MAGNETIC RESONANCE SPECTROSCOPY (MRS) of skeletal muscle has preceded other applications of human MRS in vivo (1–7). Initially, sufficient magnetic field strength for observing spectra with a good spectral resolution and reasonable signal-to-noise ratio (SNR) was available only in small-bore magnets. Although these instruments were appropriate for extremities, they did not provide sufficient space for whole-body examinations of humans. Gradient-based volume selection was introduced in spectroscopy and applied in human skeletal muscle in the early 1980s (6). However, these gradient systems were not as refined as modern shielded gradient systems, and thus eddy currents produced artifacts that were deleterious for spectral quality in many MR systems. It was advantageous that volume selection was (or at least seemed to be) less crucial in skeletal muscle than it was in other tissue. Therefore, a relatively coarse volume selection by surface coils (2) combined with free-induction decay (FID) sequences was sufficient for many applications. For some metabolites with extremely short transversal relaxation times (T2), such as glycogen or deoxymyoglobin, a combination of surface coils and FID acquisition remains the method of choice since the acquisition has to start as soon as possible after the generation of the transversal magnetization (8, 9). Thus, multiecho, gradient-based, volume-selection sequences, such as point-resolved spectroscopy (PRESS) (10) and stimulated echo acquisition mode (STEAM) (11), were not widely used in MRS of skeletal muscle for a rather long time. During that period of time, FID sequences with transmit-receive surface coils were preferred for heteronuclear MRS of human muscle.
For over a decade, 31P-MRS was the workhorse for muscle physiologists who used this method successfully, particularly for studies of high-energy phosphates and intracellular pH (3–5, 12–23). Figure 1 shows a 31P-MR spectrum of human skeletal muscle together with 1H- and 13C-MR spectra to illustrate the large number of metabolites that can be observed. High-energy phosphates and pH are important indicators for muscular metabolism, especially for muscular bioenergetics in the time range of seconds and minutes. The application of 13C-MRS for examinations of muscular metabolism attracted the attention of physiologists, since glycogen and metabolites of the tricarboxylic acid (TCA) cycle became accessible in 13C-MR spectra (24–31). Since 13C-MRS and 31P-MRS work at different frequencies compared to 1H-MRI and -MRS, standard clinical MR scanners without broadband capabilities (transmitters, coils, and receivers) could not be used for this type of study. Even more demanding is a second channel for 1H-decoupling and nuclear Overhauser enhancement (NOE) (8, 9, 32–34), which is necessary for 13C-MRS and also beneficial for 31P-MRS. Since this equipment requires extra investment and knowledge, radiology departments with standard MRI scanners were not able to perform this kind of physiological study even when collaboration with muscle physiologists was possible. Therefore, many key applications of heteronuclear musculoskeletal MRS were not carried out at clinical MR sites. This may partially explain why MRS of skeletal muscle was not as widely used as initially expected.
1H-MRS of skeletal muscle ameliorated this disadvantage for clinical MR sites to some extent because it is possible to observe intra- (IMCL) and extramyocellular lipids (EMCL) on many routine MR systems with ordinary spectroscopy packages. However, 1H-MRS of skeletal muscle has been ignored for a long time, even though 1H is the most frequently used nucleus in brain examinations. Two large peaks of water and fat were expected to hide all important features of metabolism in skeletal muscle, and only a few research groups reported 1H-MR spectra of skeletal muscle (31, 35–40). Reliable water suppression, improved gradient performance with reduced eddy currents, clean volume selection, and very careful positioning of the volume of interest (VOI) made relevant and partially unexpected features of the 1H-MR spectrum visible (41, 42). In particular, its ability to observe IMCL (43, 44) makes this method a very valuable tool for physiology and endocrinology.
Very promising applications of MRS in skeletal muscle result from multinuclear MRS, e.g., combinations of 1H-MRS for the detection of muscular lipids with 31P-MRS for the observation of phosphorylation, and/or with 13C-MRS for the observation of glycogen or the determination of TCA cycle activity (45–56). Such applications require broadband MR equipment, which is not widely available. On the other hand, modern clinical MR scanners would allow spectroscopy to be combined with all kinds of imaging methods, such as morphological methods, angiography, perfusion and blood oxygenation level-dependent (BOLD) studies, arterial spin labeling (ASL), and determination of fiber orientation (8, 57). However, this very beneficial combination of MRI and MRS is often obtained at the price of reduced X-nuclei capability, which limits 31P-MRS and particularly 13C-MRS to dedicated research systems. Other nuclei, such as 23Na, have not often been used to study muscle metabolism (58, 59). While uncommon nuclei (60) have been used in cell suspensions, isolated organs, and even human brain, applications in human skeletal muscle are rare.
As a result of these technical and methodological limitations, MRS of skeletal muscle is used more commonly to elucidate physiological and pathophysiological mechanisms than to obtain radiological diagnoses in individual patients. It remains to be seen whether the numerous ultra-high-field magnets that are currently being installed will change the situation and promote sophisticated MRS studies of skeletal muscle in clinics.
COMMONLY USED TECHNIQUES FOR MUSCLE SPECTROSCOPY
Since skeletal muscle is often close to the surface and (at least at first glance) relatively homogeneous, the application of transmit/receive surface coils (2) for volume selection is tempting, particularly for the heteronuclei 31P and 13C. In this approach, the selection of the signal-producing volume is achieved by the sensitivity profile of the surface coil, and thus no additional gradient encoding of the signal is necessary. While gradient-based volume selection requires some time for the application of the gradients, the acquisition of the FID in combination with surface coils can start immediately after excitation. This is particularly advantageous for nuclei and metabolites with short T2, such as glycogen or deoxymyoglobin.
