Exercise Induced Fluid Shifts are Distinct to Exercise Mode and Intensity - a Comparison of Blood Flow Restricted and Free Flow Resistance Exercise.

AIM
MRI can provide fundamental tools in decoding physiological stressors stimulated by training paradigms. Acute physiological changes induced by three diverse exercise protocols known to elicit similar levels of muscle hypertrophy were evaluated using muscle functional magnetic resonance imaging (mfMRI).


METHODS
The study was a cross-over study with participants (n=10) performing three acute unilateral knee extensor exercise protocols to failure and a work matched control exercise protocol. Participants were scanned after each exercise protocol; 70% 1 repetition maximum (RM) (FF70); 20% 1RM (FF20); 20% 1RM with blood flow restriction (BFR20); free-flow (FF) control work matched to BFR20 (FF20WM). Post exercise mfMRI scans were used to obtain interleaved measures of muscle R2 (indicator of edema), R2' (indicator of deoxyhemoglobin), muscle cross sectional area (CSA) blood flow and diffusion.


RESULTS
Both BFR20 and FF20 exercise resulted in a larger acute decrease in R2, decrease in R2', and expansion of the extracellular compartment with slower rates of recovery. BFR20 caused greater acute increases in muscle CSA than FF20WM and FF70. Only BFR20 caused acute increases in intracellular volume. Post-exercise muscle blood flow was higher after FF70 and FF20 exercise than BFR20. Acute changes in mean diffusivity were similar across all exercise protocols.


CONCLUSION
This study was able to differentiate the acute physiological responses between anabolic exercise protocols. Low-load exercise protocols, known to have relatively higher energy contributions from glycolysis at task failure, elicited a higher mfMRI response. Noninvasive mfMRI represents a promising tool for decoding mechanisms of anabolic adaptation in muscle.


INTRODUCTION
Despite solid evidence that resistance training induces muscle hypertrophy (1-3), a deeper physiological understanding of the underlying mechanisms is still required. Monitoring and mapping physiological factors in vivo are fundamental tools in identifying and decoding physiological stressors of specific training paradigms associated with the stimulation of muscle hypertrophy. Mechanical tension, for instance, has long been suspected to be a physiological stressor, as resistance training with high loads is known to elicit muscle hypertrophy in both healthy individuals and a wide range of patient populations of all ages (4)(5)(6)(7)(8)(9)(10). In recent years, low load resistance training performed to failure or performed with blood flow restriction (BFR) has also gained attention as similar hypertrophy is observed despite the low mechanical tension (5,7,8,(11)(12)(13)(14)(15). The diversity of these training strategies has brought focus to a larger range of physiological stressors which are suspected to trigger pathways leading to hypertrophy which include metabolic stress (16)(17)(18), myocellular swelling (16,17,19), transient hypoxia (13,20,21), as well as mechanical tension (22)(23)(24)(25).
Muscle functional magnetic resonance imaging (mfMRI) techniques can evaluate useful parameters with regard to the aforementioned physiological stressors. In a recent mfMRI study, acute low load BFR exercise induced accentuated R 2 decreases, increases in R 2 0 , reduced blood flow, and similar tissue fluid mobility (diffusion) compared to a heavier load in free-flow conditions (26). R 2 changes of muscle tissue during exercise, often referred to as a "T 2 shift" (T 2 = 1/R 2 ), correlates with acute muscle activity (27)(28)(29) and is metabolic in nature, as it is primarily driven by the changes in composition, pH, and volume of intracellular fluid following the accumulation of metabolites, such as lactate, after loading (30)(31)(32)(33)(34). R 2 data can also be used to estimate changes in the relative extra-/intracellular volume of muscle tissue. Increases in R 2 0 (indicator of deoxyhemoglobin) combined with blood flow measures can monitor transient hypoxia (35)(36)(37), and diffusion imaging can detect exercise-induced micro damage (38). Despite recent advances, however, mfMRI has seldom been applied to study exercise protocols known to elicit muscle hypertrophy.
The present study employed quantitative magnetic resonance imaging (MRI) techniques to investigate the acute physiological response of three differing resistance exercise strategies which are known to induce similar levels of hypertrophy in longitudinal ("chronic") settings (4)(5)(6)(7)(8)10). The three protocols encompassed high-load 70% one repetition maximum (1RM) and low-load 20% 1RM (8,11) in free-flow conditions, and a low-load (20% 1RM) performed in BFR condition. All of the protocols consisted of four sets of one-legged knee extension completed to contraction failure. All mfMRI measures were acquired repeatedly to quantify both the magnitude of exercise-induced changes but also the rate by which they return to pre-exercise levels, the recovery rate (k), (35,39). Recovery rates can both be used to disentangle the time course of physiological responses, thus reducing the impact of the post exercise delay time of the measurement but also serve as a separate quantitative parameter.
The specific aim of this study was to evaluate the exerciseinduced mfMRI response by low-load BFR exercise and low/ high load free-flow exercise to failure compared to a work matched controlled free-flow exercise. A cross evaluation of mfMRI results is performed using measures of macrovascular blood flow, water diffusion, muscle cross sectional area (i.e., macroscopic muscle swelling), and an estimate of extracellular/intracellular water fraction. It was hypothesized that low-load BFR resistance exercise performed to failure would lead to amplified decreases in R 2 and R 2 0 , increases in muscle cross sectional area (CSA) and intracellular volume, and slower recovery rates toward baseline conditions compared to free-flow conditions.

