The activity levels of participants were maintained throughout the duration of this study. Participants were asked to avoid caffeine for 6 h before each test. Five tests were performed on an electronically braked cycle ergometer, each on a separate day with a minimum interval of 48 h between the tests. Participants wore a nose clip to prevent nose breathing and a mouthpiece to facilitate gas exchange analysis and ventilatory measurements.

Breath-by-breath gas exchanges and ventilatory rates at the mouth were assessed using a mass spectrometer (Innovision, AMIS 2000, Lindvedvej, Denmark) and are described in detail elsewhere (Babcock et al., 1994). Briefly, flow rates during inspiration and expiration were determined with a low dead space bidirectional turbine (Alpha Technologies, Laguna Beach, CA, United States, VMM 110) and pneumotach (Hans Rudolph, Shawnee, Kansas, United States, Model 4813) calibrated with a 3 L syringe. Gas samples at the mouth were analyzed for O₂, CO₂, and N₂ concentrations. Changes in concentrations of gases were matched to the corresponding increase or decrease in volumes of the gases. There was a 20 ms interval between collection samples that were sent electronically to a computer to analyze individual breaths. Each breath started with inspiration and ended with expiration. Therefore, each 5 s BH was recorded as a single breath.

The procedures for near-infrared spectrometry (NIRS) data collection were similar to those previously described (Inglis et al., 2017). Continuous measurement of the quadriceps was performed with a NIRS device (Oxiplex TS, model 95,205, ISS, Champaign, IL, United States). Laser diodes pulsed quickly (110 MHz) at two different wavelengths near the infrared region (690 and 828 nm). These were connected to a plastic probe that was placed midway between the lateral epicondyle and the greater trochanter of the femur on the belly of the vastus lateralis muscle head. An elastic strap secured the device in place. An optically dense, black vinyl sheet was placed over the device to prevent the exposure to extraneous light. A tensor bandage was wrapped gently around the leg of the participant to secure the NIRS device to the site of interest and to further prevent the intrusion of external light into the site of NIRS measurement. Care was taken to ensure that no blood flow occlusion occurred to the leg. Deoxygenated hemoglobin concentration ([HHb]) and oxygenated hemoglobin ([HbO]) were measured, whereas [Hb_tot] and tissue oxygen saturation (S_atO₂) were derived with this apparatus. [Hb_tot] was calculated as the sum of [HHb] and [HbO], and S_atO₂ was estimated as the percentage of [HHb] to [Hb_tot]. To account for individual differences in absolute tissue absorption, [HHb] and [Hb_tot] were adjusted to baseline values (Δ[HHb] and Δ[Hb_tot], respectively).

Rubbing alcohol was applied to the left index finger of each participant before each test. Blood lactate concentrations ([La^–]) were taken 3 min pre- and post-exercise for each test. A lancet (ACCU-CHEK Safe-T-Pro Plus) exposed blood on the finger, which was examined by the SensLab GmbH Lactate SCOUT arterialized-capillary lactate analyzer.

Gas exchange and NIRS data were cleaned by the removal of aberrant data points that were at least 3 SDs from the local mean. The data were interpolated linearly to convert from breath-by-breath to 1 s intervals. Then, data points were averaged into 5 s bins. This analysis technique has been described previously (Keir et al., 2014). [HHb] and [Hb_tot] were adjusted to zero with baseline as described earlier. These baselines represented the average of 60 s before the square-wave change in PO. Δ[HHb]/VO₂ was determined as the ratio of the normalized [HHb] to VO₂. Between-group comparisons on gas exchange and NIRS variables were performed from the start of the square-wave change in PO (the exercise on-transient) to the end of the bout (0–360 s). Within-group comparisons were performed between the first 25 s and the last 5 s of each 30 s cycle for VO₂ and VCO₂. Lamarra et al. (1987) described this mono-exponential function that models the on-transient VO₂ curve during a step transition:

where y(t) represents VO₂ as a function of time during the transition to the steady state of new PO. y_BSL is VO₂ before the transition, A_p is the amplitude (increase above y_BSL), t is the dependent time variable, TD is the time delay, and τ is the time constant (time that elapses for 63% of the response to occur). The curve was fit to the data by applying the Levenberg–Marquardt algorithm for non-linear least squares analysis (Origin 9.7; OriginLab, Northampton, MA, United States).

