Exercise Intensity Effects on the Muscle Functional MRI Signal Intensity Time Course: Absolute and Relative Contributions of BOLD Effects

Abstract

The time course of exercise-induced T₂-weighted signal intensity (SI) changes contains an initial rise, early dip, and secondary rise. The purposes of this study were to test the hypothesis that the secondary rise occurs earlier during more intense contractions and to determine the contribution of BOLD contrast to the SI changes. Eight subjects performed 90s isometric dorsiflexion contractions at 30 and 60% of maximum voluntary contraction (MVC) while T₂-weighted (TR/TE=4000/35ms) images were acquired and total hemoglobin ([THb]) and oxy-Hb saturation (%HbO₂) were measured. At 30% MVC, [THb] remained constant and %HbO₂ decreased from 66.3% (SE=2.6) to 32.4% (SE=6.4). At t=88s, SI increased by ~8% and was greater than at t=8 and 56s. At 60% MVC, [THb] remained constant and %HbO₂ decreased from 70.2% (SE=2.3) to 40.4% (SE=5.4). SI increased by ~17% and at t=56 and 88s was greater than at t=8 and 20s. The absolute contribution of calculated BOLD effects was –1% at 30 and 60% MVC. The relative contribution was greater at 30% than at 60% MVC (up to –26% and –10%, respectively). We conclude that the secondary rise occurs earlier at 60% MVC and that the relative contribution of BOLD effects is greater during less intense contractions.

INTRODUCTION

The use of T₂ or T₂-weighted signal intensity (SI) changes to observe the metabolic, hemodynamic, and mechanical events associated with muscle contraction has come to be known as muscle functional MRI (mfMRI) (1). Mechanistic studies have demonstrated that the T₂ change of exercise relates primarily to changes in the intracellular environment associated with increased rates of cellular energy metabolism (2-4). These changes are thought to result from the intracellular accumulation of osmolytes such as inorganic phosphate and lactate, which draw water into the cell along osmotic pressure gradients (2,4). T₂ increases may further depend on the intracellular acidification that results from the accumulation of certain glycolytic intermediates and end-products, especially lactic acid (2,5). The observation that the T₂ change of exercise is magnetic field strength-dependent in isolated muscles experiencing large exercise-induced increases in intracellular acidification and volume suggests that these changes may affect T₂ by modulating chemical exchange pathways between sites differing in Larmor frequency (2).

During in vivo studies, the large increases in oxygen extraction associated with exercise suggest a potentially important role for blood-oxygenation level dependent (BOLD) contrast mechanisms as well. Muscle BOLD effects affect both the whole-tissue T₂ (6-9) and the effective transverse relaxation time constant (T₂*) (6,10-12). Blood oxygenation changes affect the transverse relaxation of the blood itself – the intravascular BOLD effect (13,14) – as well as that of spins in the tissue parenchyma – the extravascular BOLD effect, described in a number of analytical (e.g., (15)) and Monte Carlo (e.g., (8,16)) models. The Monte Carlo models have indicated that at 1.5T, intravascular BOLD effects are the predominant source of BOLD contrast in muscle (8,16).

These physiological phenomena occur simultaneously during an exercise bout, and their complex and sometimes opposing effects on T₂ cause the time course of mfMRI signal intensity (SI) changes to have a temporally complex character that cannot be fit to simple, one-component models (17). Phenomenologically, the mfMRI time course obtained during sustained isometric contractions includes an initial rise, an early dip, and a secondary increase in SI. Previously, we used a computer model to predict the time-varying contributions of muscle shortening, the creatine kinase reaction, anaerobic glycolysis, and increases in blood volume and oxygen extraction to the mfMRI signal time course (18). This model predicted that in single-slice image acquisitions, the initial rise in SI is determined primarily by the introduction of previously unsaturated spins into the slice plane as a result of muscle shortening, with a lesser contribution from the effects of capillary recruitment on proton density. The early dip was suggested to be determined jointly by the alkalotic effect of the creatine kinase reaction (tending to decrease T₂); increased oxygen extraction (also tending to decrease T₂); and the osmotic effects of phosphocreatine hydrolysis (tending to increase the T₂). Finally, the secondary increase in SI was predicted to result from the acidotic and osmotic effects of glycolysis, both tending to increase T₂, though offset somewhat by a negative change in SI due to the continued increased level of oxygen extraction.