Participants
Ten recreationally active healthy young men volunteered to participate in the present study. Inclusion criteria were recreational physical active males aged between 18 and 30 where all candidates having a known medical condition or previous leg injury were excluded. Before entering the study, all participants provided their written informed consent in accordance with the Declaration of Helsinki. The experimental procedures were approved by the Regional Ethics Committee of the Capital Region of Denmark (protocol no. H-1-2013-146).
One week before the experimental procedures, participants had their unilateral knee extensor 5-repetition maximum (RM) load determined for the knee extensors of each leg separately using a commercial leg extension training device (TecnoGym, Inc.). Subsequently, participants were familiarized with blood-flow restricted strength exercise by performing two sets to failure with the right leg at 20% of 1RM. Blood-flow restricted exercise was performed using a 11-cm-wide pneumatic cuff (model: 60-7600-005, Zimmer, Dover, OH). The occlusion cuff was placed proximally on the thigh and inflated to a pressure corresponding to 30 mmHg above the participants resting diastolic blood pressure measured at rest.

Exercise Protocols
The experimental procedures consisted of four different protocols of isolated unilateral knee extensor resistance exercise (Table 1) performed over two separate scanning days. In all exercise protocols, four exercise sets were completed, after which participants were placed in the MRI scanner for acquisition of data (Fig. 1). The two exercise protocols performed on the first experimental day (day 1) were: leg A: four sets of knee extension with 20% 1RM to task failure with blood-flow restriction (BFR20) and leg B: four sets of knee extension with 70% of 1RM to task failure with free-flow resistance exercise (FF70).
The two exercise protocols performed on the second experimental day (day 2) were: leg A: four sets of knee extension with free-flow 20% 1RM performing the same number of repetitions for each set as were performed during BFR20 (i.e., work-matched), on the same leg that had performed BFR20 (FF20 WM ). The FF20WM protocol was introduced to serve as a control condition to the BFR20 exercise protocol. Leg B: four sets of knee extension with free-flow 20% 1RM performed to task failure with the same leg that performed FF70 on day 1 (FF20). The first and second experimental days were separated by a minimum of 48 h. The assignment of "leg A" or "leg B" as being the participants right or left leg was random.
On both day 1 and day 2 the planned exercise protocols were performed in a random order, separated by at least one hour of passive rest. For each protocol, sets were performed in a moderate tempo ($ 3 s/repetition) separated by 45-s rest between sets. Upon cessation of each exercise protocol, participants were scanned to repeatedly measure a number of MRI parameters (see Fig. 1) over a postexercise period of 20 min. In the case of the BFR20 protocol, the cuff remained inflated throughout the four exercise sets and was released seconds before the first post-exercise MRI measurement. Since the scanning field of view covers both legs, scans obtained after the second exercise session contained not only acute data for the leg having just performed the second exercise protocol but also late timepoints ($80 min) from the leg having performed the first exercise protocol of the day as well. As the order of the two exercise protocols performed each day was randomized, late time points were obtained for five of the 10 participants for each protocol.

MRI Sequences
Diffusion, R 2 and R 2 Ã data were acquired over the same field of view (475 Â 215 mm 2 ) covering a transaxial slice of the legs at mid femur and included SPAIR fat suppression on a Philips 3 T Achieva dStream scanner using a 16-channel anterior coil strapped over the thighs of both legs. Axial scans were performed at 50% femur length (mid-thigh) as identified from a frontal scout scan covering the entire femur.
After each exercise paradigm, the placement of the coil and participant was controlled to avoid displacement.
Diffusion weighted (DWI) data were acquired for 13 b-values (0, 100, 200, 250, 300, 350, 400, 450, 550, 700, 850, 1,000, 1,300 s/mm 2 ) and six directions for three 5-mm slices using a TR/TE of 1,100 ms/98 ms. Data were acquired in a matrix of 160 Â 69 with an EPI factor of 69 reconstructed to a 256 matrix with reconstructed voxels of 1.86 Â 1.86 mm 2 . Maps of the mean apparent diffusion (MD) and fractional anisotropy (FA) were calculated from a monoexponential fit to the diffusion images comprising of b-values from 200 to 800 s/mm 2 using the Diffusion Toolbox version 3.0 from FSL (40). Data from the b-values above 800 were acquired for an analysis outside the scope of the present study. Leg A BFR20 20% 1RM with blood flow restriction completed to task failure each set. FF20 WM Free-flow 20% 1RM performing the same number of repetitions for each set as were performed during BFR20 (i.e. work matched).

Leg B
FF70 Free-flow 70% 1RM performed to task failure each set. FF20 Free-flow 20% 1RM performed to task failure each set.
All participants performed four separate unilateral knee extensor resistance exercise protocols in total, two on each scanning day (day 1 and day 2). Each protocol consisted of four sets of exercise performed at a moderate tempo and with 45-s rest between sets where participants were placed in the MRI scanner immediately after the fourth set. The assignment of the right/left leg to be "leg A" or "leg B" was randomized as was the order exercise protocols performed on a given scanning day. The BFR20, FF70, and FF20 exercise protocols have previously been documented to elicit similar degrees of hypertrophy over longer training periods as opposed to FF20 WM which served a control condition. BFR20, 20% 1RM with blood flow restriction; FF, free flow; FF20, free flow 20% 1RM; FF20 WM , free-flow control work matched to BFR20; FF70, free flow 70% 1RM; RM, repetition maximum; R 2 0 , indicator of deoxyhemoglobin; T 1/2 , half-life. Figure 1. Experimental protocol. On each experiment day, participants performed two separate exercise protocols, one with each leg, in a randomized order. Both scanning days began with a baseline scan where all measures were acquired once. After baseline, participants performed the first leg's unilateral knee extensor protocol for the day and then underwent a 20-min MRI scan with interleaved measures of R 2 0 , R 2 , flow (PC), and diffusion (DWI). After scanning, participants rested for 40 min, and then performed the second leg's exercise protocol of the day on the opposite leg followed by a second post exercise MRI scan. The scanning field of view located the mid-thigh axial region and comprised both legs centered, to allow both acute measures of the exercised leg and late measures of the opposite leg to be included in the second scan session. The scanning procedure on the second experimental day was identical to the first, except that BFR20 and FF70 protocols were substituted with FF20 WM , and FF20 protocols, respectively. BFR20, 20% 1RM with blood flow restriction; DWI, diffusion weighted imaging; FF, free flow; FF20, free flow 20% 1RM; FF20 WM , free-flow control work matched to BFR20; FF70, free flow 70% 1RM; PC, phase contrast; RM, repetition maximum; R 2 , indicator of edema; R 2 0 , indicator of deoxyhemoglobin. R 2 data were acquired using a turbo spin echo sequence (TSE factor 20) with TR 1,500 ms and 20 echo times (TE = n Â 4.9 ms). One 5-mm slice was acquired with a 68 Â 160 matrix and reconstructed to a 256 recon matrix with a pixel size of 1.75 Â 1.75 mm 2 having a sense factor of 1.5 and a scan time of $70 s.

