Fifteen French Army young male soldiers volunteered to participate in the present study. A sample size of 11 (15 subjects included) was deemed sufficient to determine statistical power. This was calculated using the sample size calculator G^∗Power 3.1.9.2. A large effect size of 0.40 was determined from previous data on the peak knee extensor (KE) torque, alongside a power of 95 % and a significance level of 0.05. Their anthropometrical and physiological characteristics were as follows: age: 29 ± 2.5 years, weight: 79.6 ± 9.1 kg, height: 174 ± 10 cm, BMI: 26.1 ± 1.9 kg^∗m⁻², and body fat percentage: 12.5 ± 3.3%. They performed regular physical activities, such as strength training, running, and/or cross training (between 6 and 15 h.w⁻¹), with no recent history of muscular, joint, or bone disorders or receiving any medication that could interfere with neuromuscular responses. This study was part of a medico-physiological follow-up performed at the request of the French Army. All the volunteers were fully informed of the experimental procedures, aims, and risks and gave their written assent before any testing was conducted. Each participant participated in an inclusion session consisting of a complete medical examination with anthropometric data collection and complete familiarization with the experimental procedures. This study was approved by the scientific leadership of the French Armed Forces Biomedical Research Institute. All experiments were conducted in accordance with the Helsinki Declaration (Schuklenk, 2001). A version of this manuscript was released as a preprint on bioRxiv at https://doi.org/10.1101/402644 (Siracusa et al., 2018).

After a 5-min warm-up at submaximal intensity, participants performed an intermittent voluntary fatigue protocol consisting of 6 × 10 repetitions of alternating isometric 5-s MVCs of the KE muscles and 5-s passive recovery periods between repetitions and 15-s between series (Figure 1). The number of contractions was chosen to generate a high level of voluntary strength loss and peripheral fatigue, as demonstrated previously in adults (Ratel et al., 2015). Double and single electrical stimulations were delivered to the femoral nerve before and every ten MVCs, during the contractions and at rest, to determine the time course of central and peripheral neuromuscular fatigue (see below). Moreover, a linear transducer was fixed to the skin, using a dynamic probe fixation system (USONO, Eindhoven, Netherlands), over the VL muscle and used in SWE mode (musculoskeletal preset) to capture the shear-wave propagation velocity and to determine in vivo an index of muscular stiffness. After each ten MVCs and electrical-stimulation series, the resting VL shear elastic modulus (VLμ) was measured during a 5-s period. Participants were not informed of the criterion of task failure (60-MVCs) but had visual feedback of the torque output during the exercise. They were also strongly encouraged by the experimenters during the entire fatiguing task.

The data were screened for normality of the distribution and homogeneity of variances using the Shapiro–Wilk normality and Levene tests, respectively. Differences in absolute values and relative changes were analyzed by one-way ANOVA (effect: number of repetitions) with repeated measures. If the ANOVA revealed significant effects or interactions between factors, Fisher’s LSD post hoc test was applied to test the discrimination between means. The intrasession repeatability of the resting shear elastic modulus was evaluated for the VL muscle between each five measurements, obtained during the 5-s movie, by calculating the intraclass correlation coefficient (ICC) and standard error of measurement (SEM) (Hopkins, 2000). Results with a P-value < 0.05 were considered to be significant. Statistical procedures were performed using Statistica 8.0 software (Statsoft Inc, United States). The results are presented in means ± SD for table and text and means ± SEM for figures.

ANOVA revealed significant absolute and relative (to the control “non-fatigued” value) repetition-dependent effects on MVC torque (P < 0.001, Figure 3). MVC decreased by 38.4 ± 12.6% by the end of the exercise (P < 0.001), with a significant reduction starting from the 10th MVC (P < 0.001).

The measurements were reproducible (five measurements obtained during a 5-s clip) throughout the fatigue protocol. The ICC (95% CI) and SEM (kPa) values were as follows for each measurement: Control (99.5%, 0.2 kPa), MVC10 (99.5%, 0.24 kPa), MVC20 (98.9, 0.27 kPa), MVC30 (99.0%, 0.26 kPa), MVC40 (96.7%, 0.27 kPa), MVC50 (99.0%, 0.26 kPa), and MVC60 (97.5%, 0.26 kPa).

The time course of the relative and absolute resting VLμ (in % and kPa, respectively) is presented in Figure 4A,B. A representative scheme of the kinetics of the 2D color map of the muscle μ is presented in Figure 2C. ANOVA revealed a significant repetition-dependent effect for both absolute and relative resting VLμ (P < 0.001). Relative resting VLμ values significantly decreased progressively from the 10th MVC (−12.4 ± 7.7%, P < 0.001) to the end of the exercise (−34.7 ± 6.7%, P < 0.001).