A major disadvantage of this method arises if absolute concentrations of metabolites should be determined (8, 61–70). Transmit/receive surface coils define an extremely complex shape of the volume that generates the signal. Even if adiabatic pulses are used for excitation, the weighting of the signals from different regions within the sensitive volume is very complicated. Unlike 1H-MRS, which can use the water signal as a relatively stable reference, the spectra of heteronuclei rarely contain a similar signal that can be used as an internal standard. The 13C-signal from creatine in muscle is an exception since the creatine concentration is constant enough for creatine to be used for absolute quantitation (71). However, the preferred application of this internal standard would be to determine absolute glycogen concentrations, which are typically acquired at very short repetition times (TRs). This requires either a lengthening of TR, resulting in an inefficient acquisition for glycogen, and/or compensation for the substantial saturation of the creatine resonance. Other methods of absolute quantitation involve replacing the muscle with a known solution in a container of similar shape, or employ little vials filled with a substance that resonates outside of the interesting part of the spectrum and can be used in connection with calculations of the sensitive volume.
While volume selection is of limited importance for the observation of heteronuclei, it is absolutely crucial for 1H-MRS of skeletal muscle (42, 44, 72, 73). The large contribution of subcutaneous fat makes it impossible to acquire reasonable 1H-spectra without careful positioning and excellent suppression of outer-volume signals. Volume selection can be achieved by single-voxel techniques, such as PRESS or STEAM sequences, or by chemical shift imaging (CSI). For single-voxel techniques, the use of regular sequences from the manufacturers in combination with very careful positioning of the voxel in regions without any visible fat in the images is sufficient to obtain spectra of reasonable quality, at least in slim volunteers. Additional outer-volume suppression (OVS) allows the voxel to be placed even closer to the subcutaneous fat in patients with very slim muscles. Unfortunately, it is extremely difficult to obtain spectra from the most important patients for these kind of studies: obese patients. Marbling of the muscle with fatty infiltration makes it almost impossible to obtain spectra without a large EMCL resonance that contaminates or even hides the IMCL peak of interest.
The limited resolution of CSI grids (74–77) and subsequent bleeding of the strong signals from subcutaneous fat due to the point-spread function (PSF) (8) into other parts of the grid is a crucial problem. One solution is to acquire large CSI data sets with very high spatial resolution, and another is to presaturate all fat-containing regions in the preselected CSI volume, including subcutaneous fat and bone marrow. Residual artifacts from strong fat signals can be removed by a lipid extrapolation procedure (75, 78).
1H-decoupling of 13C-MR spectra (27, 30, 79–84) is essential for improving the SNR of the tiny metabolite signals and simplifying the spectra. Since the lipid resonances of the subcutaneous fat are stronger by orders of magnitude than the signals from glycogen and other metabolites, it is crucial to remove the coupling of the large resonances, which results in the decoupled peaks moving away from the signals of interest. This helps to clean the baseline in the region of the metabolite signals. The irradiation of the protons coupled to the 13C-nuclei has an additional positive effect on the SNR since these signals are increased by the nuclear Overhauser effect (NOE). This very advantageous effect, however, creates additional problems for the calculation of absolute concentrations and has to be considered separately.
1H-decoupling of 31P-MR spectra (85–88) is less common than in 13C-MRS, mainly because 1) the inherent SNR of 31P-MRS is already higher than that of 13C-MRS even without decoupling, and 2) protons are not directly coupled to high-energy phosphates, and thus decoupling does not result in a collapse of multiplets into a singlet. Nevertheless, the cancellation of long-range couplings by [1H]–31P-decoupling leads to a better separation of the signals and to an additional NOE that helps to increase SNR. Therefore, decoupling of 31P-MR spectra has become rather common in MR systems that are equipped with a second channel.
Advanced techniques, such as magnetization transfer (MT) (89, 90) and diffusion measurements of metabolites (91, 92), have been used successfully in recent years; however, a detailed technical description of these applications would go beyond the scope of this review.
STRUCTURAL ASPECTS OF SKELETAL MUSCLE THAT INFLUENCE MR SPECTRA
Since some structural peculiarities of skeletal muscle influence MR spectra directly, they are listed here briefly. More extensive descriptions can be found elsewhere (57, 72).
Skeletal muscle is structurally organized on various levels, beginning with the organization and compartmentation of cell organelles, the striated structure of various types of muscle fibers, the orientation of muscle fibers, and the alignment of bulk fat along fasciae. While other organs, such as the brain, obviously have a very complex organization, they lack the structural symmetry that is typical of skeletal muscle. Since the MR spectra are influenced by this structural order and compartmentation, it is impossible to understand the spectral information without considering these effects.
In particular, the major effects of the structural organization of skeletal muscle can be observed in 1H-MR spectra, as discussed in more detail later in the text. Since the susceptibility of the laminar structures of EMCL (bulk fat) is orientation-dependent, these signals can be separated from IMCL that is stored in spherical droplets (43, 44). Despite the fact that the chemical nature of these resonances (methyl and methylene groups) is identical between the two lipid forms, the resonance frequency is slightly different. A completely different effect can be found in creatine resonances, where anisotropic motional averaging results in residual dipolar coupling effects (42, 93). In addition, taurine, carnosine, lactate, and other metabolites show similar effects in skeletal muscle (94–99). Since residual dipolar coupling is strictly orientation-dependent, it is possible to measure fiber orientation based on this effect (100). Figure 2 illustrates how much structural aspects of skeletal muscle can influence MR spectra, particularly in 1H-MRS. During exercise, the creatine signals in 1H-MR spectra reveal another unexpected peculiarity. It seems that these signals disappear during exercise despite the fact that “total creatine” (the sum of creatine and phosphocreatine) should remain constant (101, 102). This observation is not yet clearly understood, since experiments in creatine-kinase knock-out mice have produced different results (99, 103). However, limited visibility or at least compartmentation of different pools of metabolites appears to be a consequence of structural effects and has also been reported for other metabolites (104). MT experiments (89, 90) and observations of restricted diffusion (91, 92) also indicate that structural order influences the results of MRS sequences.
