The human body stores large quantities of triglycerides within adipose tissue as fat for internal cushioning or as metabolically active intraabdominal and subcutaneous fat (1, 2) that serves as energy depots. Nonadipose tissues such as muscle, liver, and kidney have only limited capacity for lipid storage, but they still contain triglycerides dispersed as small lipid droplets in cytosol to provide substrate for cellular metabolism and membrane synthesis (3). For example, skeletal muscles use intramyocellular (IMCL) fat as a supplementary substrate to glycogen during prolonged exercise (4). Excessive accumulation of triglycerides within nonadipose cells may have toxic effects and could cause pathological changes in metabolic pathways, leading to altered cell function or apoptosis (5–9). A variety of techniques have been utilized to demonstrate that intramuscular lipid stores are increased in the skeletal muscle of diabetics and correlate inversely with insulin-mediated glucose disposal in nondiabetic individuals (3, 11–14). As opposed to other techniques, magnetic resonance spectroscopy (MRS) permits noninvasive differentiation of the triglyceride stored in adipocytes from that accumulated in myocytes because resonances from fatty acids within these two cell types have different chemical shifts (10). This unique property makes MRS the only currently available method capable of quantitatively measuring lipids accumulated within nonadipocytes in vivo. Despite the potential for broad application of MRS for measuring lipids in vivo, only a single study has discussed the chemical shift differences between fat resonances from different compartments in skeletal muscle (4).
The purpose of the present study is to demonstrate directly that bulk magnetic susceptibility (BMS) is responsible for the discrimination between intra- and extramyocellular lipid (EMCL) resonances. Fat phantoms imitating the compartmentation of lipids in skeletal muscle were constructed, and the theory of Chu et al. (15), simplified to the case of fat resonances from two compartments, was adopted to interpret the experiment.
MATERIALS AND METHODS
Methods and Samples
A simple model of lipids contained within skeletal muscle was created employing two generally accepted assumptions: 1) lipids enclosed within myocytes are found in tiny droplets dispersed in the cytoplasm (3), and 2) adipocytes densely filled with triglycerides are distributed as long, approximately cylindrical clusters between fibers in skeletal muscle (4). Intralipid™ (Frenius-Kabi, Hamburg, Germany), an intravenous fat emulsion of fine droplets (no larger than 0.5 μ in diameter) of soybean oil, was chosen to represent IMCL (16). Soybean oil (Sigma-Aldrich Co., St. Louis, MO) represented EMCL. Specific phantoms were prepared for experiments in both vertical and horizontal magnets. A total of four samples were studied: three in a vertical magnet and one in a horizontal magnet. Sample 1 contained exclusively Intralipid™ in a 5-mm standard NMR tube. Sample 2 contained exclusively soybean oil in a 5-mm NMR tube. Sample 3 was prepared by placing a 1.5-mm capillary tube filled with soybean oil inside a 5-mm NMR tube filled with Intralipid™. For experiments in the horizontal magnet, sample 4 was constructed in the following manner. A 3-mm NMR tube was filled with soybean oil and sealed. The tube was then concentrically placed in a 15-mm-diameter cylindrical container filled with Intralipid™ and tightly sealed. This phantom was placed in an even larger cylindrical container filled with deuterated saline. It is essential to note that although the Intralipid™ emulsion in our experiments was accommodated within a macroscopic container of cylindrical shape, the oil in that compartment was distributed within spherical vesicles (16), mimicking the lipid droplets found in the cytosol of myocytes.
Initial high-resolution experiments were performed in a vertical magnet (300 MHz, 7 Tesla) interfaced with a 300 MHz Varian Inova spectrometer with samples positioned along the external magnetic field, B0. Individual proton spectra of soybean oil, Intralipid™ and bicompartmental fat phantoms were collected using a single pulse sequence with interpulse delay TR = 5 s, spectral width = 3 kHz, and 2048 data points.
