Compositional marker in vivo reveals intramyocellular lipid turnover during fasting-induced lipolysis

Intramyocellular lipid (IMCL) is of particular metabolic interest, but despite many proton magnetic resonance spectroscopy (1H MRS) studies reporting IMCL content measured by the methylene (CH2) resonance signal, little is known about its composition. Here we validated IMCL CH3:CH2 ratio as a compositional marker using 1H MRS at short echo time, and investigated IMCL content and composition during a 28-hour fast in 24 healthy males. Increases in IMCL CH2 relative to the creatine and phosphocreatine resonance (Cr) at 3.0 ppm (an internal standard) correlated with circulating free fatty acid (FA) concentrations, supporting the concept of increased FA influx into IMCL. Significant decreases in IMCL CH3:CH2 ratio indicated a less unsaturated IMCL pool after fasting, and this compositional change related inversely to IMCL baseline composition, suggesting a selective efflux of unsaturated shorter-chain FA from the IMCL pool. This novel in vivo evidence reveals IMCL turnover during extended fasting, consistent with the concept of a flexible, responsive myocellular lipid store. There were also differences between soleus and tibialis anterior in basal IMCL composition and in response to fasting. We discuss the potential of this marker for providing insights into normal physiology and mechanisms of disease.

The unique ability of proton magnetic resonance spectroscopy ( 1 H MRS) to non-invasively distinguish IMCL from extramyocellular lipid (EMCL) 16,17 has led to its widespread use, reporting IMCL concentrations from the signal intensity of the methylene (CH 2 ) resonance in FA chains in TG (the predominant species detected by 1 H MRS 18 ). However, CH 2 signal is also influenced by lipid composition, a notional normal value being assumed to quantify IMCL. In contrast, the methyl (CH 3 ) resonance is independent of lipid composition 18 , but measurement is hampered by its lower signal intensity and the overlapping EMCL CH 3 resonance. To address this Ren et al. 19 exploited the spectral simplification of long echo time acquisition and the enhanced signal-to-noise and spectral resolution at 7 T. They proposed the CH 3 :CH 2 ratio as a marker of IMCL TG composition 19 . However, at long echo times the reduction in signal and the need for accurate T 2 correction 19,20 pose challenges, and its practical utility has not been established.
Most muscle IMCL measurements are made at 3 T with a short echo time. We set out to test the feasibility of measuring the CH 3 resonance under these conditions, using optimal quality spectra and fitting algorithms. To evaluate the CH 3 :CH 2 ratio as a compositional marker, we measured it in vitro in phantoms simulating IMCL and EMCL. To study IMCL composition and dynamics during extended fasting we performed 1 H MRS at short echo time in healthy males at both 8 h (a standard baseline time point for metabolic studies regarding food intake) and 28 h of fast. We measured in vivo changes in CH 2 (influenced both by TG amount and composition), CH 3 (reporting only TG amount, and thus reflecting net FA influx-efflux) and CH 3 :CH 2 (the compositional marker) in two leg muscles, the more oxidative soleus (SOL) and the more glycolytic tibialis anterior (TA), and related these to circulating FFA concentrations and changes in MRI-measured adipose tissue volumes.