A particularly sensitive approach for detecting anisotropic and ordered compartments is the use of 23Na double quantum filtered (DQF) (58, 59, 105) and 1H-DQF (104, 106) MR spectra. Compartmentation of skeletal muscle can also lead to a transport phenomenon that can be observed by MRS (107).
The value of 31P-MRS had already been recognized for studies of metabolism in isolated organs and cells (1) before the introduction of surface coils allowed the localized observation of MR signals in whole animals (2). Soon after, muscular disease was investigated noninvasively for the first time in a human patient with McArdle's disease (3). Reports from different research groups on phosphofructokinase-deficient human skeletal muscle soon followed (4, 5). A combination of surface coils with gradients was used to separate the signals from different regions in rat skeletal muscle (6), and was therefore one of the first applications of CSI principles in vivo.
31P-MS became popular among muscle physiologists because it revealed major factors of muscle metabolism noninvasively. In parallel, it was expected that the method would be increasingly used in clinical routine for the diagnosis of individual patients. While some groups used this tool regularly to investigate muscular diseases (108–114), the distribution of 31P-MS in clinical routine was not widespread as expected, and 31P-MRS remained to some extent an expert tool for research in physiology and pathology.
Well-known resonances in the 31P-MR spectrum include the high-energy phosphates (Figs. 1 and 3), which are involved in the creatine kinase reaction:
Adenosine triphosphate (ATP) is restored at the cost of phosphocreatine (PCr) by this reaction, which is close to equilibrium in resting muscle. If ATP is degraded to adenosine diphosphate (ADP) during workload, the amount of PCr is reduced and the net reaction
leads to an increase of the inorganic phosphate (Pi) signal.
Since the protonated and unprotonated forms of inorganic phosphate H2PO and HPO have two different resonance frequencies and are in a fast equilibrium
the Pi resonance titrates with a pK value of 6.77 (pK is the −log10 of the equilibrium constant and represents the center of the titration curve), i.e., it moves continuously from 3.23 ppm to 5.70 ppm depending on the pH. This allows a rather accurate determination of the intracellular pH noninvasively. This information is often used to study muscle physiology, in particular the influence of H+ on biochemical reactions and parameters such as fatigue, performance, and glycolysis (115–128).
It has been noted that the Pi resonance may split during exercise, indicating that the pH in the sensitive volume is heterogeneous (12, 129–132). While some authors found that different fiber types in the sensitive volume were responsible for this heterogeneity, others questioned this explanation for the splitting of the Pi resonance.
An analogous titration curve can be found for the resonances of ATP, which depend on the concentration of Mg++ according to the equilibrium:
Since the interaction with Mg shifts the ATP resonances (γ > β > α), this titration allows for an estimation of the magnesium concentration in skeletal muscle (61, 133–137).
It is noteworthy that even concentrations of undetectable metabolites, such as ADP, can be determined if all the other partners in a chemical equilibrium are accessible. Based on the measurement of creatine and phosphocreatine concentrations, the determination of the pH, and an estimation of the equilibrium constant, it is possible to determine ADP indirectly (18, 138–140).
One of the most popular applications of 31P-MRS of human skeletal muscle is to determine oxidative capacity, which is related to high-energy phosphates (99, 113, 141, 142). Well-trained athletes with a higher oxidative capacity show higher PCr/Pi ratios at a given workload (12, 14, 16), and a faster recovery of PCr after exercise (14, 143). PCr recovery can be described by a monoexponential curve (13, 22, 113, 144–147) as long as the workload is within a range that depletes PCr sufficiently and the pH remains above ∼6.75. As compared to the effect of workload on the ratio PCr/Pi, the measurement of the PCr recovery has the advantage that it is independent from the applied workload (23, 144). It is difficult to determine the effective workload in a magnet because 1) magnetic field and radiofrequency (RF) pulses jeopardize force measurements within the magnet; 2) if the power is mechanically transmitted outside the magnet, the losses are largely unknown; and 3) the recruitment of muscle groups is heterogeneous and can change during an experiment, particularly since the restricted motion in the magnet is unfamiliar to the volunteer.
While the recovery of PCr after workload is independent from the applied workload and would therefore be a rather robust parameter, it is not independent of the reperfusion of the limb (21, 148–154). This is a major limitation of this kind of experiment and leads to large variations of the relation between oxidative capacity and PCr recovery in humans.
It has been suggested that oral intake of creatine would increase the availability of high-energy phosphates during exercise and thus lead to higher performance. It is tempting to use 31P-MRS to follow the reaction of muscle metabolism to creatine supplementation (144, 155–158). Figure 3 illustrates the effect of oral creatine supplementation on the PCr/ATP ratio during workload and subsequent recovery (144). While a slightly higher performance can be shown in some of the studies, it has also been reported that the recovery rate is somewhat slowed down, since a larger amount of available creatine (Cr) has to be transformed to PCr by an equal number of creatine kinase molecules. The absolute concentrations, however, remain higher during workload and recovery, and thus could explain the increase in performance.