Ten participants were recruited based on a sample size power calculation of the measured VO_2max. It was found that 10 participants were sufficient (80% power) to calculate pre vs. post VO_2max (α = 0.05) to within an SD of ± 30 mL with the consideration of a 20% dropout rate. Mean analyses from 0 s (start of square-wave change in PO) to 360 s (end of exercise) (n = 10) of VO₂, VCO₂, P_ETO₂, P_ETCO₂, V_E, f_B, VTI, Δ[HHb], Δ[Hb_tot], S_atO₂, and Δ[HHb]/VO₂ between conditions (CONLD, CONLD-BH, FLK, and FLK-BH) were performed by a one-way repeated measures (1-RM) analysis of variance (ANOVA). [La^–] was compared between conditions and time (PRE vs. POST) by a two-way RM (2-RM) ANOVA. The first 25 s and last 5 s of each apneic cycle were analyzed for VO₂ and VCO₂ via a 2-RM ANOVA. The Shapiro–Wilk and Brown–Forsythe tests were performed to assess the normality and heteroscedasticity of the data, respectively. The 1-RM ANOVA comparisons for VCO₂, P_ETO₂, P_ETCO₂, V_E, f_B, VTI, Δ[HHb], Δ[Hb_tot], S_atO₂, and Δ[HHb]/VO₂ were performed by a Friedman RM ANOVA on ranks due to a rejection of either or both tests, whereas mean VO₂ was evaluated with a parametric 1-RM ANOVA. Data are reported as mean ± SD. The statistical significance threshold was p < 0.05.

Mean VO₂ from time 0 to 360 s was similar between CONLD and CONLD-BH (p = 0.406) and between FLK and FLK-BH (p = 0.165) (Table 3 and Figure 2A). VO₂ was greater in FLK and FLK-BH than in CONLD (p < 0.001 for both comparisons) and greater in FLK than CONLD-BH (p < 0.001) over the same time period. Mean VO₂ for the last 5 s of each 30 s cycle was greater in CONLD than CONLD-BH (p = 0.001) and greater in FLK than in FLK-BH (p < 0.001) and CONLD-BH (p < 0.001) (Table 4). Mean P_ETO₂ was greater in CONLD than CONLD-BH (p < 0.001) but was unchanged in FLK compared to FLK-BH (p = 0.682) (Table 3 and Figure 2C). P_ETO₂ was lower in CONLD and CONLD-BH than in FLK and FLK-BH (p < 0.001 for all comparisons) (Table 3 and Figure 2C). P_ETCO₂ was lower in CONLD than CONLD-BH (p < 0.001) but was greater in FLK than FLK-BH (p < 0.001) (Table 3 and Figure 2D). VCO₂ from 0 s to the end of exercise was similar between CONLD and CONLD-BH (p = 0.914) and between FLK and FLK-BH (p = 0.641), but was greater in FLK and FLK-BH than in CONLD and CONLD-BH (p < 0.001 for all comparisons) (Table 3 and Figure 2B).

NIRS-derived normalized total hemoglobin content (Δ[Hb_tot]) was greater in CONLD compared to CONLD-BH (p < 0.001), FLK (p < 0.001), and FLK-BH (p < 0.001), and also in FLK compared to FLK-BH (p < 0.001) (Table 3 and Figure 3C). Deoxygenated hemoglobin normalized to baseline (Δ[HHb]) was greater in CONLD than in CONLD-BH (p = 0.011) and FLK (p < 0.001), and also in FLK-BH than in other three conditions (p < 0.001 for all comparisons) from the onset to the cessation of exercise (Table 3 and Figure 3A). S_atO₂ was greater in CONLD compared to CONLD-BH (p < 0.001), FLK (p < 0.001), and FLK-BH (p < 0.001), and it was not significantly different in FLK compared to FLK-BH (p = 0.434) (Table 3 and Figure 3D). The ratio of Δ[HHb]/VO₂ was similar between CONLD and CONLD-BH (p = 0.086), but lower in FLK than FLK-BH (p < 0.001) (Table 3 and Figure 3B).

