Eighteen healthy young males and females participated in this study (Table 1). The participants gave written informed consent after receiving a detailed explanation of the purposes, potential benefits, and risks associated with participation. This study was approved by the Research Ethics Committee of Chukyo University (2015-002, 2016-057) and conducted in accordance with the Declaration of Helsinki.

During 8-week intervention, all participants trained their left and right knee extensor muscles using isometric unilateral knee extension resistance training twice a week. The contraction intensity was >80% of their maximal voluntary contraction (MVC) torque, contraction time was 4 s of the sustained phase at or over the target torque with a 1-s increasing phase to the target torque, training volume was 3 sets of 10 contractions for each leg, and rest intervals between contractions and sets were 10 and 60~120 s, respectively. During training, the performed and target torques were shown on a monitor to provide visual feedback and an experimenter confirmed the performed torque and encouraged the participants to complete the training program. After the measurements of MVC every 2 weeks, the target torque was revised when MVC increased. The training was performed using a custom-made dynamometer (Takei Scientific Instruments Co., Ltd., Niigata, Japan) with a force transducer (LU-100KSE; Kyowa Electronic Instruments, Tokyo, Japan). The same training regimen was administered to older adults in our previous study (9).

The participants were divided into two groups of nine participants who ingested 5 g of fish protein (Alaska pollack protein, APP) or 5 g of casein as a control (CAS) in addition to normal daily diets. Amino acid scores for APP and CAS were both 100 and the detailed amino acid components of each protein are described previously (11). The participants took them in a cup of soup every day with either breakfast, lunch, or dinner. A cup of soup for APP/CAS was comprised of 5.2 of protein including 5 g of APP/CAS, 0.5/0.6 g of fat, and 10.7/10.2 g of carbohydrate. Total energy content of soup in APP and CAS was 68 and 67 kcal, respectively. Amounts of protein supplementations as a function of the body mass of participants for APP and CAS were 0.095 ± 0.016 and 0.086 ± 0.010 g/kg/d, respectively. These values of APP supplementation were calculated from the results of previous studies using rats (10, 11). On the other hand, the amount of CAS supplementation used in this study is relatively low in comparison with amounts used in previous studies (10–63 g/d or 0.3 g/kg/d) (18, 19). This amount of CAS could be used as a control condition since it may not be sufficient to act intrinsically via protein supplementation on muscle strength and mass. In fact, our previous study showed that ingestions of the same amounts of APP and CAS induced significant increases and no significant changes, respectively, in the muscle volume during 8 weeks of resistance training intervention in older adults (9). This suggests that these amounts of APP and CAS can lead to different muscular adaptations in each group after initiating resistance training. Nutritional supplementations of APP and CAS were applied as a randomized, double-blind, placebo-controlled treatment. To blind the study, we packed the soups for both conditions in powdered form in identical bags and the flavorings were controlled for each condition by a professional food company. A dietary survey was performed by a nutritionist with a license from the Ministry of Health, Labor and Welfare in Japan, using a brief-type self-administered diet history questionnaire (BDHQ) (20) to quantify total energy, carbohydrate, protein, and fat in daily diets and normalized values by individual body mass in order to estimate the participants' nutritional conditions during the intervention. This questionnaire was applied at around 6~8 weeks of intervention, and so the results should reflect nutritional conditions during the intervention.

During the 8-week intervention, the muscle strength, motor unit firing pattern, and anthropometry tests were conducted every 2 weeks: the day before the intervention commenced (0 wks), and 2 (2 wks), 4 (4 wks), 6 (6 wks), and 8 weeks (8 wks) after beginning the intervention. For familiarization with the motor tasks used in the measurements, the participants came to the laboratory before the first day of intervention.