This is the first of two companion papers that consider effects related to the intensity dependence of the mfMRI time course. In this paper, we consider how contributions of the two key determinants of the T₂ change – anaerobic glycolysis and BOLD effects – vary with exercise intensity. The first purpose of this study was to test the hypothesis that sustained isometric contractions at higher exercise intensities would lead to elevated SI changes at all exercise durations and, because glycolytic activation occurs earlier at 60% of maximum voluntary contraction (MVC) than at 30% MVC (19), an earlier occurrence of the secondary rise in SI. Also, our previous modeling study predicted that the combined actions of capillary recruitment and oxygen extraction would cause muscle BOLD effects to provide a small positive contribution to SI at short exercise durations, followed by a negative contribution to SI at longer exercise intensities (18); thus the second aim of this study was to test this hypothesis explicitly. Muscle BOLD effects were investigated using T₂, rather than T₂*, because the effects are more specific to the microvasculature (15) and to provide greater applicability to the large literature concerning T₂ changes during exercise. The companion work focuses on an application of the data, focusing in particular on the issue of spatial heterogeneity in mfMRI.

METHODS

Subjects

The experimental procedures were approved by the Vanderbilt University Institutional Review Board and were in accordance with the Declaration of Helsinki. Eight subjects (4 male), having no known personal history of cardiovascular, endocrine, metabolic, neurological, or neuromuscular disorders, participated in the study. The subjects had mean (standard error, SE) age=22.5 (0.4) yrs, height=177.2 (5.0) cm, and mass=73.3 (6.2) kg. After receiving an explanation of the potential risks and benefits of participation in the study, the subjects provided written informed consent to participate.

Experimental Protocol

The subjects initially reported to the lab for an orientation session, during which health and MRI screening procedures were performed, informed consent was received, the MVC force was measured, and the contraction protocol was rehearsed. To measure the MVC force, the subject performed two or more 3 second isometric dorsiflexion contractions. The greatest force was recorded during each contraction. Contractions were performed until the maximum forces recorded during two contractions were within 5% of each other; the higher of the two values was then recorded as the MVC force. The subject then rehearsed the contraction protocol, using feedback from a computer screen to maintain a desired submaximal force level (see below).

During the testing sessions, the subject performed sustained isometric dorsiflexion contractions at either 30 or 60% of MVC (one contraction per testing session). These contraction intensities were selected because data exist in the literature concerning the time courses of metabolic changes during sustained isometric dorsiflexion exercise at these intensities (19) and because they correspond respectively to blood flow-permissive and blood flow-occluded conditions (20). The order of the testing sessions was randomized. During the contractions, mfMRI and near-infrared spectroscopy (NIRS) data were acquired. Details concerning the NIRS and mfMRI data acquisitions are provided below. The subjects were instructed not to consume caffeine or use tobacco during the six hours prior to a test session and not to use alcohol or perform moderate or heavy physical activity for at least 24 hours prior to each session. To ensure compliance with these instructions, the subjects completed a survey of their diet and lifestyle activities covering this period, and a test session was postponed in the event of noncompliance.

Isometric Contractions

Contraction Protocol

All exercise involved isometric dorsiflexion of the subject's self-reported dominant foot. The subject lay supine with his/her foot strapped into a home-built isometric exercise device. For outof-magnet studies, the exercise device was bolted to a grid plate attached to an exercise table; for in-magnet studies, the device was bolted to a similar grid plate on the patient bed of the imager. The ankle angle was 90° and the knee was supported by a bolster and slightly flexed (angle of leg flexion ~7.5°). The foot was firmly strapped to the exercise device using 3.8 cm wide nylon straps; to limit the involvement of the toe extensors, the straps were placed across the foot proximal to the base of the fifth digit. To eliminate hydrostatic effects on the circulation, the heart was at approximately the same level as the tibialis anterior (TA) muscle, the primary dorsiflexor. The imaging session began with the acquisition of scout and anatomical images. Following these, the subjects performed one 90 s isometric dorsiflexion contraction at either 30 or 60% of MVC. The subjects were verbally encouraged to maintain the target force throughout the contraction.

Force Data Acquisition and Analysis

The force acquisition system has been described in detail previously (12,21). Briefly, the isometric dorsiflexion device included an Interface Force (Interface Force, Scottsdale AZ) Model SSM-AJ-500 load cell, the signals from which were amplified and conditioned using a bridge amplifier (Model SGA/A, Interface Force) and then transmitted to an analog-to-digital conversion card (Model 6036E, National Instruments, Austin TX) via a serial connector box (Model SCB-68, National Instruments). Data acquisition software written in LabVIEW 7.1 (National Instruments) was used to collect force data at 1 kHz and provide real-time, 20 Hz visual feedback to the subjects in the form of a simulated LED panel on a computer screen (during familiarization procedures) or on a pair of MRI-compatible LCD goggles (for in-magnet studies). The baseline force offset was measured immediately prior to the contraction and accounted for when calculating relative force production. The force signals from the serial connector box were also sampled by the NIRS software (ISS Oximeter software v 2.14, ISS, Inc., Champaign IL).

