The heat acclimated phenotype has been extensively described (e.g., Armstrong and Maresh, 1991; Périard et al., 2015) and is characterized by adaptations enabling an individual to better accommodate a given thermal-stressor. Typically, heat acclimation (HA) is acquired by frequently and repeatedly elevating both core (T_C) and skin (T_Sk) temperature (Regan et al., 1996) to a level challenging sudomotor and vasomotor thermoeffector responses for a sufficient duration (Fox et al., 1963). Although passive approaches have sometimes been employed (Beaudin et al., 2009), the increased thermal strain is often achieved through a combination of environmental heat-stress and increased metabolic heat-production through exercise (e.g., Lorenzo et al., 2010; Gibson et al., 2014, 2015; Keiser et al., 2015). More recently, it has been suggested that dehydration, the process of losing fluid and achieving a state of hypohydration (lower-than-normal body water volume), may also represent an important stimulus for facilitating HA (Garrett et al., 2012, 2014; Périard et al., 2015; Akerman et al., 2016), although this may be controversial (Horowitz et al., 1999; Schwimmer et al., 2006) and in contrast to traditional guidelines for maintaining fluid and electrolyte balance (Armstrong and Maresh, 1991; Bergeron et al., 2012).

Dehydration through combined exercise and heat-stress causes hyperosmotic hypovolemia, reducing thermoeffector function [lower sweating and skin blood flow (Sawka, 1992)], and increasing thermal, cardiovascular, and fluid-regulatory strain (Sawka, 1992; Kenefick et al., 2007). Whilst impaired thermoeffector activity might possibly be maladaptive in terms of sudomotor and vasomotor function, the resultant increased tissue-temperature is important; for T_Cs between 37.3°C and 38.5°C the magnitude of HA is proportional to the thermal forcing-function (Fox et al., 1963), although increasing T_C beyond 38.5°C may not confer any additional benefit (Gibson et al., 2014, 2015). Indeed, because dehydration and heat are often inter-linked in their causation and the strain they induce, demarcating their individual effects can be difficult (Akerman et al., 2016). Recent research employing an isothermal strain (target rectal temperature (T_re) = 38.5°C) HA programme suggests that dehydration can provide a thermally-independent adaptation stimulus (Garrett et al., 2014). Restricting fluid ingestion (permissive dehydration) during the five, daily, exercise-heat exposures (90 min·day⁻¹) increased plasma aldosterone concentration ([aldo]_p) over the HA programme, relative to euhydration; this correlated with an increased plasma volume (PV), while increased resting forearm perfusion and reduced exercise heart rate were also observed during a subsequent heat stress test (HST). The adaptations that appear to be most affected by permissive dehydration (e.g., PV expansion and cardiovascular stability) are among the most rapidly acquired during HA (~4–5 days) and also the quickest to decay upon cessation of HA (Williams et al., 1967; Périard et al., 2015). It remains to be established whether permissive dehydration positively influences the adaptive responses to heat over the longer timescales (~10 days) typically necessary to optimize HA (Périard et al., 2015), or whether permissive dehydration affects the retention of the heat acclimated phenotype following HA. Evidence from rodent studies indicates that severe (10% body mass loss) acute hypohydration can adversely affect the longer-term adaptive response to heat (Horowitz et al., 1999; Schwimmer et al., 2006), although the relevance of this work to humans repeatedly dehydrating to a milder hypohydration (<3% body mass loss) over the course of HA is unclear.

