The purpose of this study was to determine the effects of intense exercise on the proton transverse (T2) relaxation of human skeletal muscle. The flexor digitorium profundus muscles of 12 male subjects were studied by using magnetic resonance imaging (MRI; 6 echoes, 18-ms echo time) and in vivo magnetic resonance relaxometry (1,000 echoes, 1.2-ms echo time), before and after an intense handgrip exercise. MRI of resting muscle produced a single T2 value of 32 ms that increased by 19% (P < 0.05) with exercise. In vivo relaxometry showed at least three T2 components (>5 ms) for all subjects with mean values of 21, 40, and 137 ms and respective magnitudes of 34, 49, and 14% of the total magnetic resonance signal. These component magnitudes changed with exercise by −44% (P < 0.05), +52% (P < 0.05), and +23% (P < 0.05), respectively. These results demonstrate that intense exercise has a profound effect on the multicomponent T2 relaxation of muscle. Changes in the magnitudes of all the T2 components synergistically increase MRI T2, but changes in the two shortest T2components predominate.
- proton transverse relaxation
- skeletal muscle
- magnetic resonance
exercise-induced changes in the transverse (T2) relaxation of skeletal muscle have been extensively studied for over three decades, but the underlying mechanism is not well understood. This phenomenon was originally described in 1965, when Bratton and co-workers (5) reported an increase in the T2of isolated frog skeletal muscle after isometric contractions with electrical stimulation. The authors suggested that the apparent T2 of muscle increased because protein-bound water molecules, in rapid exchange with “free” water molecules, were released during muscle contraction (5). In the next few years, the T2 of isolated animal muscle was determined with greater accuracy by using relaxometry protocols (7, 30). The T2relaxation of ex vivo muscle was found to be multiexponential, characterized by three distinct T2 components of <5, 20–40, and >80 ms, with the intermediate component comprising the majority of the signal (4, 20). Because more than a single component was measured, Hazlewood et al. (20) interpreted these results as evidence against two fractions of cellular water in rapid exchange (5). They ascribed the three T2 components to protein-bound and intra- and extracellular water compartments, respectively, which were thought to exchange water molecules slowly but could also include fast-exchanging subcompartments (20). According to this interpretation, exercise-induced T2 changes reflect a shift of water among cellular compartments. Fung and Puon (18), however, described muscle T2 as not being multiexponential but instead reflecting a complicated function of hydrogen ion exchange between water and the functional groups of protein filaments. This was further argued to support the notion that an exercise-related intracellular pH drop accounted for the exercise-induced T2 changes.
The early relaxometry experiments involved excised muscle tissue because in vivo T2 data obtained with those protocols (7,30) would have been subject to signal contamination from extraneous tissues. The advent of magnetic resonance imaging (MRI) provided a means for in vivo T2 measurements, but with less accuracy than the relaxometry techniques (19). Nevertheless, MRI was used to demonstrate a single T2 value between 30 and 40 ms for human skeletal muscle. With exercise, T2 increases by 10–30% (12), a phenomenon similar to the original observation of Bratton et al. (5). Consequently, active muscles appear brighter on T2-weighted MRI images, which have been useful for muscle recruitment studies (11, 22, 44). Conventional imaging protocols do not show multicomponent T2 relaxation for in vivo muscle (1, 11, 12, 22, 34, 38), so investigators can only speculate whether the MRI T2 increases actually involved changes in the long (>80-ms) (12, 38) or short (20- to 40-ms) (1, 11, 22) T2 components.
We have recently reported multicomponent T2 relaxation for in vivo skeletal muscle at rest (36), with some T2components not previously observed ex vivo (4, 20). These data were obtained with a novel in vivo relaxometry technique that has measurement parameters [echoes (n) and signal-to-noise ratio (SNR)] two orders of magnitude greater than standard imaging methods (19). With this technique, at least three T2 components (>5 ms) with values of 21, 39, and 114 ms were resolved from the flexor digitorium profundus (FDP) (36). The objective of the present study was to discover which, if any, T2 components of in vivo muscle change with intense exercise and to determine their relationship to the increase in the MRI T2 with exercise.