Post-exercise arterialized capillary lactate concentration ([La^–]) was unchanged in CONLD compared to CONLD-BH (p = 1.000), and both were lower compared to FLK and FLK-BH (p < 0.001 and p = 0.007, respectively) (Table 3 and Figure 4). Post-exercise [La^–] was similar between FLK and FLK-BH (p = 1.000) (Table 3 and Figure 4).

The changes in PO in CONLD and CONLD-BH, relative to the previous related research (Δ50%-Δ20%) (Lim et al., 2018), reflect the additions of the regulated breathing to this 25 s breathing/5 s apnea protocol. The purpose of this modification was to simulate the difference in breathing patterns between swimming backstroke (25 s free-breathing) and front crawl (25 s regulated breathing). By introducing these restrictions, participants were forced to perform a much lower intensity (189 W, Δ20%) compared to that suggested by Lim et al. (2018) (218 W, Δ50%, p = 0.027), despite having similar aerobic fitness (VO_2max: 3.23 and 3.17 L/min, p = 0.82, respectively), LTs (VO₂ at LT: 1.72 and 1.77 L/min, p = 0.57, respectively; PO at LT: 155 and 156 W, respectively), and PPO (303 and 314 W, p = 0.58, respectively) to perform the 6 min trials. Moreover, the reduced PO of this study elicited lower mean V_E (68 L/min) compared to that suggested by Lim et al. (99 L/min) (Lim et al., 2018). Within the context of this PO adjustment of this study, VO₂ was unchanged with the addition of the 5 s apneic periods and matching of f_B and VTI during the 25 s breathing periods (Table 2) of CONLD-BH, demonstrating that any potential increase in O₂ cost derived from the periodic apneas had been met. The data in this study show that the aerobic metabolic demand (i.e., VO₂) was supported by the decreased S_atO₂ (Table 3 and Figure 3C) and increased muscle deoxygenation under the reduced perfusion conditions, as reflected by the much greater decrease in mean Δ[Hb_tot] (Table 3 and Figure 3D) and modest decrease in Δ[HHb] (Table 3 and Figure 3A). A similar reduction in S_atO₂ and an increase in muscle deoxygenation were observed by Kume et al. (2016) at a PO corresponding to 65% of VO_2max interspersed with 4 s exhalations followed by a maximal inhalation, simulating hypoventilation. Furthermore, previous work by Hoffmann et al. (2005) under apneic and rebreathing exercise conditions during higher intensity exercise, which generated similar hypoxia (decreased P_ETO₂ in rebreathing), demonstrated greater mean arterial pressure and lower heart rate in the apneic compared to the rebreathing condition, demonstrating that only the apneic state elicited the observed O₂ conservation or diving response that is associated with the decreased intramuscular blood flow. They concluded that the mechanical cessation of breathing was the key stimulus to this drop in perfusion. This has been corroborated by similar protocols comparing rebreathing with apnea (Lindholm et al., 1999). The data of this study confirm a similar response under the combination of the regulated breathing and BH paradigm. Moreover, the observed decrease in Δ[Hb_tot] shown in this study was not observed during hypoxic exercise (∼12% fraction of oxygen in inspired air [F_IO₂]) performed at low-moderate (single-leg knee extensions) (DeLorey et al., 2004) and supra-LT intensities (Ainslie et al., 2007), suggesting again that the reduced blood flow in this study was apnea-related. Furthermore, earlier research by Lim et al. (2018), within a comparable protocol as this study, and Kume et al. (2013), performing intermittent apneas during continuous exercise at a comparable intensity, identified similar temporal reductions in Δ[Hb_tot] resolved by similar increases in muscle deoxygenation. This decrease was observed despite previous suggestions that Δ[Hb_tot] is increased during deep diving via splenic contractions (Hurford et al., 1990).

