Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Magn Reson Med. Author manuscript; available in PMC 2011 September 1.
Published in final edited form as:
PMCID: PMC3022418

1H MRS of Intramyocellular Lipids in Soleus Muscle at 7T: Spectral Simplification by Using Long Echo Times without Water Suppression


The popular short TE 1H MRS acquisition method for detection of intramyocellular lipids (IMCL) suffers from spectral overlap due to the large, broad and asymmetric extramyocellular lipid (EMCL) signals, the time-consuming practice of selecting “lean” voxels for spectroscopy, and the overlap of the EMCL signal with the creatine methyl 1H signal at ~ 3 ppm, commonly used as an internal standard. Using an alternative acquisition strategy, spectra with well-resolved IMCL resonances were acquired from large volumes (10 to 15 mL) of human soleus muscle in less than five minutes by single-voxel 7T 1H MRS using a TE of 280 ms. From the high resolution spectra, an average IMCL concentration of 7.7 ± 3.5 mmol / kg muscle was found for 25 healthy subjects (male/female 17/8; age 29.4 ± 6.6 yr). Since water suppression was not required, the 1H signals from unsaturated intracellular triglycerides at about 5.3 ppm were easily detected which, in combination with the well-determined – (CH2)n-/CH3 intensity ratio at long TE, enabled assessment of the composition of triglycerides in the IMCL compartment. Long-echo single-voxel spectroscopy at 7T offers rapid and convenient acquisition of high resolution spectra from human soleus muscle.

Supplementary key words: intramyocellular lipids, skeletal muscle, creatine, magnetic resonance spectroscopy


Lipids in skeletal muscle are stored in both intra- and extramyocellular spaces to accommodate the body’s energy needs (1). Intramyocellular lipids (IMCL) are small fat droplets located in the cytosol, often close to mitochondria, presumably allowing for rapid transfer of long-chain acyl- groups into mitochondria for β–oxidation (25). Extramyocellular lipids (EMCL) are aggregates of adipocytes nestled between muscle fiber bundles in the form of long strands or curved plates in a variety of sizes and orientations (2,6,7). Because of the metabolic importance of fatty acid oxidation in skeletal muscle, there is considerable interest in measuring the concentration and composition of triglycerides in this compartment (816). The clinical relevance of this information was demonstrated in early studies showing that elevated IMCL correlated with insulin resistance (10). However, high concentrations of IMCL were also found in athletes who are presumably insulin-sensitive (17,18) so the precise role of IMCL in the pathogenesis of metabolic disorders remains uncertain (1922).

1H MRS is a unique tool for studies of lipid metabolism because it is the only noninvasive method that separately quantifies IMCL and EMCL. As first reported by Schick et al. (23) and Boesch et al. (24) more than a decade ago, these two lipid compartments in calf muscle were separated about ~0.2 ppm, with EMCL signal shifted downfield when the leg is approximately parallel to Bo. Translation of this data to estimates of concentration is complicated, however, by a number of factors. First, the magnitude and direction of the EMCL shift is orientation-dependent due to magnetic susceptibility anisotropy (MSA) in this compartment (2426) and consequently different EMCL lineshapes and Δδ value may be observed for the distinct muscle groups which differ in their fiber geometry and pennation angle (4). Deep soleus muscle with more oxidative fibers and large pennation angles generally displays small Δδ values and broad EMCL lineshapes, as compared to the superficial and glycolic tibialis anterior muscle. Small shifts (Δδ), dominant peak intensities, broad linewidths and the asymmetric appearance of EMCL results in a substantial resonance overlap with the IMCL resonances in spectra collected using short TE. Without sufficient separation and adequate fitting procedures with prior knowledge of lipid NMR properties (linewidth, chemical shifts and relaxation times) or more complex algorithms, the EMCL overlap may have resulted in overestimates of IMCL (27,28). Second, the methyl -CH3 resonance of total creatine (free creatine, plus phosphocreatine, PCr) at ~3.0 ppm has been proposed as an internal concentration reference in 1H MRS (29) to measure the signal intensity of IMCL and other metabolites (due to a relatively constant total creatine concentration of ~30 mmol/kg wet weight in muscle tissues (2)). However, at short TE, the creatine -CH3 signal is contaminated by the methylene protons of bis-allylic groups (=CH-CH2-CH=) in the EMCL fat pool at ~2.95 ppm. Such contamination would result in an underestimation of IMCL if the creatine -CH3 signal is used as an internal standard. Third, many authors recommend selection of “lean” voxels to improve IMCL detection. While this is a reasonable strategy, the operator becomes involved in selecting regions for analysis and, since small voxels are preferred to avoid extracellular fat, a sampling error could be introduced. Fourth, because the water signal is so large when using short echoes, sophisticated water suppression methods are used that unavoidably distort some signals near the water peak (the -CH=CH- resonances of the IMCL at 5.3 ppm) which may distort information about the chemical composition of the IMCL pool. The goal of this study was to develop an acquisition strategy that would yield high-resolution 1H MRS data at 7T with well-resolved IMCL resonances from large voxels in soleus muscle without the need for water suppression. In addition, other important factors such as total scan time, removal of overlap between the creatine -CH3 and IMCL bis-allylic groups, and SAR were taken into account. These considerations led us to an acquisition strategy that combines long echo-times (TE) with large localization voxels without water suppression (WS).

This approach offers a number of advantages aside from simplicity in selecting a volume for study and a small reduction in SAR. Existing literature methods for measuring IMCL have been largely based on resolving the bulk methylene -(CH2)n- resonance, the dominant signal in a typical spectrum at short TE (10,24,28). However, the chemical composition of fat (amounts of saturated, mono-unsaturated and polyunsaturated fats) will vary somewhat with diet, exercise, metabolic diseases and other factors (30,31) and this in turn will introduce variations in the amount of -(CH2)n- signal per mole of triglyceride. Hence, we chose to quantify IMCL by focusing on the terminal methyl –CH3 resonances instead of the – (CH2)n- resonances. Since the signal intensity depends on acquisition parameters, TR and TE (32), one must correct for T1 and T2 differences between IMCL and creatine methyl signals to accurately quantify the IMCL concentration. This correction was found to be negligible (2%) for TR and TE values equal to 2 s and 280 ms, respectively. Thus, by using these acquisition conditions, one may directly evaluate the IMCL concentration from the area of the IMCL methyl resonance compared to the area of the creatine methyl resonance. This fortuitous cancellation allows a qualitative visual estimate of the IMCL concentration from a single 1H spectrum.