Shear-wave elastography has been demonstrated to be a highly promising alternative to conventional elastography techniques, and provides a reliable and quantitative real-time assessment of muscular tissue stiffness (Brandenburg et al., 2014; Lacourpaille et al., 2015; Andonian et al., 2016). This study is the first to address the interest of μ on a non-contracted muscle (VL) at various stages of the fatigue during a series of isometric MVC of the KE muscles. Methodological factors such as slight probe motion, compression, variability of the measurement, and the ability of subjects to achieve a fully relaxed state were meticulously controlled to ensure good reliability of the measurements (Lacourpaille et al., 2012; Eby et al., 2013). There was little variability of the measurements (five images obtained during the 5-s clip) during the entire fatigue protocol. ICC and SEM values varied from 96.7% to 99.5% and from 0.20 to 0.26 kPa for the VL muscle, despite the development of fatigue. We achieved this level of reproducibility by positioning a dynamic probe fixation system over the skin onto the VL muscle (Figure 2B) avoiding any muscle deformation. In addition, the EMG VL and BF EMG activity and passive torque did not change during the SWE recordings (Table 1), verifying that the participants were in a relaxed state.

Our resting VLμ baseline values are concordant with existing published SWE data (Lacourpaille et al., 2017). The mean and standard deviation of 9.3 ± 2.0 kPa (ranging from 6.3 to 12.9 kPa) for a knee flexion of 90° (0°= knee fully extended) are in accordance with the values of <10 kPa for the quadriceps muscles for the same knee flexion (90°) in active individuals in a non-fatigued state (Lacourpaille et al., 2017). However, our control values were higher than those of other studies (∼3–5 kPa) for the VL muscle (Lacourpaille et al., 2012; Botanlioglu et al., 2013; Andonian et al., 2016). These differences may be explained by the short muscle length used in these studies (knee fully extended), as it is known that resting μ and muscle stiffness increase with increasing muscle length (Lacourpaille et al., 2014, 2017).

Most previous studies have characterized fatigue or training-induced changes in muscle-tendon stiffness by exploring the movement of human tendons and/or aponeurosis structures in vivo (Fukashiro et al., 1995; Kubo et al., 2000; Bojsen-Moller et al., 2003). Several studies have shown greater muscle-tendon compliance (increases in the elongation of connective structures for the same level of the produced force) after repeated contractions of the KE, (e.g., VL muscle), and arm muscles (Vigreux et al., 1980; Kubo et al., 2001a,b). However, one of the main limitations of these studies is that ultrasound-based techniques reflect modifications in the stiffness of several structures (muscles, tendons, nerves and skin) around a given joint and are not specific to skeletal muscle stiffness.

As mentioned above, SWE provides a quantitative and reliable measurement of individual muscular tissue stiffness (Brandenburg et al., 2014; Lacourpaille et al., 2015; Andonian et al., 2016). A strong linear relationship between individual muscle force and dynamic muscle μ, evaluated during contractions, has been demonstrated, suggesting that it may be a good index of individual force (Bouillard et al., 2011). Furthermore, Bouillard et al. (2012) confirmed this relationship during submaximal isometric contractions, even when muscle fatigue occurs. These results suggest that SWE can be used to quantify relative modifications in voluntary force, even during fatiguing conditions. In another study (Bouillard et al., 2014), the same authors showed that when fatigue was previously induced in one quadriceps muscle (VL), lower dynamic μ values were observed, both initially and during a subsequent submaximal isometric task, relative to the control non-fatigued muscle. Although the amplitude of the muscle μ appears to be affected by fatigue, the underlying mechanisms for its decline are still unknown.

In our study, we observed a progressive decrease in the absolute resting μ of the VL muscle from the 20th MVC (−12.4 ± 7.7%) to the end of the fatiguing exercise (−34.7 ± 6.7%), suggesting a progressive rise in muscle compliance (at 50% of the VL muscle length). Some studies have suggested that increases in muscle-tendon compliance can be explained, in part, by alterations of the viscoelastic properties of the intramuscular connective tissue due to repeated contractions (Warren et al., 1971; Vigreux et al., 1980). Currently, specific viscoelastic properties of the soft tissues may be quantified by real-time supersonic shear imagining, as described in the literature (Bercoff et al., 2004; Gennisson et al., 2010), but no study has investigated the effects of peripheral fatigue on the viscoelastic properties of skeletal muscle and its role in force transmission.

