Those interested in ³¹P MRS and energy metabolism.We use localized progressive saturation transfer experiments in ³¹P MRS to probe ATP synthesis rate in the rat brain et 11.7 T. Intra and extracellular inorganic phosphate (Pi) can be resolved due to the difference in pH between both compartments¹. Here we describe how not accounting for extracellular Pi can lead to significant bias in forward rate constant estimation.

In vivo ³¹P MRS: Two rats anesthetized with 1.5-2% isoflurane were used for ³¹P saturation transfer MRS². All measurements were performed using an 11.7 T Bruker BioSpec using a 1.5-cm double-tuned surface coil (Rapid Biomed GmbH®). Localisation of a large 11×8×12 mm (Fig. 1A) voxel was achieved using an adiabatic ISIS module (TR=8 sec), preceded by selective BISTRO progressive saturation of the γ-ATP peak².

We quantified Pi peak attenuation relative to an unsaturated spectrum rather than symmetric saturation, because signal loss due to RF bleed-over was minimal (~5%) during selective symmetric saturation, as assessed on preliminary experiments. Seven different saturation times (t_sat=0.55, 0.825, 1.1, 1.65, 2.2, 3.3 and 4.4 s) were used for γ-ATP saturation, as well as an unsaturated spectra. Each spectrum was acquired in 17 min, and all measurements were repeated twice for each animal.

Data Analysis: Spectra were analysed with LCModel using a simulated basis-set. Pi signal was fitted as a non-linear function of saturation time according to modified Bloch equations accounting for chemical exchange².The standard deviation for k_f-ATPase in each situation was derived from Monte Carlo simulations over 500 repetitions.

It can be seen on Fig. 1B that the Pi signal is actually composed of two peaks, one larger at 4.9 ppm (corresponding to pH=7.04), and one smaller at 5.3 ppm (corresponding to pH=7.41), representing ~20% of the larger peak. These peaks can be reasonably ascribed to a large intracellular Pi pool at lower pH, and a small extracellular Pi pool at higher pH, as already suggested in muscle³. Incidentally, the extracellular Pi signal relative to total Pi signal is similar to the extracellular volume fraction (~15-20%), suggesting similar Pi concentrations in the extra and intracellular spaces.

The saturation transfer from γ-ATP to Pi due to ATPsynthase is expected to affect only intracellular Pi, not extracellular Pi. Indeed, signal decrease was observed at 4.9 ppm when increasing selective saturation was applied on the γ-ATP resonance, while the signal at 5.3 ppm remained fairly stable within experimental error (See Fig. 2A and 2B), providing further evidence for the assignment of this peak to an extracellular compartment (e.g. rather than an intra-mitochondrial pool⁴, which should be affected by saturation). Using modified Bloch equations to fit the data (Fig. 2B) we were able to associate signal decrease at 4.9 ppm to a reaction rate of k_f-ATPase=0.30±0.03 s^-1. When fitting for the sum of both phosphate pools, as would be done in an experiment with lower spectral resolution, the forward rate constant decreases to k_f-ATPase=0.25±0.02 s^-1, closer to values reported in the literature using the same technique at lower field^5,6.

Our results strongly suggest that little to none of the extracellular Pi is in exchange with γ-ATP or intracellular Pi at the time scale of the saturation time. Consequently, when not considering this “inactive” extracellular phosphate, a bias up to ~20% towards lower values can be introduced in the measurement of k_f-ATPase. Past works have interpreted the very good agreement between measured ATP synthesis rates and glucose consumption rates as an evidence for the absence of contribution of reversible ATP synthesis by GADPH/PGK at the glycolytic level in the brain^2,7 as opposed to the muscle. The fact that ATP synthesis rates measured by ³¹P may actually be higher than previously reported when considering the total Pi pool, and thus higher than the net ATP synthesis that can be theoretically expected from complete glucose oxidation, suggests that a small contribution of GADPH/PGK to ³¹P measurement may actually exist in the brain. Further measurements, in particular comparison with glucose consumption and oxidation rates measured in the same voxel, should help to better address this question.

We demonstrate that, at high spectral resolution, we are able to resolve the pools of inorganic phosphate in both intra and extracellular space in the brain. Furthermore, the intracellular pool of Pi is the only one affected by γ-ATP selective saturation, suggesting minimal Pi exchange between extra and intracellular space, and pointing out towards possible bias in the measurement of ATP synthesis rate when extracellular Pi is not separated from intracellular Pi.