Human Subjects

The protocol was approved by the Institutional Review Board of University of Texas Southwestern Medical Center. Prior to the MRS study, informed written consent was obtained from all 25 participants (8 female and 17 male), all healthy adults with no history of peripheral vascular, myopathic or systemic diseases, ages ranging from 22 to 44 yr (average 29.4± 6.6 yr), and BMI ranging from 19 to 32 (average 23.4 ± 3.6 kg/m2). To avoid possible exercise-associated physiological variations among subjects, all subjects were asked to refrain from moderate to intense physical activities for three days prior to the study, and were given a 20 minute rest prior to entering scanning room. Subjects were instructed to move slowly in the scan room. The entire scanning session was 60 min or less and was well-tolerated by all subjects. All subjects were interviewed after the exam and again at 24 hours after the exam. All subjects specifically denied dizziness, nausea, vertigo, headaches or visual changes.

MR Spectroscopy

All subjects were positioned supine in a 7T system (Achieva, Philips Medical Systems, Cleveland, OH). The left leg of each subject was positioned parallel to the magnetic field with the foot constrained in a soft stabilizer. 1H MRS spectra were acquired with a partial volume quadrature transmit/receive coil customized to fit the shape of a human calf. Axial, coronal, and sagittal turbo spin echo (TSE) images were initially acquired of the left calf muscle. Typical parameters were: field-of-view (FOV) 180 × 180mm, time of repetition (TR) 1500 ms, echo time (TE) 75 ms, turbo factor 16, and number of acquisitions (NA) one. These TSE images, which are largely T2-weighted, provided clear visualization of fasciae separating different muscle groups which allow the placement of the voxel exclusively within the soleus.

Single-voxel 1H MR spectra were collected from soleus with a typical volume of ~ 10 – 15 mL, using STEAM sequence, rather than PRESS, to minimize chemical shift displacement effect (32). Parameters were: TR 2000 ms, TE 20 ms for short TE, and 280 for long TE, spectral bandwidth (BW) 4 kHz, number of points (NP) 4096 zero-filled to 8192, line-broadening weight factor 3 Hz prior to FT, NA = 32 (short TE) or 128 (long TE) with 16-step phase cycling. The placement of the voxel was guided by T2w images so that the geometry and orientation of the voxel were adjusted to fit the actual shape of subject’s muscle and to avoid the boundaries of muscle. The 2nd order shimming and four dummy scans were applied prior to spectral data collection. For short TE studies, water-suppression (WS) was applied using a four-pulse MOIST technique provided in the Philips software. The RF pulse used for acquisition had a flat excitation profile with a uniformly excited window in the chemical shift region of interest (from 0.5 to 6.0 ppm).

The 1H T1 relaxation times of IMCL and EMCL were measured using inversion-recovery method, with seven to nine inversion delay times (TI) in the range from 20 ms to 5000 ms with a constant TR 8 s and TE 180 ms. The T2 relaxation time was measured by varying TE in the range from 30 ms to 280 ms with ten data points and a constant TR of 2 s.

Spectral Analysis

The 1H chemical shifts of all metabolite resonances in the muscle were referenced to water protons set to 4.65 ppm. The area of each metabolite 1H resonance signal was determined by fitting the spectrum to a Voigt lineshape (variable proportions of Lorentzian plus Gaussian) using ACD software (Advanced Chemistry Development, Inc., Toronto, Canada). The asymmetry of the EMCL signal was fitted by two Voigt lineshapes with different frequency shift values. Their initial linewidth and chemical shift were set so that the summation of these two components approximately fit the EMCL lineshape profile, but then were allowed to vary freely by the program. Such a fitting procedure using a minimal number of variables yielded a consistent output for all subjects. One component had a relatively narrow lineshape and large magnitude and was centered at the maximum peak intensity of EMCL signal while the second component had a relatively broad lineshape but small magnitude and was centered between the EMCL and IMCL peaks. The fitting result of the two-lineshape model was found to be insensitive to the initial lineshape setting, but the fitting residual was significantly improved, as compared to that from a single Voigt lineshape fitting (more details are provided below). The intensity of EMCL signal was thus obtained by the summation of the area from these two Voigt lineshape components.

To facilitate the evaluation of the spectral quality on the basis of how well the IMCL signal is resolved from EMCL, an empirical, dimensionless index, termed “signal separation parameter (SSP)” was defined asΔH/H, whereΔH and H are the resolved and overall height of the observed IMCL -(CH2)n- signal, respectively, as illustrated in Figure 1d. In addition to ΔH/H, the chemical shift separation (Δδ in ppm) between IMCL and EMCL –(CH3)n- resonance signals was also measured.

Fig. 1
1H MR spectra collected from a single voxel in soleus muscle at 7T. The subject was a 32-year-old male. The selected volume is indicated by the white box on T2w images (a). Spectra with (panel b, short TE of 20 ms) and without (panel c long TE of 280 ...

To evaluate IMCL and EMCL T1 values, the signal intensity data from an inversion-recovery data set were fitted to a mono-exponential equation S(TI) = a + b*exp(-TI/ T1) where TI is the delay time after the inversion pulse; and a and b are two fitting constants. IMCL and EMCL T2 relaxation time were evaluated by fitting the TE-dependent signal intensity data to a linear equation ln(S(TE)) = a - TE/T2, where a is a fitting constant. A logarithmic fitting procedure for T2 evaluation overcomes the issue of underweighting of data points at long TEs associated with conventional exponential fitting procedure.


