To develop an efficient method for the hyperpolarization (HP) of ¹³C-bicarbonate with the added benefit of low toxicity for clinical imaging of extracellular pH (pH_e).

Inspired by the observance of short-lived glycerol-bicarbonate adducts in ¹³C MR spectra of HP sodium ¹³C-bicarbonate in glycerol, we conceived of a new method of generating nontoxic, high-signal dissolutions of HP bicarbonate:

1. Hyperpolarization of a carbonated precursor molecule

2. Base-catalysed decarbonation after dissolution to form HP ¹³C-bicarbonate and biocompatible by-products

We evaluated three precursor molecules compatible with this technique: 1,2-glycerol carbonate (GLC), diethyl pyrocarbonate, and 2,3-O-carbonyl-α-D-mannopyranose; these were carbonated analogs of glycerol, ethanol, and D-mannose, respectively (Figure 1). Each non-¹³C-enriched compound was formulated with 15 mM trityl radical and formed a glass at cryogenic temperatures. 457 μmol of each were polarized for 1 hour and then dissolved with 3.5 mL of 270 mM NaOH, heated for 5 s, and neutralized with HCl before HP ¹³C spectra were acquired at 11.7 T (n = 3 each). ¹³C-enriched GLC was also polarized, and ¹³C NMR spectra were acquired under dissolution conditions of breakdown (n = 3) or no breakdown (n = 4). The maximum solution-state polarization was determined by referencing to an NMR spectrum acquired at thermal equilibrium.

HP pH_e imaging was performed in both a phantom and in a transgenic adenocarcinoma of the mouse prostate (TRAMP) mouse model using a 14 T small-animal imaging system:

Phantom: HP dissolution was added to three tubes of phosphate buffer at pH values between 6.4 and 7.6 along with carbonic anhydrase to facilitate pH equilibration. Each pH was measured with a conventional pH electrode after imaging.

Mouse: 350 μL of HP dissolution were injected via tail-vein catheter over 12 seconds.

In both experiments, 2D chemical shift images were acquired immediately after (8 x 8 x 256 matrix, 8 mm slice, 4 mm in-plane resolution, 8013 Hz spectral window, 2 s imaging time). The excitation pulse was a sum of two phase-modulated Gaussian pulses and excited bicarbonate and CO₂ slices simultaneously within the coil center, with a different flip angle on each resonance (25º on CO₂, 2.78º on bicarbonate – see Figure 2). The pH per voxel was calculated using the ratio of the bicarbonate and CO₂ peak integrals and a modified Henderson-Hasselbalch equation (1). The pK_a was taken as 6.17 (1-4), and the ratio was flip-angle corrected assuming complete bicarbonate-CO₂ magnetization exchange between excitations. Voxels with bicarbonate or CO₂ SNR < 3 were eliminated from the analysis.

We formulated 11.9 M GLC for HP and obtained 92.7 ± 1.4% of the total HP signal as bicarbonate and CO₂ (Figure 3). The back-calculated solution-state polarizations with and without decarbonation were 16.4 ± 1.3% and 18.1 ± 2.4%, respectively; the difference was not statistically significant (p > 0.25). GLC has a T₁ relaxation time constant of 63.7 ± 5.7 s. The bicarbonate and CO₂ concentration produced was 75.1 ± 1.4 mM (n = 2), which represents a ~30% concentration loss. The phantom spectral and electrode pH values agreed within 0.1 pH unit (Figure 4). The average pH_e inside and outside the tumor, averaged across all voxels, were 7.15 and 7.36, respectively. The lowest pH_e measured in the tumor was 6.95 (Figure 5).The HP ¹³C-bicarbonate polarization and concentration were comparable to those first reported, but the process eliminated the need for Cs⁺ ion removal from the dissolution (1). While there appears to be little polarization loss during decarbonation, the concentration decreased, presumably due to CO₂ boil-off during neutralization with HCl. Performing these steps within an enclosed space may reduce concentration losses, increasing bicarbonate concentrations up to 110 mM. Catalysts for the decarbonation reaction may exist which can reduce the time of breakdown (currently 35-40 s). The phantom pH measurements were accurate to a similar degree as reported previously (1). Although the in vivo pH_e is lower in tumor tissue than in the surroundings, the tumor pH_e values may be overestimated due to incomplete ¹³C label exchange between bicarbonate and CO₂ prior to imaging (5).We have demonstrated the utility of a new hyperpolarization technique that provides large gains in HP bicarbonate signal without compromising dissolution toxicity. This technique is compatible with accurate pH_e mapping via MR spectroscopic imaging, both in vivo and ex vivo. Future work will reduce HP signal losses during decarbonation and improve pH imaging accuracy and resolution.