Muscle strength of knee extensor muscles was measured based on MVC during isometric knee extension for the right leg using a custom-made dynamometer which was used in resistance training in the present study (see Resistance training for details). During MVC, participants' hip and knee joint angles were fixed at 90° (180° corresponds to full extension). The MVC trial consisted of 2–3 s of a gradual increase in knee extension torque to maximum effort and 2–3 s of a maintained plateau phase at maximum effort with a verbal count given at 1-s intervals. Two MVC trials were performed with a ≥ 2- min rest interval. The higher MVC from the two MVC trials was chosen and used in further analysis and as a reference to set a target force in resistance exercise. These MVC measurement procedures were determined according to the methods and equipment used in our previous studies (9, 21, 22).

After MVC trials, the participants performed unilateral submaximal isometric right knee extension ramp contraction from 0 to 30% of MVC (Ramp 30) and 0 to 70% of MVC (Ramp 70). These slow ramp contractions were applied in order to identify the motor unit recruitment threshold and calculate motor unit firing rates at various force levels. Ramp 30 consisted of a 15-s increasing phase from the baseline to 30% of the MVC force level with a 2% MVC/s rate of force increase and 15-s sustained phase at a 30% MVC force level. Ramp 70 consisted of an 18-s increasing phase from the baseline to 70% MVC force level with a 5% MVC/s rate of force increase. From two trials, that showing a smaller difference between target and performed torques was chosen for further analysis for Ramp 30 and Ramp 70, respectively.

Muscle mass of the whole body and lower limb was estimated from body impedance measurements (InBody270, InBodyJapan Inc., Tokyo, Japan) and body and fat masses were also measured. This method was used in our previous study and could detect increases in the muscle volume following resistance training with APP ingestion in older adults (9). Our previous study performed this body impedance method on different days over 3 weeks in fifty older male and female in order to test its reproducibility (22). The intraclass correlation coefficient (ICC) was 0.992 for muscle mass measured by this body impedance method. We thus believe that this body impedance method could provide detectable and reliable changes in muscle mass following the intervention.

During the submaximal ramp contractions, multi-channel surface EMG signals were recorded from the vastus lateralis (VL) muscle of the right leg using a semi-disposable adhesive grid of 64 electrodes (13 rows and 5 columns with one missing electrode at the corner) with a 1-mm diameter and 8-mm inter-electrode distance (ELSCH064NM2, OT Bioelettronica, Torino, Italy). Prior to attaching the electrodes, the skin was shaved and cleaned by alcohol. Conductive gels were inserted into the cavities of the electrode grid to assure proper electrode-skin contact. Monopolar surface EMG signals were recorded by a multi-channel surface EMG amplifier (quattrocento, OT Bioelettronica, Torino, Italy) and amplified by a factor of 150, sampled at 2,048 Hz, and converted to digital form by a 16-bit analog-to-digital converter together with the force transducer signal. The center of the electrode grid was located at the mid-point of the longitudinal axis of the VL muscle, i.e., the line between the head of the greater trochanter and inferior lateral edge of the patella, and the columns were aligned along the VL longitudinal axis. A reference electrode was placed at the right iliac crest. These electrode locations were the same as in our previous studies (9, 21, 22).

From the recorded multi-channel surface EMG signals, individual motor unit firing patterns were decomposed by the Convolution Kernel Compensation (CKC) technique (23–26). We used the decomposition procedure that was previously extensively validated for signals from various skeletal muscles (21, 26, 27) and used the pulse-to-noise ratio (PNR), introduced by Holobar et al. (28), as an indicator of the motor unit identification accuracy (28). Only motor units with PNR > 30 dB, corresponding to an accuracy of motor unit firing identification > 90%, were used for further analysis; all other motor units were discarded (28). After decomposition, we performed visual inspection to identify discharge times for individual motor units based on PNR, and the discharge times were used for calculation of instantaneous motor unit firing rates for individual motor units. We excluded the discharges with inter-discharge intervals <33.3 or >250 ms, since firing rates calculated from this range of inter-discharge intervals are unusually high (>30 Hz) or low (<4 Hz) for the VL muscle (21, 26).