To explore the physical mechanisms of Signal separation parameter enhancement at long TE, a phantom was constructed to mimic IMCL and EMCL in vivo using a combination of agarose gel, vegetable oil and Kleenex (Kimberly-Clark Corp., Irving, TX). First, a Kleenex was tightly rolled into a cylinder about 3 mm in diameter and 6 cm in length which was then soaked in vegetable oil taking care not to trap air bubbles. Second, 400 mL of 1% argar gel was prepared by dissolving 4 g of agar (Aldrich) into 400 mL of DI water by gentle warming. The gel was poured into a beaker and chilled to 32°C (agar must be at 32°C, otherwise oil floats to the top if too warm). Third, using a 5 mL syringe with a 25-gauge needle, oil droplets were swirled into the gel at the center of the beaker, and then the oil-soaked tissue was placed in the zone of droplets. For 1H MRS sampling, the phantom was placed in the center of a customized RF volume coil (single-channel, T/R) with the tissue roll parallel to Bo mimicking normal leg positioning. A single voxel (2 mL) was chosen to include both oil droplets (“IMCL”) and oil-soaked tissue roll (“EMCL”), and 1H MR spectra were acquired at both short TE of 20 ms and long TE of 180 ms (TR = 2 s, same as in vivo 1H MRS) at 7T. For comparison of lineshape and orientation dependence, another two-compartment phantom was constructed using two oil-filled glass capillaries, one parallel to Bo and the other at 55° relative to Bo.


Muscle 1H MRS Spectra at Long TE

Typical 1H spectra acquired with water-suppression and a short TE of 20 ms is compared to data acquired without water suppression and a long TE of 280 ms in Figure 1. Other conditions such as shimming and voxel location were otherwise identical. Spectral resolution was clearly improved in the long TE spectra not only in the important fat region (0.7 – 1.9 ppm), but also in the “metabolic fingerprint” region (2.9 – 4.1 ppm in which multiple metabolites may appear including creatine, carnitine, taurine, lactate and glycogen), and in the lipid “double-bond” region (~5 – 6 ppm). The spectrum at long TE was greatly simplified since several short-T2 lipid resonances were attenuated below the detection limit, including the -CH2- resonance β to COO group (at 1.79 ppm for EMCL, and at 1.55 ppm for IMCL), the EMCL bis-allylic CH2 resonance (at ~2.95 ppm), and methine -CH- resonance of lipid glycerol moiety (at 5.42 for EMCL, and 5.18 ppm for IMCL).

As a result, in the upfield region of 0.7 – 1.9 ppm, the 1H spectrum collected using a TE of 280 ms (Figure 1c) yielded a set of four well-separated resonances from bulk –(CH2)n– and terminal –CH3 resonances of both IMCL and EMCL, and the total creatine –CH3 resonance at ~ 3.0 ppm became a sharp, symmetric singlet. Compared to the spectrum at short TE (20 ms, Figure 1b), the ΔH/H parameter at long TE (280 ms, Figure 1c) increased ~5-fold (15% to 80%, Figure 1d). The concentration of IMCL as estimated from the methyl proton signal intensity was 6.4 mmol/kg (= 30 × 0.64 / 3 where 0.64 is the IMCL-CH3 to Cr-CH3 intensity ratio, each triglyceride molecule has 3 methyl groups, as compared to one in the case of creatine) prior-to T1 and T2 correction. Since the correction for T1 and T2 is small (see below) this value is very close to the final determination, 6.3 mmol/kg.

The long TE spectra also illustrated some fine spectral features not obvious in the short TE spectra. For example, the taurine (Tau) –CH2–CH2–protons appear as two doublets, indicative of residual dipolar coupling (RDC, 0.07 ppm or 21 Hz) likely due to restricted tumbling of taurine molecules in the anisotropic muscle tissue. In the short TE (20 ms) spectra, however, the upfield doublet of taurine was not detectable since it was completely buried by a much larger carnitine TMA signal (Figure 1b). Furthermore, in spectra collected using a long TE (280 ms), the IMCL -CH=CH- resonance at ~5.3 ppm was also clearly resolved (Figure 1c). This might provide a mechanism to estimate the concentration of unsaturated fatty acids in myocytes.

The 1H spectrum in Figure 1c illustrates that suppression of the water signal using the long TE sequence does not result in baseline distortion and sideband artifacts commonly seen when using short TE sequences. In fact, the water signal was attenuated by a factor of ~3,000 using a long TE compared to its intensity when using a short TE as expected for a T2 of 32 ms (water T1: 1.14 s) at 7T. This contrasts sharply to the total creatine -CH3 proton signal at ~3.0 ppm which was attenuated by a much smaller factor of 17, given its long T2 of 91.6 ms (Cr-CH3 T1: 0.95 s) at 7T. Thus, the intensity of these two resonances became quite comparable in spectra collected using a long TE (280 ms). Without WS, SAR was 5% lower when using a STEAM sequence under these experimental conditions. Additionally, the additional ~1 min preparation time typically required for WS optimization prior to data sampling was also saved.

The practical value of this long echo method is illustrated in Figure 2. The subject is a 24-yr healthy female whose soleus muscle was characterized by presence of substantial EMCL deposits (Figure 2a). In this particular case, the IMCL signal in the short TE 1H MR spectrum was unresolved from the large EMCL signal (ΔH/H ~ 0%, Figure 2b), and therefore impossible to quantify. A lineshape fitting of the resolved peaks ΔH/H (56%, Figure 2c) in the long TE spectrum however allowed an estimate of IMCL concentration for this subject (3.7 mmol/kg).

Fig. 2
1H MR spectra collected from a single voxel in soleus muscle at 7T. The subject was a 25-year-old female with a high EMCL content, as shown by the dense white streaks on the MIP images, panel a. The region 0.7 – 1.7 ppm illustrates severely overlapping ...