The force data from the submaximal contractions were filtered at 25 Hz using a low-pass, fifth-order Butterworth filter. To correspond to the MRI data analysis described below, the average force production was calculated during 4 s windows centered on contraction durations of 8 s, 20 s, 56 s, and 88 s.

MRI Data Acquisition and Analysis

MRI data were obtained on a 3T Philips Intera Achieva MR Imager/Spectrometer. As the subject lay supine with the foot in the exercise device, but before being advanced into the imager, s/he contracted the dorsiflexors. During the contraction, the maximum cross-sectional area of the TA muscle was identified visually and by palpation. The probe from the MRI-compatible NIRS device was secured to the leg at the location of interest. A 6-channel SENSE RF torso coil was then positioned around both legs such that the NIRS probe fit into a small opening near the center of the coil. The subject was advanced into the magnet such that the area of interest was at the magnet's isocenter. The imaging procedures began with the acquisition of 3-plane gradient-echo scout images, used to verify to location of the TA's maximum girth. If necessary, the patient bed was moved so that this location was nearer the magnet isocenter. Then a high resolution, T₁-weighted anatomical image was obtained using a fast spin-echo sequence and slice thickness 7.5 mm, field of view 18×18 cm, acquired matrix 256×256 (reconstructed matrix 512×512), repetition time (TR)/echo time (TE)=500/18.6 ms, echo-train length=3, and number of excitations (N_EX)=2. A volume of interest was specified within the imaged portion of the leg and used for second-order shimming. Next, mfMRI data were obtained using a T₂-weighted spin-echo echo-planar imaging sequence for 20 s before, during, and 60 s after each contraction. The functional images were obtained using the same geometric parameters as the anatomical image, acquired matrix= 64×64 (reconstructed matrix=128×128), a fat-frequency selective saturation (SPIR) pulse applied immediately before the excitation pulse, N_EX=1, 60.3% k-space sampling, and the TR/TE = 4000/35 ms.

Image analysis was conducted using Matlab v. 7.0.1 (The Mathworks, Natick MA). In the anatomical images, a line was drawn normal to the skin surface and used to measure the depths of the subcutaneous fat and the muscle. Also, a region of interest (ROI) was drawn along the border of the TA muscle, carefully excluding any resolved vessels, and the TA's in-plane cross-sectional area was determined. In the functional images, a similar ROI was specified. The mean pre-contraction SI value was calculated and the subsequent values in the time series were divided by this value to calculate a normalized SI. For statistical purposes, the SI values in the images acquired at exercise durations (t) of 8, 20, 56, and 88 s were considered of particular interest. The t = 8 and t = 20 s time points correspond respectively to the times at which the initial rise and the nadir of the early dip in SI occur. The SI at end-exercise (t = 88 s) was used to characterize the maximum SI change undergone during the contraction, and the midpoint of the early dip and end-exercise times (t = 56 s) was used to characterize the early portion of the secondary rise in SI. Only the TA muscle was analyzed in this study because its location most corresponds to the depth of NIR light penetration (see Results and Discussion).

NIRS Data Acquisition and Analysis

As previously described (12,21), tissue oxygenation data were collected using a frequency domain, multi-distance NIRS oximeter (Model 96208, ISS, Inc.) and its accompanying software. Prior to each testing session, the oximeter was calibrated using a block with known absorption and scattering coefficients and the performance was verified using a second block with different optical properties. A rigid emitter-detector head was placed as near as feasible to the region of interest and held in place using an elastic strap. The strap was adjusted to be just tight enough to prevent motion of the probe during contraction. Following removal of the probe, a small impression of the probe and a slight reddening were observed in the skin, which we interpreted to reflect good contact of the probe with the skin. In one case, it was necessary to shave the hair overlying the TA muscle.

The oximeter emits light through fiber optic cables from four laser diodes at a wavelength of 730 nm and from four laser diodes at a wavelength of 860 nm. The oximeter head is arranged such that the four fibers emitting light at each wavelength are located at different distances from a single detector (2.0, 2.5, 3.0, and 3.5 cm). The light is intensity-modulated at 110 MHz and the AC, DC, and Phase components of the light at each distance are measured to determine the absorption and scattering coefficients. These data were used to calculate total hemoglobin ([THb]) and %HbO₂ according to the manufacturer's algorithms. The data were sampled at 6.45 Hz and a 7 point (1.09 s) moving average was applied during data analysis, which accurately preserves the [THb] and %HbO₂ kinetic information (21).