R2'
A monoexponential R 2 map was calculated as a least squares minimization of: where S i is the MRI signal as a function of echo time (TE), S o the unattenuated signal, and C is a constant that includes signal from spins with a range of relaxation rates much slower than that of the majority of spins in the tissue [R 2(C) << R 2(tissue) ] with limited attenuation in the given range of TE values applied in the present experiments. The fraction of the total signal that appears constant, C/ (C þ S 0 ), was used as an estimate of the fraction of water in the extracellular compartment (EC) of muscle tissue (34), and (1-EC) · CSA was used as an estimate of intracellular volume.
CSA/R 2 Ã data were acquired using a gradient echo sequence with 10 echoes TE1/DTE 8.0 ms/7.4 ms TR 266 ms, a flip angle of 40 degrees, and a fast field echo readout with an EPI factor of 5. Four dynamics of three 5-mm slices covering both legs at the mid femur with a field of view of 475 Â 215 mm 2 reconstructed to a 256 Â 256 recon matrix with a pixel size of 1.85 Â 1.85 mm 2 . SPAIR fat suppression was applied. CSA was determined by drawing a single region of interest around the quadriceps muscle.
The total rate of dephasing (R 2 Ã ) was calculated by least squares minimization of the equation: from which R' 2 could be calculated as the difference between R 2 Ã and R 2 : Muscle blood flow was calculated using phase contrast MRI (PC-MRI) with 100 cm·s À1 velocity encoding and an acquisition time of 10 s as previously described (26). A single 6-mm slice with a 256 Â 132 acquisition matrix reconstructed to voxel dimensions of 1.6 mm Â 1.6 mm was acquired using a TR/TE of 10.7/6.5 ms and turbo field echo factor 50. Macrovascular arterial and venous vessels within the quadriceps muscles were identified using k means clustering (5 bins and unrestricted cluster size using velocity maps and magnitude images as input and an adaptive threshold of 10 percent above average absolute water velocity of the quadriceps; Refs. 26, 41, and 42).
To characterize the rate of return to baseline levels following the acute exercise intervention, absolute changes in post exercise values for R 2 , R 2 0 , CSA, MD, FA, and blood flow were fitted to a function for exponential decay functions: with t denoting time from exercise stop in minutes, Q (t) the absolute difference for the given parameter from baseline as a function of t, Q post the absolute difference from baseline of the first post exercise measurement and TD post the time delay from exercise stop until post exercise values begin to decay toward baseline (TD post limited to a minimum t À TD post = 0). The parameter k denotes an exponential rate constant (units·min À1 ) reflecting the rate at which post exercise parametric values return to their baseline level following cessation of the exercise protocol. The rate of recovery is also expressed by its half-life, T 1/2 = ln(2)/k, which is the time in minutes required for the difference between the post exercise parametric value and baseline value to be reduced by half.

Statistical Analysis
Sample size was calculated based on the primary end point of R 2 changes in percent. A sample of eight participants was found necessary to detect a difference of 5 percentage points in R 2 changes between exercise protocols with >98% certainty based on a R 2 measurement uncertainty of 3%.
Values for R 2 , R 2 0 , FA, and MD of the quadriceps muscle were obtained from regions of interest in the vastus lateralis, vastus intermedius, vastus medialis, and rectus femoris muscles. A mixed-effects model was used to test for differences of measured changes in quadriceps R 2 , muscle CSA, FA, MD, and blood flow at the time of first post exercise measure between the four exercise protocols with participants set as a random effect. The first analysis tested for a significant effect of exercise protocol on the aforementioned variables (fixed effect = exercise protocol performed, dependent variable = change from baseline to first post exercise measure in variables' ROI values, random effect = participant). A second mixed model analysis was performed to compare the same variables for differences between day 1 and day 2 measures and right versus left leg to identify long-term effects from exercise protocols (fixed effect = leg A or leg B, dependent variable = difference in parameter's baseline measure between day 2 and day 1, random effect = participant). A third mixed model analysis was performed to test for correlations between each of the acute mfMRI-R 2 and R 2 0 -changes after exercise with changes in CSA, blood flow, EC, and MD. Significant correlates identified from the third mixed model analysis were entered into a linear regression model where goodness of fit was evaluated with a Pearson's adjusted R 2 value. Linear regressions having an adjusted R 2 value above 0.2 and a P value of less than 0.05 were considered minimal criteria for reporting correlations. Image coregistration, ROI analysis, calculations, and statistical analysis were performed with software created in MATLAB 2016 b (MathWorks, Natick, MA). Unless stated otherwise, values are presented as group mean ± SD. Data points in figures represent group mean values with error bars indicating standard error of the mean.

