INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

THE IMPORTANCE OF THE SPATIAL MATCHING of skeletal muscle perfusion () and metabolic demand (O₂; /O₂) is often cited but has been an elusive measurement to attain. The ability to measure /O₂ is essential to examine and truly understand skeletal muscle function and dysfunction during the challenge of muscular work. For example, taken from one of our laboratory's specific areas of interest (exercise O₂ delivery and utilization during exercise), the finding that maximal O₂ extraction in exercising muscle is limited such that effluent venous O₂ content never falls to zero has been interpreted as evidence of diffusion-limited O₂ supply (15, 26, 32, 34). However, this finding can equally well be explained by the existence of /O₂ inhomogeneity within an exercising muscle (21, 23). Indeed, there is considerable experimental evidence that heterogeneity in this scenario does exist (6, 20), but each of these studies documents blood flow heterogeneity with respect to tissue volume, not O₂. Consequently, the diffusion-limited component of O₂ transport has continued to be accepted on the basis of the assumption that the heterogeneity of /volume is equally matched by nonhomogeneous metabolic O₂/volume. Thus in areas where there is low there may be low O₂ and vice versa.

It is clear that there is a need for a technique that can accurately assess and metabolic activity in the same volume of exercising muscle, and several methodological combinations have previously been suggested. For example, the use of local PO₂ potentially may reflect O₂, measured by either PO₂ electrodes (13) or O₂ phosphorescence quenching (29) in combination with colored or fluorescent microspheres to determine local blood flow (11). However, recent observations of small to no change in calculated mean capillary PO₂ and intracellular PO₂ with progressively intense exercise to maximum cast doubt on the validity of PO₂ as an indicator of O₂ (27). These findings, in addition to concerns about spatial resolution and tissue damage from electrodes, do not promote the validity of these methods. Glucose uptake, on the other hand, offers an indication of local O₂ and has previously been coupled with microspheres (a powerful tool with very good spatial resolution of muscle blood flow) (16), but this latter technique has the fundamental problem that such studies are, by necessity, terminal in nature, thus limiting its use to research animals.

Two magnetic resonance imaging (MRI) techniques can provide noninvasive in vivo multiple localized measurements of both and O₂. In combination, they can provide accurate measurements of /O₂ in adjustable localized volumes of exercising muscle. The first of these two techniques, ³¹P chemical shift imaging (CSI) (2), allows the acquisition of comparable phosphorus spectra from multiple voxels simultaneously. Because phosphocreatine (PCr) depletion is directly proportional to ATP hydrolysis (18), its relative depletion can be used as an index of muscle O₂ (4, 12, 17). Thus CSI allows the measurement of O₂ for a given volume of muscle. The second MRI technique allows the measurement of both spatially and temporally resolved in vivo by using arterial spin labeling (ASL) (3). This technique developed in the brain (36), but now, adapted for use in muscle (10), involves magnetically tagging arterial blood proximal to an imaging slice, then observing the changes that occur as blood flows into that volume (3). With this technique, the local magnetic resonance (MR) signal is proportional to the amount of arterial blood delivered to the voxel in a short time interval, and absolute units can be calculated. Combining the results of these two MR techniques provides measurements of /O₂ in a given volume of muscle and the distribution of this ratio across and within exercising muscles.

Thus the purpose of this manuscript is to report for the first time the combination of ASL for and ³¹P CSI for O₂ to assess /O₂ in multiple voxels of human skeletal muscle during steady-state exercise.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

These studies were approved by the Human Subjects Committee of the University of California, San Diego. Representative data are presented from a single healthy physically active male volunteer aged 28 yr who gave his written, informed consent.

Exercise protocol. A plantar flexion exercise paradigm was utilized to facilitate dynamic exercise within the confines of a clinical MR scanner (14). After a 5-min warm-up period, exercise was performed at a rate of 0.5 Hz at a constant work rate (~6 W) for 10 min. The rate of muscle contraction was determined by the pulse sequence timing of the ASL data collection; however, the subject timed the contractions with a metronome (set at the same rate) during the CSI data collection because the sound of the pulse sequence was different. For the CSI studies, exercise was commenced 2.5 min before the signal acquisition to ensure that PCr levels had reached a steady state of depletion. Because CSI offers no temporal resolution, the achievement of a steady state is essential. The 10 min of steady-state exercise were performed twice to facilitate the collection of both CSI and ASL data. An additional 10 min of scanning at rest in the plantar flexion ergometer were performed to collect resting CSI data. There were at least 20 min of rest between the repeated scans. The lower leg orientation was maintained as for the imaging, with the CSI data being collected from the same axial anatomic slice location as the images.