The 1H MRS results at TE 280 ms for all 25 participants are shown in Figure 3. The signal resolution between IMCL and EMCL resonances were significantly improved in all long TE spectra. The measured ΔH/H values for the 25 participants ranged from 56% to 90%, with an average ΔH/H value of 77% (Figure 3a). This represents an improvement of 4.3-fold compared to spectral resolution seen in the short TE spectra (averaged 18%). Due to improved spectral resolution, the chemical shift separation (Δδ) between IMCL and EMCL is more accurately determined at long TE (average Δδ: 0.24 ppm, Figure 3b). Overall, for the 25 subjects, the average IMCL signal intensity relative to total creatine Cr-CH3 signal is 1.94 ± 0.83 for -(CH2)n- and 0.79 ± 3.6 for -CH3 (Figure 3c). The derived IMCL concentration was 7.7 ± 3.6 mmol/kg after correction for T1 and T2 (see below). This represents the amount of triglyceride in fat droplets in soleus muscle at rest. Interestingly, compared to the linewidth of Cr-CH3 (12.4 Hz, at TE 280 ms), there was a significant line-broadening (~50%) observed for the IMCL signals (17.0 Hz for -(CH2)n- and 18.1 Hz for -CH3, Figure 3d).

Fig. 3
Soleus IMCL and EMCL MRS results (average ± standard deviation) measured at TE 280 ms and 7T (N = 25). (a) ΔH/H of IMCL -(CH2)n- signal; (b) chemical shift separation Δδ between IMCL and EMCL -(CH2)n- signals; (c) LW1/2 ...

Data reproducibility was also evaluated on two randomly-chosen subjects. A comparison between the initial scan and the follow-up scan three weeks later showed that the variation in linewidth, peak intensity and signal separation parameter all fell within 8%, a reasonable variation considering the possible variation in subject physiology/metabolism and in instrumental and repositioning settings during the three week span. As a comparison, a CV (coefficient of variation) value of 6% was reported by Boesch et al (24), 7.9–11.8% Szczepaniak et al (33), and 13.4–14.4 Torriani et al (34) in their muscle MRS studies.

Composition of the IMCL Triglyceride Pool

Given that the (-HC=CH-)/-CH3) ratio reflects the relative content of unsaturated versus saturated fatty acids and (CH2)n-/-CH3) ratio reflects the length of saturated carbon-chain in the backbone of fatty acid molecule, an analysis of these signals provides additional information about the fatty acid composition. For the 25 subjects, the average IMCL ((-HC=CH-)/-CH3) ratio was 0.27 ± 0.11 in soleus muscle (Figure 3e) compared to 0.26 ± 0.12 measured from spectra of subcutaneous lipids from a smaller group of 7 volunteers (whose calf subcutaneous tissue was thick enough to well accommodate a localization voxel of size 0.5×0.5×0.5 mm3) under identical TR and TE conditions using the same STEAM sequence. This small group of subjects also yielded an average (-(CH2)n-/-CH3) value of 1.84 ± 0.18 for subcutaneous tissue, slightly smaller than that of IMCL (2.20 ± 0.32) and EMCL (2.10 ± 0.28) obtained for the 25 subjects. Figure 4b compares the IMCL and EMCL (-(CH2)n-/-CH3) intensity ratio for the 25 subjects. These ratios were not significantly different among the 25 individuals. After correction for small T1 and T2 differences between IMCL -(CH2)n- and -CH3 resonances, a molar ratio (- (CH2)n-/-CH3) of 6.6 ± 1.3 was found for IMCL (Figure 4a), which is very close to the value of 6.9 (= 62/9, 62 protons/molecule triglyceride resonating at 1.24 ppm, 9 protons at 0.84 ppm) calculated from the average composition of six most abundant triglycerides in human body (C14:0, C16:0, C16:1 (n-7), C18:0, C18:1 (n-9), and C18:2 (n-6)) by Boesch et al (2).

Fig. 4
The concentration of IMCL and –CH2- / CH3 ratio for 25 subjects. (a) the IMCL concentration in the soleus muscle at rest is shown in panel a, and the intensity ratio (-(CH2)n-/-CH3) at TE 280 ms for both IMCL and EMCL is shown in panel b.

IMCL and EMCL T1 and T2 Relaxation Times

Figure 5a showed a series of inversion-recovery 1H MR spectra acquired at TR = 8 s (> 5 T1) with inversion delay time varying from 20 to 5000 ms. The spectra clearly indicates that the bulk -(CH2)n- signal recovers more rapidly than the -CH3 resonance. The inversion-recovery intensity curves for these resonances were single exponential (Figure 5b). The T1 values derived from fitting these curves were summarized in Table 1. A comparison of T1 values of the different lipid stores shows variations among the terminal -CH3 resonances, IMCL (1.38 s) > EMCL (1.20 s) ~ bone marrow (1.16 s) > subcutaneous tissue (1.08 s). In contrast, the lipid -(CH2)n- resonance was constant among these same lipid pools (Table 1). A T1 value of 0.95 s was measured for creatine –CH3 resonance by inversion-recovery.

Fig. 5
Inversion-recovery 1H MR spectra and exponential curve fitting of the signal intensity of the bulk CH2 and terminal CH3 1H resonances for evaluation of IMCL and EMCL T1 relaxation time in soleus muscle. The red and blue traces in the spectra represent ...
T1 and T2 Relaxation Times of Methyl and bulk Methylene Proton Resonances in EMCL and IMCLa

Figure 6a shows a series of soleus muscle 1H MR spectra of fat in the 0.5–1.7 ppm region for a series of echo times (TE) from 30 to 280 ms. As described above, IMCL resonances shown improved resolution with an increase in TE, from 45% at TE 30 ms to 85% at TE 280 ms. While the IMCL signals were relatively sharp and symmetric at all echo times, the EMCL signals were clearly asymmetric. The asymmetric EMCL lineshapes were further revealed by the relative large residuals resulting from fitting with a single symmetric Voigt lineshapes (Figure 6b). The fitting residuals were substantially improved by using a fitting procedure that included two symmetric Voigt lineshapes (Figure 6c). The resulting two EMCL components had a frequency span of 0.07 ± 0.01 ppm, one situated at the peak position of the observed EMCL signal and the other centered 0.07 ppm upfield of the main EMCL peak. The downfield EMCL component was relatively narrow and dominated the observed EMCL signal at long TE while the upfield component had a broad lineshape and decayed more rapidly with an increase in TE. As a consequence, the observed EMCL signals, both -(CH2)n- and -CH3, became more symmetric and less overlapping with IMCL at long TE. The linewidth (LW1/2) narrowing effect is different between EMCL and IMCL. A much larger effect was observed for EMCL (from 46.2 Hz at TE 30 ms to 32.2 Hz at TE 280 ms for –(CH2)n–signal) than for IMCL (from 17.3 Hz at TE 30 ms to 14.6 Hz at TE 280 ms).

