The primary innovation of the presented experimental platform for investigation of upper-extremity cycling exercise in humans is that it enables dynamic, quantitative in vivo assay of ATP metabolism and pH balance in working muscles of the upper arm in addition to conventional physiological measures such as whole-body oxygen consumption. Thus, it can uniquely inform on contributions of oxidative versus anaerobic ATP production of individual muscles of the upper arm during execution of any particular arm-cranking task. Below, we will discuss how this new information impacted the outcome of the practice intervention study in healthy naïve subjects showcasing the platform, as well as technological and practical aspects of the platform including recommendations for future upgrading and use.
Outcome of the practice intervention study.
The practice intervention study that we conducted to showcase the new platform tested the hypothesis that three weeks of daily training would improve the gross mechanical efficiency of execution of a physical task consisting of supine asynchronous arm-cranking at 90 rpm for 6 min against a workload of 15W. Tested solely against the combined results of the conventional physiological measures that we collected, this hypothesis was not rejected: the mean gross ME of ACT execution significantly improved after training in the study test-population (Table 2). The in vivo 31P MRS recordings from biceps muscle during ACT execution that we additionally collected, however, showed that, with respect to this particular physiological measure, this outcome was biased by results in subjects #4 and #5. Specifically, these two individuals had recruited more fast-twitch fibers with low oxidative capacity (FG fibers; (21)) to perform the ACT trial post-training than pre-training, as evidenced by increased biceps muscle acidification (Fig. 4). Neither of these subjects had appreciably changed their arm-cranking strategy towards any ‘pull’ mechanism in response to practice; if any, subject #5 had rather adopted more of a ‘push’ strategy (Fig. 6). Therefore, the outcome of the ratio of whole arm power-output and whole-body energy expenditure during ACT execution was skewed towards a higher ratio post-training (Table 2). This particular finding was similar to results of a recent study of the effect of a low-intensity training intervention in manual wheelchair propulsion in naïve subjects (22). Specifically, the authors reported improvement of gross ME of wheelchair propulsion concomitant with opposite rather than parallel changes in power-output over time for the biceps and brachialis muscles of the upper arm. Notably, power-output of the biceps muscle was found to increase, not decrease with training (22) Together, these studies indicate that findings of increased ME after training of any form of upper body exercise based solely on whole body measurements of oxygen consumption should be interpreted with caution. The new experimental platform for upper-extremity cyclic exercise presented in this report uniquely affords to gather complementary data on energy expenditure in upper-extremity muscles to strengthen objective evaluation of the outcome of upper body training interventions.
The results of the practice intervention study also provide new insight into the contribution of aerobic versus anaerobic ATP metabolism in muscles recruited during cyclic upper-body exercise. As discussed in the above, the data showed that the assumption of strictly aerobic muscular ATP metabolism at low-intensity dynamic upper-body exercise implicit in the use of the parameter ‘gross ME’ to evaluate training outcome (23,24) did not hold for the majority of untrained, able-bodied, healthy individuals that participated in the present study. Secondly, we found that the contribution of anaerobic motor units of the biceps to power-output during arm-cycling was not uniformly affected by the training intervention across test subjects. Inter-individual differences in learning efficacy have previously been described (7,25–29) and may become more explicitly characterized using this platform.
The novelty of the current approach comes with some technological limitations that should be considered when assessing the results. First, subjects performed the ACT trial twice – i.e., inside the mock-up MRI scanner and inside the MRI scanner, respectively – both pre- as well as post-training in order to obtain a complete set of experimental data consisting of sEMG, mechanical force and PO, spirometry and 31P MRS recordings, respectively (Fig. 1). As such, it was assumed that individual participants performed the physical task in both environments in identical manner. Past findings that motor skill acquisition takes place on a timescale of multiple minutes (25) suggests that this assumption may perhaps have been problematic. However, our results showed that individual arm-cycling strategies were consistent over the entire course of the experiment and did not majorly change during training. Secondly, our practice intervention study employing the new platform was limited in its sample-size resulting in low statistical power of evaluation of the study outcome for the group. Here, both the high cost as well as limited availability of MRI scanner time played a role. However, the observed heterogeneity in individual ACT strategy within these five subjects in and by itself suggests that a common group pattern of task execution may perhaps not be expected. On the other hand, past evidence suggests that motor skill acquisition is a general principle of motor functioning and, as such, should hold for groups of subjects (23,30,31). Therefore, it should be of interest to study in a larger population of test subjects if the results of the present study can be reproduced.
Technological and practical aspects of the platform.
sEMG and force sensor recordings.
Upper-arm muscle recruitment patterns in participants performing cyclic upper-body exercise were well reflected by the results of both sEMG and force-sensor recordings during ACT execution (Figs. 5 and 6, respectively). Of these, sEMG is commonly available in movement research labs evaluating cyclic upper-body exercise (32–34). The force-sensor/pole ensemble was custom-built using an ‘off the shelf’ one-dimensional sensor incorporated in the pole, whereby the pole handles were constructed such that no torque was possible (Methods; Fig. 3). Although the force-sensor recordings lacked the detail on individual muscle contributions to overall arm power-output provided by sEMG (Fig. 5), the former adequately distinguished flexor from extensor movements with superior signal-to-noise and superior dimensionality (i.e., N; (Fig. 6). Moreover, while similar pictures on arm muscle recruitment during ACT execution in individual subjects emerged from the sEMG and force-sensor recordings in the present study, the latter suggested training induced a small shift from ‘mixed push/pull’ to ‘pull’ mode in subject #5 that was less evident from the sEMG recordings.