³¹P CSI. Two series of ³¹P CSI studies were performed on a 1.5-T MRI scanner (GE Medical Systems, Milwaukee, WI), one at rest and one during 10 min of steady-state plantar flexion exercise. A dual-tuned ³¹P/¹H flexible coil (Medical Advances, Milwaukee) was used to obtain ¹H axial images to confirm anatomic localization and ³¹P CSI data. ³¹P CSI data were acquired by using the commercially available pulse sequence provided by GE Spectroscopy Research Accessory, in a similar fashion to previous studies (2). Specifically, a radio frequency (RF) pulse to excite magnetization within a slice was applied in the presence of a gradient. To obtain both spectral and spatial information, phase-encoding gradients were applied before the data readout. The spatial resolution is determined by the number of phase-encoding steps and by the field of view (FOV). A 14-cm FOV was used with an acquisition matrix size of 14 × 14 phase-encoding steps and a slice thickness of 1 cm. This resulted in an in-plane resolution of 1 cm² and 1 ml volume of tissue in each voxel. Magnetic field (Bo) homogeneity was adjusted for each subject by using the autoshim capabilities of the scanner with the transmitter frequency on resonance for the water signal. The transmitter frequency was switched to phosphorus and adjusted to be on resonance for PCr to ensure that there was no misregistration between the anatomic images and the CSI data set due to chemical shift errors (these anatomic images, using pixel registration, ensure an accurate combination of the CSI and data sets). The difference spectra (rest and exercise) for each voxel were determined off-line by use of a Silicon Graphics INDIGO equipped with SAGE (GE Medical Systems).

imaging using arterial spin labeling. All images were acquired on a standard 1.5-T clinical imaging system (GE Medical Systems) fitted with a local gradient knee coil of our own design (37) and built in our laboratory (9). This coil produces 6 G/cm at 100 A on all three axes with gradient rise times of 100 µs from zero to full scale and so provides resolution and signal-to-noise ratio superior to that achievable with a standard extremity coil, allowing the acquisition of images with spatial resolution high enough to easily identify the different muscle groups in the lower leg even with an echo-planar imaging sequence.

Calculation of the variance in and metabolism. In an attempt to better describe the matching of with metabolism, we examined the variance in both ASL-measured and ³¹P-measured metabolism. This analysis, similar to that employed in the assessment of ventilation-to- matching in the lungs (31), necessitates the calculation of the second moment of the distribution and PCr distribution on a log scale (log SD and log SD_PCr, respectively). The mathematical basis of this calculation is illustrated in Eq. 1 
where and V are and O₂, respectively, to piece i of n pieces, and ln is the natural logarithm of the O₂-to- ratio. The calculation of the second moment of the metabolism-weighted PCr- distribution was calculated in the same manner.

Studies of validity and reliability of ASL and ³¹P CSI. A series of additional studies examining the validity and reliability of both techniques were also performed. Specifically, to test the reliability of the ASL technique, a subject performed two identical plantar flexion exercise bouts of 6 min at 5 W (gated to the ASL data collection) punctuated by a period of 6 min of rest before and after each exercise bout. The coefficient of variation for these repeated measurements was then calculated. To demonstrate both the validity and reliability of the ASL technique, a subject performed plantar flexion, again gated to the ASL data collection, at 6 W for an 8-min period. For the first half of these 8 min we allowed normal , whereas the second half was completed under partial ischemia by inflating a blood pressure cuff around the thigh at 100 mmHg. Additionally, the effect of this protocol on blood velocities in the popliteal artery, supplying the gastrocnemius, was assessed with ultrasound flowmetry (Cerebrovascular Diagnostics System, Medasonics, Freemont, CA).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Figure 1 was collected from the single representative subject reported here and is typical of a image attained by the ASL technique. The color scale from dark blue to bright red represents areas of low to high , respectively. The localized recruitment, related to , during plantar flexion exercise is apparent in the medial head (left side of figure) and the lateral head of the gastrocnemius (lower right of figure). These data clearly illustrate the heterogeneity that exists within a single muscle. Interestingly, in this subject, also increased quite significantly in the extensor digitorum longus muscle (right side of figure). This is indicative either of muscle recruitment in this area or of to an inactive area, with the former possibility being most likely. The three centrally located bright red regions are artifacts due to large conduit vessels (center of figure).