Fig. 6
Comparison of spectral resolution and the lineshape fitting in the fat region from 0.5 – 1.7 ppm acquired at different TEs from 30 to 280 ms from soleus muscle of a 27-yr healthy male. (a) experimental spectra; (b) the fitted spectra (red trace) ...

Another factor contributing to the increase in signal separation parameter at long TE was that EMCL also had a shorter T2 value than IMCL. Figure 6d shows plots of ln(S) (S: signal intensity) versus TE; the slopes of these lines show that the T2 of EMCL is ~24% smaller than IMCL (74 versus 97 ms for -CH3 and 51 versus 66 ms for -(CH2)n- resonances, Table 1). In comparison, the creatine -CH3 signal at ~3.0 ppm had a T2 value of 92 ± 6 ms, significantly longer than the T2 of EMCL -CH3 resonance (74 ms) but slightly shorter than that of IMCL -CH3 (97 ms). Because of the T1 and T2 difference between creatine and IMCL -CH3 resonance, a correction factor was needed to convert their intensity ratio to molar ratio by the formula exp(-TE/T2)(1-exp(-TR/T1). This yielded a correction factor of 0.02 or 2% using the experimental parameters TR = 2 s and TE = 280 ms as well as the T1 and T2 values of creatine and IMCL -CH3 resonances. For the same reason, the average IMCL -(CH2)n- to -CH3 intensity ratio of 2.2 (N = 25, Table 1) corresponded to a molar ratio of 6.6 when the difference in T1 and T2 values between -(CH2)n- to -CH3 was taken into account.

Phantom Study

It is not clear why the triglycerides in the extracellular compartment have shorter T2 values compared to triglycerides in the IMCL compartment. To explore the physical mechanisms for this observation, two different two-compartment phantoms were constructed as shown Figure 7. The first phantom consisted of oil-soaked cylinder-shaped tissue (mimicking EMCL) and oil droplets (mimicking IMCL) scattered randomly in agar (in the mimicking tissue environment in vivo). As shown in Figure 7a, the phantom was placed in the magnet with the oil-soaked cylinder-shaped roll of tissue parallel to Bo and 1H MR spectra were collected from a single voxel containing both the cylinder and oil droplets at TE 20 and TE 180 (Figures 7b and 7c respectively). The 1H MR spectra bore similar characteristics as those in vivo spectral data, an indication of similar geometric and susceptibility properties For example, the “EMCL” signals at short TE were large, broad and asymmetric with a long upfield tail overlapping with the relatively sharp and narrow “IMCL” signals. This contrasts sharply to the spectrum at long TE of 180 ms which showed a clearly-resolved “EMCL” and “IMCL” resonances. As expected, an analysis of the -(CH2)n- 1H signal intensity data at different TEs (6 data points, from 100 to 280 ms) indeed yielded distinct T2 values for these two compartments: 69.4 ± 5 ms for “EMCL” and 81.3 ± 6 ms (linear correlation coefficient R2: 0.994 for “EMCL” and 0.991 for “IMCL”)

Fig. 7
Experiments with phantoms. A two-compartment phantom in mimicking IMCL and EMCL in vivo, with vegetable oil droplets in mimicking “IMCL” and oil-soaked cylinder-shaped tissue roll in mimicking “EMCL”; The 1H MRS from phantom ...

The second two-compartment phantom consisted of two oil-filled glass tubes (o.d. 1 mm, length 6 cm). Again, the 1H MR spectra were collected from a single voxel containing both tubes at TE of 20 (short, Figure 7e) and 180 (long, Figure 7f), respectively. The lineshape and intensity from both tubes are very similar, though they have a frequency separation of 0.24 ppm. In both orientations, very narrow line- width was observed without significant asymmetric frequency dispersion, as compared to the oil-soaked cylinder-shaped tissue roll in Figure 7b and 7c. Indeed, longer T2 values were found for the oil filled in these two cross-placed glass tubes, and the -(CH2)n- T2 difference between these two orientations are minimal: 104.1 ± 7 ms for the tube parallel to Bo, (R2: 0.976) and 100.2 ± 8 ms for the tube 55° to Bo (R2: 0.969). This indicated that the broad and asymmetric linewidth observed on “EMCL” (oil-in-tissue) might be related to the reduced molecular motion in triglycerides in the tightly-rolled tissue layers. However, the longer TE value allowed more time for the molecular to exchange among MSA-distinct regions and led to a more effective average on the field heterogeneity.


A limitation of 1H MRS for noninvasive quantitation of IMCL is the lack of high spectral resolution since the EMCL signal is separated by only ~0.2 ppm and it is about one order of magnitude larger than the IMCL signal and often asymmetric. Consequently, small voxels, careful positioning of the leg, limitation to studies of the tibialis anterior muscle, and complex data analysis schemes are needed. The current work uses a spectral acquisition strategy using long echo times combined with a large voxel STEAM sequence without water suppression. This approach contrasts to the current popular method of using smaller voxels combines with short a TE and PRESS sequence (to offset the low sensitivity of a small voxel. Note that the method of taking smaller voxels to avoid fat inclusion may not show a significantly suppression effect on EMCL). There are at least three significant benefits of using the long echo approach.