The drop in P_ETO₂ (Table 4 and Figure 2C) and increased VO₂ (Table 4 and Figure 2A) observed immediately post-apnea in this study reflect continued alveolar to capillary O₂ diffusion facilitating the unchanged VO₂. This response has also been observed under 20 s apneas (Lindholm and Linnarsson, 2002) and also reflected the actual arterial O₂ (P_aO₂) under these transient apneic conditions (Suskind et al., 1950).

The maintenance of mean VO₂ between CONLD and CONLD-BH after the exercise on-transient response was supported not only by the microvasculature hemodynamic changes outlined earlier, but also by the increased VO₂ during the 25 s of regulated breathing (Table 4 and Figure 2A) and the reduced P_ETO₂ immediately post-BH (Table 4 and Figure 2C) suggested earlier (Yamamoto et al., 1987). Specifically, Woorons et al. (2007) found that intermittent apneas performed at a high pulmonary volume (i.e., near total lung capacity) during higher intensity exercise (∼70% VO_2max), similar to this study, maintained pulmonary arterial S_atO₂.

The unchanged mean VCO₂, between CONLD and CONLD-BH, suggests that the H⁺ buffering, associated with the lactate production of anaerobic glycolysis, has been reflected in the observed increase in P_ETCO₂ and that the BH did not impose a greater anaerobic glycolytic contribution. The unchanged post-exercise [La^–] between CONLD and CONLD-BH trials is in contrast to other studies that have reported increased [La^–] accumulation during incremental (Seo et al., 2020) (50 W + 12.5 W/min to exhaustion) and intermittent (Lim et al., 2018) exercise under hypoxic or apneic condition, respectively. However, these studies were performed at much higher supra-LT PO, which would elicit much greater blood [La^–].

From a swimming front crawl perspective, at this Δ20% PO intensity, the addition of the regulated breathing paradigm and the 5 s BH, while maintaining a similar intensity of kicking, may be performed without negative physiological consequences. Notably, this is within the context of the legs only paradigm of this experiment as opposed to the simultaneous arm and leg action that is performed after the underwater portion of the swim.

The FLK condition mimics the strategies of those swimmers who believe that sprint kicking during the underwater portion after the turn, along with their own superior hydrodynamics compared to their opponents, will give them a competitive advantage. In contrast to the study by Lim et al. (2018), utilizing a similar protocol, albeit at much higher PO, our results showed that implementing the intermittent 5 s periods of apnea to FLK (FLK-BH) did not affect mean VO₂. Similar to CONLD-BH and CONLD, f_B and VTI were also regulated in FLK-BH to FLK; however, although VTI was statistically greater in FLK-BH (Table 2), the unchanged V_E suggested that this ∼80 mL/min change in VTI (Δ1.6% over the 6 min) had no physiological effects. It was notable that VTI was lower in FLK compared to CONLD (2.86 ± 0.30 L and 2.93 ± 0.21 L, respectively). However, coupled with the increased f_B (25.2 ± 5.2 breaths/min and 22.0 ± 4.7 breaths/min), a higher V_E, in conjunction with the higher VCO₂ in FLK vs. CONLD, was observed (Table 2). The lowered P_ETO₂ immediately post-apnea (Table 4 and Figure 2A) suggests that the observed pulmonary alveoli to capillary diffusion continued during the 5 s facilitating O₂ transport. Moreover, the increased Δ[HHb] and Δ[HHb]/VO₂ in FLK-BH compared to FLK, which reflected a greater reliance on muscle deoxygenation, at the active muscle was responsible for this result (duManoir et al., 2010; Murias et al., 2011).

