Does altitude level of a prior time‐trial modify subsequent exercise performance in hypoxia and associated neuromuscular responses?

Abstract We examined the influence of prior time‐trials performed at different altitudes on subsequent exercise in moderate hypoxia and associated cardiometabolic and neuromuscular responses. In normobaric hypoxia (simulated altitude 2000 m; FiO2: 0.163), 10 healthy males performed (1) an incremental test to exhaustion (VO 2max_2000) and (2) a test to exhaustion at 80% of the power output associated to VO 2max_2000 for a reference time (947 ± 336 sec). Thereafter, two sessions were conducted in a randomized order: a cycle time‐trial corresponding to the reference time (TT 1) followed 22 min later (passive rest at 2000 m) by a 6‐min cycle time‐trial (TT 2). TT 1 was either performed at 2000 or 3500 m (FiO2: 0.135), while TT 2 was always performed at 2000 m. As expected, during TT 1, the mean power output (247 ± 42 vs. 227 ± 37 W; P < 0.001) was higher at 2000 than 3500 m. During TT 2, the mean power output (256 ± 42 vs. 252 ± 36 W) did not differ between conditions. Before and after TT 1, maximal isometric voluntary contraction torque in knee extensors (pooled conditions: −7.9 ± 8.4%; P < 0.01), voluntary activation (−4.1 ± 3.1%; P < 0.05), and indices of muscle contractility (peak twitch torque: −39.1 ± 11.9%; doublet torques at 100 Hz: −15.4 ± 8.9%; 10/100 Hz ratio: −25.8 ± 7.7%; all P < 0.001) were equally reduced at 2000 m or 3500 m. Irrespective of the altitude of TT 1, neuromuscular function remained similarly depressed after TT 1 both before and after TT 2 at 2000 m. A prior time‐trial performed at different altitude influenced to the same extent performance and associated cardiometabolic and neuromuscular responses during a subsequent exercise in moderate hypoxia.


Introduction
Fatigue is a disabling symptom in which physical and cognitive function is limited by interactions between performance fatigability (i.e., the decline in an objective measure of performance such as contractile function and/or muscle activation) and perceived fatigability (i.e., sensations that regulate the integrity of the performer) (Enoka and Duchateau 2016). Under this frame, prior exercise that challenges multiple physiological (i.e., cardiorespiratory, metabolic, neuromuscular) regulatory systems and thereby modifies effort perception and influences performance during the completion of a subsequent exercise bout (Karlsson et al. 1975;Hogan and Welch 1984;Amann 2011). Currently, there is renewed interest in the methodological approach that consists in manipulating the amount/severity/type of pre-established fatigue levels through the induction of localized fatigue of specific muscle groups (e.g., neuromuscular electrical stimulation protocol of the quadriceps; Hureau et al. 2014) or following the completion of an initial whole-body exercise bout (Amann and Dempsey 2008;Girard et al. 2015).
In one study, Amann and Dempsey (2008) demonstrated that the induction of different levels of pre-existing locomotor muscle fatigue [following the completion of constant-load cycling of~10 min at 347 W (83% of peak power output) vs.~276 W (67%)] had a substantial dose-dependent influence (À6%) on performance time during a subsequent (i.e., 4 min later) 5-km time-trial (TT). Specifically, the higher the level of pre-existing locomotor muscle fatigue, as assessed via preand post-exercise magnetic femoral nerve stimulation, the lower the central motor drive and power output during the subsequent TT. A striking finding, however, was that the end of the subsequent TTs coincided with an almost identical level of peripheral fatigue, independent of the level of pre-existing fatigue and/or the marked differences in exercise performance. Authors claimed that feedbacks from fatiguing muscles play an important role in the determination of central motor drive and force output, and that the development of peripheral muscle fatigue is confined to a certain level (also referred as a "critical" threshold), so as not to surpass a sensory tolerance limit.
Increasing the degree of environmental stress such as the ambient hypoxia severity [i.e., a reduction in environmental oxygen (O 2 ) availability] is known to exacerbate exercise-induced demands (and thereby recovery requirements) in turn leading to excessive fatigue levels (Amann 2011). With this in mind, we recently manipulated hypoxia severity during an initial set of repeated sprints (eight 5-sec sprints; 30 sec rest) under normoxia, moderate, or severe hypoxia and examined the effects on performance and lower limbs neuromuscular activity during a subsequent set (four sprints) performed 6 min later in normoxia (Girard et al. 2015). Despite sprint performance and neural alterations were largely influenced by the hypoxia severity in the initial set, hypoxia had no residual effect during the subsequent normoxic set (i.e., similar fatigue pattern across conditions), yet the absence of negative "carry-over" effects would still need to be confirmed under hypoxic conditions. In addition, central and peripheral mechanisms underpinning neuromuscular fatigue, likely influenced by hypoxia severity, were not assessed in this latter study.
To date, the majority of studies investigating the interplay between central and peripheral mechanisms of fatigue through hypoxic perturbations have employed exhaustive, whole-body continuous exercise (Amann et al. 2006a) and repeated, brief submaximal (Millet et al. , 2012Goodall et al. 2010) or maximal (Christian et al. 2014) isometric contractions. In comparison to exercising at fixed work rate to fatigue (i.e., open-loop design), selfpaced exercise (i.e., time-trial) of predetermined duration/distance/work (i.e., closed-loop designs) involves pacing strategies for achieving optimal performance. It well described that TT performance and associated physiological responses vary between hypoxic and normoxic conditions (Amann et al. 2006b;P eriard and Racinais 2016;Saugy et al. 2016). The extent to which the neuromuscular consequences also differ between normoxic and a range of moderate-to-severe hypoxic conditions is not clear.
Our intention was therefore to examine the influence of prior TT performed at different altitudes on subsequent exercise performance in moderate hypoxia and associated cardiometabolic and neuromuscular responses. We anticipated that severer hypoxia during a prior TT would exaggerate the demands placed on various physiological regulatory systems, in turn resulting in lower power output. We were unsure, however, whether specific recovery requirements and fatigue-related residual or "carry-over" effects would differently influence performance and associated cardiometabolic and neuromuscular responses during a subsequent exercise performed in same moderate hypoxia.