First, the IMCL –CH3 resonance rather than bulk -(CH2)n- resonance can be used for assessing the IMCL concentration. The advantage of such an approach is straightforward: one does not need to assume a particular lipid composition to translate the area of the –CH3 resonance into concentration, as in the case of the -(CH2)n- resonance. The long TE strategy provided the basis for such a method since the T2 of the lipid –CH3 resonance is significantly longer than the T2 of the -(CH2)n- resonances (Table 1). IMCL quantification was made easier at a practical level by adopting TR = 2 s and TE = 280 ms where intensity corrections are small. From spectra collected using TE = 280 ms, and analysis of IMCL for 25 individuals shows a large variation in IMCL concentrations from 3.0 to 15.2 mmol/kg, with an average concentration of 7.7 ± 3.5 mmol / kg (Figure 4a). In comparison, Cui et al (35) reported an average value of 7.7 ± 1.2 mmol/kg for soleus IMCL concentration at 4T using 1.0 mL voxels for non-obese sedentary healthy subjects (N = 7, BMI = 24.9 ± 3.8 kg/m2, age 36.0 ± 11.7 yr). Hwang et al (22) recently reported an average value of 8.8 ± 8.3 mmol/kg (individual values ranged from 2.8 to 33.2 mmol/kg) for a group of twelve nondiabetic subjects (age 34.0 ±15.7 yr ; 10 males, 2 females; BMI 25.6 ±3.2 kg/m). In a previous study (28) using spectra collected for small voxels (1 mL) and a short TE (13 ms) estimated a soleus concentration of IMCL = 4.8 ± 2.2 mmol/kg for a group of nine volunteers aging from 21 – 68 y (average 44 y). The current data for healthy sedentary adults is consistent with earlier estimates of the concentration of IMCL.

A second advantage of this approach is that one can remove the contamination of the EMCL =CH-CH2-CH= signal to the total creatine –CH3 at ~3.0 ppm. Although there is no ideal internal concentration standard in skeletal muscle, the concentration of creatine is relatively constant. The long echo experiment precludes underestimation of IMCL as a result of contamination of the creatine internal standard. In addition, the long TE measurement also reduced the intensity of the EMCL -(CH2)n- resonances relative to the IMCL OOC-CH2-CH2 signal and probably other unidentified macromolecules in that chemical shift region. These spectral simplifications make automatic lineshape fitting procedures simpler.

A third advantage is that water suppression was not used. Consequently, the preparation time need to begin data collection is reduced by ~ 1 min, SAR is reduced by at least 5% (as reported by the scanner) and other metabolites with resonances near water are not attenuated (24). The water signal in the 1H MR spectrum thus can be explored in several different ways, for example to establish the relationship between muscle function and dehydration - an interesting yet challenging topic in athletic communities (36).

The long TE used here to gain spectra resolution inevitably compromises sensitivity. One can use a larger voxel to make up the loss of sensitivity, but this may be a problem for subjects with smaller muscle mass such as children. In such cases, one may need to use smaller voxels and longer scan times. When the chemical shift displacement artifact (CSD) (32) is not a major concern, PRESS sequence can replace STEAM to boost S/N ratio under the long TE scheme with smaller voxel. It is known that echo-based localized 1H MR spectrum intrinsically suffers from CSD, an artifact that causes two different resonances within a given spectrum originate from two different geometric regions, shifted by a distance which is a function of the size of the localization voxel, the strength of the magnetic field, the chemical shift difference (in ppm) between these two resonances, and the pulse sequence used. PRESS gives rise to significantly larger CSD than STEAM due to the use of two 180° pulses. Additionally, TE can be reduced to enhance sensitivity or increased to further enhance resolution dependent on the actual need during the study. However, in such cases, one may have to correct the T1 and T2 difference between IMCL and reference signals due to deviation from the advantageous setting of TR = 2 s and TE = 280 ms for good spectral resolution.

The advantages of collecting long TE spectra at 7T can be primarily attributed to the shorter T2 of the EMCL protons, the relaxation mechanism of which remains unclear. The phantom studies show that the proton linewidths of oils in capillary tubes is not orientation-dependent (Figure 7e and 7f) yet the appearance of asymmetric “EMCL” signal in the short T2 spectrum (see Figure 7b) of an oil soaked Kleenex oriented in agar suggests that EMCL signals in human muscle may also be broadened by magnetic susceptibility anisotropy (MSA) in a magnetically heterogeneous environment. In fact, the increased EMCL chemical shift dispersion at higher magnetic field strength was thought to be one of the drawbacks of going to higher field. However, the phantom study demonstrated that the triglyceride molecules trapped in the tightly-rolled Kleenex cylinder (“EMCL”) do appear to have shorter T2 than the oil molecules in droplets (“IMCL”) due to more restricted motion. Based on MRI, a significant percentage of EMCL is in fairly large depots, one may not consider that such lipid stores constitute a motion-restricted environment for the fatty acid and triglyceride molecules. This implies that the improved resolution at long TE is due to the attenuation of EMCL component with much short T2 and dispersed chemical shift.

If the same strategy for resolving EMCL and IMCL signals of skeletal muscle fats also works as well in insulin-resistant patients as illustrated here for healthy subjects, then one should be able to use this technique to confirm the early work on the correlation between insulin resistance and elevated IMCL and to clarify the issue of the “metabolic paradox” (11,37). This in turn may help position 1H MRS of skeletal muscle fat as a useful diagnostic clinical tool. The long TE strategy may also provide new venues for studies of lipid metabolism in skeletal muscle using 1H MRS. For instance, the IMCL and EMCL - (CH2)n-/-CH3 signal intensity ratios, which can only be reliably obtained using a long TE, is a index worthy of further exploration. It is anticipated that this index will be sensitive to fatty acid composition, since the three most abundant fatty acids in human tissue have very different -(CH2)n-/-CH3 ratios: palmitic acid 9.3 (= 14 × 2/3), oleic acid 6.7 (= 10 × 2/3) and linoleic acid 4. 7 (= 7 × 2/3). Given that skeletal muscle is an essential tissue for whole-body energy metabolism, including insulin-stimulated glucose uptake and fatty acid oxidation, the feasibility and non-invasiveness of the long TE 1H MRS in determining the fatty acid composition of fat droplets in myocytes may profoundly impact future muscle metabolism studies. Evidence derived from skeletal muscle biopsies and blood samples indicates that the fatty acid composition of skeletal muscle phospholipids is related to peripheral insulin sensitivity and obesity in several human populations (3841). One may use this index to study factors that influence deposition and mobilization of fats within and between individuals. (39,4245).