In elite swimmers, similar maintenance of VO₂ has been demonstrated during 4 min of submaximal intensity swimming with apnea induced by regulated breathing conditions of every two arm strokes up to a maximum of five-arm strokes, notwithstanding the reduced V_E in the latter condition (Dicker et al., 1980). The consequences of the imposed regulated breathing and comparable intensity in the current study were apparently resolved by the greater alveolar to pulmonary capillary O₂ diffusion. The muscle deoxygenation response was not recorded in the earlier work by Dicker et al. (1980). However, a similar increase in Δ[HHb] was observed by Billaut and Buchheit (2013) during 10 repeated 10 s maximal sprints followed by 30 s rest, under hypoxic conditions (13% F_IO₂) compared to normoxia. Conversely, others have studied the effects of repeated 3 s loadless cycling recovery periods, as opposed to the 5 s of high PO as in this study, and observed an improved microvascular O₂ delivery reflected by decreased Δ[HHb]/VO₂ (Belfry et al., 2012). It is suggested that the contrasting results of this study were attributed to the apneic O₂ conservation effect (Kume et al., 2013), as reflected by the decreased Δ[Hb_tot] in FLK-BH compared to FLK. This result is similar to the CONLD-BH compared to CONLD, of this study and others, reflecting a similar redistribution of blood flow away from working muscles during apnea (Kume et al., 2013; Lim et al., 2018). Furthermore, it is suggested that the unchanged mean VCO₂ and [La^–] during FLK-BH compared to FLK despite the apneic periods was a consequence of the continued buffering and pulmonary capillary to alveolar diffusion and buffering during this 5 s BH (Table 4). Conversely, Lim et al. (2018), under a similar apnea protocol (5 s), but with free-breathing (25 s), as opposed to regulated breathing of this study, showed lower mean VCO₂ in FLK-BH compared to FLK and higher lactates associated with the higher mean PO (Δ50% vs. Δ20%), suggesting a relative decrease in ventilatory buffering at this much higher relative PO.

To our knowledge, this was the first study that regulated breathing during supra-threshold PO exercise on a cycle ergometer following periods of apnea, to a free-breathing protocol at the same PO similar to the lower extremity work associated with swimming in the prone position. Consequently, mean VO₂ was maintained in these apneic conditions (CONLD-BH and FLK-BH) compared to their free-breathing counterparts (CONLD and FLK, respectively) by greater muscle deoxygenation, despite decreased intramuscular blood perfusion. The regulated f_B and V_E during the 25 s breathing periods of this study eliminated the transient increases in V_E that previously corresponded with greater post-apneic VO₂ and VCO₂ (Lim et al., 2018). VO₂ was sustained in CONLD-BH compared to CONLD, and in FLK-BH compared to FLK, as a function of the continued pulmonary diffusion during apnea, as is reflected by the immediate increases in P_ETO₂ post-apnea (Table 4). At the level of the muscle, the preserved arterial O₂ content along with the increased muscle deoxygenation facilitated the unchanged VO₂. Although these protocols were designed to replicate the breathing opportunities afforded during regulated compared to free-breathing conditions, the cardiorespiratory responses to facial immersion supine exercise, and the combined arm and leg muscle action specific to swimming (Guyatt et al., 1965; Christie et al., 1990; Leahy et al., 2019) were not studied, due to the inherent difficulties of calculating gas exchange while submerged, and the unavailability of a recumbent ergometer that would interface successfully with our data collection equipment. Additional research is needed to clarify the role of these specific swimming characteristics in these physiological outcomes. Furthermore, our results may not reflect the responses of well-trained competitive swimmer, and as such, future studies should compare the physiological resolution of expert swimmers, compared to non-swimmers, to these breathing restrictions. Finally, only healthy participants were tested; therefore, no comparisons to diseased populations should be made.

Moreover, only male participants were included in this study to best match participant characteristics to the previous study (Lim et al., 2018). Therefore, these observations may not apply to females because of sex differences in body composition, fluctuations in reproductive hormones (estrogen and progesterone) (Arora et al., 1998), and lactate production during exercise (Jurkowski et al., 1981). Also, the sample size (n = 10) of this study adequately powered the statistical analyses used; a larger sample size may be required to generalize these results.