The data presented in this study are part of the MEGA training study of which partial data have been presented (Inglis et al. 2024). As previously indicated (Inglis et al. 2024), ninety-nine young healthy females (n = 49) and males (n = 50) volunteered and provided written informed consent to participate after completing the physical activity readiness questionnaire (PAR-Q +) and being cleared for exercise by a certified exercise physiologist. Due to illness/injury (n = 5), COVID-19 related precautions (n = 4), or other personal reasons (n = 6), fifteen of these participants were unable to complete the study. Thus, eighty-four participants (42 F, 42 M) ranging from sedentary to recreationally active volunteered to participate in the study. Stratified randomization based on sex was used to equally assign participants to one of five cycling exercise intervention groups (n = 70) or control (CON) (n = 14). \(}\text}_}\) kinetics data for three participants were not collected due to technical issues and thus their data for the other measured variables are not presented. The sample size for the MEGA project was calculated considering the main outcome variable (i.e., maximal \(}\text}_}\).). However, the post analysis statistical power for \(\tau }\text}_}\) in this study indicated values greater than 0.8 for the main effects of time and group. All procedures included in this study were approved by the Conjoint Health Research Ethics Board at the University of Calgary (REB19-0452) and complied with the principles in latest version of the Declaration of Helsinki.

A metabolic cart and mixing chamber (Quark CPET; COSMED, Rome, Italy) were used to measure gas exchange and ventilatory variables continuously during maximal and MMSS testing sessions. A breath-by-breath setup was used for the submaximal protocol designed to evaluate \(}\text}_}\) kinetics. Prior to each testing session, gas and flow calibrations were performed according to manufacturer recommendations. Heart rate was also monitored continuously during all testing and training sessions (Garmin, Chicago, IL). Blood lactate concentration ([La−]) was measured from capillary blood samples obtained from a prick of the finger (Biosen C-Line, EKF Diagnostics, Barleben, Germany).

All maximal, MMSS and training sessions (apart from sprint interval training (SIT)) were performed using the Tacx NEO Smart Trainers (Garmin, Chicago, IL, USA) with custom, adjustable bike frames. A tablet with the Golden Cheetah software (version 3.4; https://www.goldencheetah.org/) was used to control the protocols. Participants were individually assigned to a training station that was kept consistent throughout the study. The \(}\text}_}\) kinetics protocol and SIT was performed on an electromagnetically braked cycle ergometer (Velotron Dynafit Pro, Racer Mate, Seattle, WA, USA) using the Velotron CS and Wingate software, respectively. For all training and testing sessions with the NEO system, participants were asked to maintain their preferred cadence between 70 and 95 rpm. This range was selected to accommodate the individual preferences of participants with limited cycling experience to make sure that they did not cycle at a cadence that was not comfortable for them. The majority of the participants had a preferred cadence between 75 and 85 rpm. Once a preferred cadence was established this was kept consistent throughout the study.

Participation in the study lasted approximately 11.5 weeks, with testing phases occurring before (PRE), between (MID), and immediately following (POST) exercise training. The intervention was composed of two, 3-week training phases (6 weeks total). Each testing phase consisted of one ramp incremental maximal session, 2–3 sessions to determine MMSS, and one fasted submaximal testing session. Time of day for the testing sessions was kept consistent (± 1 h) for each participant at PRE, MID and POST. A minimum of 24 h separated the maximal session from the first MMSS session, 48–72 h separated subsequent MMSS trials, and the fasted submaximal session was performed following a minimum of 48 h without testing/training. The CON group performed PRE and POST testing.

Maximal session. To measure \(}\text}_}}}\), HRmax, PPO, θLT, and the respiratory compensation point (RCP), participants performed a cycling ramp incremental test (Females: 20 W·min−1, Males: 25 W·min−1) to task failure. Prior to the ramp, participants performed 4 min of cycling at 30 W followed by 8 min of cycling in the moderate intensity domain (60–80 W individual dependent). Then, following 3 min of rest, participants performed a 4-min bout of cycling at 30 W before the ramp portion of the test began. Participants cycled until task failure, defined as the point at which there was a drop in cadence below 65 rpm for more than 5 consecutive seconds or participants could not continue to cycle, despite strong verbal encouragement. Then, after a brief recovery period (~ 7–10 min), participants performed a constant work rate trial that consisted of a 2-min baseline at 30 W followed by square wave step transition to ~ 80–85% of PPO performed to task failure. The purpose of this trial was to induce maximal responses for a separate aim of the MEGA study. Immediately (within 30 s) following both the ramp-incremental and the constant work rate trial, blood lactate ([La−]) samples were taken.