Proton decoupling during acquisition of phosphorus spectra was not used for a long period of time, since it does not have the same dramatic decoupling effects found in 13C-MRS. In 31P-MRS, long-range couplings are canceled, which leads to smaller lines and cleaner spectra. In particular, decoupling allows signals such as glycero-phosphoryl-choline and glycero-phosphoryl-ethanolamine to be distinguished, which may help to improve our understanding of membrane metabolism (31, 85, 86, 88). In addition, the observation of glucose-6-phosphate is improved by proton-decoupled 31P-MRS (159). Investigations of glucose-6-phosphate have elucidated the transport mechanisms and metabolism of glucose in skeletal muscle (46, 160–162).
In principle, MR spectra can follow metabolic changes with a temporal resolution that is defined by the TR if single spectra are analyzed. If the SNR requires more acquisitions, this temporal resolution is reduced accordingly. However, using a technique called “saturation transfer” (ST), one can observe reactions that are orders of magnitude faster than the time between two acquisitions, e.g., in the creatine kinase reaction (89, 90, 163–165). If a nucleus is involved in a fast exchange reaction (9), it may be saturated in one chemical status and transfer this saturation (or “magnetization”) to the new chemical status. As an example, if a phosphate group in the γ position of the ATP molecule is irradiated, the exchange with the phosphate groups in PCr is so fast that the saturation survives the chemical exchange and, as a consequence, the signal of PCr is also partially saturated, depending on the exchange rate. Such observations of equilibria on the time scale of chemical reactions are very attractive. In addition, the recent development of genetically manipulated mice (e.g., creatine-kinase knock-out mice) may help to elucidate many aspects of 31P-MRS in more detail (103).
Various techniques (166–170) have been proposed to increase the spatial resolution of 31P-MR spectra without exceeding the acquisition times necessary to observe changes of high-energy phosphates during exercise.
Together with MT techniques, diffusion-weighted (DW) 31P-MRS of PCr and Pi may provide more information about the creatine-kinase reaction and the role and nature of creatine/phosphocreatine as energy shuttle, reservoir, or buffer in the myocyte (15, 92, 171). While DW imaging (DWI) is widely and very successfully used, unfortunately diffusion MRS is applied by only a few expert groups, even though it provides a unique insight into molecular motion and the processes of metabolites.
The investigation of muscle pathology (109) has led to many clinically oriented 31P-MRS studies of skeletal muscle in McArdle's disease (3, 17, 19, 124), glycogen storage disease type XIII (172), phosphofructokinase deficiency (4, 5), adenylosuccinate lyase deficiency (173), dermatomyositis (174, 175), fibromyalgia (176), muscular dystrophy (177), Friedreich ataxia (178), poliomyelitis (179), mitochondrial disease (180–182), Parkinson's disease (183, 184), peripheral arterial disease (PAD) (150, 185), severe burn trauma (186), cardiomyopathy (187), type 2 diabetes (188), malignant hyperthermia (189), and aging (53, 190, 191).
While 31P-MRS has been the MR workhorse in muscle physiology for more than a decade, 1H-MRS took over this function to a certain extent in the late 1990s, as discussed below. One may speculate that 31P-MRS will regain its crucial role in musculoskeletal spectroscopy when it is combined in multinuclear studies. Together with 1H-MRS for lipid metabolism and 13C-MRS for glycogen, 31P-MRS would be extremely valuable for characterizing mitochondrial function and integrity.
Despite the enormous potential of 13C-MRS for studies of muscle physiology (192–195), this nucleus has been observed by only a few expert groups. This is mainly because of the technical and methodological complexity of this approach, since 13C-MRS requires extra equipment and knowledge depending on the type of study conducted.
The MR scanner has to be equipped with broadband transmitters and receivers in order to emit and detect frequencies that are one-quarter that of 1H. Many clinical MR scanners are equipped for 1H only.
Decoupling and NOE-build-up (Fig. 1) are crucial for 13C-MRS in order to overcome the inherently low SNR (8, 9, 30, 31, 79–82, 84, 196). To decouple the 13C spectrum, one must use a second frequency channel, which is not available in the vast majority of clinical MR scanners. As an alternative, sequential transmission of 13C and 1H frequencies is possible; however, this requires switching of the frequency and the RF path within a few microseconds.
Depending on the decoupling scheme used, as discussed above, the RF coils must have either two circuits (one tuned to the 13C frequency and the other to 1H) or a coil, which has to be double-tuned (one circuit with two resonance modes). Typically, surface coils are used for transmission and reception, for two reasons: 1) volume selection in 13C-MRS is often based on the sensitive volume of the surface coil, and 2) no manufacturer builds a body coil tuned to the 13C-frequency, even though a homogeneous excitation would be desirable for easier quantitation of the metabolites.
The SNR of 13C-MRS is inherently low due to the lower resonance frequency that induces less voltage in a detector (8, 9). In particular, the observation of naturally abundant 13C (1.1% of all carbons) suffers from this low SNR, and therefore an increased sensitivity at very high magnetic field strength is desirable. While it is possible to obtain reasonable 13C-MR spectra at 1.5 Tesla, an increased SNR at higher field strength would be advantageous. The recent installation of many clinical 3 Tesla systems may help 13C-MRS in that respect.
As long as naturally abundant 13C is observed, and the spectra are acquired by pulse-and-acquire techniques under relaxed conditions with [1H]-decoupling, the resonance area still does not represent the concentration of a metabolite since the NOE (8, 9) has to be determined and considered.
