Introduction

Hyperpolarized (HP) [1-¹³C]pyruvate is the most widely utilized molecular contrast agent for MR metabolic imaging to characterize abnormal metabolism in patients with tumors, based on the measured glycolytic metabolism of [1-¹³C]pyruvate-to-[1-¹³C]lactate conversion upregulated in tumors.¹ The application of [1- ¹³C]pyruvate, however, is limited for assessing the flux of pyruvate to TCA cycle only through pyruvate dehydrogenase (PDH) because the ¹³C label is lost as CO₂ during the conversion to Acetyl-CoA. Alternatively, labeling the C2 position of pyruvate provides access to the TCA cycle as the labeled carbon is carried over to acetyl-CoA and eventually to the glutamate pool, but is more difficult to image because of a shorter T₁, a less sparse spectrum, and peak-splitting due to J_CH coupling. The safety and feasibility of using HP [2-¹³C]pyruvate for measuring its conversion to [2-¹³C]lactate and [5-¹³C]glutamate in the brain of healthy volunteers was previously demonstrated with non-localized dynamic ¹³C MR spectroscopy.² In this work, multi-slice, dynamic HP [2-¹³C]pyruvate echo-planar imaging (EPI) was developed and applied in healthy volunteer studies with whole brain coverage to investigate its application for human brain studies using HP [2-¹³C]pyruvate.

Methods

Following phantom testing, HP [2-¹³C]pyruvate metabolic data of the brain were acquired from healthy volunteers (N = 5) using a 2D multi-slice EPI sequence,³ following the injection of 241 ± 7 mM [2- ¹³C]pyruvate hyperpolarized for ~ 3hrs in a 5T GE SPINlab at 0.8 K. The pulse sequence consisted of a single-band spectral-spatial RF pulse that was sequentially applied to the resonance frequencies of [2- ¹³C]pyruvate, [5-¹³C]glutamate, and downfield and upfield peaks of [2-¹³C]lactate doublet with 20°, 60°, 60, and 60° flip angles, respectively. The images of downfield and upfield lactate signals were acquired separately to avoid artifacts due to J-coupling.⁴ A multi-resolution approach was applied, yielding an in- plane spatial resolution of 7.5 x 7.5 mm² for pyruvate and 22.5 x 22.5 mm² for glutamate and lactate, for 5 slices with 3 cm thickness.⁵ Twenty timeframes were acquired per metabolite with a 3 s temporal resolution,yielding a total scan time of 60 s. All experiments were performed on a 3T clinical MR scanner using a 24-channel receiver (Rapid Biomedical, Germany) in combination with birdcage RF transmit coils. The acquired ¹³C data were denoised using a patch-based higher order singular value decomposition method.⁶ Voxel-wise calculations of k_PG (apparent pyruvate-to-glutamate conversion rate) and k_PL (apparent pyruvate-to-lactate conversion rate) were performed with two-site exchange models.⁷

Results and Discussion

Figure 1 shows representative dynamic images of [2-¹³C]pyruvate, [5-¹³C]glutamate, and the upfield and downfield [2-¹³C]lactate signals from the brain of a healthy volunteer. The maximum apparent signal-to- noise ratio (SNR) in this dataset were 844, 32, 97, and 72 respectively for pyruvate, glutamate, and the downfield and upfield peak of lactate. The high-resolution pyruvate data acquired at 7.5 x 7.5 mm² show strong signals from the arteries and veins, which were well separated from surrounding tissues. For glutamate and lactate, coarser resolution of 22.5 x 22.5 mm² achieved high SNR to allow confidence in signal quantification. The lactate distribution was similar between the images of downfield and upfield signals, and no artifacts arising from J-coupling were observed.
In Figure 2, signal intensities of downfield and upfield lactate images were compared on a voxelwise basis for each slice. Overall, it showed good agreement between the two resonances, which demonstrated the feasibility of acquiring two peaks of lactate doublet separately. Thus, the total lactate signals can be obtained by combining the downfield and upfield lactate images without any correction, improving the lacate SNR for quantification.
Figure 3 displays representative ¹³C sum-over-time (AUC) images of pyruvate, glutamate, and lactate from another volunteer. The lactate AUC images represent the sum of the downfield and upfield lactate signal intensities. The metabolite ratios of [5-13C]glutamate to [2-¹³C]pyruvate ([5-¹³C]Glu:[2-¹³C]Pyr) and [5- ¹³C]glutamate to Σ[2-¹³C]lactate ([5-¹³C]Glu: Σ[2-¹³C]Lac) for the whole brain were calculated from the AUC images. The averaged values for [5-¹³C]Glu:[2-¹³C]Pyr and [5-¹³C]Glu: Σ[2-¹³C]Lac from 5 datasets were found to be 0.045 ± 0.010 and 0.35 ± 0.048, respectively. Also, the average k_PG and k_PL values were determined as 0.003 ± 0.001 and 0.012 ± 0.003 s^-1, respectively. Figure 4 summarizes the measured metabolic ratios, k_PL, and k_PB values obtained from each dataset. ROI-based quantification of metabolic conversion and the assessment of its accuracy is currently being investigated. The capability of assessing both pyruvate-to-lactate and pyruvate-to-glutamate conversion with spatial resolution indicates a potential for using HP [2-¹³C]pyruvate to investigate cerebral energy metabolism and characterize brain tumors with isocitrate dehydrogenase (IDH) mutations.⁸

Conclusion

This work demonstrates the feasibility of volumetric and dynamic EPI of HP [2-¹³C]pyruvate metabolism to [5-¹³C]glutamate and to [2-¹³C]lactate in the human brain probing glycolytic and oxidative metabolism simultaneously. The strategy of separately exciting the two peaks of the lactate doublet was shown feasible for imaging metabolites with signal splitting. In combination with a variable resolution approach, the hyperpolarized metabolite signals were utilized for quantifying k_PL and k_PB throughout the brain of healthy volunteers providing new measures of cerebral energy metabolism.

Acknowledgements

This work was supported by NIH grants P41EB0135898, P01CA118816, U01EB026412, and the UCSF NICO project.