RESULTS
All recruited participants (age 24 ± 3 yr, exercise/week 7.4 ± 4.2 h) completed the four exercise protocols and MRI scans. All baseline measurements and absolute changes induced by the exercise protocols are presented in Table 2. The average 1RM for leg A (BFR20 and FF20WM) and leg B (FF70 and FF20) was 34 ± 7 and 35 ± 6 kg. Participants' right leg had a higher 1RM (0.8 ± 0.3 kg, P = 0.023), CSA (3.0±0.4 cm 2 , P = 0.013), and both lower R 2 (À1.1 ± 0.2 s À1 , P = 0.035) and FA (0.04 ± 0.01, P = 0.014) baseline values than the left leg. Due to right-left randomization, however, no significant differences between baseline measures of the participants' leg A and leg B at the first experimental day were observed. Average elapsed time between the first and second experimental day was 7 ± 3 days. Time between cessation of exercise to the first measure time point was 3.3 ± 0.5 min for muscle blood-flow, 3.5 ± 0.5 min for CSA, 3.8 ± 0.5 min for R 2 , and 6.9 ± 0.5 min for diffusion. The mean exercise volume (load Â repetitions) for FF20 was 850 ± 242 (7 ± 1 kg lifted a total of 120 ± 6 repetitions during the four sets) and 840 ± 154 (25 ± 4 kg Â 34 ± 5 repetitions) for FF70. With BFR20 the exercise volume was significantly lower (P = 0.003) with a mean exercise volume of 644 ± 158 (7 ± 1 kg Â 95 ± 5 repetitions) during BFR20. The same load and repetitions performed under BFR20 were repeated under freeflow conditions during the FF20 WM protocol. Average cuff pressure during the BFR protocol was 102 ± 5 mmHg.

R 2
Averaged across all exercise protocols, R 2 decreased by À3.6 ± 1.8 s À1 after exercise which corresponds to a relative decrease of 12.4% from baseline ( Table 2, Fig. 2). Mean recovery rate for R 2 was 0.11 ± 0.07 min À1 corresponding to a halflife (T 1/2 ) of 6.1 min. Changes in R 2 were protocol dependent (P = 0.001) with almost identical post exercise reductions for BFR20 and FF20 (À4.0 ± 1.5 s À1 and À4.2 ± 1.8 s À1 ) exceeding R 2 reductions after FF20 WM and FF70 by $30% (Table 2, Fig.  3). Rate of recovery was slower for BFR20 and FF20 compared to FF20 WM with a similar difference in T 1/2 of 2.3 min between them and the control protocol. FF70 had the same rate of recovery as FF20 WM . The same differentiation was observed for extracellular volume fraction, EC, (fraction of muscle water having low R 2 values). A higher fraction of total extracellular muscle water volume was observed after the BFR20 and FF20 exercise protocols (16.4 ± 0.7% and 16.6 ± 0.8%) than FF20 WM and FF70 (15.8 ± 0.7% and 15.7 ± 0.7%; Fig. 4). A significant group right-left leg discrepancy was observed where mean R 2 decreases in the exercised right leg exceeded those of the left leg by 1.2 ± 0.3 s À1 across protocols (P < 0.001). At baseline, the extracellular water fraction was 10.6 ± 2.9% with no significant differences between participants' legs.
Mixed effects modeling indicated absolute changes in R 2 (s À1 ) to be dependent on changes in muscle diffusion, MD (mm 2 /s), and EC (%). Linear regression analysis using these parameters corrected for participant and right versus left leg yielded the following equation: The goodness of fit values were R 2 = 0.79 and adjusted R 2 = 0.70.

Muscle Cross Sectional Area
Baseline quadriceps muscle CSA was 77.6 ± 13.4 cm 2 which increased across exercise protocols by an average of 9% or Values presented are mean values from quadriceps ROIs ± SD. For the recovery rates of macrovascular blood flow, the mean of participants arterial and venous blood flow were used to yield a combined recovery rate. Baseline values are averaged from the baseline measures of both scanning days. P values represent significant differences between the exercise protocols known to induce hypertrophy and the "control" exercise protocol FF20 WM where "n.s" denotes not significant (P > 0.05). BFR20, 20% 1RM with blood flow restriction; CSA, cross sectional area; MD, mean apparent diffusion; FA, fractional anisotropy; FF, free flow; FF20, free flow 20% 1RM; FF20 WM , free-flow control work matched to BFR20; FF70, free flow 70% 1RM; RM, repetition maximum; ROI, region of interest; R 2 , indicator of edema; R 2 0 , indicator of deoxyhemoglobin.
7.0 ± 4.4 cm 2 ( Table 2). Following BFR20 Quadriceps muscle CSA increased by 10.2 ± 3.1 cm 2 corresponding to 13.1% gain, which was significantly more than the other exercise protocols (P = 0.01; Fig. 4). Acute changes in muscle CSA were lowest for FF70 with changes of 4.7 ± 4.8 cm 2 . Collectively, acute increases in CSA remained higher throughout the first 20 minutes of post exercise recovery when participants performed BFR20, and lower after FF70, exercise than the control FF20 WM protocol (P < 0.001 and P = 0.023, respectively). Average recovery rate in muscle CSA was 0.08 ± 0.09 min À1 which corresponds to an average T 1/2 of 9 min. There were no significant differences in CSA recovery rate between exercise protocols.
Upon cuff release after BFR20 exercise, a rapid decrease in CSA was noted where dilated veins were observed to decrease in circumference. This post release effect typically was normalized within the initial 30 s of the MRI scan for R 2 Ã /CSA data (the first post exercise measure) and were not observed in the following interleaved MRI measures. Post exercise changes in intracellular volume only reached statistical significance following BFR20 exercise (2.3 ± 1.4%, P = 0.018; Fig. 4).