This study involved two experiments, Exp 1 and Exp 2. Exp 1 was conducted to determine the changes in muscle T2 relaxation after intense exercise. Both MRI and in vivo relaxometry were used to measure the T2 of the FDP before and after a single bout of a maximal-intensity handgrip exercise to fatigue. To analyze the relaxometry data of Exp 1, three common techniques were used: nonnegative least squares using all even echoes [NNLS (n= 1,000)] (43), NNLS on even echoes that approximate logarithmic sampling [NNLS (n = 195)] (3), and the Marquardt-Levenberg nonlinear minimization using all echoes [MQ (n = 1,000)] (29).
Exp 2 was conducted on a different day to determine the rates of change in the magnitudes of each T2 component throughout rest, exercise, and recovery. Relaxometry data from the FDP were acquired every 30 s over a 32-min period. During this time there were 5 min of rest followed by a handgrip exercise identical to that in Exp 1 (∼2 min) and 20 min of recovery. The relaxometry data of Exp 2 were analyzed with NNLS (n = 1,000) as it did not require the formulation of initial guesses.
Twelve healthy and active men with a mean age of 26 ± 2 yr volunteered to participate in Exp 1 and gave informed consent to the procedures. Five of the subjects (mean age 25 ± 2 yr) also participated in Exp 2. The experiments outlined in this paper were approved by the University of Western Ontario Review Board for Research Involving Human Subjects.
Experimental protocol for Exp 1.
Proton nuclear magnetic resonance (NMR) data were acquired by using a 30-cm bore, 1.89-Tesla magnet (SMIS, Oxford Research Systems) interfaced to a computer console, and a custom-built gradient/shim system (Magnex). Within the bore of the magnet, a 14-cm-diameter, high-pass, 16-rung-quadrature birdcage coil functioned as both a transmitter and receiver. Placed inside the birdcage coil was an armrest, on which the subject's right forearm was secured with Velcro straps to reduce involuntary motion. The magnetic resonance (MR) sequences used in this study have been described in greater detail elsewhere (36).
The armrest was positioned to ensure that the FDP was at the isocenter of the magnetic field, and this was verified with “scout” MR images. The scout images had a localization scheme (identical to that of the relaxometry technique used next) combined with a single spin-echo imaging sequence [1-cm slice thickness; field of view (FOV) = 15 cm; 128 × 128 matrix; repetition time (TR) = 200 ms; echo time (TE) = 20 ms; total scan time = 26 s]. The scout images allowed for a quick determination of the suppression quality, which is the ratio between the mean signal intensity within the localized volume to the extraneous outer volume (39). In addition, the scout images were used to determine whether the localized volume for the T2measurements was located within a region of the FDP, excluding visible fat and blood vessels. If this was not the case, repositioning of the arm was necessary. Once the arm positioning was deemed acceptable, the magnet was shimmed and the 90 and 180° radio-frequency pulse settings were determined.
Resting muscle T2 data were then acquired with the in vivo relaxometry technique, Projection-Presaturation-Carr-Purcell-Meiboom-Gill (PP-CPMG) (36). This technique eliminates the MR signal from tissue extraneous to a cylindrical volume (2-cm diameter, 5-cm length) within the FDP to acquire T2 data with 2,000 echoes, TE = 600 μs, TR = 15 s, and 6 averages. A single T2 measurement, therefore, was attained in 90 s. Only the even echoes were used for subsequent analysis, so the data in its final form were characterized by n= 1,000 and an effective TE of 1.2 ms.
Next, a conventional MRI T2 measurement of resting muscle was obtained in the form of an axial T2-weighted forearm image. The MRI technique (1-cm slice thickness; FOV = 128 mm; 128 × 128 matrix, n = 6, TE = 18 ms, TR = 1 s, and 1 average) took 128 s to completion.
The subject was then instructed to keep the forearm still and perform a maximal-intensity handgrip exercise, with contractions at 1 Hz, until fatigue (typically 1 min). When the subject could no longer perform the exercise, the relaxometry and T2-weighted imaging sequences were repeated. Because the relaxometry sequence was executed immediately after exercise and the imaging sequence was executed 90 s later, the two techniques measured T2 values at different times in the recovery from exercise.
The total experiment time per subject, including arm positioning, MRI, and T2 relaxometry, for Exp 1 was ∼25 min.
Experimental protocol for Exp 2.