View larger version (128K): [in this window] [in a new window]   Fig. 1.   Skeletal muscle perfusion image of the left leg (left of image = medial) collected by the arterial spin labeling (ASL) technique, which facilitates the measurement of perfusion in the exercising lower leg during plantar flexion. Color scale, from blue to red, represents variations in perfusion from low to high.

Figure 2 illustrates the local origin of the CSI data on an axial gradient echo MR image of the lower leg. These data were acquired simultaneously and provided comparable phosphorus spectra from multiple voxels (1 cm³, voxels A-M) in the same slice location shown in Fig. 1. The gray PCr spectra were collected in resting conditions whereas the black spectra were attained during steady-state exercise. Only the PCr peak is visible because the transmitter frequency was switched to phosphorus and adjusted to be on resonance for PCr. These data clearly document the heterogeneity of O₂ within an exercising muscle even between two adjacent voxels; for example, in voxel N there was an 80% reduction in PCr concentration, whereas voxel M exhibited zero depletion. Thus a wide range of PCr depletions was recorded (0-92%). The mean value for all voxels was 60% PCr depletion. This is in agreement with our previous global assessment of muscle PCr changes using magnetic resonance spectroscopy and the same work rate. Hence, if this 15 ml of muscle had all been under a surface coil, the mean would be have been 60%, which at 6 W is consistent with our previous observations (14).

View larger version (50K): [in this window] [in a new window]   Fig. 2.   Phosphocreatine (PCr) data collected by chemical shift imaging during submaximal plantar flexion exercise with the left leg. As illustrated on the anatomic image (left side of image = medial), this method allows acquisition of comparable phosphorus spectra from multiple voxels (1 cm3) simultaneously (A-M). The gray PCr spectra were collected in resting conditions, whereas the black spectra were attained during exercise. Only the PCr peak is visible because the transmitter frequency was switched to phosphorus and adjusted to be on resonance for PCr.

Figure 3 illustrates the local measurements of and O₂ plotted against the ratio of these variables on a log scale. This analysis, similar to that employed in the assessment of ventilation-to- matching in the lungs (31), revealed a similar log SD and log SD_PCr (Eq. 1). Unlike measurements of ventilation-to- matching in the lungs, for which reference data are available, the practical implications of these values in muscle are not readily apparent because of the novelty of these data. In an attempt to provide greater insight into this matter, we modeled the effect of variations in log SD and log SD_PCr in terms of attainable O₂, assuming no diffusion limits. The results of this analysis are presented in Table 1 and indicate that the log SD and log SD_PCr of 0.47 would allow only 92% of the target O₂ to be achieved. This suggests that the mismatch of /O₂ does play a role in determining O₂ transport and utilization during exercise, even in young, healthy muscle.

View larger version (15K): [in this window] [in a new window]   Fig. 3.   Assessment of perfusion and metabolic demand in 15 voxels within the gastrocnemius (see Fig. 2) utilizing the analytical methods employed to study ventilation-perfusion matching in the lung by the multiple inert gas elimination technique (30). Note that the second moment of the perfusion distribution and the PCr distribution on a log scale (log SD and log SDPCr, respectively) is almost identical. There is a small (6%) indication of shunt within the muscle, indicating an area of infinite mismatch between blood flow and metabolic demand.

View this table: [in this window] [in a new window]   Table 1.   Modeled effect of changes in log SD and log SDPCr on muscle O2 (target 150 ml/min) and venous PO2

Figure 4 illustrates the frequency distribution for both and PCr depletion within the 15 voxels measured. From this analysis, it appears that is evenly distributed across a wide range whereas O₂, as measured by PCr depletion, was more unimodal with over 50% of the voxels falling to a similar extent. This analysis indicates that there are voxels that are either vastly over- or underperfused for their O₂.

View larger version (17K): [in this window] [in a new window]   Fig. 4.   Frequency distributions of regional blood flow (A) and regional PCr depletion (B). Data were collected from 15 voxels of 1-cm3 size.

Figure 5 illustrates that there is no relationship between PCr depletion and blood flow increase from rest to exercise. It is apparent that there is a large variation in ratio of PCr depletion to blood flow across the 15 voxels. These data and this analysis provide evidence of large variation in matching of /O₂ and again suggest areas of both under- and overperfusion for a given metabolic turnover. That is, in the same voxels varied to a great extent (from <40 to >140 arbitrary units, 350% range) and the majority of voxels depleted PCr to a similar extent (20-25 mM, 25% range).