In experiments designed to demonstrate the importance of the orientation relative to the field direction, sample 4 was rotated in the horizontal magnet (4.7 Tesla) from a parallel to a perpendicular orientation relative to the external magnetic field B0 in increments of 15°. Spectra were collected at each position using spectral parameters identical to those of the previous vertical-magnet experiments.
Pure soybean oil contains a mixture of saturated and mono- and polyunsaturated triglycerides, with a typical spectrum consisting of 10 resonances originating from protons in the glycerol backbone and the different elements of fatty acyl chains as presented in Fig. 1. The most prominent peak (labeled B) represents a signal from the methylene protons of the fatty acid chains, and its position will be discussed throughout this work. The results of the experiment in the vertical magnet are summarized in Fig. 2. The upper trace represents the spectrum from Intralipid™. When the resonance of water is standardized at 4.8 ppm, the signal from the methlyene protons of Intralipid™ appears at 1.4 ppm (Fig. 2a). The middle trace illustrates that the resonance lines for the components of soybean oil are shifted by +0.2 ppm in respect to Intralipid™ (Fig. 2b). The frequency shift is even more readily apparent when a capillary tube filled with soybean oil is placed within an NMR tube filled with Intralipid™ (Fig. 2c). The resulting spectrum resembles spectra from skeletal muscle obtained in previous in vivo studies (11–14) and demonstrates the validity of our model of IMCL and EMCL.
Another experiment was designed to assess whether the chemical shift changes in a predictable manner as the phantom is rotated from a parallel (0°) to a perpendicular (90°) orientation relative to the external magnetic field. These results are shown in Fig. 3. The resonance from Intralipid™ appears at the same frequency (1.4 ppm) regardless of sample orientation in the magnet, while the resonance from soybean oil shifts between 1.63 and 1.31 ppm, and crosses with the resonance of Intralipid™ at the orientation of approximately 55°. These experimental results were tested against the theoretically predicted chemical shifts induced by BMS (15).
Calculation of Chemical Shift Induced by BMS in Spherical and Cylindrical Compartments
Whenever matter is placed in a magnetic field, electromagnetic interactions “disperse” or “concentrate” magnetic field lines. Magnetic susceptibility of the bulk medium influences individual magnetic resonance frequencies so that resonances of a sample experience an extra frequency shift called the BMS shift. That shift in resonance frequency (δχ) is determined by magnetic induction (B) within specific compartment and external magnetic induction (B0), (δχ = ((B – B0)/B0) × 106). For highly symmetrical compartments such as spheres and cylinders, magnetic induction is simplified to B = (1 + D)B0, and factors amending the element D define the shift introduced to resonances by BMS. Element D is modulated by the volume magnetic susceptibility of a medium inside (χi) and outside (χe) of a compartment, as well as by the shape and orientation of the compartment wall relative to external magnetic induction. In general, gradients of magnetic induction (ΔB) are set up near the compartment walls due to differences of magnetic induction inside (Bi) and outside (Be) of the wall (ΔB = Bi – Be).
Analytical expressions for magnetic induction gradients (ΔB) in parallel and perpendicular orientations of a compartment wall relative to B0, for spherical and cylindrical symmetries were derived by Chu et al. (15). The adaptation of these formulas to our needs with the assumption that Intralipid™ and soybean oil retained spherical and cylindrical macroscopic symmetries, respectively, is listed below.
According to Eq. , BMS does not introduce changes to the chemical shift of Intralipid™ emulsion. Equations  and  demonstrate that BMS chemical shift of soybean oil is anisotropic, different for the parallel and perpendicular orientation of the sample in the magnet. BMS shift for soybean oil is defined by susceptibility of the media inside (χf) and outside (χw) of the compartment wall (susceptibility of oil and water, respectively). Using standard values for the magnetic susceptibility of protons in fat (χf = –8.44) and water (χw = –9.05) (17), and Eqs.  and , BMS shifts of soybean resonance in parallel and perpendicular orientations were calculated.