Results and Discussion
Phantom validation. The IMCL CH 3 :CH 2 ratio is sensitive to unsaturation, and to a lesser degree the chain length in intramyocellular TG (Fig. 1A). In agar phantoms simulating the combined IMCL and EMCL pools, the CH 3 :CH 2 ratio correctly distinguished 4 different lipid compositions by their relative order (Fig. 1B, right). Bias due to factors including J-modulation, fitting prior knowledge, and T 2 effects likely account for the systematic difference between observed and theoretical values, so ratio values were used comparatively. Figure 1B and C show representative spectra (and the corresponding fits) acquired from a phantom and from SOL in vivo, respectively. The CH 3 signal from omega-3 FA resonates at different frequencies to the CH 3 resonance shown in Fig. 1, but as shown by biopsy, muscle TG contains very little (~1%) omega-3 lipids 21 and so the fitting models we used disregard this.
In vivo fasting effects. The 24 healthy male participants were of mean ± SEM age 34.9 ± 1.8 years and BMI 23.4 ± 0.5 kg.m −2 . Spectra with no clear distinction between EMCL and IMCL CH 2 resonances were eliminated, yielding 22 and 19 complete pre-and post-fasting sets for TA and SOL respectively. IMCL CH 2 relative to the internal standard creatine plus phosphocreatine (Cr) increased on extended fasting (Table 1), consistent with observations in longer fasts in human vastus lateralis 7 , TA and SOL 8 . FFA concentrations correlated with the fasting increment (Δ) in IMCL in both TA (ΔCH 2 : r = 0.57, p = 0.006; ΔCH 3 : r = 0.39, p = 0.07) and SOL (ΔCH 2 : r = 0.52, p = 0.02; ΔCH 3 : r = 0.47, p = 0.04), supporting the concept that FA influx into the cell controls IMCL pool size in these circumstances, either by directly altering the influx into the IMCL pool or indirectly by suppressing IMCL lipolysis. Subcutaneous adipose tissue (SCAT) decreased whilst other fat depots remained unchanged ( Table 1), suggesting that SCAT is the main source of increased FFA in these non-obese males. This is consistent with studies of regional FFA release in vivo in humans using isotope techniques 22 where upper body subcutaneous fat was by far the greatest contributor to systemic FFA release under basal conditions and during prolonged fasting. Direct validation studies would be needed to support this application of MRI in determining the dynamics of fat depots conveniently in vivo.
IMCL CH 3 /Cr (which unlike CH 2 /Cr, the increment in which purely reflects net FA flux) increased by ~12% in TA (p = 0.03) implying net FA influx during the fast, but not significantly in SOL. IMCL CH 2 /Cr increased by 26% in SOL (p = 0.007) and 31% in TA (p < 0.001) ( Table 1). The IMCL CH 2 /Cr increase was therefore more than the increase in CH 3 /Cr and thus represents significant decreases in the compositional marker CH 3 :CH 2 ( Table 1, Fig. 2A), indicating a change towards decreased unsaturation and possibly increased chain length of IMCL during fasting. These IMCL compositional changes on fasting correlated inversely with baseline IMCL composition in both muscles ( Fig. 2A), suggesting a selective efflux of unsaturated and/or shorter chain FAs from the IMCL pool during fasting. This matches the mobilisation profile in a rat biopsy study of muscle TG 14 , in which the reduction in CH 3 :CH 2 ratio was mainly due to the efflux of unsaturated and/or shorter chain FA. Supporting this notion, net efflux of FFA from the IMCL pool during the fast were associated with decreases in the compositional marker in both muscles (Fig. 2B). IMCL compositional changes were not associated with circulating FFA or IMCL ΔCH 2 /Cr in either muscle (all p > 0.23). Together our observations indicate that during the fast an efflux of FA with a high CH 3 :CH 2 ratio from the IMCL pool is the main determinant of compositional changes in IMCL .

Differences between muscles.
There was a striking difference in baseline IMCL CH 3 :CH 2 between muscles ( Table 1, Fig. 2C), TA having a more unsaturated shorter-chain TG pool than SOL. Modelling based on phantom results suggests that SOL lipid CH 3 :CH 2 is similar to or more saturated than olive oil, and TA similar to sunflower oil. A possible contribution to these differences may result from inconsistencies in spectral fitting between muscles due to slightly different EMCL resonant frequencies that depend on fibre orientations; however, this is unlikely to fully explain these differences. Biopsy TG samples processed by normal methods are prone to EMCL contamination, and there is little published data on the composition in these muscles; we cautiously suggest that such a difference in IMCL composition might explain reported differences in determinants of IMCL 23 .  The middle panel shows how introduction of a single double bond increases this ratio to 3/16 = 0.188, due not only to the desaturation of two CH 2 , but also to the alteration in the chemical environment of the neighbouring CH 2 . For FA chains of equal number of double bonds, the CH 3 :CH 2 ratio will also be scaled by chain length, although this will have a proportionally smaller effect (lower panel). SFA, saturated FA; MUFA, monounsaturated FA; PUFA, polyunsaturated FA. (B) Results in phantoms. The graph on the right of the panel shows data from agar phantoms simulating IMCL (oil droplets) and extramyocellular lipid (EMCL; oil-soaked tissue roll) made using olive, rapeseed, sunflower and cod liver oils. Three 1 H spectra were acquired from each phantom and the IMCL CH 3 :CH 2 ratio correctly distinguished the order of four oil phantoms. On the left of the panel is a representative 1 H spectrum (solid black) acquired from a voxel containing both simulated IMCL and EMCL compartments, also showing the overall fit (dashed grey) and individual fit components (solid grey).  Figure 3B shows the correlation with oleic acid concentration, the main circulating MUFA 24 . This suggests a differing net uptake of FA into IMCL during fasting, with the influx of FA with a higher CH 3 :CH 2 ratio into the TA muscle IMCL pool, this is a profile which matches the baseline composition. It is unknown whether this differing uptake is a result of selective FA transporters or of a differing FA profile remaining after short-chain and unsaturated FA are preferentially oxidised 13,25 . Summary and implications. We have shown by practical validation in IMCL/EMCL simulated phantoms, that the IMCL CH 3 :CH 2 ratio in good-quality spectra showing clear distinction between EMCL and IMCL CH 2 resonances, acquired at 3 T using the PRESS sequence with a short echo time of 35 ms is a comparative marker of TG composition. These acquisition parameters are commonly used, and therefore this method has potential for widespread use. Although this marker does not distinguish saturated from mono-or poly-unsaturated FA components of TG, as recently achieved using a specialised sequence 26 , it is particularly sensitive to differences in the degree of unsaturation and chain length. We stimulated lipolysis using an extended fast in healthy adult males and showed evidence of IMCL turnover, such that IMCL is synthesized from circulating FFA and hydrolysed causing a selective efflux of unsaturated and/or shorter chain FA. This combination of in vivo MRS measurements of IMCL CH 2 , CH 3 and CH 3 :CH 2 with measurements of plasma FFA has the potential to inform on both the composition and net dynamics of this metabolically-active pool. Although isotopic tracer studies such as pulse-chase approaches performed in vivo 27 may provide more definitive mechanistic information underlying the turnover of a specific FA species in intramuscular triglyceride, 1 H MRS offers the ability to detect the intramyocellular triglyceride pool non-invasively, allowing repeated measures from the same sample conveniently in a timeframe of a few minutes, permitting investigation of net changes of all FA species both within short-term and over longitudinal intervention studies. The 1 H MRS method outlined here, together with isotopic tracer studies, have the potential to better define the involvement of IMCL in insulin-resistant states and their treatments.