Median firing rates of individual motor units were calculated from instantaneous firing rates during each 5% of MVC for Ramp30 and 10% of MVC for Ramp70. For example, the median firing rate at 15% of MVC for Ramp30 was calculated from the instantaneous firing rates in the interval from 12.5 to 17.5% of MVC, and the median firing rate at 50% of MVC for Ramp70 was calculated from instantaneous firing rates in the 45–55% MVC interval. Median firing rates with >30% coefficient of variation were excluded for further analysis (29). We divided the detected motor units into three groups by the recruitment force: motor units recruited at <10% (MU10), 10–20% (MU20), and 20–30% of MVC (MU30) for Ramp30, and 0-20% (MU20), 20–40% (MU40), and 40–60% of MVC (MU60) for Ramp70. Averaged values were calculated for each motor unit group with varying recruitment thresholds in different intervention periods. It is well-known that firing patterns are inconsistent among motor unit groups with different recruitment thresholds (30, 31). Our previous studies demonstrated recruitment threshold-dependent changes in motor unit firing patterns following resistance training (9, 22). Also, since the relationship between motor unit firing rate and exerted force is not linear (21, 32), firing rate should be analyzed at various force levels in order to assume the detailed neural adaptations. Therefore, the present study performed analyses of individual motor unit groups with different recruitment thresholds at various force levels. These procedures for calculating motor unit firing rates were employed in our previous studies (9, 21, 22).

The results are reported as the mean ± SD. To determine whether data showed a normal distribution, we performed the Shapiro-Wilk test before statistical analysis. We decided to use non-parametric statistical tests, since our results included non-normally distributed data and the sample size was not large enough to use parametric analysis.

For comparisons of age, height, body mass, muscle mass, fat mass, and lower limb muscle mass estimated by InBody, and MVC between the groups at 0 wks, the Mann-Whitney test was used. The body mass, muscle mass, fat mass, and lower limb muscle mass estimated by InBody, and MVC at 2, 4, 6, and 8 wks were normalized by the values at 0 W. The normalized values were analyzed by the Friedmann test to assess the effect of intervention, and the Bonferroni-Dunn test was used to compare the values at 2, 4, 6, and 8 wks with the value at 0 wks when a significant effect of intervention was detected (33). The Mann-Whitney test was performed to compare the nutritional parameters calculated from BDHQ between the groups. We also compared the change in absolute values (Δ) from 0 to 2, 4, 6, and 8 wks in body mass, muscle mass, fat mass, and lower limb muscle mass, and MVC by Mann-Whitney test.

Motor unit firing rates for each force level and motor unit group in different time periods were tested by the Kruskal-Wallis test to analyze the effect of intervention. Dunn's test was used to compare the values at 2, 4, 6, and 8 wks with the value at 0 wks when a significant effect of intervention was detected by the Kruskal-Wallis test (34). Motor unit firing rates at each force level were also compared among the motor unit groups with different recruitment forces by the Kruskal-Wallis test. When this test showed a significant effect of the motor unit group, firing rates at a specified force level were compared among the motor unit groups with different recruitment forces by Dunn's test (34).

Statistical analysis was performed using SPSS (version 21.0, SPSS, Tokyo, Japan). Since we used multiple comparisons in post-hoc tests, the level of significance was partly modified by Bonferroni correction to minimize the familywise error rate and maintain statistical power (35). The modified level of significance was 0.05/the numbers of comparisons for each analysis. Therefore, for example, the level of significance in the post-hoc test to compare the variables among the periods was set at 0.0125 (0.05/4) when the variables at 2, 4, 6, and 8 wks were compared with 0 wks (4 pairs).