Vascular Blood Flow
Post exercise macrovascular venous flow and arterial flow did not differ significantly between exercise protocols. Large interindividual variations were noted for both venous and arterial flow (Table 2, Fig. 5). Intraindividual variations in flow depending on the exercise protocol performed were considerably smaller. The average recovery rate of venous and arterial blood flow combined was 0.13 ± 0.09 min À1 corresponding to a T 1/2 of 5.4 min.

Diffusion
Baseline MD and FA values were 1.2 ± 0.1 mm 2 /s and 0.4 ± 0.09, respectively. MD values increased on average by 0.33 ± 0.1 mm 2 /s after performing exercise corresponding to an increase of 27.5% (Fig. 6). There were no significant differences between exercise protocols in the post exercise increases and decreases in MD and FA, respectively. The average recovery rate of MD determined from 5 min and 25 min after exercise was 0.013 ± 0.03 min À1 which corresponds to a T 1/2 of 54 min. and similarly for R 2 0 (B) for a representative participant. On the first day, this participant performed FF70 first and then, after scanning and a rest, performed BFR20 exercise with the other leg. Note that at the time of the second post exercise scan, R 2 and R 2 0 values in leg B quadriceps have almost returned to baseline from the first training. On the second day the participant performed FF20 exercise first on the same leg as FF70 (leg B) and then FF20WM on the same leg that performed BFR20 (leg A) after having rested. The right-left assignment of legs to be leg A or B was randomized as was whether leg A or leg B exercised first. Thus, late timepoints in the second scans were acquired for FF70 and FF20 (as is the case for this participant) in the participants (n = 5) having performed these exercises first, and for BFR20 and FF20WM in the remaining participants (n = 5). BFR20, 20% 1RM with blood flow restriction; FF, free flow; FF20, free flow 20% 1RM; FF20 WM , free-flow control work matched to BFR20; FF70, free flow 70% 1RM; RM, repetition maximum; R 2 , indicator of edema; R 2 0 , indicator of deoxyhemoglobin; D, change.
Mean recovery rate for R 2 0 was 0.068 min À1 corresponding to a half-life (time for the difference from baseline to be reduced by one half) of 10.2 min. FF20 decreased more than FF20 WM (P = 0.0273) and all three protocols (FF20, BFR20, and FF70) had a lower rate of recovery than FF20 WM ( Table  2, Fig. 7). There were no significant right versus left leg differences. Mixed-effects model fitting revealed absolute changes in R 2 0 (s À1 ) to be dependent on changes in muscle macrovascular blood flow (V flow , mL/min), CSA (%), and EC (%). A linear regression using these parameters gave the following equation with a goodness of fit of R 2 = 0.69 (adjusted R 2 = 0.51). DR 2 0 = 0.26 DCSA À 0.017 DVflow À 0.86 DEC À 5.9(P < 0.001)

Differences between Day 1 and Day 2
On the second scanning day (day 2) both legs demonstrated lower mean baseline R 2 (reduced by À0.5 s À1 ± 0.22; P = 0.034) and higher mean baseline MD (elevated by 0.05 ± 0.024 mm 2· s À1 (P = 0.030). By separating diffusion into directions along the longitudinal axis of the muscle fiber and radial to the muscle fibers cross section, the increase in radial diffusion, 0.046 ± 0.025 mm 2 /s (P = 0.048) reached statistical significance.

DISCUSSION
Our main finding was that the present MR-based imaging protocols were able to differentiate the exercise-induced mfMRI response of these fundamental diverse resistance exercise protocols. Low load type of resistance training performed to failure either with blood flow restriction (BFR20) or without (FF20) elicited a higher mfMRI response. The acute mfMRI measures include changes in muscle R 2 (indicator of metabolically driven fluid shifts), R 2 0 (indicator of deoxyhemoglobin concentrations), thigh cross sectional area as well as muscle blood flow and diffusion. Both low load resistance exercise protocols to failure (FF20 and BFR20) differed from the low load work matched control (FF20 WM ) which wasn't performed to task failure and is known to induce little hypertrophy. Interestingly, the high load to failure protocol, FF70, also induced smaller R 2 and R 2 0 reductions, smaller acute increases in CSA and smaller extracellular increases compared to the low load BFR20 and FF20 protocols. Moreover, mfMRI measures obtained exhibited longer recovery periods after BFR20 and FF20 compared to the heavy type FF70 resistance exercise. BFR20 differed from all free-flow exercise protocols in that it induced larger acute gains in muscle CSA, which included significant increase in intracellular volume as well as extracellular volume. Furthermore, the intracellular volume increase induced by BFR was sustained over a longer post exercise period. The effect of BFR versus free-flow training was not the same for all participants. Notably, the observed changes in R 2 and R 2 0 values in the BFR condition compared to the freeflow conditions to failure (FF20, FF20 WM and FF70) varied greatly between individuals.