Participants
Ten healthy men (Mean AE SD: age 34.4 AE 6.8 years; height 180.4 AE 7.7 cm; body weight 78.6 AE 10.9 kg; body fat 15.0 AE 5.9%) volunteered to participate in the study. All participants were born and raised at <1000 m and had not traveled to elevations >1000 m in the 3 months prior to investigation. They gave their informed, written consent preceding the commencement of the experiment. Experimental protocol was conducted according to the Declaration of Helsinki for use of Human Subjects and approved by the Ethics Committee of Valais, Switzerland (CCVEM 007/10).

Experimental design
Participants visited the laboratory on two occasions, separated by 3-7 days, before the two main experimental trials. During the first preliminary visit, they were initially familiarized with neuromuscular testing and performed an incremental test in normobaric hypoxia at a simulated altitude of 2000 m (FiO 2 0.163) for maximal oxygen uptake determination (VO 2max_2000 ). During a subsequent preliminary visit, again at 2000 m, participants cycled at constant workload (80% of the power output associated with their VO 2max_2000 : 245 AE 42 W) to exhaustion (Tlim = 947 AE 336 sec). Exercise was terminated when pedaling rate dropped below 60 rpm for >5 sec (exhaustion). Thereafter, two sessions were conducted in a randomized order: a cycle time-trial (TT 1 ) with Tlim as individualized reference duration followed 22 min later (passive rest at 2000 m) by a 6-min cycle time-trial (TT 2 ). TT 1 was either performed at 2000 or 3500 m (FiO 2 0.135), while TT 2 was always performed at 2000 m. After preliminary tests, the duration of 22 min between the two time-trials was chosen as a good trade-off between the times needed to allow significant perceptual recovery from TT 1 , with the parallel methodological requirement to keep recovery time short enough for a partial recovery of neuromuscular function, both likely to influence subsequent efforts (Minett and Duffield 2014).

Testing procedures
Testing was conducted in a normobaric hypoxic chamber (SL -400, ATS, Sydney, Australia) of~30 m 3 (2.4 m 9 5.0 m 9 2.5 m) maintained at a constant temperature of~25°C and~40% relative humidity. Prior to each cycling bout, FiO 2 (Oximeter Gox 100, Greisinger, Germany) as well as room temperature (°C) and humidity (%) were measured. This chamber allowed modifying the simulated altitude between 3500 and 2000 m in less than 5 min. All tests were performed on a computer-controlled electrically braked cycle ergometer (Lode Excalibur Sport, Groningen, the Netherlands). Participants performed their trials at the same time of the day (AE1 h) and wore similar sports gear (cycling shoes, short and jersey). They were instructed to refrain from any strenuous physical activity and maintain their normal diet (i.e., avoiding any nutritional supplements, caffeine, or alcohol consumption) and sleeping habits (≥7 h/night) for the 24 h before each test. They were encouraged to drink 4-6 mL of water per kilogram of body mass every 2.5 h on the day before each experimental session to ensure euhydration at the start of exercise.

