7T allows correct measurement of PDE in the human liver without decoupling
Lucian A. B. Purvis1, William T. Clarke1, Michael Pavlides1, Matthew D. Robson1, and Christopher T. Rodgers1

1OCMR, Department of Cardiovascular Medicine, University of Oxford, Oxford, United Kingdom


Phosphomonoesters (PME) and phosphodiesters (PDE) are emerging biomarkers in cirrhosis and fatty liver disease. In this study, phosphorus (31P) liver metabolite concentrations were quantified at 7T for 10 healthy volunteers using a 28 minute, 3D UTE-CSI sequence and an endogenous 2.5mM ATP reference. The inorganic phosphate concentration was 1.95 ± 0.18mM (1.25-2.43mM in the literature). Summing individual ester peaks, the total PME and PDE concentrations were 1.90 ± 0.29mM (cf. 1.98-3.89mM) and 4.31 ± 0.80mM (cf. 8.01-11.40mM). We attribute this apparent reduction in PDE to a broadening of the underlying phospholipid bilayer resonance and the increased spectral resolution at 7T.


Hepatic disease affects over 10% of people in Western populations1, 2. Phosphomonoester (PME) and phosphodiester (PDE) concentrations are emerging biomarkers indicating the presence of non-alcoholic fatty liver disease3 or liver cirrhosis4. The region of the liver phosphorus magnetic resonance spectroscopy (31P-MRS) spectrum containing PME and PDE has several other resonances which make accurate quantitation difficult. Increasing field strength from 3T to 7T improves spectral resolution for in vivo 31P-MRS5, which should allow for improved quantitation. As yet, only inorganic phosphate (Pi) concentrations have been reported at 7T6. In this study, we use our 16-element receive array coil7 to report PME and PDE concentrations from the livers of 10 healthy volunteers.


Each voxel can be corrected to yield an absolute concentration8:

$$[M] = \frac{S_MF_M\nu_R\eta_M[R]}{S_RF_R\nu_M\eta_R}$$

in which S is the signal per voxel, F is the saturation correction factor, ν is the filling factor, and $$$\eta$$$ is the sensitivity correction of the tissue or reference sample in the voxel. The subscripts stand for reference R and metabolite M.

When using an endogenous reference (i.e. concentration of adenosine triphosphate (ATP) = 2.5mM), $$$\eta_M = \eta_R$$$ and $$$\nu_M =\nu_R$$$ reducing Eq.1 to:

$$[M] = \frac{S_MF_M}{S_RF_R}\times2.5\tt{mM} $$


10 fasted, healthy volunteers (6M/4F, 26.7 ± 4.8 years, 72.4 ± 10.3 kg, recruited ethically) were scanned in a Magnetom 7T (Siemens) using a 16-channel 31P receive array coil (Rapid Biomedical, Germany)9.

A 3D UTE-CSI pulse sequence, with 1s TR, excitation bandwidth covering all metabolites from uridine diphosphoglycerate to phosphocholine (-12 to 10 ppm), acquisition weighting with 10 averages at k=0, and one B1-insensitive train to obliterate signal (BISTRO) saturation band to suppress overlying skeletal muscle10, was used to acquire a 16x16x8 matrix of spectra over a 27 x 24 x 20 cm3 field of view covering the liver in a total of 28 min.

Spectra from each receive element were combined with whitened singular value decomposition (WSVD) combination11. The combined spectra were analysed following our established protocol7; in brief: peaks were fitted with linewidth-constrained AMARES12, fiducial markers were used to localize each coil element, and saturation correction factors were computed using the Biot-Savart law and literature metabolite T1 values13. Concentrations were then calculated for this work using Eq. 2, assuming a 2.5mM concentration of ATP14.


A typical liver spectrum is shown in Figure 1. The individual PME and PDE peaks were quantified separately and then summed together for comparison with the literature. The concentrations for individual subjects are plotted in Figure 2. The average concentrations from our study are presented alongside comparators from the literature in Table 1 and Figure 3.