No significant differences were observed in body mass, muscle mass, fat mass, or lower limb muscle mass estimated by InBody and MVC at 0 wks (Table 1) nor in their normalized values at 2, 4, 6, or 8 wks between APP and CAS groups (Table 2; p > 0.05). We detected a significant effect of intervention on MVC in both APP and CAS groups (p < 0.05), and MVCs at 4, 6, and 8 wks were significantly greater than those at 0 wks in both APP and CAS groups (p < 0.0125; Figure 1, Table 2). There were no significant effects of intervention on body mass, muscle mass, fat mass, or lower limb muscle mass estimated by InBody (Table 2; p > 0.05). No significant differences were found in changes in absolute values (Δ) from 0 to 2, 4, 6, and 8 wks in body mass, muscle mass, fat mass, and lower limb muscle mass, and MVC (p > 0.05; Table 2).

There were no significant differences in total energy (1,938.0 ± 602.1 kcal for APP and 1,988.5 ± 704.7 kcal for CAS), carbohydrate (279.5 ± 100.4 g for APP and 279.4 ± 92.2 g for CAS), protein (68.7 ± 26.4 g for APP and 71.3 ± 37.6 g for CAS), or fat (53.8 ± 20.6 g for APP and 58.0 ± 26.1 g for CAS) in daily diets and the values normalized by individual body mass in total energy (36.9 ± 12.9 kcal/kg for APP and 34.3 ± 12.0 kcal/kg for CAS), carbohydrate (5.2 ± 1.7 g/kg for APP and 4.8 ± 1.6 g/kg for CAS), protein (1.3 ± 0.6 g/kg for APP and 1.2 ± 0.6 g/kg for CAS), and fat (1.1 ± 0.5 g/kg for APP and 1.0 ± 0.4 g/kg for CAS) in daily diets between APP and CAS groups (p > 0.05 for all).

In total, 1,048 motor units were identified in this study: 396 in APP and 344 in CAS groups for Ramp 30 and 161 in APP and 147 in CAS groups for Ramp 70.

There were no significant changes in motor unit firing rates following the intervention for Ramp 30 in either the APP or CAS group (p > 0.05; Figure 2). No significant changes were noted in motor unit firing rates following the intervention for Ramp 70 in the CAS group (p > 0.05; Figure 3). In the APP group, significant changes in firing rates for MU40 and MU60 were noted at 70 and 60% of MVC, respectively (p < 0.05; Figure 3).

For Ramp 30, at 0 wks significant differences in firing rates were observed between MU10 and MU20 at 20% of MVC (p < 0.05) in the APP group and between MU10 and MU20 at 20, 25, and 30% of MVC (p < 0.05) and between MU10 and MU30 at 30% of MVC (p < 0.025) in the CAS group (Figure 4). Significant differences in firing rates among different motor unit groups that were not observed at 0 wks were noted during the intervention in the APP group: MU10 vs. MU20 at 25% of MVC at 2 wks (p < 0.05), MU10 vs. MU30 at 30% of MVC at 2 and 4 wks (p < 0.025), MU10 vs. MU30 at 30% of MVC at 2, 4, 6, and 8 wks (p < 0.025), and MU20 vs. MU30 at 30% of MVC at 4, 6, and 8 wks (p < 0.025) (upper panels in Figure 4), but not in CAS (lower panels in Figure 4).

For Ramp 70, significant differences in firing rates were observed between MU20 and MU40 at 60% of MVC (p < 0.025) in the APP group and between MU20 and MU40 at 60 and 70% of MVC (p < 0.025) in the CAS group (Figure 5) at 0 wks. Significant differences in firing rates among different motor unit groups that were not observed at 0 wks were noted during the intervention in the APP group: MU20 vs. MU40 at 70% of MVC at 2, 4, and 8 wks (p < 0.025) and MU20 vs. MU40 at 50% of MVC at 6 wks (p < 0.025), in addition to significant differences in firing rates among different motor unit groups that were observed at 0 wks (upper panels of Figure 5). Significant differences in firing rates among different motor unit groups that were observed at 0 wks were also noted at 4, 6, and 8 wks in the CAS group: MU20 vs. MU40 at 60 and 70% of MVC (p < 0.025) and no additional pairs of different motor unit groups with significant differences in firing rates during the intervention (lower panels of Figure 5).