Preliminary sessions
During the first preliminary session, participants' anthropometrical parameters (body height, body weight) and body composition (Bod Pod, Cosmed US Inc., Concord) were first measured. Then, they were requested to perform maximal voluntary contractions (MVC) of the knee extensors until they felt accustomed to the equipment; the coefficient of variation in three successive trials was <5%. Afterwards, the optimal stimulation intensity for one single stimulus was determined by increasing the current gradually from 10 mA until there was no further increase in peak twitch torque and concomitant maximal M-wave amplitudes. This intensity was further increased by 30% (e.g., supramaximal) and subsequently maintained for the entire session. Thereafter, participants performed the complete procedure of neuromuscular tests (see below; total duration~3 min). Then, they entered the normobaric hypoxic chamber at a simulated altitude of 2000 m, and after~30 min rest, the incremental test was undertaken. The protocol consisted of cycling for 5min warm-up at 60 W, then at a starting power output of 90 W and increasing every minute by 30 W until volitional exhaustion, despite strong verbal encouragement. The highest 30 sec average value of VO 2 (see below) was defined as VO 2max_2000.
The second preliminary session was conducted inside the hypoxic chamber at a simulated altitude of 2000 m as follows: (1) rest in a seated position for 30 min; (2) Participants cycled at constant workload (80% of the power output associated with their VO 2max_2000 ) to exhaustion.

Cycling time-trials
During each time-trial, participants were asked to maintain the highest sustainable effort, while receiving strong verbal encouragements. The starting work rate was 80% of individual power output at VO 2max_2000 for all exercise trials. Participants were informed of every minute elapsed during time-trials. They were able to continuously selfregulated power output (AE10 W). This research was run in double-blinded, controlled manner. Participants were told that the overall goal of the experiment was to test the reproducibility of their cycling time-trial performance in hypoxia, yet without any accurate information about the randomized simulated altitude levels inside the hypoxic room that were set and controlled by an independent research assistant.

Neuromuscular function
The neuromuscular assessment consisted of a 4-sec MVC of the knee extensors with a superimposed 100 Hz doublet (Db100) applied to the peripheral motor nerve when torque had reached a visible plateau. This was followed after pre-TT 1 neuromuscular assessment, participants were warmed up by completing 5 9 4-sec MVC with progressively increasing subjective effort (starting at 50% of subjective maximal effort with increments of 10%) followed by 2 9 4-sec MVC (separated by 1 min of passive rest).

Cardiopulmonary and respiratory responses
Heart rate (HR), monitored via a wireless Polar monitoring system (Polar Electro Oy, Kempele, Finland) and pulse oxygen saturation (SpO 2 ), estimated noninvasively via pulse oximetry using an earlobe probe (Nonin, Wristwatch, McAllen), were recorded immediately prior to entering the chamber and every 60 sec during exercise. Rating of perceived exertion (RPE) was obtained using the 6-20 Borg scale. The following respiratory variables and pulmonary gas exchange parameters were measured breath-by-breath at rest and throughout all exercises using the portable analyzer Metamax 3b (Cortex Biophysik, Leipzig, Germany). The following parameters were measured: oxygen uptake [VO 2 (mL kg À1 min À1 )], minute ventilation [VE (L min À1 )], carbon dioxide production [VCO 2 (L min À1 )], respiratory exchange ratio (RER), ventilatory equivalent for O 2 (VE/VO 2 ), and breathing frequency [BF (breath min À1 )]. Finally, a capillary blood sample was taken from the fingertip and analyzed for blood lactate concentration with the Lactate Pro (LT-1710, Arkray, Japan) portable analyzer, 2 min before and exactly 2 min after TT 1 and TT 2 .