How Measured MRI Parameters Differentiate the Exercise Protocols
The present study confirms our hypothesis that BFR exercise induces larger acute increases in muscle CSA and estimated intracellular volume than free-flow protocols. With regard to the remaining MRI parameters, BFR either induced similar changes as FF20 (R 2 , R 2 0 , and EC) during the studied time period or there was no differentiation between exercise protocols at all (MD, FA, and blood flow). Despite a difference in exercise volume, BFR20 and FF20 induced similar changes in R 2 , R 2 0 and identical acute shifts in the extravascular fraction of muscle volume. Likewise, FF70 and FF20 WM induced a similar attenuated acute response in these parameters despite FF70 involving larger exercise loads, and greater exercise volume since being performed to task failure as opposed to FF20 WM . The calculation of recovery rates revealed a similar trend between the different exercise protocols, where accentuated decreases in R 2 and R 2 0 values Figure 3. Recovery of post exercise R 2 values toward baseline. A: mean relative post exercise changes from baseline (mean ± SE) in knee extensor R 2 are presented for all four exercise protocols. In this timeframe, changes induced by BFR20 and FF20 differed from that of FF20 WM and FF70 (P < 0.001) with no significant differences within these subgroups. Initially, post exercise ($3.9 min) R 2 changes were 0.92 s À1 greater for FF20 than FF20 WM (P = 0.035). Recovery rate coefficients (k R2 ) for quadriceps ROIs are presented in B with BFR20 and FF20 demonstrating a slower recovery rate than FF20 WM and FF70 (P < 0.001). BFR20, 20% 1RM with blood flow restriction; FF, free flow; FF20, free flow 20% 1RM; FF20 WM , free-flow control work matched to BFR20; FF70, free flow 70% 1RM; RM, repetition maximum; R 2 , indicator of edema.
induced by BFR20 and FF20 returned to baseline at a slower rate than FF70 and FF20 WM . Notably, similar exercise volumes were reached during the high and low load free-flow exercise protocols (FF70 and FF20) performed to contraction failure. The application of BFR (BFR20) reduced the exercise volume to contraction failure significantly (Table 1). In contrast, diffusion data, MD and FA, did not reveal any significant post exercise differences between exercise protocols suggesting that overall mobility in the targeted muscle tissue water is similar during this time frame regardless of the exercise paradigm performed. Likewise, macrovascular arterial and venous blood flow during the studied post exercise time period did not differ significantly between exercise protocols. The similarity of these measured parameters indicate that even though the overall movement of fluid, both in blood flow and diffusion within the tissue, is elevated following exercise, the magnitude and rate of normalisation is not sensitive to the differences between the exercise protocols studied in this time frame. This indicates that the observed differences in post exercise R 2 , R 2 0 , CSA, EC, and intracellular volume changes between protocols are not driven by post exercise blood flow and fluid mobility. In a recent acute fMRI study (Ref. 26), we examined exercise protocols differing in exercise loading (low versus high) and BFR versus free-flow conditions, comparable to the conditions of the present study. Although all exercise was performed inside the MR scanner, no significant differences were observed between exercise protocols in terms of diffusion and blood flow $3 min after exercise stop, respectively, despite observations that macrovascular flow was elevated by BFR or high-load free-flow conditions within the first minute following exercise stop (26). Collectively, these and the present observations indicate that the observed differences between exercise protocols in post exercise R 2 , R 2 0 , CSA, EC, and intracellular volume changes are not determined by differential changes in post exercise flow and fluid mobility. Again in our preceding study (Ref. 31), even the elevated hyperemia following the release of the cuff after BFR had little effect on R 2 contrary to the decreases it provoked in R 2 0 . The first post exercise measures of R 2 0 changes and venous flow measures . Post exercise changes in muscle cross sectional area and extracellular space volume fraction. Post exercise knee extensor CSA (A) was increased more after BFR20 (after cuff release $3.5 min post exercise) than the other exercise protocols (P < 0.001). As the order of the two exercise protocols performed on a given day was randomized, each exercise protocol was allocated as the first exercise of the day in five out of 10 participants. Late timepoints of the first exercised leg obtained from the second scan (n = 5 each) are included to indicate progression. After 80 min, acute muscle CSA gains induced by BFR20 remained elevated above baseline (P = 0.02), but all other quadriceps muscles having performed free-flow protocols were below baseline (P < 0.001). After exercise ($ 3.8 min post exercise), the fraction of extracellular muscle water (B) increased for all protocols and after 80 min, the estimated extracellular volume fraction was lower than baseline values regardless of the exercise protocol (P < 0.001). Post exercise changes in intracellular volume (C) only reached statistical significance following BFR20 exercise. Since the first CSA measure precedes the first EC measure, the first time point for the estimate of intracellular volume changes was not included for BFR20, as it would be affected by the rapid changes in blood volume occurring immediately after cuff release. BFR20, 20% 1RM with blood flow restriction; CSA, cross-sectional area; FF, free flow; FF20, free flow 20% 1RM; FF20 WM , free-flow control work matched to BFR20; FF70, free flow 70% 1RM; RM, repetition maximum.
were found to have a weak correlation across all protocols as did changes in R 2 and MD values post exercise. During the post exercise time period (>3 min), recovery to baseline conditions occurred quite independently of one another as MD returned to baseline at a much slower rate than R 2 values and venous flow returned to baseline at a faster rate than R 2 0 values. It is important to note that in this post exercise time frame, R 2 0 values contain differing information compared to during exercise where R 2 0 increases have been verified to correlate with deoxygenated hemoglobin levels (36). This study confirms a post exercise increase in R 2 0 which could indicate higher venous blood oxygenation than during rest conditions.