View larger version (16K): [in this window] [in a new window]   Fig. 5.   Evidence of no relationship between PCr depletion and exercise-induced muscle blood flow. Letters correspond to each individual 1-cm3 voxel, as labeled in Fig. 2. This figure also illustrates the wide coexisting range of ratios between PCr depletion and blood flow.

Studies of validity and reliability of ASL and ³¹P CSI. Figure 6A illustrates the reproducibility of repeated measurements during repeated exercise bouts at the same level of exercise. The coefficient of variation for these measurements was 9%. Figure 6B illustrates that the local ASL measurements tend to reflect both the Doppler assessment of velocities and the expected bulk flow changes in muscle blood flow indicative of the validity of this technique, and the similar response in the repeated protocol demonstrates the reproducibility of the ASL technique. Specifically, the ultrasound flowmetry assessment revealed blood velocities of ~2 cm/s at rest, ~20 cm/s during exercise, ~5 cm/s during exercise with the cuff inflated, and a substantial hyperemia of ~30 cm/s for 20 s at the end of exercise when the cuff was released.

View larger version (32K): [in this window] [in a new window]   Fig. 6.   A: example of the reliability of the ASL perfusion measurements in a single typical 1.5-cm3 voxel during 2 identical exercise (EX) bouts of 6 min with 6-min rest periods. B: further evidence of the validity and reliability of ASL perfusion measurements collected in a 0.35-cm3 voxel in the gastrocnemius during 2 repeated periods of rest, plantar flexion exercise, plantar flexion exercise with partial ischemia, and hyperemic response after this ischemic exercise. It should be noted that continuous data are collected every 6 s; consequently, both panels reflect raw nonaveraged data. Thus fluctuations in these data may be a result of real fluctuations in muscle perfusion (e.g., systolic and diastolic pressure changes).

View larger version (66K): [in this window] [in a new window]   Fig. 7.   Illustration of the validity and reliability of 31P chemical shift imaging measurements at rest and 2 different intensity levels of plantar flexion exercise with the right leg. In this case, each voxel is 1.5 cm3. Alphabetical labels in voxels relate each voxel to its place of origin in the anatomic image, whereas the numerical values represent the PCr depletion from resting levels. It is interesting to note that PCr depletion increases with increasing exercise intensity, as expected, but there are clear exceptions. However, the mean values of 44 ± 7% PCr depletion at 5 W and 63 ± 6% PCr depletion at 9 W are in agreement with previous global assessments of PCr depletion using traditional spectroscopy (14).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The results of the present study demonstrate a novel method for assessing the matching of /O₂ in human skeletal muscle in vivo. The need for such an assessment has been recognized for a significant period of time, but advancements have been hampered by the methodological complexities of such data collection. Utilizing this method, we have recorded considerable heterogeneity in both muscle and muscle O₂ during steady-state exercise and no correlation between these two variables. However, it is important to recognize that the intention of this manuscript is to document a methodological advance and not to definitively characterize the relationship between and O₂ in exercising human skeletal muscle. This novel, noninvasive MRI method may ultimately provide this assessment in both health and disease. We provide a limited discussion of our preliminary results to place this method in the context of the current literature.

Heterogeneity of and metabolism. Although there have been many studies that have reported heterogeneous muscle blood flow (6, 20, 24), to our knowledge there is only one other documented attempt to measure local blood flow and metabolism within the same skeletal muscle. This work by Iversen and Nicolaysen (16) studied rabbit muscle at rest and during electrical stimulation by using microspheres to assess and glucose uptake as a marker of metabolic activity. Their conclusion that regional blood flow within single skeletal muscles was not strongly linked to regional metabolic activity is in line with our initial data collection using this new noninvasive MRI method (Figs. 4 and 5).