MMSS Sessions. Participants completed 2–3 constant PO exercise trials to determine the power output (PO) associated with MMSS. Each trial began with 4 min of cycling at 30 W, followed by an instantaneous increase to a pre-determined PO for 30 min during which [La−] was measured at 5 min intervals and \(}\text}_}\) was measured continuously. The PO for the initial trial was determined using a predictive equation based on measured RCP (Iannetta et al. 2018). For subsequent trials, and to obtain measures of MMSS PO with a small margin of error, the PO was increased or decreased by 5% of PO based on the [La−] and \(}\text}_}\) responses. Blood [La−] samples were obtained at baseline and at 5-min intervals thereafter in duplicate or triplicate (15th and 30th min), and each timepoint was represented by the average of the two available samples or the average of the two closest samples when taken in triplicate. As previously defined (Iannetta et al. 2022b), MMSS was determined to be the highest PO at which stable [La−] and V̇O2 responses were achieved between the 15th and 30th min of exercise.

\(}\text}_}\) kinetics. The \(}\text}_}\) kinetics protocol consisted of three cycling moderate step transitions (Spencer et al. 2011) from a 20-W baseline period (6 min) to 80–85% of the PO associated with the θLT (6 min) at PRE using a square wave transition. The PO was kept consistent at PRE, MID and POST for each participant. Throughout the protocol, participants were asked to maintain a steady cadence in the range of 65–75 rpm.

NIRS derived Oxidative Capacity. A NIRS probe (PortaMon, Artinis Medical Systems, Elst, The Netherlands) was used to provide an indirect measure of mitochondrial oxidative capacity as previously described (Hamaoka et al. 1996; Motobe et al. 2004; Rasica et al. 2024). Briefly, the NIRS probe was placed on the lower third of the vastus lateralis (VL) and covered with an elastic bandage to prevent movement of the probe and the intrusion of external light. A cuff connected to a pneumatic automatic rapid inflation system (Hokanson E20, Bellevue, WA, USA) was placed on the proximal portion of the right thigh (above the VL NIRS probe) to occlude blood flow to the leg. Ankle weights (3.5 kg for females and 4.5 kg for males) were placed on the ankle of the right leg. First, participants performed 20 repetitions of knee extension in a seated position at 0.5 Hz (1 s extension, 1 s flexion). Immediately after, the cuff was inflated to 300 mmHg for 5 s and subsequently deflated for 10 s and this inflation-deflation procedure was repeated 20 times over a 5-min period. Following 2-min of rest, once the NIRS signals had stabilized, a second trial was performed. NIRS variables were recorded continuously (10 Hz) throughout the repeated occlusion protocol.

Exercise training was separated into two 3-week phases, each consisting of nine sessions (three sessions per week) for a total of 18 sessions at the end of the program. Exercise training sessions were completed Monday through Saturday with ≥ 24 h between sessions and at least one 48 h rest period within each week (i.e., participants were not allowed to exercise on three consecutive days). All training sessions were supervised and adherence across all intervention groups was 98 ± 5%, as previously described (Inglis et al. 2024). Detailed information on reasons for the selected intensities and durations for the five exercise interventions have been presented elsewhere (Inglis et al. 2024). Briefly, the groups performed cycling at a constant-PO in the (i) moderate-intensity domain (MOD) (50-min of cycling at 90% of the PO at θLT); (ii) lower boundary of the heavy-intensity domain (HVY1) (~ 41-min at 110% of the PO at θLT); (iii) upper boundary of the heavy-intensity domain (HVY2) (~ 30-min at 100% of the PO at MMSS); or interval training in the form of (iv) high-intensity interval training (HIIT) in the severe-intensity domain (5 to 6 intervals with a 4:3-min work:rest ratio); and (v) SIT in the extreme-intensity domain (up to 6, 30 s maximal-effort sprints against a fixed resistance (0.06 to 0.075 kg-kg−1 body mass)). Participants in the MOD, HVY1, HVY2, and HIIT groups performed a standardized 2-min warmup at 55% of θLT whereas the SIT group performed a 10-min warmup at 50 W with three 3-s practice sprints interspaced within the later half. SIT participants started with three sprints per session, with the number of sprints per session progressively increasing by a minimum of one sprint per week until six sprints were performed by week four. The total work performed in HVY1, HVY2, and HIIT were individually work-matched to the total work of each participants’ MOD equivalent (i.e., 50-min of cycling at 90% of the PO at θLT). The training intensities in phases 1 and 2 were based on the results of testing at PRE and MID, respectively (i.e., the training load was adjusted following MID to ensure adequate classification within the domain-specific regions).

Raw ventilatory and gas-exchange data from the ramp incremental test were independently evaluated by two experts to identify the \(}\text}_}\) corresponding to the θLT and RCP. In case of a disagreement of > 100 mL/min, both experts re-evaluated the profile together until an agreement was reached. The θLT and RCP were established as previously detailed (Beaver et al. 1986; Gaesser and Wilson 1988; Whipp et al. 1989; Keir et al. 2022).