Using mfMRI to Assess Hypertrophic Potential of Exercise Protocol
It has been widely argued that muscle hypertrophy is stimulated by two primary mechanisms: mechanical tension and metabolic stress (6,9,16,17,24). Although no one parameter measured in the present study differentiated the FF70 protocol-with higher tension-from the low load protocols, some metabolically driven observations were made. The amplified R 2 decreases observed after the BFR20 and FF20 exercise protocol support our initial hypothesis that R 2 changes would be sensitive to the elevated levels of metabolic stress associated with these exercise strategies as opposed to load, exercise volume or total metabolic activity. Metabolic stress levels are often evaluated from increased myocellular lactate concentrations, reduced PCr concen-trations and lower pH. Shifts in R 2 have also been reported to correlate with metabolic markers in muscle tissue such as glucose uptake (28), decreases in phosphocreatine concentrations (43), and pH (32). During resistance exercise, a higher exercise load (>70% 1RM) requires more ATP hydrolyzed per contraction and can produce faster changes in PCr, lactate and [H þ ] concentrations than lifting a lower load. When exercising to fatigue, however, end point metabolic stress tends to be higher after resistance training with lower loads (<35% 1RM) (44)(45)(46)(47)(48)(49) than high loads. The use of lower relative training loads allows for more repetitions to be performed before reaching contraction failure, hence resulting in more sustained metabolic stress (due to more pronounced glycolysis) in a larger part of the active myofibers at the point of contraction failure (50). In the present study BFR20 and FF20 exercise to failure were employed to represent protocols with higher metabolic stress and were therefore predicted to cause larger R 2 decreases than FF20 WM and FF70. This suggests that metabolic stress is a primary factor driving acute R 2 changes, given that anaerobic energy systems are reported to be more strongly involved in the BFR exercise conditions (BFR20) compared to matched free-flow conditions (FF20wm) (51,52) and to be more heavily taxed in the low-load (FF20) compared to high-load (FF70) free-flow exercise performed to failure (44,45). Fleckenstein et al. (27) were the first to report that changes in R 2 reach a plateau with continued repetitions and that although the rate at which R 2 changed with increasing contractions was load dependent, the point at which R 2 decreases plateaued was not load dependent. Jenner et al. (29) later confirmed that the plateau is not determined by a maximal limit to R 2 changes, but instead reflects a steady state determined by the exercise intensity independent of the load or total work performed. Even after repeated sets of exercise, measured R 2 decreases have been shown to reflect the same heterogeneous distribution as relative muscle glucose uptake during exercise (28). Recently, R 2 changes have been demonstrated to quickly plateau in the first of four sets of conventional free-flow resistance exercise to failure establishing a new steady state that was maintained throughout the remaining exercise sets (26). In the same study, it was observed that although CSA increases and R 2 decreases already reached a steady state after the first of four sets for free-flow exercise to failure, R 2 values continued to decrease and muscle CSA increased over the subsequent sets of BFR exercise (26). This would appear to result from the effect BFR has on muscle tissues ability to meet metabolic demands as venous occlusion alone, without exercise, induces only minor R 2 changes that recover quickly (36,37,53). In the present study, not only were acute R 2 decreases accentuated by BFR, but the recovery rate to preexercise levels was also slower than observed in the FF20 WM protocol. Our findings match those of Fisher et al. (53) who reported that R 2 recovery rates were not determined by the values in all muscle groups (P < 0.001) but did not vary between the four exercise protocols. During 20 min of rest, no differences in recovery rate was observed across exercise protocols as MD values decreased (P = 0.001) and FA values increased (P = 0.001) toward baseline. Changes in MD and FA for hamstring muscles, antagonists to the knee extension exercises performed, were reciprocal to those of the quadriceps although of a lower magnitude. BFR20, 20% 1RM with blood flow restriction; FF, free flow; FF20, free flow 20% 1RM; FF20 WM , free-flow control work matched to BFR20; FF70, free flow 70% 1RM; RM, repetition maximum. 0 values decreased more after the FF20 exercise protocol than FF20 WM (P = 0.0273) and all hypertrophic protocols (FF20, BFR20, and FF70) had a lower rate of recovery than FF20 WM . BFR20, 20% 1RM with blood flow restriction; FF, free flow; FF20, free flow 20% 1RM; FF20 WM , free-flow control work matched to BFR20; FF70, free flow 70% 1RM; RM, repetition maximum; R 2 0 , indicator of deoxyhemoglobin; T 1/2 , half-life. load applied during resistance exercise and did not observe a significant correlation between acute exercise-induced increases in CSA and changes in R 2 .

Interpretation of R 2 , EC, and Intracellular Volume Changes
The observed relationship between CSA and R 2 in the present study challenges simplistic models regarding how exercise related factors, such as cell swelling, drive R 2 changes. The intracellular buildup of several metabolites in response to intensive regimes of exercise loading is in itself believed to be a primary factor creating an elevated osmotic gradient thought to drive the intracellular water retention observed by Armstrong (54). The metabolically driven intracellular water retention contributes to both acute R 2 decreases and CSA increases interlinking the two measures.
Baseline CSA values were similar to previous reports of 77.5 ± 3.0 cm 2 for young men (55) and the observed post exercise increases in CSA corresponded closely to the acute post exercise increases of 8%-10% reported previously (56). The more pronounced increases observed in post exercise CSA with BFR compared to free-flow resistance exercise also aligns with previous observations (11) with acute increases after BFR exercise exceeding that of free-flow exercise by $50%. However, despite that the CSA increases observed post BFR20 exercise exceeded that of FF20 by $50% and the intracellular volume increases were higher, post exercise R 2 changes were similar for BFR20 and FF20. Likewise, after FF20 exercise, muscle CSA increases and estimated intracellular volume changes were similar to those observed after FF20 WM , yet changes in R 2 values were significantly higher. This could indicate that not all of the BFR induced fluid shifts contribute to the acute increase in muscle CSA have a substantial effect on R 2 . Interestingly, Farup et al. (11) demonstrated that when exercise bouts were repeated over several weeks, the elevated acute changes in muscle CSA (i.e., swelling) observed with BFR compared to free-flow exercise conditions were diminished. The primary mechanism leading to changes in R 2 during exercise involves the modulation of pathways regarding protein-water interactions (57). Although the osmotic water shifts during exercise are a primary factor (30,33,57), pH, metabolite concentrations, or other physical factors within the intracellular space that influence this interaction may contribute as well (32,57). Acute increases in CSA also include the increased volume of the extracellular space which has been demonstrated to have little impact on muscle R 2 values as determined by either both the mono exponential fit used in this study or from multi exponential fitting (30,34,58,59). Sjøgaard et al. (60,61) found that intracellular water volume remained relatively stable during submaximal exercise, while increasing only during maximal intensity exercise which was accredited to a fluid shift driven by an increasing osmotic gradient between the extra-and intracellular space. In the same study, the authors reported acute gains in intracellular water increases to be both smaller than the concurrent increases in the extracellular compartment and more varied between participants, with actual decreases being noted in two of their six participants. A similar pattern was observed in the present study, with significant increases in the estimated extracellular volume across all protocols with smaller estimated intracellular volume increases that only reached statistical significance following BFR20 exercise. The increases noted were comparable to the 2% increases of muscle water reported after acute low-load (60 W) knee extensor exercise to exhaustion (62). Even larger increases of intracellular water have been reported during isolated knee extensor exercise (>10%) but were observed to decrease within minutes of exercise stop towards values from the earliest time points measured in the present study obtained $4 min post exercise (60). Using muscle biopsy sampling, Sjøgaard et al. (61) estimated the extracellular water fraction to be 12% for the human quadriceps and also observed that absolute extracellular water volume increased by 82% following intense exercise which was concluded to be driven by hydrostatic forces from increased blood flow. These values correspond well with the estimates of the extracellular compartment in the present study of 10.6% at rest and 16.5% $4 min after exercise cessation. Previous estimates using the slow R 2 compartment to estimate the extracellular fraction for resting muscle match the present study (10.6%) with values ranging from 8% to 12% have been reported using various models (63)(64)(65). The baseline R 2 values in this study of 29 s À1 are not only similar to reported values from previous studies which applied monoexponential fits (29 s À1 and 33 s À1 ) (53,63), but are also similar to the primary "intracellular" component reported from studies using multiexponential fits (31.2 s -1 , Ref. 63, or 31s -1 , Ref. 58). Likewise, the present changes in R 2 after the four different acute exercise protocols were in the same range as reported earlier for freeflow muscle exercise (27,28,34,53).