In conclusion, the current 1H MRS study at 7T demonstrated that the use of long echo times (TE), together with large voxel sizes without water-suppression greatly enhances the resolution of skeletal muscle EMCL and IMCL and thereby allows a more reliable quantitative measure of IMCL content. The long TE strategy also has the advantage of lower SAR, large sampling volume, short scan time and inclusion of water in the spectra. The successful application of this long TE strategy may provide new opportunities to explore the fat composition and metabolism in skeletal muscle and to better characterize metabolic disorders in diseases such as obesity, insulin resistance and diabetes.


The authors are grateful to Ivan Dimitrov (Philips Medical Systems) for his expert advice, Deborah Douglas for assistance with data acquisition, and Charles Storey for preparation of the phantoms. Jeannie Davis and Sonya Rios recruited and managed the human subjects. This study was supported by the National Institutes of Health (RR002584 and DK081186) and the Department of Defense (Contract number W81XWH-06-2-0046).


magnetic resonance spectroscopy
number of averages
number of points
stimulated echo acquisition mode
point-resolved spectroscopy
echo time
repetition time
magnetic susceptibility anisotropy


1. Stehno-Bittel L. Intricacies of Fat. Phys Ther. 2008;88(11):1265–1278. [PubMed]
2. Boesch C, Machann J, Vermathen P, Schick F. Role of proton MR for the study of muscle lipid metabolism. NMR Biomed. 2006;19(7):968–988. [PubMed]
3. Malloy C, Ren J, White C. Intramyocyte Lipids May Impair Insulin Signaling. Am J Psychiatry. 2007;164(10):1475. [PubMed]
4. Schrauwen-Hinderling VB, Hesselink MKC, Schrauwen P, Kooi ME. Intramyocellular Lipid Content in Human Skeletal Muscle. Obesity. 2006;14(3):357–367. [PubMed]
5. Hoppeler H. Exercise-induced ultrastructural changes in skeletal muscle. Int J Sports Med. 1986;7(4):187–204. [PubMed]
6. Machann J, Steidle G, Thamer C, Mader I, Schick F, GAW . Annual Reports on NMR Spectroscopy. Vol. 50. Academic Press; 2003. In Vivo Proton NMR Studies in Skeletal Musculature; pp. 1–74.
7. Machann J, Stefan N, Schick F. 1H MR spectroscopy of skeletal muscle, liver and bone marrow. Eur J Radiol. 2008;67(2):275–284. [PubMed]
8. Torriani M, Thomas BJ, Barlow RB, Librizzi J, Dolan S, Grinspoon S. Increased intramyocellular lipid accumulation in HIV-infected women with fat redistribution. J Appl Physiol. 2006;100(2):609–614. [PMC free article] [PubMed]
9. Kelley DE, Goodpaster BH. Skeletal Muscle Triglyceride: An aspect of regional adiposity and insulin resistance. Diabetes Care. 2001;24(5):933–941. [PubMed]
10. Krssak M, Falk Petersen K, Dresner A, DiPietro L, Vogel SM, Rothman DL, Shulman GI, Roden M. Intramyocellular lipid concentrations are correlated with insulin sensitivity in humans: a 1H NMR spectroscopy study. Diabetologia. 1999;42(1):113–116. [PubMed]
11. Goodpaster BH, He J, Watkins S, Kelley DE. Skeletal Muscle Lipid Content and Insulin Resistance: Evidence for a Paradox in Endurance-Trained Athletes. J Clin Endocrinol Metab. 2001;86(12):5755–5761. [PubMed]
12. Muoio DM, Koves TR. Lipid-induced metabolic dysfunction in skeletal muscle. Novartis Found Symp. 2007;286:24–38. [PubMed]
13. Newcomer BR, Lawrence JC, Buchthal S, Hollander JAd. High-resolution chemical shift imaging for the assessment of intramuscular lipids. Magnetic Resonance in Medicine. 2007;57(5):848–858. [PubMed]
14. Thamer C, Machann J, Bachmann O, Haap M, Dahl D, Wietek B, Tschritter O, Niess A, Brechtel K, Fritsche A, Claussen C, Jacob S, Schick F, Haring H-U, Stumvoll M. Intramyocellular Lipids: Anthropometric Determinants and Relationships with Maximal Aerobic Capacity and Insulin Sensitivity. J Clin Endocrinol Metab. 2003;88(4):1785–1791. [PubMed]
15. Toledo FGS, Menshikova EV, Azuma K, Radikova Z, Kelley CA, Ritov VB, Kelley DE. Mitochondrial Capacity in Skeletal Muscle Is Not Stimulated by Weight Loss Despite Increases in Insulin Action and Decreases in Intramyocellular Lipid Content. Diabetes. 2008;57(4):987–994. [PubMed]
16. van Loon LJ, Goodpaster BH. Increased intramuscular lipid storage in the insulin-resistant and endurance-trained state. Pflugers Arch. 2006;451(5):606–616. [PubMed]
17. van Loon LJC, Koopman R, Manders R, van der Weegen W, van Kranenburg GP, Keizer HA. Intramyocellular lipid content in type 2 diabetes patients compared with overweight sedentary men and highly trained endurance athletes. Am J Physiol Endocrinol Metab. 2004;287(3):E558–565. [PubMed]
18. Hoppeler H, Lüthi P, Claassen H, Weibel ER, Howald H. The ultrastructure of the normal human skeletal muscle. A morphometric analysis on untrained men, women and well-trained orienteers. Pflugers Arch. 1973;344(3):217–232. [PubMed]
19. Watt MJ. Storing up trouble: does accumulation of intramyocellular triglyceride protect skeletal muscle from insulin resistance? Clin Exp Pharmacol Physiol. 2009;36(1):5–11. [PubMed]
20. Guo Z. Intramyocellular lipids: maker vs. marker of insulin resistance. Med Hypotheses. 2008;70(3):625–629. [PMC free article] [PubMed]