The mean pre-contraction [THb] and %HbO₂ values were calculated for the 10 s period before each contraction. Two analyses were conducted using the data from the exercise periods. The first was a discrete analysis, in which the average [THb] and %HbO₂ values were calculated during 4 s time windows centered on contraction durations of 8, 20, 56, and 88 s. The second was a continuous analysis of the %HbO₂ data, which were fit to:

$$\%Hb{O}_2=A_0+A_1\cdot \exp (-(t-TD)∕\tau ),$$ [1]

where A₀ is the baseline value, A₁ is the amplitude of a primary decay component, TD is a time delay before the start of %HbO₂ decay, and τ is an exponential time constant. The data were fit using Matlab's built-in Nelder-Mead algorithm, which iteratively adjusts the parameter estimates in order to minimize the residual variances in an ordinary least-squares sense.

Calculation of Intra- and Extra-Vascular BOLD Effects

Intravascular BOLD effects on the SI were modeled by using the mean %HbO₂ time courses for the 30 and 60% MVC contractions and by assuming a constant 3% blood volume fraction. As described previously by Meyer et al. (8), the blood transverse relaxation rate, R_2,Blood, was calculated as:

$$R_{2,Blood}=3.4+6.1\cdot (1-Y)+15\cdot {(1-Y)}^2,$$ [2]

where Y is the fractional oxyhemoglobin saturation (Y = %HbO₂/100). In order to calculate the effect of blood oxygenation changes on the image SI, the intracellular (V_Intra) and interstitial (V_Inter) volume percentages were assumed respectively to be 88 and 9% and to have intracellular (R_2,Intra) and interstitial (R_2,Inter) R₂ values of 28.6 and 8 s^–1 ((2) and references therein). The R₂ of the tissue parenchyma (R_2,Paren) was calculated as the average of R_2,Intra and R_2,Inter, weighted by cell volume fraction. The resulting SI changes were calculated using:

$$SI=V_{Blood}\cdot \exp (-TE\cdot R_{2,Blood})+(1-V_{Blood})\cdot \exp (-TE\cdot R_{2,Paren}),$$ [3]

with the experimental TE value of 35 ms. The SI changes were normalized such that pre-exercise SI=1.0 and the SI change was calculated as the variation from this value.

Extravascular BOLD effects on the SI were calculated as follows. First, the magnetic susceptibility difference between the blood and tissue (Δχ) was calculated using calculated magnetic susceptibility values for the blood χ_Blood) and tissue parenchyma (χ_Paren). To calculate χ_Blood, Eq. 6 of Spees et al. (14) was used. The hematocrit (Hct) was assumed to be 0.3 and the values of Y were determined from the measured %HbO₂ time courses. To calculate χ_Paren, the magnetic susceptibility of interstitial fluid (χ_Inter) was assumed to be similar to that of plasma and the magnetic susceptibility of intracellular fluid (χ_Intra) was assumed to depend on the fractional oxy-myoglobin (oxy-Mb) saturation, Y_Mb, in a similar manner to the magnetic susceptibility dependence of the red blood cell on Y. Y_Mb was calculated as:

$$Y_{Mb}=\frac{P{O}_2}{P_{50}+P{O}_2}$$ [4]

where PO₂ is the tissue partial pressure of oxygen determined from Y and P₅₀ is the PO₂ corresponding to half-maximal oxy-Mb saturation (=2.8 mmHg; (22)). Equation 1 of Spees et al. was then modified to account for Mb's intracellular concentration ([Mb] = 200 μM), single heme group (such that the paramagnetic contribution to the molar magnetic susceptibility of deoxy-Mb is 12,020×10^–6 ml/mol), molar volume in solution (assumed to be 12,069 ml/mol, or one-fourth that of Hb), and molecular weight (16,113 g/mol). χ_Paren was calculated as the average of χ_Inter and χ_Intra, weighted for cell volume fraction. Next, the parallel cylinders model presented by Stables et al. (15) was used to determine changes in R_2,Paren as a function of exercise duration. The parallel cylinders model was assumed to be applicable and the mean capillary orientation was assumed to be similar to the mean fiber orientation because only ~10% of muscle capillaries’ lengths are accounted for by branching and tortuosity; otherwise, they run parallel to the fiber axis (23). To determine the R_2,Paren changes, Δχ values were calculated from the %HbO₂ time course, making the following assumptions: V_Blood=3%, a 6 μm capillary radius, a 25° capillary orientation to the B₀ field (24), and a diffusion coefficient perpendicular to the capillaries of 1.3×10^–5 cm²/s (25,26). Finally, the SI change due to extravascular BOLD effects was calculated by substituting V_Blood, the resting value for R_2,Blood, and the time-dependent values of R_2,Paren into Eq. 3.

