Magnetic resonance techniques

Nuclear magnetic resonance (NMR) has been widely applied to metabolic studies in vivo. It relies on the magnetic properties of nuclei which have a net spin by virtue of the spins of their constituent protons and neutrons (Morris et al., 1994).

NMR involves the interaction of nuclear spins with a static magnetic field and has the unique ability to differentiate between different metabolites due to

‘chemical shift’. The resonance frequency of the signal produced is distinct for nuclei in different chemical environments, enabling individual structures to be distinguished. The strength of the NMR signal (the area under a peak in the NMR spectrum) is proportional to the number of spins contributing to it (Morris et al., 1994). The wide chemical shift range (>200 ppm) of the ¹³C nucleus permits metabolites to be identified directly and unambiguously from their ¹³C NMR spectra (Morris et al., 1994). The low natural abundance (1.1%) of the stable isotope ¹³C means that only metabolites such as glycogen, that is highly concentrated in tissues such as liver and muscle, can be directly observed at

natural abundance in vivo (Alger et al., 1981). Despite the ubiquitous presence of the proton, the development of high-field instrumentation and water suppression techniques has permitted the widespread use of ¹H NMR in the study of lipid metabolism (Morris et al., 1994).

MRS was performed on a Philips 3T system at the Sir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, University Campus. The MRS scans were performed by the following collaborators: Miss Mary Stephenson (PhD student, School of Physics and Astronomy), Miss Elisa Placidi (PhD student, School of Physics and Astronomy) and Dr. Luca Marciani (Senior RCUK Academic Fellow, Nottingham Digestive Diseases Centre NIHR Biomedical Research Unit). A transmit/receive body coil was used for ¹H imaging and MRS for measurements of liver volume and hepatic and skeletal muscle (soleus) lipid concentrations, respectively. A 14 cm ¹³C surface probe with quadrature proton decouple coils was used for measurement of hepatic and calf (gastrocnemius) muscle glycogen. We studied the gastrocnemius muscle, in line with previous investigators (Carey et al., 2003), as the positioning and aligning of the lower leg and calf on the ¹³C coil was easier and more reproducible than placing the coil on top of the thigh quadriceps muscle. Liver volumes were measured using T1-weighted, breath-hold turbo field echo scan with resolution 2×2×7 mm³, 36 slices, matrix 180×182, TR = 3.11ms, total scan time 14.4s. The images were analysed by drawing regions of interest in Analyze6 (Biomedical Imaging Resources, Rochester, MN) and values are reported as volume (litres) and % change from baseline. The coefficient of variation (CV) for repeated

measurement of liver volume was 0.8%. Hepatic ¹H spectra were acquired from a 30×30×30 mm voxel positioned in the right lobe using a respiratory triggered PRESS sequence and the following parameters: echo/repetition time 40/5000 ms, 16 averages, bandwidth 2000 Hz, 1024 samples. Two ¹H spectra were acquired in muscle: water-suppressed for intramyocellular lipid (IMCL, 32 averages) and non water-suppressed for total calf lipid (16 averages). PRESS localization was used with echo/repetition time 40/7000 ms, voxel 30x30x50 mm, bandwidth 2000 Hz, 1024 samples. ¹H non water-suppressed spectra were post-processed using jMRUI and peak areas were calculated using in-house software built in Matlab®. Water suppressed spectra were analysed using the AMARES algorithm (Vanhamme et al., 1997) in jMRUI, fitting to Gaussian lineshapes. Lipid values are reported as % change from baseline. The CV for repeated measurements of liver lipid, IMCL and extramyocellular lipid (EMCL) concentrations were 4%, 6% and 21%, respectively. The EMCL CV was higher than that for IMCL as the content of extramyocellular lipid is more variable and more dependent on the positioning of the voxel. ¹³C spectra (Figure 4.2) were acquired using a proton-decoupled pulse acquire sequence with bandwidth 7000Hz, 512 samples, ¹³C adiabatic pulses and narrowband decoupling (3 spectra with repetition time 2150 ms, 288 averages, total duration 30 min for the liver and 2 spectra with repetition time 1300 ms, 336 averages, duration 15 min for the calf). jMRUI was used for post-processing and peak areas were determined using in-house software built in Matlab®. Spectral peaks were selected using the AMARES algorithm and were fitted to Lorentzian lineshapes. The integral of the

glycogen peak was expressed as a fraction of the formate peak derived from a phantom containing formate placed at the centre of the ¹³C coil. Quantification of glycogen concentrations was performed using a phantom replacement method (Leverton, 2009). In brief, a cylindrical phantom containing a known glycogen concentration was placed at various distances from the coil and images and spectra were acquired for each position. The images were used to identify pixels corresponding to the phantom and those corresponding to the marker at the centre of the coil. These were then used to find a relationship between the distance between glycogen containing pixels, the coil and the size of the resulting glycogen signal. The same was performed for the in vivo images thus producing scaling factors for the subjects and phantoms which were used for quantification. The subject glycogen concentrations were calculated using the following equation:

[Subject] = the glycogen concentration in the subject, [Phantom] = the glycogen concentration in the phantom,

Sglygogen = the integral of the NMR peak from the subject due to glycogen, Sformate = the integral of the NMR peak from the subject due to formate marker, Pglycogen = the integral of the NMR peak from the phantom due to glycogen,

Pformate = the integral of the NMR peak from the phantom due to formate marker, Scalesubject = scaling factor for the subject,

Scalephantom = scaling factor for the phantom.

Figure 4.2: Example of the dynamic changes in ¹³C (used to calculate glycogen concentrations), during the course of the study, as determined by magnetic resonance spectroscopy. ‘Baseline’ scan was performed 4 hours after the standardised mixed-meal. The subsequent scans were performed after 12 and 24 hours of fasting (decreased area under ¹³C peak) and 2 hours following ingestion of the carbohydrate-based study drink (ONS).

Liver glycogen is reported in mmol and as % change from baseline. Muscle glycogen is reported as % change from baseline. The CV for repeated measurements of liver and muscle glycogen were 14% and 11% respectively.

Magnetic resonance spectra were blinded prior to analysis to avoid operator bias.