This experiment was conducted on a separate day from Exp 1 and had an identical procedure for positioning the subject's arm in the magnet. Once the scout MRI confirmed that the FDP muscle was in the isocenter of the magnetic field, the PP-CPMG sequence was executed. For this experiment, the parameters of the PP-CPMG sequence were modified so as to collect a T2 relaxation curve every 30 s (2,000 echoes, TE = 600 μs, TR = 30 s, and 1 average) from the localized volume in the FDP (2-cm diameter, 5-cm length). As in Exp 1, only even echoes were kept for subsequent analysis, so the data were also characterized by n = 1,000 and an effective TE of 1.2 ms. The PP-CPMG sequence was executed for a 32-min period, which involved 5 min of rest followed by a handgrip exercise identical to that in Exp 1 (∼2 min) and 20 min of recovery.
The total experiment time per subject, including arm positioning and T2 relaxometry, for Exp 2 was ∼45 min.
The T2-weighted MR images acquired in Exp 1 were analyzed to determine the forearm cross-sectional area (CSA) and T2value. The forearm CSA was calculated with Image Display software (SMIS) and excluded bone and subcutaneous fat. T2 values were determined from a region (0.5 × 0.5 × 1 cm) within the FDP. T2 values were calculated by plotting the average signal intensity of this region at each of the six echoes as a function of TE (18 ms, 32 ms, …, 108 ms) and fitting the data to a monoexponential curve.
The SNR of the raw relaxometry data from both Exp 1 and Exp 2 were calculated as the ratio of the first data point (first even echo amplitude) to the SD of the baseline (the last 50 data points).
The relaxometry data from Exp 1 were analyzed with three standard multiexponential algorithms: NNLS (n = 1,000), NNLS (n= 195), and MQ (n = 1,000). First, the NNLS algorithm was employed because it is robust in the presence of noise and has the advantage of not requiring a priori knowledge. The fitting parameters used for the NNLS analysis were 500 linearly spaced values, or data bins, between 0 and 1,000 ms, with a bin width of 2 ms. The value of the NNLS regularizer term (43) was determined from the results of previously reported phantom studies (36). Next, the NNLS analysis was repeated with the same fitting parameters and regularizer term using only selected data points from the T2 results. The data sets were reduced from 1,000 to 195 points by including every echo up to 50 ms, every second echo up to 200 ms, every fourth echo up to 400 ms, every eighth echo up to 600 ms, and every sixteenth echo up to 1.2 s. Selecting data points in this manner approximates logarithmic sampling and thereby gives equal weighting to the relaxation times of each component (2). Finally, the NNLS results were used to formulate the model function for the next method of analysis, the MQ algorithm. The data were modeled to three, four, and five components, with the best fit determined by that which minimized the χ2 statistic without increasing the Cramér-Rao SD.
All three multiexponential algorithms resolved several T2components from the relaxometry data of Exp 1 that were labeled A, B, C, D, and E in order of increasing time constant. Component A, with a T2 value of <5 ms, was disregarded because components in this range cannot be resolved with the same accuracy as those with longer values (36). The magnitude of each component (MB, MC, MD, and ME) was expressed as a percentage of the total component magnitude at rest (%Mt, r). For the NNLS analysis, the total component magnitude was the integral of the resulting T2-NNLS spectrum (43). For the MQ analysis, the total component magnitude was determined by the magnetization at t = 0.
The relaxometry data of Exp 2 consisted of 64 T2measurements for each subject, whereas Exp 1 only had two T2 measurements per subject: rest and postexercise. The data of Exp 2 were analyzed with the NNLS (n = 1,000), and the results were expressed as described above, with one exception. In this experiment there was more than a single T2 measurement at rest, so Mt, r was the average of the total magnitudes from each T2 measurement in the initial 5-min resting period. The results were plotted as component magnitude (%Mt, r) vs. time (min), and time constants of the changes over time were determined by exponential fits of the data.
The statistical testing involved analysis of variance for repeated measures, and the results were considered significant at P < 0.05. The values of MB and MC in Exp 2 were compared by using Pearson's correlation. All data are presented as means ± SE.