Log SD and Log SD_PCr in skeletal muscle. Unlike for the lungs, where the characterization of ventilation-to- matching is routinely measured and the functional implications are known (31), the determination of skeletal muscle log SD and SD_PCr is at present somewhat abstract. What is normal, and when does this match become abnormal? To objectively offer initial insight into the functional consequence of a log SD and log SD_PCr of 0.47, we have considered the possible scenario in which all venous PO₂ was the result of heterogeneity, with no diffusion limitation. In this scenario, if there is zero heterogeneity, O₂ is equal to the product of blood flow and arterial O₂ content, and venous PO₂ must be zero. Thus, with this conceptual starting point and data drawn from our previous assessment of plantar flexion O₂ (12), numerous mass-specific flow measurements (25), and others muscle volume data for the gastrocnemius (7), we calculated the theoretical O₂ (138 ml O₂/min) that could be achieved with the measured log SD and log SD_PCr (0.47) and the venous PO₂ that would result (27.1 mmHg). In this model, we assumed blood flow to be 1.00 l/min and target O₂ to be 150 ml/min with the measured value of 6% shunt. Hence, this level of log SD and log SD_PCr in the model permitted only 92% of the anticipated O₂ (Table 1). Although preliminary in nature, this model suggests that the measured level of /O₂ matching and relatively small shunt do play a role in attenuating metabolic capacity and elevating venous PO₂, even in young, healthy active muscle working at a submaximal work rate. To better put these data in perspective, we have tabulated the calculated effect of altered log SD and log SD_PCr through the spectrum of 0.1 to 1.2, including the measured value of 0.47, on both the achievable muscle O₂ and venous PO₂ (Table 1). To provide an additional comparison with our measurements, we used the same SD calculations to produce an estimate of muscle mass- distribution in an exercising in situ canine gastrocnemius preparation (34) [modified to reduce surgical procedures and tissue traumatization (35)], injected with colored microspheres and sectioned into 12 pieces after exercise. The log SD for this preparation was only 0.13. Unfortunately, here we had no measure of local O₂ in these tissue volumes. Again, although preliminary in nature, this significantly greater log SD in human skeletal muscle compared with the canine is certainly intriguing and may be related to the obvious species differences in terms of aerobic capacity exhibited between human and dog.

Validity and reliability of ASL and ³¹P CSI. As with any new method, it is important to demonstrate a certain degree of both validity and reliability. Consequently, during the development of this method we performed several series of studies to determine the validity and reliability of the two techniques. Typical data sets are presented here in Figs. 6 and 7. From Fig. 6A, the reproducibility of this technique is clearly demonstrated. Figure 6B illustrates that that the local ASL measurements tend to reflect the measured (Doppler) and expected bulk flow changes in muscle blood flow indicative of the validity of this technique, whereas the similar response in the repeated protocol again demonstrates the reproducibility of the ASL technique. Additionally, Fig. 6B represents nonaveraged raw data from a voxel sized 0.35 cm³, which is substantially smaller than the presently utilized 1-cm³ voxel necessitated for matching with CSI data.

Present limitations to this methodology. As with any new method, there are presently several limitations to this noninvasive assessment of local and metabolism. First, these two measurements are not simultaneous. As described, the determination of local and local metabolism can be assessed on the same subject in the same magnet with the same position (both acquisition setups can be prepared at once); however, both assessments cannot, at present, be measured simultaneously. This necessitates two exercise periods and the superimposing of the data from each bout in a single interpretation. Thus differences in either response may occur between the two exercise bouts. This potential effect should be minimized by adequate warm-up before each exercise bout. Second, because of the need to average several minutes of CSI data to attain an average metabolic cost in each voxel examined, this method is limited to spatial measurements and cannot detect, although it may be influenced by, temporal heterogeneity. With the present approach this limitation cannot be removed, but with continued optimization of the CSI methods (or an increase in voxel size), the time resolution can be improved. Third, again related to resolution, this method (in its present form and with a reasonable exercise period of 10 min) is limited to a voxel size of 1 cm³. Although this tissue volume is relatively small compared with several previous studies of muscle and heterogeneity, this is still a substantial voxel compared with muscle fiber size. It is at present unclear as to the optimum voxel size for this type of investigation. Suffice to say, the smaller the possible tissue volume the better as this can easily be increased, if necessary. As noted, the present method is limited to 1-cm³ volumes only by the CSI technique, which requires a significant increase in data collection time during steady-state exercise to reduce each discrete sample volume (see Fig. 6B for ASL data collected in a 0.35-cm³ voxel). Unfortunately, a 50% reduction of the voxel size assessed by CSI requires a 400% increase in length of data collection necessary to achieve the same signal-to-noise ratio. With the ultimate goal of creating a clinically useful tool and temporal concerns about long duration exercise, this is not a practical approach. Consequently, efforts are being focused on optimizing the CSI signal collection. Despite these limitations, the present manuscript clearly documents the feasibility of a method that holds significant promise to address the age-old question of homogeneity of and metabolism and to address this issue noninvasively in vivo in humans.