The ergogenic potential of HA under more temperate conditions is currently under debate (Minson and Cotter, 2016; Nybo and Lundby, 2016). Lorenzo et al. (2010) demonstrated improved exercise performance in a cool environment (13°C; 30% RH) following a 10-day exercise-heat acclimation programme (40°C; 30% RH) compared to the same training in the cool conditions, possibly related to PV expansion and its influence on VO_2max by a Frank-Starling effect. Studies also provide indication that HA elicits improvements in VO_2max (Sawka et al., 1985; Lorenzo et al., 2010), exercise economy (Sawka et al., 1983) and lactate threshold (Lorenzo et al., 2010) in temperate conditions; together these are key determinants of endurance performance (Joyner and Coyle, 2008). However, many of these studies have been criticized for inadequate control (Corbett et al., 2014) and this ergogenic effect has not been replicated in recent experiments employing more appropriate controls (Karlsen et al., 2015; Keiser et al., 2015). Moreover, the influence of PV expansion on VO_2max depends on the balance between increased cardiac output and the haemodilution effect on O₂-carrying capacity, which may be unfavorable in an already hypervolemic population. Recently, Keiser et al. (2015) showed no effect of PV expansion on VO_2max or exercise performance among a well-trained cohort, whether induced through HA, or by albumin-solution infusion, although there was considerable inter-individual variation. Given that dehydration may augment the hypervolemic aspect of HA (Garrett et al., 2014), understanding the resultant effects on VO_2max and exercise performance is important, particularly as these programmes are often used by athletes and individuals undertaking heavy physical work. Interestingly, there is some evidence of an ergogenic effect of short-term HA programmes with permissive dehydration amongst trained individuals in hot (Garrett et al., 2014) and temperate conditions (Neal et al., 2016), but these studies must be interpreted cautiously due to the lack of an appropriate comparison group.

Accordingly, the primary aim of this study was to investigate the influence of permissive dehydration on the time-course and magnitude of the acquisition and decay of HA over a short- and longer-term, using a matched thermal-strain HA programme. An ancillary aim was to investigate the ergogenic potential of HA and specifically to examine whether permissive dehydration augmented any ergogenic effects of HA, particularly those effects related to PV expansion.

The main findings of the present study were: (i) there was substantial evidence of adaptation to heat over both the short- and longer-term phases of the present study, but when thermal strain is matched, the time course and magnitude of the acquisition and decay of HA are largely unaffected by permissive dehydration, compared to maintaining euhydration; (ii) permissive dehydration did not notably influence the effect of HA on key parameters related to endurance performance (VO_2max, LT, GME) and although there was a small ergogenic effect [4.6(5.8)% increased PPO], this was not affected by the drinking regimen.

Our primary finding does not support the suggestion that dehydration provides an additional stimulus for the induction of HA (Garrett et al., 2012, 2014; Périard et al., 2015; Akerman et al., 2016). The data from the short-term phase are somewhat at odds with recent work indicating that dehydrating during 90 min daily exercise-heat stress within a 5-day isothermal HA programme facilitated some aspects of HA (Garrett et al., 2014), but the reason for these discrepant findings is unclear. Aerobic fitness reduces the strain induced by mild hypohydration (Merry et al., 2010) and aerobically fit individuals require a greater stimulus to challenge the fluid-regulatory processes than less fit individuals (Merry et al., 2008). However, the fitness of our participants [VO_2max 57(7) mL·kg⁻¹·min⁻¹; PPO 338(49) W)] was comparable to Garrett et al. (2014) [VO_2max 60(7) mL·kg⁻¹·min⁻¹; PPO 340(30) W] and greater hypohydration lacks ecological validity, could impair some training adaptations (Judelson et al., 2008) and in rodents at least, might impair aspects of the genomic (Schwimmer et al., 2006) and phenotypic (Horowitz et al., 1999) adaptation to heat. A more sustained stimulus might be required to optimize the rebound hypervolemic response (Akerman et al., 2016), but the drinking regimes were virtually identical and earlier, rather than later, carbohydrate-electrolyte fluid replacement is crucial for recovering PV following ~3% body weight loss (Kovacs et al., 2002). Alternatively, because fluid consumption may need to exceed fluid losses by ~50% to restore euhydration in a hypohydrated individual (Shirreffs and Maughan, 1998), the ad libitum intake of fluid, electrolyte and protein following the permissive dehydration may have been insufficient to enable any additional hypervolemic adaptation (Kay et al., 2005), but this is not supported by the stable daily baseline body mass and (euhydrated) urine osmolality and while there was some evidence for reduced blood volume change in HA_De, this appeared to be during the decay, rather than induction, phase.