The decline of the resting μ of the VL muscle found in our study is consistent with those observed for locomotor muscles after strenuous long-distance running, evaluated by invasive techniques, such as tension-myography and muscle belly deformation (Garcia-Manso et al., 2011; Giovanelli et al., 2016), as well as SWE (Andonian et al., 2016). However, these exercise-models are clearly very different from that used in our study and the extent of induced fatigue may be different than that in our model. For example, Andonian et al. (2016) observed significant decreases in the resting μ of the quadriceps muscles (without distinguishing between the heads, but mainly in the VL muscle) after an extreme mountain ultra-marathon, which were still reduced after more than 45 h. However, the relationship between changes in muscle μ and neuromuscular parameters of fatigue was not explored. In contrast to these results, Akagi et al. (2017) observed an increase in resting μ (∼+7%) of the medial gastrocnemius, but not soleus or lateral gastrocnemius, muscle using another model of fatiguing exercise (sustained 10% MVC during 1 h). Moreover, Lacourpaille et al. (2017) showed that intense, non-damaging exercise (3 × 10 concentric MVCs at 120^∘∗s⁻¹) did not modify the resting μ of the elbow flexor muscles. They suggested that the resting μ would not be influenced by peripheral factors originating from fatiguing contractions, contrary to those observed in our study. However, in their study 30 concentric MVCs did not induce a significant decrease in voluntary torque, limiting the conclusions concerning the relationship between fatigue and changes in resting μ. It is possible that in our study the significant decline in resting VLμ during the fatigue protocol may be closely related to the greater observed extent of exercise-induced peripheral fatigue.

The mechanisms underlying changes in muscle μ by SWE after fatiguing exercise are not well described. As expected, in our study the significant reductions in the KE MVC were mostly explained by alterations of peripheral factors. The large extent of peripheral fatigue observed was mostly reflected by a significant decrease in the amplitude of the single- and double-stimulated responses at end of the exercise. These results may suggest modification of the excitation-contraction coupling (Bigland-Ritchie, 1981; Edwards, 1981; Enoka and Stuart, 1992; Gandevia, 2001; Debold, 2012; Millet et al., 2012), more specifically, impairments in calcium homeostasis. Moreover, the EMD (+21.9 ± 6.8%) and CT (+20.4 ± 10.4%) increased significantly until the end of the exercise and the MRTD (−51.0 ± 12.7%) was lower than control values. These results are suggesting that any alteration on the electromechanical properties of the skeletal muscle are mainly associated with elastic rather than electrochemical processes, [e.g., alteration in the propagation of action potentials along the muscle membrane (Ce et al., 2013, 2015)], due to the lack of modification in the potentiated VL M-wave amplitude during the fatigue protocol. Under fatiguing conditions, changes in electromechanical properties of the skeletal muscle would result in a longer time to stretch a more elastic muscle and to transfer tension to the tendon insertion point (Zhou et al., 1995, 1998). These findings could suggest a relationship between the decline in resting VLμ and alteration of quadriceps muscle elastic properties (higher muscle compliance) with fatigue. For instance, studies have suggested that fatigue-induced modification of the viscoelastic properties in the muscle-tendon complex could be strongly associated with a longer EMD and higher CT (mechanical component) (Taylor et al., 1997; Zhang and Rymer, 2001). However, no findings allow us to determine this association in this study limiting conclusions. Thus, in the present study one head of the quadriceps muscle complex was evaluated limiting therefore interpretations about changes of the whole quadriceps neuromuscular properties with fatigue. The evaluation of the viscoelastic properties of muscles could help us to better understand the higher muscle compliance experienced after fatiguing isometric contractions. However, modification of the viscoelastic properties of skeletal muscle due to fatigue is yet to be investigated by supersonic imaging approaches.

This study provides scientific evidence related to changes of VL muscle μ and peripheral fatigue. However, it presents some methodological limitations: (i) in the present study the whole quadriceps complex was not evaluated limiting interpretation of our results concerning others synergist muscles (rectus femoris and vastus medialis) and their association with the whole quadriceps neuromuscular assessments. A recent study (Heales et al., 2018) indicated that muscle stiffness of RF, VL and VM was not similarly affected after series of eccentric contractions since it only increased in RF. These results may indicate that biarticular muscles such as RF are more affected by a damaging exercise than their monoarticular synergists (VL and VM). Then, (ii) it is possible that the magnitude of changes in muscle stiffness was different at other muscle lengths. Indeed, muscle stiffness was higher at long compared to short muscle lengths after eccentric contractions (Lacourpaille et al., 2014). However, the relationship between muscle fatigue, stiffness and muscle length need therefore to be more investigated.

Finally, given the characteristics of the population (well-trained young male soldiers) and the fatigue protocol (maximal isometric contractions), the results of the present study must be cautiously interpreted. It is possible that our results cannot be transferred to other populations (females, untrained men, older volunteers) or mode of contraction. Muscle fatigue is a complex phenomenon and variables such as sex (Hunter, 2009), age (Ratel et al., 2015), pathology (Garcia-Vicencio et al., 2016), task (Enoka and Stuart, 1992) have been shown to affect the magnitude of strength loss and the time of course of central and peripheral mechanisms. This intrinsic variability may have an important effect on the magnitude of the shear elastic modulus when fatigue is installed.