Force and electromyographic recordings
During all neuromuscular assessments, participants were seated upright on a custom-built adjustable chair with the hips and knees flexed at 90°(0°corresponding to full knee extension). Restraining straps placed across the chest and hips secured the participants in the chair to prevent extraneous movement, while the dynamometer (Captels, St Mathieu de Treviers, France) was attached 3-5 cm above the tip of the lateral malleoli.
Electromyographic (EMG) signals of the VL and rectus femoris (RF) muscles (cycling and neuromuscular assessment) were recorded via bipolar Ag/AgCl electrodes (Ambu Blue sensor T; Ambu A/S, Denmark) with a diameter of 9 mm and an interelectrode distance of 25 mm. Before electrode placement, the skin was lightly abraded and washed to remove surface layers of dead skin, hair, and oil. The ground electrode was attached to the right wrist. The position of the EMG electrodes was marked with indelible ink to ensure that they were placed in the same location during subsequent trials. To ensure low levels of movement artifact, electrode cables were fastened to participant' bodies with medical adhesive tape and wrapped in net. The myoelectric signal was amplified (Octal Bioamp, ML138, ADInstruments, Oxfordshire, UK; input impedance = 200MΩ, common mode rejection ratio >96 dB, gain = 1000), band-pass filtered (bandwidth frequency = 5 to 500 Hz), digitized (sampling frequency = 2,000 Hz), acquired, and later analyzed (LabChart v7.0, ADInstruments Inc, Oxfordshire, UK) with force signal.

Femoral nerve stimulation
A high-voltage (maximal voltage 400 V) constant current stimulator (Digitimer DS7AH; Digitimer, Hertfordshire, UK) was used to deliver square-wave stimuli of 1 ms duration. The femoral nerve was stimulated percutaneously via a 10 mm diameter self-adhesive cathode electrode (Skintact, Austria) pressed manually by the investigator onto the skin at the femoral triangle. The anode, a self-adhesive pad (5 9 10 cm, Medicompex; Ecublens, Switzerland) was applied to the gluteal fold.

Data analysis
Power output, cardiorespiratory and pulmonary as well as EMG data were averaged over each entire cycling bout to obtain one value for TT 1 (TT 1_2000 vs. TT 1_3500 ) and TT 2 (TT 2_2000 vs. TT 2_3500 ) For each neuromuscular test sequence, voluntary force (MVC force) was recorded over the highest 1-sec plateau preceding the superimposed twitch. The peak potentiated twitch force (i.e., the highest value of twitch tension production, Pt) was determined from the mechanical response of the three evoked twitches (and averaged to obtain one Pt value) and one paired doublet at 10 Hz and 100 Hz (Db10 and Db100, respectively). The peak-to-peak amplitude of the concomitant VL and RF M-waves during the three resting twitches was measured and averaged across the three stimulations to obtain one representative M-wave value. Voluntary activation (VA) was assessed using twitch interpolation and defined as follows: VA (%) = {[1 -(superimposed Db100 doublet/resting potentiated Db100)] 9 100}. From the mechanical response induced by paired high-frequency [100 Hz (i.e., 10-ms interstimulus interval)] and low-frequency [10 Hz (i.e., 100 ms interstimulus interval)] supramaximal electrical stimulation, the low-to high-frequency torque ratio was calculated (10/100 Hz) and used as a surrogate of low-and high-frequency tetanic stimulations (Verg es et al. 2009 to the quadriceps both before and after "fatigue" has been reported elsewhere (Place et al. 2007).

Statistics
Values are expressed as mean AE SD. For TT 1 and TT 2 separately, paired samples t-tests were used to compare cardiorespiratory and pulmonary and EMG data. Two-way analysis of variance (ANOVAs) [Time (pretests and posttests) 9 Condition (TT 1 and TT 2 )] were also used to compare neuromuscular responses. To assess the assumptions of variance, Mauchly's test of sphericity was performed using all ANOVA results. A Greenhouse-Geisser correction was performed to adjust the degree of freedom if an assumption was violated, while post hoc pairwise comparisons with Bonferroni-adjusted P values were performed if a significant main effect was observed. For each ANOVA, partial etasquared was calculated as measures of effect size. Values of .01, .06, and above .14 were considered as small, medium, and large, respectively. All statistical calculations were performed using SPSS statistical software V.21.0 (IBM Corp., Armonk, NY). The significance level was set at P < 0.05.