Methodological Considerations
A number of potential methodological limitations should be taken into consideration for the present study. First of all, this study only included young males in an attempt to minimize variation between participants at the cost of the study being less representative of the broader population. Even then, ensuring the same conditions surrounding each of the exercise protocols to avoid variations in the physiological stimulation to participants was challenging. To ensure a uniform BFR stimulus we regulated cuff occlusion pressure to participants based on their individual diastolic blood pressure. Still inter-individual variations were evident, as the number of repetitions performed until failure with 20% 1RM in free-flow (FF20) and BFR (BFR20) varied greatly. By using a cuff pressure of participants' diastolic blood pressure plus 30 mmHg, the resulting mean cuff pressure of 102 mmHg was slightly below levels of $110 mmHg which have been recommended since the completion of this study (66).
As previously discussed, the identification of physiological compartments using R 2 values lacks a methodologically proven model for the present exercise conditions. In the present study, spin echo times of less than 100 ms were employed to reduce scan time and increase the number of interleaved data recordings. This limits the ability to characterize the R 2 values of the slower relaxing components which instead are collected as a constant in a monoexponential fit. Multiexponential models to include fast R 2 components have been found to provide better fits to T 2 -weighted signal on a voxel basis by identifying a fraction of muscle water (8%-15%) having fast relaxation (R 2 > 200 s À1 ), believed to represent water bound to macromolecules (67). Diffusion data which include a wider range of b values have also shown promise separating vascular, extracellular and intracellular compartments (68). A last consideration regarding R 2 estimation is the choice of repetition time due to the R 1 relaxation. Sharafi et al. (67) evaluated this effect using Bloch simulation finding a R 2 estimation error between using a TR of 1,500 ms (as was used in this study) compared to a TR of 5,000 ms to be 5%. This error was prefered to the considerably longer acquisition time of using a TR of 5,000 ms.

Future Perspectives
An improved understanding of the physiological stressors that stimulate muscle hypertrophy plays an important role in the ability to personalize rehabilitation for patients as well as training for healthy individuals including top-level athletes (69). Should metabolic stress and associated fluid shifts drive hypertrophy, or even be significant correlation to it, magnetic resonance imaging (MRI) would be a highly suited modality to quantify and map the degree of stimulation in muscle tissue. The MRI sequences applied in the present study are clinically accessible and can be interleaved with other measures to further investigate other physiological parameters of interest such as oxygen tension (70), perfusion or specific solute concentrations (71) to study different training strategies. There are several other desirable adaptations that skeletal muscles undergo in response to exercise that can be accentuated by adjusting the training protocol used. For instance, heavy loads are consistently shown to yield maximal increases in strength (10,72) and improve neural recruitment (73) compared to BFR (74). On the other hand, BFR exercise has been reported to induce a high proliferation of myogenic stem cells (13,75,76), stimulate angiogenic signaling (77), and even induce positive exercise effects in remote muscle tissues distally to the BFR trained limb (78,79). The use of multimodal MRI to examine selected peripheral physiological variables during ongoing exercise represents a promising tool not only for decoding important mechanisms for these adaptations but also for optimizing the design of effective training modalities.

Conclusions
The present MR imaging protocols were able to differentiate the acute exercise-induced response of differing training protocols to failure that induce skeletal muscle hypertrophy. The mfMRI measures obtained exhibited larger changes and longer recovery periods in the exercise protocols designed to cause elevated metabolic muscle stress. The increase of the intracellular space was higher following exercise with BFR compared to free-flow exercise. These findings contribute to explain the myogenic potential of low-load resistance exercise, with and without BFR, in human skeletal muscle. Thus, use of multimodal MRI to examine selected peripheral physiological variables during and following acute muscle exercise represents a promising tool not only for decoding important stressors influencing myocellular adaptation but also for optimising the design of effective anabolic exercise modalities.