carbon backbone during ¹³C-pyruvate to ¹³C-lactate interconversion, with the surrounding molecular structure changing from ¹³C-pyruvate to ¹³C-lactate, and then that same ¹³C label cleaved off to create ¹³C-bicarbonate. Thus, any effect from RF saturation of ¹³C-lactate on ¹³C-bicarbonate signal is due to particular ¹³C-lactate molecules that have been converted to ¹³C-pyruvate and then to ¹³C-bicarbonate. Therefore, even if the forward or reverse reactions of the ¹³C-labeled molecules described above are offset by equal and opposite pyruvate-to-lactate conversion (i.e., “exchange”), the conclusions from this experiment remain the same: that the ¹³C-lactate pool is “turning over” as it is consumed and replenished. Regardless of whether there is any net change in total (labeled and unlabelled) lactate and pyruvate pool size within the brain during this experiment, the observed change in ¹³C-bicarbonate signal must have resulted from a subset of ¹³C-pyruvate molecules that were previously ¹³C-lactate molecules at some point during the 1-min experiment, and were thus exposed to the frequency-selective RF saturation that was only at the [1-¹³C]lactate frequency.

When exploring the variation in this effect between brain regions, regional ¹³C-lactate signal correlated with ¹³C-bicarbonate percentage change (Figure 6A), consistent with higher ¹³C-lactate oxidation in regions with higher ¹³C-lactate signal. There was no significant correlation between these same regions and either ¹³C-bicarbonate or ¹³C-pyruvate signal (Figure 6B,C). In contrast, the regional ¹³C-lactate signal was inversely correlated to the change in pyruvate signal postsaturation (Figure 7A). Multiple recent studies in humans suggest that the generation of labeled lactate from HP pyruvate is rate limited by monocarboxylate transporter family (SLC16A) proteins. It is known that MCT1 (SLC16A1) and MCT2 (SLC16A7) are expressed on the blood–brain barrier with approximate K_ms of 1 mM and 0.1 mM for pyruvate, respectively.²⁷ Given the concentrations injected, this suggests that it is sufficient to drive the same state as it would in other organs. If the major mechanism that drove bicarbonate generation from labeled lactate was near equilibrium exchange in the same compartment, then saturating more lactate would result in a larger change in the pyruvate signal. These data suggest a two-compartment model, one in which the equilibrium constant (K_eq) for generation of lactate in one compartment heavily favors the LDH forward reaction to generate labeled lactate (K_eq ≫ 1) and a second compartment which converts this lactate to pyruvate and then bicarbonate which favors the reverse reaction (K_eq ≪ 1). While this preliminary study suggests these findings, further mechanistic studies are necessary to confirm this phenomenon and to what degree these compartments are within a single cell, or are in different cells, reflecting shuttling of lactate between cells.

There were study limitations that should be addressed. The coarse 15 mm isotropic spatial resolution caused partial volume effects that likely contributed to the variance observed, although such effects would not cause a systematic bias between the two conditions. Secondly, raw ¹³C signal values were used in the analysis without normalizing for any difference in polarization or substrate transit time between subsequent injections. However, any systematic effect of differing polarization between the two conditions was mitigated by varying the temporal ordering of the two scans. Furthermore no significant differences between polarization or time to injection were found between conditions. Lastly, this experiment provides evidence that ¹³C-lactate is converted back to ¹³C-pyruvate and metabolized oxidatively producing ¹³C-bicarbonate in the process, but it does not provide any direct evidence about the particular cellular (or intracellular) compartments involved. Adding diffusion weighting to future experiments may provide further evidence regarding compartmentalization and lactate transport.

5 |. CONCLUSIONS

This study provides evidence that oxidative metabolism of ¹³C-lactate is occurring in the human brain at a level that is readily observed with hyperpolarized ¹³C MRI. Injected ¹³C-pyruvate is converted to ¹³C-lactate for some time, and then converted back to ¹³C-pyruvate and then to acetyl-CoA resulting in ¹³C-bicarbonate. It remains to be shown whether multiple compartments are involved in the ¹³C-lactate-to-¹³C-pyruvate interconversion, and future experiments with diffusion weighting and other pulse-sequence manipulations may shed light on this question.

Supplementary Material

Supplementary Information

Table S1. The polorization recorded from the quality-control module on the SPINLab polarizer and the time between the dissolution and the injection (Δt-injection) for each of scan.