Mechanisms of in vivo muscle fatigue in humans: investigating age-related fatigue resistance with a computational model

Abstract

Key points: Muscle fatigue can be defined as the transient decrease in maximal force that occurs in response to muscle use. Fatigue develops because of a complex set of changes within the neuromuscular system that are difficult to evaluate simultaneously in humans. The skeletal muscle of older adults fatigues less than that of young adults during static contractions. The potential sources of this difference are multiple and intertwined. To evaluate the individual mechanisms of fatigue, we developed an integrative computational model based on neural, biochemical, morphological and physiological properties of human skeletal muscle. Our results indicate first that the model provides accurate predictions of fatigue and second that the age-related resistance to fatigue is due largely to a lower reliance on glycolytic metabolism during contraction. This model should prove useful for generating hypotheses for future experimental studies into the mechanisms of muscle fatigue.

Abstract: During repeated or sustained muscle activation, force-generating capacity becomes limited in a process referred to as fatigue. Multiple factors, including motor unit activation patterns, muscle fibre contractile properties and bioenergetic function, can impact force-generating capacity and thus the potential to resist fatigue. Given that neuromuscular fatigue depends on interrelated factors, quantifying their independent effects on force-generating capacity is not possible in vivo. Computational models can provide insight into complex systems in which multiple inputs determine discrete outputs. However, few computational models to date have investigated neuromuscular fatigue by incorporating the multiple levels of neuromuscular function known to impact human in vivo function. To address this limitation, we present a computational model that predicts neural activation, biomechanical forces, intracellular metabolic perturbations and, ultimately, fatigue during repeated isometric contractions. This model was compared with metabolic and contractile responses to repeated activation using values reported in the literature. Once validated in this way, the model was modified to reflect age-related changes in neuromuscular function. Comparisons between initial and age-modified simulations indicated that the age-modified model predicted less fatigue during repeated isometric contractions, consistent with reports in the literature. Together, our simulations suggest that reduced glycolytic flux is the greatest contributor to the phenomenon of age-related fatigue resistance. In contrast, oxidative resynthesis of phosphocreatine between intermittent contractions and inherent buffering capacity had minimal impact on predicted fatigue during isometric contractions. The insights gained from these simulations cannot be achieved through traditional in vivo or in vitro experimentation alone.

Figure 1. Computational approach and literature sources for model components

A control function (step 1) dictates excitation to a pool of 60 motor neurons (step 2). Each motor neuron is associated with corresponding fibre‐type‐specific muscle activation kinetics (step 3) and a Hill muscle model (step 4), whose function is modified by predicted concentrations of intracellular metabolites (step 4a). The sum of forces predicted by the Hill muscle models is used in a musculoskeletal model to predict joint torque (step 5), which is compared with a task defined a priori. The result of this comparison is then used to modify excitation (step 6). Abbreviation: MU, motor unit.

Figure 2. Torque–frequency relationships in young and older men

Comparison of experimental torque in response to electrical stimulation of the peroneal nerve with that simulated using the model. Experimental data are indicated by filled (young males, YM) and open circles (older males, OM) representing mean values ± SD at several frequencies. Continuous and dashed lines are five‐parameter sigmoid curves fitted to simulated torque predicted by the Y‐Model and O‐Model, respectively.

Figure 3. Simulated metabolic response to a 12 s maximal voluntary contraction

A–C, simulated data from Y‐Model (continuous line) are compared with in vivo observations in young men (filled circles; means ± SD); data are shown for PCr (A), pH⁺ (B) and H₂PO₄ ⁻ (C). D–F, the same data are provided for older men (open circles; means ± SD) and simulated by the O‐Model (dashed line). Shaded rectangles indicate duration of the contraction.

Figure 4. Simulated repeated maximal voluntary contraction protocol

A–C, model output from a simulation of repeated maximal voluntary contractions reveals predicted metabolic responses of intracellular metabolites PCr (A), H⁺ (B) and H₂PO₄ ⁻ (C). Output from Y‐Model (continuous line) is compared with that from O‐Model (dashed line). D–F, experimental data (Lanza et al. 2007) for PCr (D), H⁺ (E) and H₂PO₄ ⁻ (F).

Figure 5. Fatigue predicted from model simulation

A, fatigue predicted by the Y‐model (continuous line) is compared with previously published (Lanza et al. 2007) in vivo observations in young males (filled circles; means ± SD). B, corresponding data from O‐model (dashed line) are compared with previously published (Lanza et al. 2007) in vivo observations in older males (open circles; means ± SD).