It is obvious that the 1.1% natural abundance of 13C primarily is primarily a disadvantage because it results in a poor SNR of natural-abundance 13C-spectra. On the other hand, the replacement of 12C- by 13C-nuclei (i.e., “13C-enrichment”) allows a substantial signal gain of specific carbon positions in a molecule. If such a 13C-enriched substrate is infused, the additional signal strength can be followed. However, since a signal increase can be a result of the increasing metabolite concentration or the increasing labeling, it is necessary to determine the degree of isotope labeling and to use mathematical models to resolve the additional complexity of the signal changes. Using this technique, one can follow the specific label in the metabolic pathways and identify newly synthesized metabolites even if a specific metabolite is naturally present in the tissue. One of the most popular applications is labeling of glucose at the C1 position and follow-up by 13C-MRS during its incorporation into muscular glycogen in clamp studies (26, 197). Another very successful application is isotope enrichment by labeled acetate and investigation of the TCA cycle flux by monitoring the 13C incorporation into glutamate (53).
A very sophisticated means of looking at metabolic pathways is isotopomer analysis (198–201), i.e., the analysis of the homonuclear coupling patterns of adjacent 13C-13C in mixtures of the same compound with different degrees of labeling. One example in skeletal muscle is the observation of [1-13C]glucose label precursor incorporation into intramuscular [1-13C]glycogen, [3-13C]lactate, and [3-13C]alanine (202). Since such studies are usually performed at very high fields in order to separate the different coupling patterns, most have been performed in animals, perfused organs, and plasma samples (202, 202–207).
The problems of 13C-MRS include volume selection and absolute quantitation. Due to the low SNR, large volumes are usually needed to generate a reasonable signal, which makes it impossible to examine small anatomical structures. Figure 1 illustrates how a typical sensitive volume for 13C-MRS covers multiple muscles while 1H-MR spectra (thanks to other volume selection schemes and to smaller sensitive volumes) can be obtained from a much better defined part of a specific muscle. The use of surface coils for volume selection also complicates absolute quantitation. To calibrate the metabolite levels, bottles filled with known solutions that simulate conditions in the human body can be measured (208), or creatine can be used as an internal standard (28).
13C-MR spectra have an exceptionally large chemical dispersion of about 200 ppm (Fig. 1). If gradients are used for spatial encoding of such spectra, the chemical shift displacement is extremely large, which makes ordinary gradient encoding almost impossible for 13C-MRS. In addition, volume selection based on echoes is not feasible for some of the most interesting metabolites, such as glycogen, that can be observed by 13C-MRS due to the short T2 values. The use of image-selected in vivo spectroscopy (ISIS) overcomes the problem of fast transverse relaxation since the magnetization is kept in the longitudinal direction during the localization procedure (209). To limit the effect of the large chemical shift dispersion, localization of 13C nuclei can also be accomplished indirectly via the coupled 1H nuclei and subsequent polarization transfer (30, 210–212). With this method, the smaller chemical shift range and thus smaller chemical shift displacement of 1H allows much better definition of the selected volume.
Improved volume selection resulting in a reduction of subcutaneous signals would be desirable since the metabolites (e.g., glycogen or glucose) are tiny signals between huge contributions from subcutaneous fat. In addition, the ability to separate the 13C-signals from subcutaneous fat and IMCL would be very attractive, since 13C-MRS allows for investigations of the degree of saturation and chain length (71). This would give important insights into the specific metabolism of IMCL (24, 213). However, signals from subcutaneous adipose tissue and IMCL cannot be separated by current volume selection methods, which limits 13C-MRS to studies of bulk fat (24, 25) without further differentiation of subcutaneous fat and IMCL.
Many applications of 13C-MRS in human skeletal muscle observe glycogen, which represents (together with IMCL) the main energy storage within muscle. It is surprising that glycogen is 100% visible in 13C-MRS (29, 197), because it has a very high molecular weight (107–109). Large molecules with a long correlation time have very short relaxation times and thus cannot be observed by the usual MR techniques that are used in solution or tissue. Obviously, the internal molecular mobility leads to shorter correlation times and thus to longer relaxation times. Applications involving the observation of glycogen in human skeletal muscle have been used in both physiological and pathophysiological studies. In particular, glycogen levels in skeletal muscle have been investigated in exercising subjects (45, 195, 214–217), as an effect of diet (50, 208, 218–226), and in patients with McArdle's disease (28), glycogen type IIIA storage disease (27), acid maltase deficiency (227), and non-insulin-dependent diabetes mellitus (NIDDM), as well as the offspring of such patients (47, 48, 228).
A misunderstanding arises surprisingly often when stable isotopes methods are considered: The sensitivity for variations in the isotope enrichment of 13C-MRS is far below that of a 13C-isotope determination by means of gas chromatography, mass spectrometry, or infrared spectrometry (229). While the latter methods can detect differences in isotope ratios of ≤0.1%, 13C-MRS requires an isotope enrichment of typically ≥10% to be accurate for the smaller resonances. In other words, 13C-MRS is not influenced by naturally occurring differences in isotope enrichment, such as the use of corn from different continents. The strength of 13C-MRS is not its sensitivity, but the organ selectivity and chemical specificity. Other, much more sensitive methods mentioned above typically analyze the substances in secreted or expired material, such as breathing air or urine, which represent whole-body averages without further localization.
1H-MRS of skeletal muscle has been neglected for quite a long time, even though 31P- and 13C-MRS have been successfully applied in physiology. The first observations of 1H-MR spectra showed two large peaks of water and lipids that covered the resonances of other metabolites. While special acquisition and editing sequences showed the potential of 1H-MRS (e.g., for deoxymyoglobin (230–235) or lactate (36, 104, 106, 236–238)), it was not believed that it would gain broad acceptance. High-resolution spectra in excised muscles and animals led to many assignments of 1H-MRS resonances (7, 35–38, 74). However, there have been only a few applications in humans and, in particular, two extremely characteristic features of 1H-MR spectra in vivo—susceptibility effects visible in the lipid region, and residual dipolar coupling of many other resonances—have been ignored.