Preliminary sessions
At a simulated altitude of 2000 m, the maximal oxygen uptake (VO 2max_2000 ) was 54.9 AE 7.3 mL kg À1 min À1 and peak power output was 306 AE 53 W. Tlim performed at 245 AE 42 W was 947 AE 336 sec.

Summary of main findings
The novel findings were as follows: (1) Neuromuscular adjustments resulting from the completion of the initial cycling bout of same duration were independent of the hypoxia severity with, in particular, profound alterations in muscle contractility (i.e., low-frequency fatigue); and (2) There was no influence of the altitude of a prior time-trial on performance and associated cardiometabolic responses, with no additional muscle fatigue development, during a subsequent 6-min time-trial performed in moderate hypoxia. We further demonstrated that performance is impaired during a time-trial (approximately 16 min) conducted at a simulated altitude of 3500 m versus 2000 m, while active muscle (quadriceps) activation responses did not differ significantly. Also, after 22 min of passive rest, recovery of neuromuscular function was minimal in both conditions. Thus, we dispute the fact that altitude level of prior time-trial influences performance and associated cardiorespiratory and neuromuscular responses during completion of a subsequent exercise of similar nature in moderate hypoxia. not differ between the two levels of hypoxia. A substantial part of this fatigue was due to neural factors, consistent with previous time-trial studies (Dahlstrom et al. 2013;Saugy et al. 2016). Hence, voluntary activation values dropped similarly by~4%, signifying that muscle activation became suboptimal at the end of both TT 1_2000 and TT 1_3500 . This was further accompanied by near-identical level of peripheral muscle fatigue, whereas M-wave amplitudes of VL and RF muscles were not affected by the time-trial. This observation confirms that the measured changes in twitch amplitude are mainly due to changes occurring within the quadriceps and that neuromuscular propagation failure might be excluded. An increased rate of accumulation of metabolites (e.g., H+, Pi) can directly inhibit the contractile apparatus or disrupt the Ca 2+ release and uptake pathways in the sarcoplasmic reticulum (Allen et al. 2008). A valid method to identify excitation-contraction coupling as a major factor of fatigue is the examination of force loss in response to low-(10 Hz) and high-(100 Hz) frequency paired stimuli (Verg es et al. 2009). As confirmed in this study, low-frequency fatigue has already been identified as one of the main fatigue-causing factor for high-intensity exercises, irrespectively of hypoxia severity (Christian et al. 2014).
No negative "carry-over" effects on subsequent exercise in moderate hypoxia (TT 2 ) The duration of the recovery period after prior exercise (i.e., SpO 2 recovery to initial values) may at least partially dictate performance during completion of a subsequent exercise bout. In this study, hypoxia severity during TT 1 had no negative "carry-over" (or residual) effect on performance fatigability and associated cardiorespiratory and EMG variables during TT 2 completed after 22 min of rest in moderate hypoxia. In support, we recently observed that despite differing hypoxic severity levels (FiO 2 = 0.133, 0.168 and 0.209) during an initial set of eight 5-sec sprints, performance and neuromechanical patterns did not differ during four additional sprints performed, 6 min later (passive rest), in normoxia (Girard et al. 2015). These results somehow contrasts with those of Amann et al. (2007) who indicated that after constant load cycling exercise to exhaustion in normoxia (~171 sec), moderate (~278 sec) and severe (~171 sec) hypoxia (FiO 2 = 0.209, 0.15, and 0.10, respectively), hyperoxygenation via acute O 2 supplementation (FiO 2 = 0.30) caused participants to prolong exercise time at task failure in severe hypoxia (+171%), but not in normoxia and moderate hypoxia. Important methodological differencesthat is, the range of SpO 2 values for the more severe hypoxic conditions (67 vs. 80%), the nature of the exercise (constant load vs. time-trial), the recovery duration between the two subsequent exercises (few seconds vs. 22 min), the O 2 condition of the second exercise bout (hypoxia, normoxia vs. hyperoxia)therefore only lead to anecdotal comparisons of performance and physiological responses between the aforementioned studies and the present one. A further interesting observation was that the additional neuromuscular consequences of completing TT 2 were minimal (i.e., no difference between pre-TT 2 and post-TT 2 ) in both conditions. The short duration (6 min) of the second time-trial might partially explain why we failed to observe the further development of significant decrement in muscle activation (at least as for TT 1 ), as greater central fatigue has been detected after longer time-trials only (Thomas et al. 2015;Froyd et al. 2016); even though these observations have been obtained near sea level with no prior exercise. In this study, since the level of locomotor muscle fatigue pre-TT 2 versus post-TT 2 was also identical, the individual critical threshold of peripheral fatigue (Amann et al. 2006b) may have already been achieved when TT 2 started. Despite exercise duration was approximately three times shorter, average power output maintained during TT 2 resembles those produced during TT 1, with also similar power values between conditions.