A clear separation of hydration state was achieved; in HA_Eu body mass was maintained consistent with euhydration [−0.56(0.71)% body mass change]; in HA_De body mas was reduced [−2.71(0.82)% body mass change] to a degree consistent with hypohydration (Cheuvront et al., 2010; Cheuvront and Kenefick, 2014) and similar to previous studies employing a HA_Deprogramme (−1.8 to −3.1% average body mass change (Garrett et al., 2012, 2014; Neal et al., 2016). Likewise, baseline plasma osmolality was within the normative range (Cheuvront et al., 2010) and was maintained in HA_Eu, but increased in HA_De to a level consistent with mild dehydration (Cheuvront et al., 2010), although this was not measured in all ISO sessions. Nevertheless, assuming a constant sweating rate, hypohydration (body mass change >−2%) will only have been achieved for the final ~23 min of each ISO and maintained for a further 10 min rest period before fluid consumption, which may have been insufficient to influence the fluid-regulatory mechanisms that are hypothesized to be integral to any effects on HA (Garrett et al., 2012, 2014; Périard et al., 2015; Akerman et al., 2016), although once dehydration is achieved [aldo]_p does not appear to further increase with time (Kenefick et al., 2007). Nevertheless, the increased plasma osmolality in HA_De did not surpass the threshold 2% increase in osmolality that may be obligatory for compensatory renal water conservation (Cheuvront and Kenefick, 2014) and although [aldo]_p was increased by the exercise-heat stress, this was not affected by permissive dehydration, at least within ISO1. Overall, given the substantial similarities in study-design, the reasons for differences between Garrett et al. (2014) and the short-term phase of the present study remain largely unclear. The lack of effect of permissive dehydration over a longer-term HA is, perhaps, less surprising given the modest degree of hypohydration induced, the minimal influence that this likely had on fluid-regulatory mechanisms (Cheuvront and Kenefick, 2014), the rapid time-course over which hematological and cardiovascular adaptations to heat manifest (Armstrong and Maresh, 1991; Périard et al., 2015), and the isothermal strain.

Because some aspects of HA develop rapidly (Périard et al., 2015), there has been considerable interest in short-term HA programmes (Garret et al., 2012, 2014; Neal et al., 2016), particularly for trained individuals who are typically partially heat acclimated and may adapt more rapidly (Périard et al., 2015), as well as for logistical reasons. A recent meta-analysis suggests there is little difference in some aspects of HA over the short and longer-time scales that we studied (Tyler et al., 2016), although few of the studies included repeated measures on the same participants and most employed a controlled work-rate regimen (66%), rather than isothermal-exercise approach (11%), meaning that the adaptation stimulus would have reduced over time. In the present study, which employed an isothermal exercise-heat stress approach, significant hypervolemia, increased whole-body sweat rate and reductions in indices of thermal and cardiovascular strain were evident at HST_mid, indicating that notable adaptation was achieved within this brief timescale, as others have also demonstrated (Garrett et al., 2012, 2014; Neal et al., 2016). For some indices, such as plasma volume expansion, exercise heart rate and whole body SR, there was no further significant change beyond HST_mid. In contrast, further reduction in thermal strain, including exercise T_re, T¯Sk, and T¯b, was evident from HST_mid to HST_post, whereas reduced PSI and perceptual benefits (improved thermal comfort and sensation) did not manifest until HST_post. Taken together, this indicates that the heat acclimated phenotype was not fully developed by HST_mid. The temporal pattern of adaptation was broadly consistent with the general consensus regarding the time-course of human HA, particularly with respect to the rapid accrual of plasma volume and associated improvement in cardiovascular function (Armstrong and Maresh, 1991; Périard et al., 2015). In contrast, sudomotor adaptations are typically regarded as being slower to develop (Armstrong and Maresh, 1991; Périard et al., 2015), but in the present study whole body sweat rate was unchanged beyond HST_mid. However, the reducing sweat [Na⁺] will have facilitated sweat evaporation and the progressive reductions in T_re and T¯b observed in the HSTs would reduce the thermoafferent sudomotor drive. Moreover, our participants displayed high initial sweating rates, presumably as a consequence of frequent exposure to high endogenous thermal load through their habitual training; fitter individuals have smaller scope for adaptation, but tend to adapt more rapidly than less fit individuals (Périard et al., 2015) and pronounced sudomotor adaption has previously been documented with short-term HA (Neal et al., 2016). Resting [aldo]_p also increased over the HA regimen, which is in keeping a recent meta-analysis indicating a small effect of HA on resting [aldo]_p, (Tyler et al., 2016) but e[HSP70] was unchanged following HA. The e[HSP70] response was somewhat surprising since we repeatedly exceeded the proposed endogenous temperature threshold for e[HSP70] release (Gibson et al., 2014), although results from meta-analysis suggests that the effect of HA on e[HSP] is trivial, relative to intracellular [HSP] (Tyler et al., 2016) and basal values may be unchanged during HA (Magalhães et al., 2010). Moreover, the responses could have been blunted by the aerobic training habitually undertaken by our participants and the associated frequent elevations in T_C, which would likely render them partially heat acclimated.