Axial forearm images acquired in Exp 1 are shown in Fig.1, A (rest) and B(postexercise). The MR images were used to calculate an average CSA of 4,300 ± 230 mm2 that increased by 4.5% after exercise to 4,500 ± 220 mm2. The FDP muscle exhibited increased signal intensity after the handgrip exercise in all 12 subjects. The images yielded a single MRI T2 value for the FDP of 31.9 ± 1.2 ms at rest, which increased to 37.9 ± 1.1 ms after exercise.
In Exp 1, scout images were used to determine that the suppression quality was 95.8 ± 0.8%. From the relaxometry data, the SNR was shown to be ∼4,000. The relaxometry data [analyzed with the NNLS (n = 1,000), NNLS (n = 195), and MQ (n = 1,000)] revealed at least three T2 components (>5 ms) for each subject. The component magnitudes are displayed in Table1, and the corresponding T2values are shown in Table 2. After exercise, there were significant changes in the magnitudes of all the T2 components, and these results are illustrated in Fig.2.
In Exp 2, despite a lower SNR of ∼1,500, the same T2components observed in Exp 1 were usually resolved for all five subjects. Data from the few T2 measurements where the components could not be resolved were discarded.
The total component magnitude of all the subjects throughout rest, exercise, and recovery is shown in Fig. 3. The duration of the exercise period was 48 ± 12 s and is indicated as the area between the two dotted lines (also in Fig.4 and Fig. 6). The total component magnitude reached a maximal value after the exercise had ceased. After exercise, return to baseline values appeared to be biexponential, with a fast (time constant, τ = 1 min) and a slow (τ = 74 min) decay.
Figure 4 represents MB and MC throughout rest, exercise, and recovery. The changes with exercise were fit to monoexponential curves that were both characterized by a τ value of 1 min. During the recovery period, the rate of change in MBand MC could be described with single exponential fits with a τ value of 9 min. Figure 5 shows the strong negative correlation observed between the magnitudes of these two T2 components (r = −0.878, P < 0.01). Finally, Fig. 6 illustrates the changes in MD throughout the experiment. The recovery was fit to an exponential with τ = 13 min.
This study used a method for in vivo T2 relaxometry (36) that permits a comprehensive examination of proton T2relaxation in human skeletal muscle. The relaxometry method was critical to this study because many data points with a high SNR are required to observe multicomponent T2 decay [which cannot be achieved with conventional MRI methods (19, 34)]. With an appropriate analysis technique (29, 43), the relaxation value and magnitude of each T2 component can be extracted from the data. Conceptually, a component magnitude (Mj) is the apparent population, or concentration, of the proton spin group that gives rise to the T2 component j. At the slow exchange limit (where no exchange occurs between spin groups), the component magnitude reflects the actual population of the proton spin group. Moreover, the sum of all the component magnitudes likely reflects the hydration state of the tissue because the vast majority of protons producing the signal are found in water molecules (16).
The most prominent finding of Exp 1 was that intense forearm exercise appreciably alters the multicomponent T2 relaxation of skeletal muscle. Immediately after the handgrip exercise, large changes were observed in MB (34–44% decrease), MC(52–63% increase), and MD (45–66% increase). The decrease in MB (which has the shortest T2value) and the increase in MC and MD (which are characterized by longer relaxation times) would synergistically increase the apparent MRI T2. Indeed, the MRI data showed a 19% increase in the T2 value of the FDP. The MRI T2 increase and the 4.5% increase in forearm muscle CSA also demonstrated that the handgrip exercise was appropriate for recruiting the FDP muscle in all subjects.
Although the absolute percent changes in MB, MC, and MD were similar (Fig. 2), the resting values of MB and MC were considerably larger than that of MD. This means that the actual changes in MB and MC (and therefore their contributions to the increase in MRI T2) were the most substantial. This finding is important because in previous years investigators have questioned whether MRI T2 increases involve changes in long (>80-ms) (12, 38) or short (20- to 40-ms) (1, 11, 22) T2components. Before our recent in vivo muscle study (36), there was only evidence of one “short” T2 component in the range of 20–40 ms (4, 20), so the effects of exercise on two components in that range had not been previously considered.