These calculations predict that the resonance of soybean oil should be shifted +0.2 ppm in respect to the resonance of the Intralipid™ when our phantom is parallel to the magnetic field. In perpendicular orientation, the resonance frequency of soybean oil resonance will shift –0.1 ppm relative to that of Intralipid™. These calculations precisely agree with our experimental results and are the basis for the conclusion that BMS effects are responsible for the shift of the resonance of soybean oil from the resonance of Intralipid™. As shown in Fig. 3, the position of Intralipid™ resonance is independent of the orientation of the sample relative to the external magnetic field, and appears at 1.4 ppm relative to water referenced at 4.8 ppm. Resonance of soybean oil shifts with the sample orientation. Methylene protons of soybean oil resonate at 1.6 ppm (0.2 ppm relative to Intralipid™) when the phantom is parallel to the external magnetic field and at 1.3 (–0.1 ppm relative to Intralipid™) when the phantom reaches perpendicular orientation.
DISCUSSION AND CONCLUSIONS
Assuming that Intralipid™ and soybean oil represent lipid contained within myocytes (IMCL) and adipocytes (EMCL), respectively, the results for our experimental model can be compared to those obtained in vivo. As first observed by Schick et al. (10) and confirmed by many others (4, 11–14), lipids in skeletal muscle from the lower extremities exhibit two sets of resonances separated by 0.2 ppm. It has been suggested (4, 10) that the resolution between resonance lines from lipids in an IMCL and EMCL compartment is determined by BMS. This hypothesis was primarily tested by rotation of a single subject's leg within a clinical magnet (4). The shift of resonances with rotation was approximately demonstrated although the spectral resolution for orientations of the human limb not parallel to the magnetic field was limited, probably due to spatial restrictions. Rotating a bicompartmental phantom in the magnet provides a more direct demonstration and is free of spatial limitations. Therefore, inspired by the initial experiment of Boesch et al. (4), we designed a model to replicate the distribution of fat in skeletal muscle in vivo and demonstrated effects of BMS on chemical shift.
According to our experiment, the magnetic induction of lipid droplets in a cytosolic environment is defined exclusively by aqueous magnetic susceptibility and remains independent of relative orientations. For adipocytes distributed along muscle fibers, magnetic induction is defined by susceptibilities of both fat and water, and is altered by the orientation of muscle fibers relative to the magnetic field direction. This has important implications for the interpretation of fat resonances measured in the intact skeletal muscle tissue: 1) IMCL always resonates at the same frequency (1.4 ppm relative to water at 4.8 ppm), while the EMCL resonance shifts with respect to that of IMCL. That shift is maximized (+ 0.2 ppm) at parallel orientation and ranges from +0.2 ppm to –0.1 ppm between parallel and perpendicular orientations, with a crossover at the magic angle of 55° (Fig. 3). 2) The parallel arrangement of the sample in respect to the magnetic field maintains the optimal separation of resonances from IMCL and EMCL compartments. Such orientation is natural for muscles in the lower and upper extremities, in which all fibers run in a single direction, but it may be impossible to achieve for some other muscle groups. However, when muscle fibers are only slightly off the parallel orientation relative to the direction of the magnetic field, the resonances of IMCL and EMCL can still be separated by deconvolution. 3) The presence of EMCL in a voxel of interest may increase the variability and obscure the quantitative precision of IMCL estimation. Errors are diminished when the spectroscopic volume of interest contains minimal amounts of EMCL. 4) The precision of the evaluation of the intracellular lipids from the NMR spectrum significantly increases with separation of IMCL and EMCL resonances within a higher magnetic field (18).
The authors are grateful to Dr. Don Woessner for thoughtful comments, and to Katie Szczepaniak for linguistic assistance. We acknowledge the support of the following grants: ADA Career Development Grant (to R.L.D.), NIH R01DK 53358 (to D.T.S.), and the New York City Speakers Fund for Biomedical Research (to D.T.S.).