Introduction

Hyperpolarized ¹³C MRI is a metabolic imaging approach that can form images of contrast agent (i.e. [1-¹³C]pyruvate) and its downstream metabolites, [1-¹³C]lactate and ¹³C-bicarbonate¹. Signals of [1-¹³C]lactate or ¹³C-bicarbonate have been generally considered as a marker for glycolytic or oxidative activities, respectively. However, the astrocyte-neuron lactate shuttle model proposes that lactate can be transferred and used as a fuel to supplement oxidative phosphorylation in neurons -- and generate bicarbonate in the process². The fraction of [1-¹³C]lactate that contributes to the ¹³C-bicarbonate signal has not been investigated.

In this study, [1-¹³C]lactate signal was selectively saturated in brain regions with 2 different slab thicknesses and positions and compared to the condition without [1-¹³C]lactate saturation to measure the effect of [1-¹³C]lactate saturation on ¹³C-bicarbonate signals in human brains. Acquired [1-¹³C]lactate and ¹³C-bicarbonate images were analyzed by converting the ¹³C-signals into z-scores via an atlas-based approach previously described³, which segments the brains into the 56 structural regions included in the LPBA40 atlas.

Methods

Image Acquisition Written informed consent was obtained from 2 control subjects. Following the protocol described in Lee et al.³, ¹H multi-echo reference data and ¹³C-signals were acquired following the injection of HP [1-¹³C]pyruvate (0.1mmol/kg dose at 4mL/s followed by a 25mL normal saline flush at 5mL/s) in a GE MR750 3.0T MRI scanner (GE Healthcare, Waukesha, WI). For this study, 2 subjects were scanned on different days using one of the ¹³C-data acquisition protocols described below.

Non-saturation: Using the single-tuned home-built ¹³C birdcage, ¹³C-signal was acquired with the same sequence and protocol previously described³ using spectral-spatial excitation⁴ of lactate/bicarbonate/pyruvate resonances (sequentially) followed by a 3D dual-echo EPI (DE-EPI) readout⁵ (5s TR; axial; 1.5cm isotropic resolution with FOV 24x24x36cm³, total of 60s acquisition).

[1-¹³C]lactate saturation: A modified version of the sequence (Fig. 1) was applied to selectively saturate [1-¹³C]lactate signals prior to each volume acquisition (applied once every 1.25 seconds). An axial saturation slab was used with thickness 9 cm for subject 1 and 4 cm for subject 2.

Following the ¹³C-data acquisition, a standard 8-channel ¹H neurovascular array (Invivo Inc.) was used to acquire fast spoiled GRE (FOV 25.6×25.6 cm², 1 mm-isotropic resolution, TR 7.6 ms, TE 2.9 ms, flip angle 11°) images.

Image Reconstruction ¹³C images were reconstructed offline using MATLAB R2013a (The MathWorks Inc., MA, Natick, Massachusetts). Geometric distortion artefacts were corrected via data-driven approach⁵. The final ¹³C images were generated by summing odd/even echoes and summing individual time-resolved images over full 60s acquisition window to improve SNR, which then were resampled to match the coordinates and matrix size of T1-weighted images (Fig. 2).

Image Analysis Regional segmentation of ¹³C-signals were conducted in the LONI pipeline processing environment^3,6. To compare saturation to the non-saturated condition in the same participant, the average ¹³C-signals from each of 56 structural regions embedded in the LPBA40 atlas were converted into z-scores³ (the number of standard deviations from the mean of all regions within a single subject).

Results and Discussion

Sagittal images of lactate, bicarbonate, and pyruvate from non-saturation and both saturation subjects are shown in Fig. 2. Regions that lie partially or fully within the [1-¹³C]lactate saturation band (i.e. regions above cerebellum for subject 1 and occipital or temporal lobe regions for subject 2) showed lower regional z-scores compared to the z-scores from the same region in the non-saturation subject (Fig. 3 & 4).

As expected, lactate z-scores showed poor agreement to non-saturation subject (Kendall’s coefficient of concordance (W) = 0.56 and 0.70 for subject 1 and 2, respectively), as expected. On the contrary, bicarbonate z-scores with saturation were in good agreement to non-saturation subject (W = 0.90 for both), showing that the saturation of [1-¹³C]lactate does not significantly affect ¹³C-bicarbonate signal. This suggests that the infused [1-¹³C]pyruvate is directly converted to ¹³C-bicarbonate in the brain, and not shuttled through the [1-¹³C]lactate pool that is visible in this experiment.

Conclusion

Hyperpolarized [1-¹³C]lactate and ¹³C-bicarbonate images were acquired with and without spectral-spatial lactate saturation in human brains. Lactate z-scores were changed in the saturation regions, as expected. The saturation of [1-¹³C]lactate signals did not significantly affect ¹³C-bicarbonate signals, suggesting that infused [1-¹³C]pyruvate is directly converted to ¹³C-bicarbonate in the brain.

Acknowledgements

The authors thank Julie Green and Sumeet Sachdeva for coordinating the study, and Ruby Endre and Garry Detzler for MR technical support. Funding support from the Brain Canada and Canadian Institutes for Health Research.