Quantitative interpretation of NMR relaxation data from complex systems such as in vivo muscle is a nontrivial problem that requires a precise determination of the distribution of relaxation times in the sample. For this reason, the data of Exp 1 were analyzed with three different methods: NNLS with n = 1,000 (43), NNLS withn = 195 (3), and MQ with n = 1,000 (29). The similarity in the T2 values, component magnitudes, and changes with exercise derived from all three methods indicates the relaxometry results are independent of the chosen algorithms.
Exp 2 was conducted to obtain more data pertaining to the T2 components of in vivo muscle, specifically the rates of change in their magnitudes throughout rest, exercise, and recovery. The improved time resolution that made this experiment possible came at the expense of data averaging, so the SNR of this data was consequently lower. Nevertheless, all the components observed in Exp 1 were also found in Exp 2. MB and MC were found to have identical rates of change during and after exercise and were inversely correlated (r = −0.878, P < 0.01), which suggests that they both were involved in a similar physiological process. Another interesting finding of Exp 2 was that the maximal value of the total component magnitude occurred after the exercise had ceased.
The 8–9% increase in total component magnitude (which is influenced by the hydration state of the tissue) was likely due to an uptake of fluid from the vasculature. This is because fluid transfer does not occur from inactive to active muscles (40), and fluid gain by metabolic water production is negligible in short-term exercise (27). The finding in Exp 2 that the tissue hydration reached a maximal value after exercise is suggestive of exercise hyperemia, when local blood vessels were no longer compressed by contracting muscle fibers. Moreover, the recovery of the total component magnitude in the minutes after exercise was biexponential, in agreement with previous observations of tissue hydration recovery after exercise (14,27).
The exact reasons certain tissues exhibit multicomponent T2relaxation are not known. Perhaps the simplest (and most widely cited) way to interpret the data is in terms of the water compartment model (5, 20, 33), where T2 components are thought to originate from anatomic compartments that exchange water molecules slowly (relative to their T2 relaxation times). This model does not require physical barriers such as a semipermeable membrane to distinguish the water compartments; nor does it rule out intermediate exchange (28). The in vivo muscle T2 relaxation that can be measured for all subjects, therefore, may be described as Equation 1where y(t i) is the signal intensity at the ith echo, TEi is the echo time of the ith echo, n is the number of echoes, and Mj and T2 , jare the magnitude and T2 relaxation value of thejth component, respectively. In a compartment model,components B and C would be considered intracellular because their T2 values and combined magnitude (MB + MC = 83–88% Mt, r ) were similar to those of the single component determined in excised muscle attributed to intracellular water, (4, 15, 20). Also, the magnitude (11–15% Mt, r) and T2 value (>100 ms) of component D closely resemble those of the component attributed to extracellular fluid (4, 15, 20). Critics of the water compartment model have suggested that the ex vivo T2component attributed to extracellular fluid was a postmortum artifact (17), but the present observation of a similar component in vivo contradicts that argument. The relative volumes of the both the extra (MD)- and intracellular fluid (MB + MC) in resting muscle were in close agreement with values determined with radio tracers (8, 26, 36, 40). Finally, component E, with the longest T2 value and smallest magnitude, was not included in Eq. 1 because it is relatively small and could not be resolved for all subjects. In a recent study, Noseworthy et al. (31) used a compartment model to analyze MRI T2 data of muscle. Although the authors could not resolve this long T2 component, they speculated that it did exist and attributed it to lymphatic fluid (31). Interestingly, the magnitude ofcomponent E is in excellent agreement with the plasma space found for in vivo rat muscle tissue (8).
The application of the compartment model to the in vivo relaxometry data may also explain the effect of exercise on MRI T2. The ∼7% increase in intracellular fluid (MB + MC) and ∼40% increase in extracellular fluid (MD) with intense exercise are in accord with results from radio tracer studies (26, 36, 40). However, it is generally accepted that gross changes in intra- or extracellular water cannot fully explain the increase in MRI T2 (1, 10, 11, 44). This has led several investigators to suggest the MRI T2 change is induced by a complex reorganization, rather than net increase, in intracellular fluid during exercise (1, 10). It is possible that the individual changes in MB and MC indicate this reorganization. One speculation is that component B includes water associated with heavily hydrated glycogen molecules andcomponent C was cytoplasmic fluid, so MB would decrease and MC would increase as glycogen is utilized during heavy exercise. Other factors must also be operant, as the changes in magnitude of components B and C are similar but not identical (Fig. 4). Indirect evidence to link component B with glycogen can be found in patients with myophosphorylase deficiency. Glycogenolysis is blocked in these individuals, and they do not exhibit the typical increase in MRI T2 with exercise (13). Moreover, the rapid depletion of glycogen postmortem (6) may explain why components B and C were not resolved in excised muscle preparations (4, 20). A recent study by Price and Gore (35) has shown a correlation between muscle MRI T2 and a glycolytic intermediate, glucose-6-phosphate. Future experiments need to be conducted to evaluate the relationship between multicomponent T2 relaxation and glycogen.