In order to interpret 1H-MR spectra it is crucial to understand these two effects, which characterize the spectra and are based on entirely different physical phenomena. Schick et al (43) observed two separate groups of resonances for lipids in voxels of skeletal muscle, shifted by approximately 0.2 ppm, despite the fact that both compartments contain fatty acids or triglycerides of similar compositions. Boesch et al (44) were able to explain the physical background of this effect by tilting the muscles relative to the static magnetic field. These authors introduced the terms IMCL and EMCL, and identified the two compartments, based on physical effects, as lipid droplets close to the mitochondria in the cytosol of the muscle cell (IMCL) and as fatty infiltrations in the shape of plates and tubes (EMCL). This explanation was later confirmed by Szczepaniak et al (239) in a model system of Intralipid™ and soybean oil in vitro. It is important to understand that the bulk magnetic susceptibility of the geometrical arrangement of EMCL causes an additional chemical shift. This shift has nothing to do with coupling effects that can also be observed in 1H-MR spectra, even if both effects are orientation-dependent. While chemical shifts are introduced on the macroscopic scale, residual dipolar coupling is an effect that occurs within molecules. Residual dipolar coupling can be canceled at the so-called “magic angle” (MA) of 54.7° (93), while bulk magnetic susceptibility shifts depend on the exact geometrical superposition of magnetic fields. Since residual dipolar coupling is based on incomplete averaging of magnetic fields induced in molecules, it leads to a splitting of resonance lines in a manner similar to J-coupling, which is well known in several compounds (e.g., the doublet of lactate). This splitting of the resonance lines without a shift of the center can be used to demonstrate the effect in two-dimensional spectroscopy (42). Unlike coupling effects, bulk magnetic susceptibility shifts of lipids cannot be used for editing. This is unfortunate because it is often difficult to separate IMCL and EMCL in obese patients, as discussed below.
The potential for observing IMCL noninvasively was soon recognized by exercise physiologists. Early studies of IMCL (44) pointed out that IMCL levels vary within and between individuals. In the first reports on IMCL observed by 1H-MRS, it became clear that IMCL is a metabolically very active pool that can be recruited by the skeletal muscle during exercise. Figure 4 illustrates how the human organism uses IMCL during strenuous exercise (55). Pre- and post-marathon spectra show a distinct reduction of IMCL during the activity. The difference of the two spectra reveals an almost pure IMCL spectrum. The more or less identical EMCL signals pre- and post-exercise should not be interpreted physiologically; rather, they show that the placement of the voxel was identical and resulted in a constant contamination with EMCL. Thanks to the relatively good accuracy of the measurement (about 6%) for intra- and interindividual comparisons, it was speculated that 1H-MRS could largely replace biopsy. However, these applications in exercise physiology would not have been sufficient to make 1H-MRS generally known and applied. The observation of increased IMCL in insulin-resistant subjects (240–244) supported the hypothesis that IMCL may play a role in the pathogenesis of skeletal muscle insulin resistance and type 2 diabetes mellitus. The possibility of studying this important relation noninvasively and with high accuracy boosted the interest in 1H-MRS studies on IMCL and the number of subsequent publications considerably (55, 245–250). For a while, it was anticipated that 1H-MRS would provide a simple measurement of insulin sensitivity in a single examination, in contrast to a glucose clamp, which takes almost a day. However, a number of reports showed that IMCL levels were dependent on many other factors, such as diet, exercise, and muscle type. In addition, it was found that IMCL levels were not directly correlated with insulin sensitivity in all groups of patients and volunteers, since a paradoxical situation of high insulin sensitivity and high IMCL levels was found in athletes (54, 251, 252). Several reviews regarding the different aspects of IMCL methodology and applications have been published (248, 253, 254).
Another metabolite that is involved in lipid metabolism can be observed in 1H-MR spectra after heavy workload, when a peak at 2.13 ppm can be assigned to the acetylgroup of acetylcarnitine (255), which is typically missing in spectra from resting muscle (Fig. 5). Carnitine is involved in the transport of fatty acids into mitochondria for β oxidation, and is important as a buffering system for a potential surplus of acetyl groups. Since the nonacetylated form of the cofactor is needed in the TCA cycle and in pyruvate dehydrogenation, acetyl groups are exported in the form of acetylcarnitine out of the mitochondria in order to keep the different forms of coenzyme A (CoA), in particular the ratio acetyl-CoA/CoA-SH, balanced. The observation of acetylcarnitine by means of 1H-MRS has not yet been widely used in physiology.