Statistical Analysis

Statistical analyses were performed using SPSS v. 15 (SPSS, Chicago IL). Descriptive data include the mean and SE. To compare the mean force production at the four exercise durations of interest, a repeated measures analysis of variance (ANOVA) was computed. The data were analyzed separately for each contraction intensity. The SI values at t=8, 20, 56, and 88 s were compared using the General Linear Model (GLM; Intensity×Duration). For discrete analysis of the NIRS variables (%HbO₂ and [THb]), a similar two-factor GLM was employed, but also including the pre-contraction data (represented as t=0 s). The %HbO₂ kinetic parameter estimates for the 30% and 60% MVC conditions were compared using a two-tailed, paired Student's t-test. Statistical comparisons were considered significant at p<0.05.

RESULTS

Anatomical Characteristics

The mean cross-sectional area of the TA muscle was 6.01 (SE 0.77) cm². The mean subcutaneous fat layer thickness was 0.50 (SE 0.12) cm and the mean depth of the TA muscle was 2.64 (SE 0.20) cm below the subcutaneous fat layer.

Force Production

The mean MVC force was 298.6 (SE 26.8) N. At 30% MVC, the subjects were able to maintain the target force level at all exercise durations (p=0.997; Figure 1). During the 60% MVC contractions, four subjects were unable to maintain the target force at the end of the exercise bout (mean contraction intensity = 40.0; SE = 4.4 %) and four subjects were able to maintain the target force (mean contraction intensity = 59.6; SE = 0.8 %). The four subjects unable to maintain the target force were the four strongest subjects. Overall, there were no significant changes in force production during the 60% MVC contractions (p=0.307).

Figure 1.

Actual contraction intensity during 4 s windows centered on exercise durations of 8, 20, 56, and 88 s during the 30% and 60% MVC tasks. Mean, SE is shown; no significant differences were observed for the group, but four subjects underwent considerable fatigue (mean intensity at end-exercise=40.0% MVC) and four subjects were able to maintain the target force at end-exercise (mean intensity at end- exercise=59.6% MVC).

NIRS Observations

The mean pre-contraction %HbO₂ values at 30 (66.3%, SE 2.6%) and 60 (70.2%, SE 2.3%) percent MVC did not differ significantly (p=0.06). Example NIRS data for 30% and 60% MVC contractions are shown in Figures 2A and 2B, respectively; also shown are the lines of best fit of the %HbO₂ data to Eq. 1. The mean %HbO₂ values from the discrete analysis (t = 0, 8, 20, 56, and 88 s) are shown in Figures 2C. For the %HbO₂ data there was a significant Duration main effect (p<0.001), as illustrated in Figure 2C, and a significant Intensity×Duration interaction (p=0.018) such that the decline in %HbO₂ began sooner during the more intense contractions. Consistent with this, the mean kinetic parameter estimates for the %HbO₂ data in Table 1 show that at 60% MVC, the increase in oxygen extraction occurred with a shorter time delay (9.6 s at 30% MVC vs. 7.4 s at 60% MVC; two-tailed p=0.025) and with more rapid kinetics (the mean values for τ at 30 and 60% MVC were 12.4 and 5.0 s, respectively; p<0.0005). However, the amplitude terms (A₀ and A₁) were similar at 30 and 60% of MVC (p=0.19 and 0.41, respectively), causing the end-exercise values to be similar. As a result, the %HbO₂ levels differed most between the two exercise intensities between the exercise durations of ~10 and ~30 s. Figure 2D shows %HbO₂ time courses generated from the mean kinetic parameter estimates. The [THb] did not change significantly (the p values for the Intensity main effect, the Duration main effect, and the Intensity×Duration interaction were 0.672, 0.052, and 0.2, respectively).

Figure 2.

NIRS observations during the 30 and 60% MVC tasks. A and B. Example %HbO₂ (gray points) and [THb] (black points) data during the 30% MVC (A) and 60% MVC (B) tasks. The line indicates the best fit of the %HbO₂ data to Eq. 1. C. Mean, SE of %HbO₂ values for the 30% and 60% MVC tasks from the discrete analysis. Asterisks indicate significant differences from the t=0 and 8 s exercise durations. D. %HbO₂ time courses for the 30% MVC (gray lines) and 60% MVC (black lines), generated from the kinetic parameter estimates in Table 1. E. Mean, SE of [THb] values for the 30% and 60% MVC tasks from the discrete analysis. No significant differences were observed.

Table 1.

Kinetic parameter estimates for %HbO₂, obtained from best fit to Eq. 1.

Intensity (% MVC)	A0 (%)	A1 (%)	TD (s)	τ (s)
30	30.8 (6.7)	35.0 (6.1)	9.6* (0.6)	12.4* (1.8)
60	34.9 (5.9)	34.2 (4.1)	7.4 (0.5)	5.0 (0.8)

^*Indicates a significant difference between the mean values for contractions at 30 and 60% of MVC. Mean (SE) is given.