Prior time-trial (TT 1 ) performance was lower at severe versus moderate hypoxia As expected, average power output was lower during TT 1_3500 versus TT 1_2000 . Our results are in line with previous studies that showed that aerobic performance is altered with altitude: Compared to normoxia, this decrease was reported to be between 10 and 20% at 3000 m (Ventura et al. 2003) in normobaric hypoxia. In this study, we did not measure performance or VO 2max in normoxia and therefore cannot compare directly our data. Many studies (Peltonen et al. 2001;Wehrlin and Hallen 2006;Mollard et al. 2007;P eriard and Racinais 2016) demonstrated that the reduced VO 2max was induced by a decrease in SpO 2 , leading to aerobic performance impairment. It is possible that the type of hypoxic condition (e.g., normobaric hypoxia) may explain the relatively moderate 8% difference in performance between the two TT 1 since we recently showed that the performance is less altered in normobaric than hypobaric hypoxia (Saugy et al. 2016). The main mechanism for endurance performance impairment in hypoxia might be the decrease in VO 2max with a mean decrease of 7.7% per 1000 m increase in altitude [À4% at 1000 m (Wehrlin and Hallen 2006;Mollard et al. 2007); À10% (Mollard et al. 2007) to À15% (Peltonen et al. 2001;Wehrlin and Hallen 2006;P eriard and Racinais 2016) at 2500-3000 m; or À30% (Mollard et al. 2007) at 4500 m, all in normobaric hypoxia]. This reduction in convective O 2 transport during hypoxic cycling precipitate a decrement in peak exercise capacity and therefore a shift of a given absolute work load to a higher relative intensity. Alterations in convective O 2 transport to the working muscles are the result of changes in arterial oxygen content and/or limb blood flow . Obviously, the VO 2max was strongly linked to SpO 2 levels also tainted by altitude (Chapman 2013). Reportedly, a > 3% reduction in SpO 2 from rest has a significant detrimental effect on VO 2max (Harms et al. 1997). SpO 2 maintenance, and not baseline VO 2max levels per se, is a primary limiting factor determining VO 2max decline with exposure to acute altitude (Chapman 2013). In this study, 2000 m and 3500 m simulated altitudes corresponded to SpO 2 values of~80 and~90%, respectively, highlighting differences in the severity of hypoxic conditions. Hyperventilation of heavy sustained exercise (>85% VO 2max ) causes substantial increases in respiratory muscle work, leading to diaphragm and expiratory muscle fatigue  in turn reducing blood flow, and thus O 2 delivery, to the working limb (Harms et al. 2000). To which extent this phenomenon would explain the observed lower performance during TT 1_3500 compared to TT 1_2000 could not be ascertained here.
In this study, TT 1 was a performance test whereby the participants were continuously able to adjust their pace as they attempt to sustain the highest average power output for a predetermined reference time. Under more severe hypoxic conditions (>10% lower SpO 2 values during TT 1_3500 versus TT 1_2000 ), reductions in power output occurred together with elevated cardiovascular load despite similar skeletal muscle recruitment (EMG). This probably ensured that the rate of peripheral fatigue development was slowed in the more severe hypoxic condition. To date, experimental data are contradictory, with some studies reporting no effect of severe hypoxia on EMG amplitude during dynamic muscle contractions (Donnelly and Green 2013), and other reporting increased activity with a muscle-specific pattern (Fulco et al. 1996;Torres-Peralta et al. 2014). During 5-km cycle time-trials, increased systemic O 2 transport from hypoxia to hyperoxia (via wide changes in FiO 2 ) resulted in parallel increases in central motor drive (skeletal muscle recruitment) and power output ultimately leading to improved time-trial performance; yet, the magnitude of peripheral muscle fatigue developed at end-exercise was identical  (Amann et al. 2006b). In this study, perceived fatigability was not different between conditions, as evidenced by similar RPE values between TT 1_2000 and TT 1_3500 .