The assumption of a 2.5mM ATP concentration is reasonable in the fasted, healthy human liver14, 15, and is consistent with exogenous reference studies16-18 (see Fig. 3). PME is significantly smaller in this study than in several previous studies16, 18, 19 (at a p=0.05 level for Student’s t-test, see Fig 3). However, the difference for PDE is more significant in all cases (at a p=0.0001 level, 3.7-7.1mM difference).

At low field strength, it is difficult to distinguish the phosphoenolpyruvate (PEP) / phosphatidylcholine (PtdCh) peak from the PDE peak because the peaks strongly overlap5. At 3T it is only possible with proton decoupling20 (see Fig. 4). Another contributing factor is an underlying resonance in the PDE region due to endoplasmic reticulum. This broadens from 210Hz at 1.9T to 7000Hz at 7T and becomes indistinguishable from baseline noise21. In a direct comparison between 1.5T and 3T, there was a small reduction in the underlying broad peak and a small positive difference between PME/PDE ratios20. This effect is likely to have become still more prominent at 7T, as the underlying peak has broadened enough to be below the noise threshold for detection.

There are two possible sources of additional ATP, which would reduce the apparent PDE concentration: skeletal muscle and blood. The average phosphocreatine (PCr) contamination was 0.66 ± 0.12mM. Given the 4:1 ratio of PCr:ATP in skeletal muscle, the average ATP contamination is less than 0.17mM. Correcting for this would give a PDE concentration of 4.43mM. The 2,3-diphosphoglycerate (2,3-DPG) to ATP ratio is about 9:122 in blood, so any contamination would increase the 2,3-DPG signal, overlapping PDE, more than ATP and the apparent PDE concentration would be increased.


The resolution of PDE into separate peaks at 7T allows more accurate quantitation and presents an opportunity to determine which metabolites are the true biomarkers for hepatic diseases. We anticipate that this will improve the predictive accuracy of the method compared to lower field strengths.


Funded by a Sir Henry Dale Fellowship from the Royal Society and the Wellcome Trust [098436/Z/12/Z] and the Medical Research Council. We would like to thank Stefan Neubauer for helpful discussion.


1. Younossi ZM, et al. Clinical Gastroenterology and Hepatology. 2011 2. Blachier M, et al. J Hepatol. 2013 3. Abrigo JM, et al. J Hepatol. 2014 4. Dezortova M, et al. World J Gastroenterol. 2005 5. Chmelík M, et al. Eur Radiol. 2015 6. Valkovic L, et al. Eur Radiol. 2014 7. Purvis LAB, et al. ISMRM, 2015 8. Bottomley PA. Encyclopaedia of Magnetic Resonance. 2009. 9. Rodgers CT, et al. ISMRM. 2014 10. Luo Y, et al. Magn Reson Med. 2001 11. Rodgers CT, Robson MD. Magn Reson Med. 2010 12. Purvis LAB, et al. ISMRM, 2014 13. Chmelik M, et al. NMR Biomed. 2014 14. Sijens PE, et al. Magn Reson Imaging. 1998 15. Hultman E, et al. Scand J Clin Lab Invest. 1975 16. Chmelik M, et al. Magn Reson Med. 2008 17. Laufs A, et al. Magn Reson Med. 2014 18. Buchli R, et al. Magn Reson Med. 1994 19. Menon DK, et al. Hepatology. 1995 20. Wylezinska M, et al. NMR Biomed. 2011 21. Murphy EJ, et al. Biochim Biophys Acta. 1992 22. Neubauer S, et al. Circulation. 1992


Figure 1: A typical liver 31P-MRS spectrum acquired at 7T from the right lobe of the liver using our 3D UTE-CSI protocol. There is no PCr in the healthy human liver, so any signal near 0ppm is contamination from skeletal muscle.

Figure 2: Concentrations of 31P metabolites. Individual subjects are given in colour. The mean ± standard deviation is given in black.

Figure 3: Comparison of liver 31P metabolite concentrations from this study against the literature. Standard deviations are shown by the error bars on each bar. Stars indicate level of significance of the difference from this study: * p=0.05, ** p=0.01, *** p=0.001.

Figure 4: Typical liver spectra at 1.5 and 3T with and without proton decoupling and NOE. From Wylezinska et al.17

Table 1: Concentrations of 31P liver metabolites from various studies.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)