In vivo 31P MRS.
The principal practical problem we had to overcome to robustly collect 31P MRS data from the human biceps muscle during high-frequency arm-cycling with ~ 10 s time resolution and adequate signal-to-noise and resonance linewidth quality for quantitative analysis, was the anatomy of human subjects in relation to MRI scanners. Specifically, the biceps muscle is physically on the periphery of the human body, typically some 25 cm out of the central body axis in lean adults, whereas the default ‘sweet spot’ of static field magnetic homogeneity of 3 T MRI scanners is typically a 20–30 cm diameter sphere in the center of the magnet. In addition, the inner-diameter of the bore of commonly available clinical MRI scanners including the MRI scanner used in the present study, is 60 cm. Together, this both limited the attainable quality of local static magnetic field homogeneity (‘shim’) over the biceps muscle (typical PCr linewidth 30–40 Hz compared to 5–10 Hz in in vivo 31P MR spectra of medial quadriceps of the upper leg (35) as well as constrained the physical space for arm exercise inside the magnet. Any additional signal quality deterioration introduced by motion artefacts during high-frequency arm-cycling (i.e., 90 rpm) was surprisingly minor. We attributed this to the fact that the upper arm remains in one and the same position during arm-cycling around the elbow (Fig. 2). In addition, 31P MRS data acquisition from the biceps during exercise was synchronized with cycling phase using a custom-built triggering setup described elsewhere (8). Together, this enabled robust dynamic 31P MRS data gathering from the biceps muscle during high-frequency arm cycling with 11 s time resolution in all subjects (Fig. 7).
Recommendations for future platform upgrading and use.
All sEMG, spirometry and mechanical force data recordings were conducted in a parallel series of pre- and post-training tests in a mock-MRI scanner due to issues of MR-compatibility of equipment. Ideally, all these recordings should be done concomitantly with the MR measurements. This should, in fact, be feasible. MR-compatible approaches to collect VO2 and VCO2 data are available including Douglas bag-based methodologies (36–38) or mass-spectrometry(39). Likewise, simultaneous acquisition of sEMG and 31P MRS data, respectively, from the upper arm muscles are well feasible provided non-gradient-based data acquisition sequences are employed such as used in the present study(40). Similarly, the applied torque over the cycle of the crank of the MRI-compatible handcycle ergometer should ideally also be measured in the MRI scanner. While such systems are available and have been used for arm-cranking and hand-cycling studies (12,33,41), these are as of yet MRI-incompatible. The sole available industry-standard MRI cycle ergometer (Lode BV, Groningen, the Netherlands) could potentially be fitted with hardware and software to render such measurements feasible in MRI scanners in the future (Jan-Reinder Franssen, Lode BV, personal communication).
In vivo 31P MR spectra were collected from the biceps muscle using a single element 31P surface coil that was available in our laboratory. In light of our findings of inter-individual differences in pull- versus push- handcycling strategies, spectra should also be collected from the triceps muscle, preferably simultaneously from both muscles. In principle, use of a two-element 31P surface coil in combination with dual receive channels should render simultaneous collection of in vivo 31P MR spectra from both upper arm muscles feasible. Modern clinical 3T MRI scanners with multinuclear capability from major vendors typically readily support use of such advanced coil designs. Moreover, these scanners are typically also available in a 70-cm diameter bore size. This both greatly enhances room for in-magnet exercise as well as relaxes some of the constraint on physical dimensions of study subjects that we encountered.
Lastly, the new platform for upper-extremity cycling exercise presented here may be useful for clinical investigations of a range of study populations and objectives. Firstly, it may contribute to guide training of athletes relying on upper-body activities including rowing, arm-cycling, sailing, rock-climbing or wheelchair track athletes. In these sports the upper-body is highly trained, yet little remains known about fibre-type distributions and relative contributions of aerobic and anaerobic energy production optimal for each particular athletic activity, for upper- and lower-extremities alike (42). Secondly, the methodology may also be useful in the field of rehabilitation of individuals with spinal cord injury. These often-traumatic injuries render individuals heavily dependent on their upper-body for daily activity, while there may be a large heterogeneity in upper-body fitness and skill prior to the accident. Of specific interest are any adaptation processes taking place at muscular levels as well as to local and bodily physiology as a consequence of continuous practice and use of their upper body (43). Other potential clinical application of the methodology may be in care for patients with neuromuscular disease, primary metabolic myopathies including mitochondrial myopathies as well as secondary myopathies including heart failure and COPD. Here, the key asset is the fact that arm-cranking constitutes a relatively low intensity dynamic exercise paradigm (dynamic range 3–25 W at 90 rpm; Supplemental Materials, Fig. 1) compared to conventional testing using bicycle ergometers. As such, it offers a platform to evaluate the quality of their muscles with respect to mechanical and metabolic functions and monitor the effects of training, pharmaceutical and dietary therapy.