Methods
Phantom validation of IMCL compositional marker. Four agar phantoms simulating IMCL (oil droplets) and EMCL (oil-soaked tissue roll) 19 were made using lipids of differing fatty acid (FA) compositions viz. olive, rapeseed, sunflower and cod liver oils. These are composed of triglycerides which, on average, increase in degree of unsaturation from predominantly 18:1 (olive) to predominantly 18:2 (sunflower), to a mixture of highly unsaturated and shorter-chain FA (cod liver oil). The same oil was used in both droplet and tissue roll Participants and protocol. 24 healthy non-obese Caucasian male volunteers were recruited, aged 18 to 50 years without diabetes and taking no medication 28 . They were asked to refrain from alcohol and vigorous physical activity and to follow their normal diet for 3 days prior to the study. Their meal on the evening prior to the study    and fitted with the AMARES 31 algorithm using identical prior knowledge parameters: a single EMCL resonance, Gaussian lineshapes, soft constraints on EMCL/IMCL CH 2 frequencies and linewidths, CH 3 resonant frequencies and linewidths determined from known and inferred prior knowledge relative to the CH 2 resonance 5 , and with all amplitudes estimated. IMCL CH 2 and CH 3 are expressed relative to the creatine and phosphocreatine CH 3 resonance at 3.0 ppm.

Free Fatty Acids (FFA).
Plasma samples were extracted with MTBE using an automated liquid handler. This organic phase was diluted with a mixture of isopropanol/methanol with 7.5 mM NH 4 Ac solution. Lipid profiling was performed on the extract using chip-based nanoelectrospray with an Advion TriVersa Nanomate (Advion, Ithaca, USA) interfaced to the Thermo Exactive Orbitrap (Thermo Scientific, Hemel Hampstead, UK), using a mass acquisition window from 200 to 2000 m/z and acquisition in positive and negative mode, as described elsewhere 32 . Negative mode data were used to measure relative concentrations of free fatty acids (FFA), and absolute FFA concentrations were calculated by comparison to the added internal standard undecanoic acid. Circulating FFA composition was reported from plasma samples at 12, 16 and 24 h fast, combined as the AUC.
Adipose tissue. Abdominal subcutaneous (SCAT abd ) and visceral (VAT) adipose tissue volumes were measured using magnetic resonance imaging over 6 cm superior-inferior distance from 5 water-suppressed T 1 -weighted transaxial slices above the L5 vertebral level. Due to the higher reproducibility of leg subcutaneous adipose tissue (SCAT leg ) and intermuscular fat (IMF) (2 or 3% CoVs), these measures were assessed from a single T 1 -weighted transaxial slice central to the 1 H MRS voxels, and hence these measures have units of area (cm 2 ). Regional MRI analysis was performed in Analyze (AnalyzeDirect, Overland Park, KS).
Intrahepatic lipid (IHL). IHL was measured by 1 H MRS as previously described 33 , and expressed as the methylene peak at 1.3 ppm relative to the water peak.
Statistics. All statistical tests were performed in IBM SPSS Statistics 23 (IBM, Armonk, NY: IBM Corp.), with significance set at P < 0.05. Normality was assessed by the Shapiro-Wilk test and non-normally distributed data were log-transformed prior to statistical testing. A 2-tailed paired-samples t-test was used to compare differences between 8 h and 28 h of fasting and Pearson's correlation coefficient for analyzing relations. Data are mean ± SEM.
Study approval. The study was approved by the Cambridge Local Research Ethics Committee and conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent.
Data Availability. The datasets generated during the current study are available from the corresponding author on reasonable request.