\(}\text}_}\) data were filtered by removing aberrant points that fell 3 standard deviations from the local mean prior to being linearly interpolated on a second-by-second basis. The moderate step transition performed prior to the ramp-incremental portion of the test was used to calculate the \(}\text}_}\) mean response time as previously described (Iannetta et al. 2019). Briefly, a linear fit was applied to the ramp-incremental test V̇O2 data from the ramp onset to the established θLT. The \(}\text}_}\) of the moderate step transition (average of the last 2-min) was then superimposed on the \(}\text}_}\) vs PO relationship from the ramp. Then, the difference in PO corresponding to the abscissa identified during the ramp linear fit versus that of the moderate-step transition was converted to time and then used to align the ramp-incremental \(}\text}_}\) with the corresponding PO, allowing for the identification of PO at the θLT. The highest 20-s rolling average from either the ramp incremental test or the subsequent constant work rate trial was defined as \(}\text}_}}}\).

\(}\text}_}\) kinetics. As previously described (Keir et al. 2014), the \(}\text}_}\) data for each step transition were cleaned, time aligned, and interpolated before being ensemble-averaged into a single time-averaged response. Then, each individual ensemble-averaged profile was time-averaged into 5 s bins (Keir et al. 2014) and fit using the following equation:

$$}\text}_} \left( } \right) \, = }\text}_}}} + }\text}_} \cdot }} \left( - }^} - }} \right)/\tau }} } \right)$$

where \(}\text}_} \left( } \right)\) represents the \(}\text}_}\) at any given time (t) during the transition, \(}\text}_}}}\) is the steady-state baseline value of \(}\text}_}\) before the moderate step-transition, \(}\text}_}}}\) is the amplitude of the increase in \(}\text}_}\) above \(}\text}_}}}\), TD is the time delay of the response, and τ is the time constant of the response (defined as the time required to attain 63% of the steady-state amplitude). To account for the phase I (i.e., the cardiodynamic phase) of the \(}\text}_}\) response, the first 20 s of the ensemble-averaged \(}\text}_}\) profile were not included in the fitting window of the phase II \(}\text}_}\) (i.e., the primary component reflecting the adjustment of muscle \(}\text}_}\)) across all participants as previously recommended (Murias et al. 2011b). Data were modelled from the beginning of phase II up to 240 s of the step-transition, after ensuring that steady-state \(}\text}_}\) had been attained within this time window. The parameter estimates were computed by least squares non-linear regression using the whippr open-source R package, with the best fit defined by minimization of the residual sum of squares and minimal variation around the Y-axis (\(}\text}_}\) = 0). The 95% confidence interval for the estimated τ was determined after preliminary fit of the data with \(}\text}_}}}\), \(}\text}_}}}\), and TD constrained to the best fit values and the τ allowed to vary.

Oxidative Capacity. \(}\text}_}\) at the level of the muscle (\(}\text}_} }\)) was estimated by calculating the slope of saturation (StO2) over a 3-s span of data, excluding the data points from the first and last second of each occlusion period to avoid any potential influence from cuff inflation or release.

\(}\text}_} }\) values were respectively fit by the following monoexponential function:

$$}\left( } \right) = }_}}} - } \times }^}}$$

where \(}\left( } \right)\) = \(}\text}_} }\) at a given time (t); \(}_}\) = \(}\text}_} }\) immediately after the cessation of the exercise; A = amplitude of the response; τ = exponential recovery rate constant (τ, τ = \(\frac}\)).

As previously indicated (Beever et al. 2020), prior to fitting, the data were visually inspected to remove invalid values or outliers possibly caused by partial occlusions or disturbances in the NIRS signal. After identifying the start of the monoexponential decay curve, data points preceding this point were not considered for the curve fitting. Points within the plateau were only removed if they dissociated enough from the curve or plateau to suggest that the point was invalid (Beever et al. 2020). A single value of τ (τOxCap) was reported for each participant, which was determined by taking the average τ obtained from the two trials (Rasica et al. 2024). Due to technical issues (i.e., adipose tissue thickness causing NIRS signal disturbance), the OxCap analysis was completed in 70 of the 84 participants (CON = 10, MOD = 13, HVY1 = 12, HVY2 = 13, HIIT = 11, SIT = 11). The r2 of the fits for each group ranged from 0.92–0.96, indicating a high quality fit.

A one-way ANOVA was used to compare baseline group characteristics groups. A paired-samples t-test was used to evaluate changes within the CON group from PRE to POST. To compare the effect of time (PRE, MID, and POST) and group on the variables of interest in the intervention group, a mixed model ANOVA was performed. Tukey’s post-hoc tests were used to confirm significance among multiple comparisons. Effect sizes are reported as partial eta-squared (partial η2), where values of 0.01, 0.06 and 0.14 correspond to small medium and large effects, respectively (Lakens 2013). A Pearson product moment correlation was used to determine the strength of relationship between changes in \(\tau }\text}_}\) and performance variables (\(}\text}_}}}\), θLT, MMSS and PPO). The strength of association for the correlation values was defined as strong (|r|> 0.5), (medium 0.3 >|r|), or weak (0.3 >|r|> 0.1). Statistical significance was set at p < 0.05. Statistical analyses were performed using SPSS Statistics v. 26.0 (SPSS; IBM, Chicago, IL).