MRI Observations

Example MRI data from the TA during the 30% and 60% MVC contractions are shown in Figure 3A; the corresponding group-mean data for the 30 and 60% MVC conditions are shown in Figure 3B. Also shown are the lines of best fit to higher order polynomials (30% MVC: 3^rd order polynomial; 60% MVC: 8^th order polynomial); in each case, the minimum polynomial order required to avoid structured residuals was used. The GLM revealed that the SI changes during 60% MVC contractions were greater than during 30% MVC contractions (p=0.002). There was also a significant main effect for Duration (p=0.011) and a significant Intensity×Duration interaction (p=0.002). For the 30% MVC contractions, the SI change at t = 88 s was greater than the SI change at t = 8 and 56 s. For the 60% MVC contractions, the SI changes at t = 56 and 88 s were greater than those at t = 8 s and 20 s.

Figure 3.

SI from the TA muscle during the 30% MVC (gray lines and points) and 60% MVC (black lines and points) conditions. A. Sample data from the same subject whose data were illustrated in Figure 2. B. Group-mean data from 8 subjects. Lines indicate the best fit to 3^rd and 8^th order polynomials (for 30 and 60% MVC< respectively). Error bars indicate the SE. C. Pre-exercise SI and SI from the four points of the discrete analysis, illustrating significant Intensity, Duration, and Intensity×Duration main and interactive effects. Within each intensity level, ^Aindicates a significant difference between the indicated time point and t = 8 s; ^Bindicates a significant difference between the indicated time point and t = 20 s; ^Cindicates a significant difference between the indicated time point and t = 56 s.

Calculated Intra- and Extra-Vascular BOLD Effects

Figure 4A shows the R_2,Blood values (calculated using Eq. 2) corresponding to the mean %HbO₂ data presented in Figure 2D. Figure 4B shows predicted changes in R_2,Paren due to extravascular BOLD effects, expressed as a function of tissue PO₂. The inset in the lower right portion of the figure shows a detail of the portion of this curve from 80-100 mmHg, and predicts that at 95 mmHg, a match in the magnetic susceptibility of the blood and tissue parenchyma occurs such that there is no extravascular BOLD effect. The inset in the upper left portion of the figure shows a detail of the portion of the curve from 0-10 mmHg. These data predict that as Mb starts to deoxygenate, the magnetic susceptibility match between the blood and tissue parenchyma improves such that extravascular BOLD effects start to reverse at a tissue PO₂ near to the P₅₀ of Mb. Finally, Figure 4C shows the results of using the data in Figures 2D and 4B to predict the R_2,Paren values (as a function of extravascular BOLD effects only) as a function of exercise duration.

Figure 4.

BOLD effects on muscle transverse relaxation rates. A. Predicted R_2,Blood as a function of exercise duration. B. Dependence of R_2,Paren on tissue PO₂. Insets show details of this relationship at low tissue PO₂ (upper left) and high tissue PO₂ (lower right). In both cases, a reversal of the extravascular BOLD effect is predicted due to improved magnetic susceptibility matching between blood and tissue parenchyma. C. Predicted R_2,Paren as a function of exercise duration. The Methods and Results sections provide details concerning the assumptions and experimental data used in modeling these effects.

The SI changes corresponding to these R₂ changes are shown in Figure 5A (which considers intravascular BOLD effects only); Figure 5B (which considers extravascular BOLD effects only); and Figure 5C (which treats the total BOLD contribution to SI changes as the sum of the intravascular and extravascular BOLD effects). Owing to the similar end-exercise %HbO₂ values at 30 and 60% MVC, the predicted absolute BOLD SI contributions were similar for both intensities, and were ~0.009, ~0.001, and ~0.01 normalized SI units for the intravascular, extravascular, and total BOLD contrast mechanisms, respectively. Figure 5D shows the lines of best fit for the group-mean data from 30 and 60% MVC (i.e., from Figure 3B) as thick lines. Also shown (as thin lines) are the lines of best fit with the total BOLD signal change subtracted from them, illustrating the non-BOLD signal change. Finally, Figure 5E shows the relative contribution of the total BOLD SI changes to total SI changes as a function of exercise duration, for the 30 and 60% MVC contractions.

Figure 5.

BOLD effects on MRI signal intensities. In all panels, 30% and 60% MVC contraction data are indicated with gray and black lines, respectively. A-C. Predicted effects of intravascular, extravascular, and total BOLD contrast mechanisms. D. mfMRI SI lines of best fit from experimental studies (thick lines) and with total BOLD effects subtracted out (thin lines). E. Relative contribution of total BOLD effects on SI, shown as a function of exercise duration for 30 and 60% MVC tasks.