Post-exercise recovery of neuromuscular function
Limited research is available to show how long a diminished neuromuscular response to intense cycling exercise will last, notably after hypoxic tasks. In this study, impairment in muscle function persisted at least 22 min into recovery, with remarkably similar post-TT 1 and pre-TT 2 peak twitch values. Although the extent to which muscle force decreased in response to TT 1 varied according to the stimulation frequency considered (1 and 10 Hz vs. 100 Hz), at all three stimulation frequencies, there was a lack of recovery between post-TT 1 and pre-TT 2 time points. Contrastingly, VA values returned near baseline at pre-TT 2 . Such apparent disconnect in time-course changes in peripheral and central factors contributing to the ensuing (partial) recovery of neuromuscular performance (MVC force) before the completion of TT 2 reinforces the contribution of the brain to performance recovery after strenuous exercises (Minett and Duffield 2014).

Additional considerations/Limitations
One major flaw of almost all whole-body exercise studies, including the present one, is that we did not measure fatigue during exercise per se nor right at exercise cessation since a short delay (~1 min) was needed to install the participants on the ergometer before testing. Although no statistically significant differences were found in MVC and VA values between mild hypoxic (FiO 2 = 0.15) and normoxic (FiO 2 = 0.209) environments, Dahlstrom et al. (2013) made the interesting observation that VA values collected at 1, 2, 3, and 4 min post-20-km cycle time-trial were all significantly reduced compared to baseline. Contrastingly, substantial recovery in skeletal muscle function occurred within the first 1-2 min after an intense bout of self-paced, high-intensity dynamic exercise (Froyd et al. 2013). While significant recovery of markers of neuromuscular function should not be overlooked, the time taken to assess fatigue was consistent between trials and our results further indicate that fatigue resulting from self-paced exercise was not hypoxia severity-dependent. During hypoxic isometric knee extensions, hypocapnia impairs cerebral oxygenation and central drive, but has some protective effects on fatigability in the muscle (Rupp et al. 2015). To which extent CO 2 breathing (i.e., clamping) is effective in preventing the development of hyperventilation-induced hypocapnia during intense hypoxic exercise as performed here and what could be the "carry-over" effects on subsequent performance, physiological responses, and neuromuscular consequences need to be researched. Although standard, our neuromuscular assessment based on MVC of an isolated muscle group (knee extensors) pre-versus postcycling might not truly be representative of effective cerebral functioning (and to a lower extent of muscle mechanics) during an actual whole-body dynamic exercise. Rather, future research would benefit from a combination of methodologies (i.e., perfusion, oxygenation, metabolism, neuronal excitability, and electrical activity) offering complementary insights into the brain in hypoxia (Verg es et al. 2012).
An important consideration when measuring stationary cycling performance using normobaric hypoxic gas mixtures in a laboratory setting is that it does not simulate the terrestrial altitude environment where there is a decrease in air density (i.e., drag forces) due to the decrease in barometric pressure (Peronnet et al. 1991). Interestingly, decreases in time-trial performance (~8%) and SpO 2 (~2%) values from normoxia to hypoxia were greater in hypobaric versus normobaric hypoxia for a terrestrial/simulated altitude of 3450 m (Saugy et al. 2016), corresponding to the present more severe hypoxic condition of TT 1 . Our conclusions must therefore remain specific to the context of our study.

Conclusion
During an initial cycling time-trial (approximately 16 min), power output was lower and cardiorespiratory load higher in more severe hypoxia, whereas quadriceps muscle activation (twitch interpolation) trend differed between moderate and severe hypoxia. This resulted, however, in the attainment of similar neuromuscular fatigue characteristics at exercise cessation, thus independent of the hypoxic severity. After 22 min of rest, muscle activation returned near baseline, while peripheral function remained depressed. There was no influence of the altitude of a prior time-trial on performance and associated cardiometabolic responses, with also no additional muscle fatigue development, during a subsequent 6-min timetrial performed in moderate hypoxia. We conclude that prior time-trial performed at higher altitude did not influence further performance and associated cardiometabolic and neuromuscular responses during completion of a subsequent exercise of similar nature in moderate hypoxia.
Olivier D eriaz and the staff at the Clinique romande de r eadaptation (Sion, Switzerland) for their invaluable support and access to facilities.