While IMCL is metabolized on the order of hours, three physiological metabolites that vary within a few seconds can also be observed by 1H-MRS: lactate, deoxymyoglobin, and hydrogen ions (pH, detected by the titration of the histidine resonances in carnosine). All three parameters are highly relevant for muscle physiology; however, they require more than the usual acquisition of spectra between 1 ppm and 4.5 ppm. A moderate improvement in common acquisition techniques that would allow the observation of the aromatic signals at the left-hand side of the water signal would make the aromatic resonances from carnosine visible. While pH is usually determined by the frequency difference between inorganic phosphate and phosphocreatine in the 31P-MR spectrum, the aromatic signals of histidine in the carnosine molecule observed in 1H-MR spectra can be used to determine the pH by 1H-MRS (36, 256). A comparison of 31P- and 1H-MRS showed a good agreement between the two methods (36). Since 1H-MR spectra in human muscle typically can be observed in voxels of about 2 mL, the spatial resolution is significantly better than that of the surface-coil-selected volumes in 31P-MRS. However, since more acquisitions are necessary, the temporal resolution is not as high as that of 31P-MRS, in which the large volume allows the pH to be determined with a temporal resolution of typically one or two scans (i.e., within seconds). The signal of the methyl group of lactate is typically not visible because it is hidden underneath much larger signals from lipids in human muscle. Special editing techniques that make use of hetero- or homonuclear spin-spin interactions, including editing techniques by inversion and decoupling (36), zero-quantum filters, or DQF (104, 106, 236–238, 257), can detect lactate. Since lactate is one of the most relevant metabolites in muscle physiology, it is not surprising that these techniques were developed and applied by expert groups in the early phase of 1H-MRS in human muscle. They are not more widely used, however, because they require special sequences that are not available on typical whole-body scanners. A similar fate can be observed for deoxymyoglobin, which allows for an estimation of the degree of intracellular oxygenation during exercise or ischemia. At very high magnetic field strength, the γ-methyl group of Val-E11 can be observed at −2.4 ppm (258); however, this signal from oxymyoglobin cannot be detected at moderate field strengths due to massive overlapping by other resonances. In ischemic muscle, the resonance of the F8 proximal histidyl Nδ proton shifts to 78 ppm (230–235) and thus becomes visible even at moderate field strengths if the acquisition parameters are optimized accordingly. Since the histidine resonance has dramatically shortened T2 relaxation times, it is important to use localization schemes without long delays. Figure 6 illustrates how 1H-MRS of deoxymyoglobin was used to monitor successful treatment of a patient with PAD by multiple applications of plasmid DNA encoding the vascular endothelial growth factor (pVEGF)-C gene (235). Deoxymyoglobin was measured at two different locations of the lower leg during occlusion by an external cuff and subsequent reperfusion. In distal and proximal locations, an examination after multiple applications of pVEGF-C revealed clear shortening of the recovery time due to improved perfusion compared to an earlier examination just after a single dose of pVEGF-C.
1H-MR spectra from a carefully localized voxel with a more or less uniform orientation of the muscle fibers show features that could not be explained by the concept of solution spectra (41, 42, 93, 101, 102). While these features were first observed for methyl and methylene resonances of creatine/phosphocreatine, it became clear in subsequent studies that many metabolites in skeletal muscle experience these effects (94–99). At first, a particular doublet at 3.96 ppm (the chemical shift position of the creatine methylene group) could not be explained with the conventional view of in vivo MRS, since one would expect a singlet from that metabolite. It was assumed that tissue in vivo represents a solution, and in particular that molecules would be mobile with sufficient freedom to average all dipolar couplings. Other molecules, such as membrane proteins, were expected to be in a kind of solid state and therefore invisible for common acquisition sequences. In a series of experiments involving creatine loading (41), rotation of the muscle relative to the main magnetic field (93), and various one- and two-dimensional spectroscopic sequences (42), it was shown that in fact this doublet at 3.93 ppm represents creatine and/or phosphocreatine with residual dipolar coupling, and that other resonances (e.g., the methyl group of creatine at 3.03 ppm) also show dipolar splitting. This explanation was later confirmed at higher field strength in animals (101). It seems that creatine and other metabolites cannot tumble freely such that all directions relative to the main magnetic field are equally probable. This limitation of the motion is extremely weak because the coupling is just a few Hertz, whereas a solid-state-like limitation of the tumbling would result in couplings on the order of several thousand Hertz. Theoretically, various mechanisms can be postulated that would result in such an effect, including temporal binding to a large receptor, “floating” of the creatine molecule in an ordered membrane, and restricted tumbling in small elongated spaces between the actin/myosin molecules. The last explanation seems to be the most probable one because the effect is relatively unspecific and is observed in various metabolites (94–99). In addition, the space between actin and myosin is small for molecules since all structures and metabolites are covered by a considerably large hydration sphere, which limits the rotational and translational freedom. These 1H-MRS experiments opened the door for a kind of “molecular view” into skeletal muscle. In particular, since creatine and phosphocreatine take part in the CK equilibrium at different places in the cell, the equilibration of the CK reaction between different regions of the cells could be limited in vivo by the restricted motional freedom of creatine/phosphocreatine.
The understanding of residual dipolar coupling became even more complicated when it was observed that the methyl resonance of creatine changed during heavy exercise (102) and postmortem in an animal study (101). Since a correlation of the creatine-methyl resonance in 1H-MRS with the PCr resonance in 31P-MRS was found, it was tempting to speculate that free creatine is invisible and that the 1H-MRS resonance represents just the phosphorylated PCr content. However, this explanation has been brought into question by findings in creatine-kinase knock-out mice (103). Since findings of limited visibility of creatine/phosphocreatine would have theoretical effects for the molecular environment in skeletal muscle (72) and practical consequences for the indirect calculation of ADP concentrations based on ATP, free creatine, pH, and the creatine-kinase reaction constant, it is important to gain a more complete understanding of these phenomena.
Two very promising MR tools for investigating muscle metabolism on a molecular scale are MT from water on the resonances of metabolites, and DW spectroscopy (90–92, 99, 164, 259, 260). A study that observed the anisotropy of PCr diffusion in 31P-MRS of rat skeletal muscle revealed restricted diffusion (91). This finding may support and complement observations of residual dipolar coupling based on limited motional averaging, as summarized above for creatine. Unfortunately, both methods are technically demanding and are not available on common whole-body MR scanners.
MULTINUCLEAR MRS, OTHER NUCLEI, AND COMBINATIONS WITH IMAGING TECHNIQUES
MR is one of the most versatile tools available for investigating physiological and pathological mechanisms in situ. This is particularly true if multiple MR methods (i.e., heteronuclear MRS or MRI) are applied during the same session.