21. Goodpaster BH, Kelley DE. Skeletal muscle triglyceride: marker or mediator of obesity-induced insulin resistance in type 2 diabetes mellitus? Curr Diab Rep. 2002;2(3):216–222. [PubMed]
22. Hwang J-H, Stein DT, Barzilai N, Cui M-H, Tonelli J, Kishore P, Hawkins M. Increased intrahepatic triglyceride is associated with peripheral insulin resistance: in vivo MR imaging and spectroscopy studies. Am J Physiol Endocrinol Metab. 2007;293(6):E1663–1669. [PubMed]
23. Schick F, Eismann B, Jung W-I, Bongers H, Bunse M, Lutz O. Comparison of localized proton NMR signals of skeletal muscle and fat tissue in vivo: Two lipid compartments in muscle tissue. Magnetic Resonance in Medicine. 1993;29(2):158–167. [PubMed]
24. Boesch C, Slotboom J, Hoppeler H, Kreis R. In vivo determination of intra-myocellular lipids in human muscle by means of localized 1H-MR-spectroscopy. Magnetic Resonance in Medicine. 1997;37(4):484–493. [PubMed]
25. Boesch C, Kreis R. Dipolar coupling and ordering effects observed in magnetic resonance spectra of skeletal muscle. NMR in Biomedicine. 2001;14(2):140–148. [PubMed]
26. Szczepaniak LS, Dobbins RL, Stein DT, McGarry JD. Bulk magnetic susceptibility effects on the assessment of intra- and extramyocellular lipids in vivo. Magn Reson Med. 2002;47(3):607–610. [PubMed]
27. Steidle G, Machann J, Claussen CD, Schick F. Separation of intra- and extramyocellular lipid signals in proton MR spectra by determination of their magnetic field distribution. J Magn Reson. 2002;154(2):228–235. [PubMed]
28. Khuu A, Ren J, Dimitrov I, Woessner D, Murdoch J, Sherry AD, Malloy CR. Orientation of lipid strands in the extracellular compartment of muscle: Effect on quantitation of intramyocellular lipids. Magnetic Resonance in Medicine. 2009;61(1):16–21. [PMC free article] [PubMed]
29. Ren J, Dimitrov I, Murdoch J, Sherry AD, Malloy CR. Issues with creatine/phosphocreatine quantification: Lipids contamination and water-suppression. Vol. 2008. Toronto, Canada: May 3–9, 2008.
30. Harris WS, Poston WC, Haddock CK. Tissue n-3 and n-6 fatty acids and risk for coronary heart disease events. Atherosclerosis. 2007;193(1):1–10. [PubMed]
31. Baylin A, Kabagambe EK, Siles X, Campos H. Adipose tissue biomarkers of fatty acid intake. Am J Clin Nutr. 2002;76(4):750–757. [PubMed]
32. Ren J, Dimitrov I, Sherry AD, Malloy CR. Composition of adipose tissue and marrow fat in humans by 1H NMR at 7 Tesla. J Lipid Res. 2008;49(9):2055–2062. [PMC free article] [PubMed]
33. Szczepaniak LS, Babcock EE, Schick F, Dobbins RL, Garg A, Burns DK, McGarry JD, Stein DT. Measurement of intracellular triglyceride stores by H spectroscopy: validation in vivo. Am J Physiol Endocrinol Metab. 1999;276(5):E977–989. [PubMed]
34. Torriani M, Thomas BJ, Halpern EF, Jensen ME, Rosenthal DI, Palmer WE. Intramyocellular Lipid Quantification: Repeatability with 1H MR Spectroscopy. Radiology. 2005;236(2):609–614. [PubMed]
35. Cui M-H, Hwang J-H, Tomuta V, Dong Z, Stein DT. Cross contamination of intramyocellular lipid signals through loss of bulk magnetic susceptibility effect differences in human muscle using 1H-MRSI at 4 T. J Appl Physiol. 2007;103(4):1290–1298. [PubMed]
36. Edwards AM, Noakes TD. Dehydration: cause of fatigue or sign of pacing in elite soccer? Sports Med. 2009;39(1):1–13. [PubMed]
37. Dube JJ, Amati F, Stefanovic-Racic M, Toledo FGS, Sauers SE, Goodpaster BH. Exercise-induced alterations in intramyocellular lipids and insulin resistance: the athlete’s paradox revisited. Am J Physiol Endocrinol Metab. 2008;294(5):E882–888. [PMC free article] [PubMed]
38. Kriketos AD, Pan DA, Lillioja S, Cooney GJ, Baur LA, Milner MR, Sutton JR, Jenkins AB, Bogardus C, Storlien LH. Interrelationships between muscle morphology, insulin action, and adiposity. Am J Physiol Regul Integr Comp Physiol. 1996;270(6):R1332–1339. [PubMed]
39. Baur LA, O’Connor J, Pan DA, Storlien LH. Relationships between maternal risk of insulin resistance and the child’s muscle membrane fatty acid composition. Diabetes. 1999;48(1):112–116. [PubMed]
40. Borkman M, Storlien LH, Pan DA, Jenkins AB, Chisholm DJ, Campbell LV. The Relation between Insulin Sensitivity and the Fatty-Acid Composition of Skeletal-Muscle Phospholipids. N Engl J Med. 1993;328(4):238–244. [PubMed]
41. Clore JN, Li J, Gill R, Gupta S, Spencer R, Azzam A, Zuelzer W, Rizzo WB, Blackard WG. Skeletal muscle phosphatidylcholine fatty acids and insulin sensitivity in normal humans. Am J Physiol Endocrinol Metab. 1998;275(4):E665–670. [PubMed]
42. Andersson A, Sjodin A, Olsson R, Vessby B. Effects of physical exercise on phospholipid fatty acid composition in skeletal muscle. Am J Physiol Endocrinol Metab. 1998;274(3):E432–438. [PubMed]
43. Arab L. Biomarkers of Fat and Fatty Acid Intake. J Nutr. 2003;133(3):925S–932. [PubMed]
44. Andersson A, Nalsen C, Tengblad S, Vessby B. Fatty acid composition of skeletal muscle reflects dietary fat composition in humans. Am J Clin Nutr. 2002;76(6):1222–1229. [PubMed]
45. Baur LA, O’Connor J, Pan DA, Kriketos AD, Storlien LH. The fatty acid composition of skeletal muscle membrane phospholipid: Its relationship with the type of feeding and plasma glucose levels in young children. Metabolism. 1998;47(1):106–112. [PubMed]