The present study also sought to investigate the extent to which any adaptation to heat was maintained over a 7-day decay period, and whether this was affected by the fluid consumption regimen employed during the HA. Relative to the time-course of induction, the decay in adaptation following HA is poorly documented, but it is generally believed that the hematological and cardiovascular adaptions are among the quickest to decay (Williams et al., 1967; Périard et al., 2015); aspects of the adaptive response most likely to be affected by permissive dehydration (Garrett et al., 2014). Nevertheless, the multitude of approaches used for the induction and assessment of HA and use of limited sample sizes of varying fitness means that there is considerable variation within the published literature regarding the time course of decay of HA. For instance, Williams et al. (1967) reported that, among a group of South African miners who had undertaken a 16 day HA regimen in hot-humid conditions, adaptations in heart rate and mean sweat rate declined by ~50% within 1 week, with a 25% loss in the adaptation in T_re. In contrast, Pandolf et al. (1977) showed little decline in heart rate or T_re in fit young men up to 18 days after a 9-day dry-heat acclimation regime and Weller et al. (2007) showed little decay in T_re or heart rate 12 days after completing a 14-day dry-heat acclimation regimen. Indeed, it has been suggested that the retention of HA benefits is superior in aerobically fit individuals and with acclimating to dry heat (Pandolf, 1998). The results of the present study are broadly in keeping with this assertion as there was no significant decay in most of the typical indices of physiological strain HA over the 7-day decay period; although SR and RER were diminished relative to HST_post, they remained above baseline values and no differences were evident between the drinking conditions. However, these assertions should be tempered by reduced metabolic heat production evident at HST_post (discussed subsequently), which occurred despite a fixed external work rate and would have reduced heat-loss requirements during the HST. Moreover, there was a trend for blood volume to decay to a greater extent with HA_De, but this did not notably influence indices of thermophysiological strain and should be interpreted cautiously given that it was under free-living conditions.

An ancillary aim of the present study was to investigate the ergogenic potential of HA and whether permissive dehydration augmented any ergogenic effects of HA. However, irrespective of drinking regimen, there was no effect of HA on VO_2max, LT, or GME, but given the similarity in the adaptive response to heat, the lack of between-groups differences is unsurprising. This finding is in contrast to a number of studies that have shown an effect of HA on these parameters (Sawka et al., 1983, 1985; Lorenzo et al., 2010), although these studies have often lacked adequate control and often a simple training effect cannot be excluded (Corbett et al., 2014). The possibility of a training effect was reduced in the present study by the recruitment of competitive athletes, although this may have diminished the adaptation potential due to a ceiling effect, whilst the perception based prescription of work rate during the ISO session and modest hypohydration resulted in similar cardiovascular strain and training stimulus in each group. Although pronounced PV expansion was evident in both drinking conditions, there was no evidence of any change in VO_2max. This is in contrast to Lorenzo et al. (2010), who demonstrated increased VO_2max concomitant with HA induced PV expansion, but is consistent with recent work showing no effect of HA induced PV expansion on VO_2max (Karlsen et al., 2015; Keiser et al., 2015). The reason for these equivocal findings is not entirely clear, although in Lorenzo et al. (2010) the relative intensity of training sessions in the heat was higher than for a control group undertaking training under cool conditions and the possibility of an additional training stimulus cannot be excluded. Cardiovascular strain was matched between control and experimental groups in Keiser et al. (2015), although it may have been higher in the experimental group of Karlsen et al. (2015). Alternatively, while the effect of PV expansion on VO_2max appears unfavorable at the population level for trained individuals, there appears to be substantial inter-individual variation (Keiser et al., 2015), possibly due to individuality in the balance between increased cardiac output and the haemodilution effect on O₂-carrying capacity. When pronounced inter-individual variation is combined with relatively small sample sizes, the data may not reflect population characteristics, although at the elite performance level these individual differences may be important.