Although the water compartment model appears to explain the results of this study well, it is by no means a unique interpretation and has never been directly proven. Perhaps the NMR literature has devoted so much attention to whether the T2 relaxation curve is dictated by the physical distribution of cellular water that other effectors have been ignored. The liberation of bound water molecules has remained a compelling possibility since the original observation of the change in T2 with muscle stimulation (5). This effect is more likely to be prevalent in the myoplasm, which contains ∼23% protein by weight (16), but may not be limited to the intracellular environment. This might mean that T2 components represent water molecules of similar translational and rotational freedom regardless of their physical location. In vivo relaxometry, therefore, can potentially be used to study the net liberation of cellular water molecules during exercise, which is thought to play a critical role in muscle contraction (32).
Cellular water molecules that contribute to the T2 signal may be located in erythrocytes as well as myofibrils. In this study, T2 data were acquired from a localized volume of skeletal muscle that did not include any visible blood vessels but could not exclude capillaries. The relative quantity of erythrocytes at the capillary level is likely to be quite small, as would be their influence on T2. Nevertheless, it has been shown that T2 relaxation of in vitro blood samples is profoundly dependent on erythrocyte volume (41), which changes with exercise (42). This dependence may account, in part, for the dramatic changes in the T2 components with exercise.
This discussion has focused on water molecules because the vast majority of protons that give rise to the T2 signal comprise tissue water (15). However, other cellular moieties such as intracellular lipids may also be important. The in vivo relaxometry technique was effective in eliminating the MR signal from fatty tissue located outside the localized volume of muscle (36), but intracellular triglycerides from within the volume warrant further consideration. Early lipid mobilization during exercises of various intensities has been reported (21), and it is quite possible that would influence T2 component magnitudes. This would involve both intracellular triglyceride stores and increased delivery of plasma lipids to active muscle. Some investigators, however, argue that glycogenolysis is a more important fuel than lipid mobilization during intense short-duration exercise and that the delivery of plasma lipids is dependent on blood flow, but the T2 effect is not (2).
There is evidence that muscle T2 relaxation is affected by the level of oxygenation in the blood (24, 31). Elevated levels of paramagnetic deoxyhemoglobin during exercise cause perturbations in local magnetic fields that dephase spins and thereby could contribute to a decrease in the measured T2 signal of muscle (24). This phenomenon could contribute to the changes in component B, which was the only component to exhibit a decreased magnitude (and T2 value) with exercise. Such a reduction in this relatively short component would therefore be expected to prolong the MRI T2 of muscle.
In summary, this study demonstrates that intense exercise has a profound effect on the multicomponent T2 relaxation of muscle. The decrease in magnitude of the component with the shortest T2 value (component B) and an increase in magnitude of those with longer relaxation times tend to synergistically increase MRI T2. Because of the large initial magnitudes of the two shortest T2 components (components B andC), their changes with exercise and therefore their contribution to the MRI T2 effect, would predominate.
The authors thank Dr. Robert Bartha and John Potwarka for assistance with the MQ analysis.
Address for reprint requests: T. Thompson, Dept. of Nuclear Medicine and Magnetic Resonance, St. Joseph's Health Center, 268 Grosvenor Street, London, Ontario, Canada, N6A 4V2.
This research was supported by grants from the Natural Science and Engineering Research Council of Canada and SMIS (UK).
Address for other correspondence: G. Saab, Dept. of Nuclear Medicine and Magnetic Resonance, St. Joseph's Health Center, 268 Grosvenor Street, London, Ontario, Canada N6A 4V2 (E-mail:).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.
- Copyright © 2000 the American Physiological Society