Transport processes and phosphorylation of glucose in insulin-resistant individuals were studied with a combination of 31P-MRS for observing glucose-6-phosphate (G6P) and 13C-MRS for determining glycogen synthesis (20, 47–49). With the use of these techniques, it was possible to characterize impaired glucose transport across the cell membrane and decreased phosphorylation leading to a reduced synthesis of muscular glycogen. Another study (50) combined 13C- and 31P-MRS, and found that plasma amino acid elevation induces skeletal muscle insulin resistance in humans by inhibiting glucose transport/phosphorylation, resulting in a marked reduction of glycogen synthesis.
A combination of 31P-MRS to assess the rates of mitochondrial oxidative-phosphorylation activity in muscle and 1H-MRS to measure IMCL and intrahepatic triglyceride content was used to investigate impaired mitochondrial activity in the insulin-resistant offspring of patients with type 2 diabetes (51). 31P- and 1H-MRS have also been used to observe high-energy phosphates and deoxymyoglobin in an interleaved scheme (52), as well as high-energy phosphates and creatine (261).
A study of insulin-resistance in elderly patients used a combination of 1H-MRS (IMCL), 13C-MRS (TCA cycle), and 31P-MRS (ST) to investigate mitochondrial activity (53).
In nutrition and sports medicine, knowledge about the use and replenishment of the two major forms of muscular energy storage—IMCL and glycogen—is important. With a combination of 1H- and 13C-MR, these metabolites can be investigated within one hour or less (54–56, 220).
Aside from 1H, 13C, and 31P, not many nuclei can be observed in vivo in muscular tissue, and those that can be have been studied mainly in animals or cells at higher field strengths. In addition, experiments that use these other nuclei are often acquired in an imaging mode (262), i.e., without resolution of the spectral axis, or in combination with multiquantum filters. 23Na imaging has been used to quantify total [Na] concentration in the skeletal muscle of healthy volunteers after exercise, and in patients with myotonic dystrophy and osteoarthritis (263). Changes in tissue sodium content vs. sodium-macromolecular interaction have been investigated by relaxation-time measurements (264) in volunteers, and triple quantum-filtered spectra have shown properties of sodium ions in ordered skeletal muscle (58) and disease progression in myotonic dystrophy patients (59). In general, 23Na is more frequently applied in kidney and heart muscle than in skeletal muscle. Since there is almost no 19F signal from human tissue, anticancer drugs and/or anesthetics can be followed noninvasively without contamination from other sources (95). Other nuclei (e.g., 2H and 39K) can be used, mainly in isolated organs and animals. Multinuclear MRS with up to four nuclei has been reported (265); however, a complete overview of this subject would go beyond the scope of this review, which concentrates on skeletal muscle.
The number of combinations of MRS with other imaging techniques is increasing, since most of the MR scanners for humans are primarily used for imaging anyway. The use of sequential or even interleaved acquisitions by multiple MR modalities can produce an enormous amount of physiological data in a single session. The most obvious application is the acquisition of localizer images for exact positioning of the voxel in 1H-MRS, since a careful placement of the voxel will result in a correct 1H-MR spectrum. MRS is also useful in morphological imaging to determine the muscle volume in order to calculate the total metabolite content (57, 266, 267). Since MRI is able to depict activated muscle groups, it can provide valuable complementary information to MRS, which has limited spatial resolution (120, 175, 268–270). A comparison of MRI diffusion experiments (271) and residual dipolar coupling in 1H-MR spectra (100) shows good agreement in the determination of fiber orientation (272). While relaxation times of water measured in imaging sequences have often been used to characterize skeletal muscle, a combination with 31P-MRS has been used to investigate the effect of heavy exercise on the relationship between transverse relaxation properties and metabolism (273, 274). Combining MRS with imaging modalities that show muscle perfusion can provide information on both the supply and availability of metabolites (275–278). Human skeletal muscle perfusion, oxygenation, and high-energy phosphate distribution were measured simultaneously by interleaved 1H and 31P-NMR spectroscopy and 1H-MRI (276, 279). Since single-voxel 1H-MRS is valuable for determining IMCL but not fatty tissue (EMCL), the combination of spectroscopy with imaging of adipose depots is promising (280–283).
CONCLUSIONS AND OUTLOOK
When in vivo MR was first applied with small-bore magnets, it was expected that 31P-MRS of skeletal muscle in particular would become a major tool for diagnosing myopathies and elucidating muscular physiology. This proved to be only partially true; however, the application remained a research tool for expert groups. A major reason is the fact that typical clinical MR scanners do not support 13C-MRS and 31P-MRS, and especially not proton-decoupled heteronuclear spectra. When the differentiation of IMCL and EMCL became possible, physiologists and diabetologists became interested in 1H-MRS, which can be implemented on regular clinical scanners. In addition, groups combined 1H-, 13C-, and 31P-MRS for a comprehensive determination of muscle physiology or used MRS together with MRI methods for volumetry, flow, and BOLD measurements. Special applications, such as diffusion MRS and MT MRS, obviously have great potential for observing muscle physiology on a molecular scale; however, these methods also remain in the hands of only a few expert groups.
Therefore, it is difficult to present an outlook on the future development of muscular MRS. While the potential of modern MRS methods and their combination with imaging modalities is obvious, many other factors will determine the fate of in vivo MRS in skeletal muscle. If clinical MR scanners will be regularly equipped with broadband channels for X-nuclei, and if decoupling can be achieved on the same systems, muscular spectroscopy may become a general tool. Since this is not very probable, MRS of skeletal muscle will probably remain a powerful tool in the hands of experts for designing and performing prospective studies to address precise pathophysiological questions.