Although our data from the HST indicate that the O₂ cost of exercise was diminished 1 week post exercise, this was not evident in the GME data obtained during the GXT. Because the improved economy was specific to performance in a hot environment it could simply represent the effect of reduced thermal strain. Alternatively, a move to a more efficient phenotype has been demonstrated in rodents undergoing prolonged HA (Kodesh et al., 2011); this could explain why this effect had not developed at HST_mid or HST_post. Results from a recent meta-analysis have also concluded that there may be a small effect of HA on GME during exercise in the heat (Tyler et al., 2016), but with the exception of studies lacking appropriate control (Sawka et al., 1983), there appears to be little evidence for an effect of HA on GME in humans under temperate conditions (Karlsen et al., 2015). Nevertheless, a small ergogenic effect was apparent as indicated by a 4.6% increase in PPO achieved at the end of the GXT, irrespective of drinking condition, but the mechanisms underpinning this ergogenic effect are unclear given the lack of change in VO_2max, LT and GME. The effect of ambient temperature on aerobic exercise is a continuum, with an exponential performance decline at temperatures above ~10°C (Galloway and Maughan, 1997). Although it is clear that HA attenuates the performance decrement in hot environments, it has been hypothesized that the improved thermoregulatory capability with HA should also attenuate the heat-related performance decrement evident under more temperate conditions (Corbett et al., 2014). Indeed significant reductions in thermal-strain were evident in the sub-maximal exercise preceding the GXT, but none of these changes were correlated with the performance improvement, and the T_re at exercise termination was similar pre vs. post HA, and below the levels associated with impaired performance. Alternatively, we cannot exclude a simple placebo or learning effect on PPO, as we did not include a sham treatment or temperate training group; the primary purpose of the present study was to examine the influence of hydration on HA and performance, rather than the effect of HA per se. This assertion is strengthened by our (unpublished) observation of a similar magnitude of improvement in PPO (6.0%) for 8 trained individuals following an identical protocol to the present study, but with all ISO session undertaken with exercise at a matched RPE, under cool conditions (13°C; 60% RH).

In summary, the present study is the first to examine the influence of dehydration on short- and longer-term HA and its subsequent decay, as well as the effect of a longer-term HA regimen with permissive dehydration on key endurance performance parameters. Our data demonstrate that, when thermal strain is matched, the time course and magnitude of the acquisition and decay of HA are largely unaffected by permissive dehydration, compared to maintaining euhydration. Furthermore, neither HA regimen affected VO_2max, LT, or GME. PPO was increased consistent with a small ergogenic effect of HA, but this was not affected by the drinking regimen and should be interpreted cautiously in the absence of a plausible mechanism. However, it is important to note that no notable negative effects of permissive dehydration were evident either, and traditional guidance to maintain hydration during HA (Armstrong and Maresh, 1991; Bergeron et al., 2012) may be unnecessary when trained individuals commence exercise in a euhydrated state, when thermal strain is matched, and where a transient mild hypohydration is induced.

RN, HM, MT, JY, and JC were involved in conceptual design, data collection, interpretation, and manuscript preparation. All authors approve the submission of this work and agree to be accountable for all aspects of the work.

RN was funded by a joint English Institute of Sport and University of Portsmouth research bursary. Additional project costs were supported by the English Institute of Sport.