DISCUSSION

This study presents simultaneously acquired data concerning the time courses of [THb], %HbO₂, and T₂-weighted SI changes in isometrically contracting muscle. The data provide novel physiological information about the blood volume and %HbO₂ responses to isometric contractions, revealing that the blood volume does not change during sustained submaximal isometric contractions and that the time course of %HbO₂ changes is exercise intensity-dependent. Moreover, the data provide new insights into the physiological basis for the mfMRI time course. First, the data demonstrate that the temporal profile of the time course depends on exercise intensity, with an earlier occurrence of the secondary rise in SI at higher exercise intensities. Also, BOLD effects begin to reduce SI at exercise durations of ~5-10 s, being somewhat earlier with higher exercise intensities. This reduction in SI is predicted to occur along the %HbO₂ time course and at later exercise durations results in a ~1% loss of SI.

Effect of Exercise Intensity on the mfMRI Time Course

It has previously been shown – using a large number of different exercise modalities, muscle groups, and definitions of exercise intensity – that exercise-induced T₂ changes increase with exercise intensity (17,27,28). In particular, it is the relative exercise intensity that is most important in determining the T₂ change (28). These observations follow logically from the indirect dependence of the T₂ change on the accumulation of osmolytes and protons produced by cellular energy metabolic pathways, and the direct relationship of the rates of metabolism on exercise intensity. In concert with experimental evidence that the T₂ change of exercise is related to electromyographic activity (29), these observations have also been used to support the use of T₂ as a non-invasive, imaged-based measure of the extent of neural activation (30,31)

It is therefore unsurprising that the 60% MVC contraction intensity elicited a greater T₂ change than the 30% MVC intensity; at 30% MVC, the end-exercise SI increase was ~8%, whereas at 60% MVC it was ~17%. Of greater interest is the observation that the temporal profile of T₂-weighted SI changes with increased exercise intensity. At 30% MVC, the SI time course had the visual appearance of an initial rise, early dip, and secondary rise; but only the end-exercise SI value was significantly elevated relative to the other time points in the discrete analysis. At 60% MVC, however, the SI change at the t = 56 s and t = 88 s time points were both greater than those at the t = 8 s and t = 20 s time points. This observation, coupled with the earlier glycolytic activation at 60% MVC (~30 s) than at 30% MVC (~45 s) (19), is consistent with the hypothesis that T₂-weighted SI changes at later exercise durations relate mostly to the accumulation of the end-products of anaerobic glycolysis (18).

Contributions of BOLD Effects to the mfMRI Time Course

Anatomical Features Potentially Affecting the NIRS Observations

The calculation of BOLD-dependent SI changes relies heavily on the validity of the NIRS data. One potentially confounding factor arises from the observation that in optical spectroscopy, the light penetration depth is equal to one-half of the light emitter-detector spacing (32). This produces a strong correlation between the thickness of the subcutaneous fat layer and the measured NIRS signals (21,33). For our device, the emitter-detector spacings of 2.0-3.5 cm would correspond to light penetration depths of 1.0-1.75 cm. Notably, the mean subcutaneous adipose layer thickness was 0.50 cm (less than the minimum light penetration depth) and the TA muscle extended an additional 2.64 cm deep (greater than the maximum light penetration depth). Thus, the NIRS data correspond primarily to the TA muscle. More specifically, because the TA's central aponeurosis divides the muscle into two compartments of approximately equal depth, the 1.0 and 1.25 cm maximum light penetration depths of our device correspond to the superficial compartment and the 1.5 and 1.75 cm maximum light penetration depths correspond the deep compartment. When the overall light path is considered, however, it is likely that there is a slightly greater weighting of the superficial compartment than of the deep compartment in the measured signals.

Contributions of Other Heme-Containing Molecules to the NIRS and MRI Signals

A second potentially confounding factor is the existence of Mb and other heme-containing molecules in muscles and their potential contributions to the NIRS signals and BOLD effects. We argue that Hb is a much more important determinant of skeletal muscle NIRS and BOLD signals than Mb. The first reason for this that there is a greater dynamic range of oxy-Hb saturation values than oxy-Mb saturation values. The %HbO₂ values that we observed ranged from 30 to 75%, reflecting tissue PO₂ values of ~20-40 mmHg. At these PO₂ values, Mb would be 88-93% saturated with oxygen, which is almost 10-fold less than the range of %HbO₂ values. Therefore, BOLD SI changes would be more influenced by Hb than by Mb. Also, concentration differences between Hb and Mb cause Hb to be the main contributor to skeletal muscle NIRS signals. This statement is supported by Seiyama et al.'s study of the perfused rat hindlimb preparation (34), which is applicable because the [Mb] is similar in human and rat muscles (35-38). In this study, >90% of the NIRS signal disappeared when the perfusate was changed from blood to a fluorocarbon (34), and so Hb must contribute to skeletal muscle NIRS signals much more than Mb. With respect to the relative contributions of Hb and Mb to BOLD effects, Lebon et al. (10) reported that during arterial occlusion, T₂* changes correspond to the calculated Hb desaturation time course, not the observed Mb desaturation time course. Thus we conclude that both the NIRS data and calculated BOLD signals reflect changes in Hb, not Mb.

[THb] and %HbO₂ Changes During Sustained Isometric Dorsiflexion and Their Effects on T₂-Weighted Signal Intensity

In contrast to our expectations, a rise in [THb] resulting from capillary recruitment was not in fact observed; as a consequence, the initial positive BOLD contribution to SI that we had predicted did not occur. What was observed was a non-significant (p=0.052) tendency for the [THb] to decrease (by 10%; Figure 2E). Because the effect was, in fact, non-significant, no blood volume changes were modeled. If the apparent fall in [THb] was a real effect, however, it would reflect the ejection of blood from the venous circulation (39) due to intramuscular pressure increases (40). Even if we had accounted for this hypothetical 10% decrease in blood volume at end-exercise, it would have produced practically non-significant differences in R_2,Paren (0.0077 s^–1 greater than that shown in Figure 4C) and the total BOLD contribution to SI (0.0001 normalized units less negative than that shown in Figure 5C).

There were predictably large decreases in %HbO₂ during exercise, which were similar at end-exercise for the 30 and 60% MVC contractions. However, the smaller time delay and time constant associated with the 60% MVC contraction than with the 30% MVC contraction (Table 1) had the effect of causing a more rapid decrease in %HbO₂ at 60% MVC. This faster fall was probably the combined result of a higher oxygen demand and a reduced oxygen supply for the more intense contractions. The latter effect would result from the higher intramuscular pressure level associated with the 60% MVC contractions (40), restricting blood flow more at 60% MVC than at 30% MVC. In flow-limited systems maintaining high oxidative metabolic rates, oxygen extraction occurs more quickly than in those with higher rates of flow (41).

By design, the calculated time courses of intravascular and extravascular BOLD effects follow the %HbO₂ time courses. However, the two types of BOLD effects have very different relative magnitudes, with the SI changes resulting from intravascular BOLD effects being nearly 10-fold greater than those due to extravascular BOLD effects. This observation is similar to those in previous reports concerning the relative contributions of intra- and extra- vascular BOLD effects in skeletal muscle at 1.5T (8,16). The total BOLD effect produces an absolute SI change of –1% of the baseline SI at the end of exercise (regardless of intensity), which is of similar order to the BOLD signal transients observed following brief MVC's (8,12). When expressed relative to the total SI change, however, the total BOLD SI contributions at each exercise intensity are quite large. At 30% MVC, the relative BOLD SI contribution has a maximal value of –26%, which occurs at t = 40 s. At 60% MVC, the relative BOLD SI contribution evolves more rapidly than at 30% MVC. Because the non-BOLD SI change is greater at 60% MVC, the relative contribution is smaller than at 30% MVC, having a maximal value of –11% at t = 20 s. BOLD effects therefore make an important, but exercise-intensity dependent, contribution to T₂-weighted SI changes during exercise.

Summary and Conclusions

In this study we have presented and analyzed simultaneously acquired NIRS and T₂-weighted SI data acquired during sustained isometric dorsiflexion contractions at 30 and 60% of MVC. The NIRS data revealed no significant changes in blood volume during the contractions and pronounced decreases in %HbO₂. The end-exercise %HbO₂ values were similar at each intensity. However, the decreases occurred more quickly at 60% MVC, an observation that we attribute to a greater level of blood flow restriction. The %HbO₂ values were used to predict the relative contributions of intra- and extra- vascular BOLD contrast mechanisms to the SI changes during mfMRI. This analysis revealed that while the absolute contributions of BOLD effects to SI changes do not differ between sustained dorsiflexion contractions at 30 and 60% of MVC, the greater T₂ change in the tissue parenchyma at 60% MVC causes the relative contribution of BOLD effects to be smaller at this intensity. Finally, we observed that the secondary rise in SI, which we have previously suggested is determined primarily by the accumulation of the end-products of anaerobic glycolysis, occurs earlier during contractions at 60% MVC than during contractions at 30% MVC. This indicates that not only the SI change but also the temporal profile of mfMRI SI changes is exercise intensity dependent.