Hypoxia gradually augments metabolic and thermoperceptual responsiveness to repeated whole‐body cold stress in humans

New Findings What is the central question of this study? In male lowlanders, does hypoxia modulate thermoregulatory effector responses during repeated whole‐body cold stress encountered in a single day? What is the main finding and its importance? A ∼10 h sustained exposure to hypoxia appears to mediate a gradual upregulation of endogenous heat production, preventing the progressive hypothermic response prompted by serial cold stimuli. Also, hypoxia progressively degrades mood, and compounds the perceived thermal discomfort, and sensations of fatigue and coldness. Abstract We examined whether hypoxia would modulate thermoeffector responses during repeated cold stress encountered in a single day. Eleven men completed two ∼10 h sessions, while breathing, in normobaria, either normoxia or hypoxia (PO2: 12 kPa). During each session, subjects underwent sequentially three 120 min immersions to the chest in 20°C water (CWI), interspersed by 120 min rewarming. In normoxia, the final drop in rectal temperature (T rec) was greater in the third (∼1.2°C) than in the first and second (∼0.9°C) CWIs (P < 0.05). The first hypoxic CWI augmented the T rec fall (∼1.2°C; P = 0.002), but the drop in T rec did not vary between the three hypoxic CWIs (P = 0.99). In normoxia, the metabolic heat production (M˙) was greater during the first half of the third CWI than during the corresponding part of the first CWI (P = 0.02); yet the difference was blunted during the second half of the CWIs (P = 0.89). In hypoxia, by contrast, the increase in M˙ was augmented by ∼25% throughout the third CWI (P < 0.01). Regardless of the breathing condition, the cold‐induced elevation in mean arterial pressure was blunted in the second and third CWI (P < 0.05). Hypoxia aggravated the sensation of coldness (P = 0.05) and thermal discomfort (P = 0.04) during the second half of the third CWI. The present findings therefore demonstrate that prolonged hypoxia mediates, in a gradual manner, metabolic and thermoperceptual sensitization to repeated cold stress.


INTRODUCTION
In a thermally stressful environment, thermal homeostasis is preserved via the seamless recruitment of certain thermoeffectors. Particularly during cold exposure, peripheral vasoconstriction attenuates heat loss, muscular shivering enhances heat production and conscious thermobehavioural actions motivated by perceived thermal discomfort promote heat balance. Inefficiency, or perhaps failure, of these thermoregulatory adjustments to maintain thermal stability results in hypothermia. metabolic heat production, and hence to a predisposition to hypothermia (Castellani, Young, Sawka, & Pandolf, 1998;Tikuisis, Eyolfson, Xu, & Giesbrecht, 2002). Firstly, Castellani et al. (1998) compared the thermoregulatory responses to three 2 h, 20 • C water immersions repeated sequentially within a day, with the responses evoked by single immersions conducted in three separate days. The authors found that the serial immersions gradually blunted the thermogenic response to cold, because of a delay in the shivering onset threshold. Tikuisis et al. (2002) have also determined that the overall shivering intensity may remain constant in the later stages of lengthy cold-water immersion, but the shivering sensitivity to falling deep-body temperature appears to be diminished. Although in the aforementioned studies such metabolic downregulation may reflect a condition of 'shivering fatigue' , it is plausible that it instead alludes to a rapid development of cold habituation (Tipton et al., 2013;Young, 1996), especially considering that, along with the thermogenic alterations, the sensation of coldness was alleviated as well . Of interest in this regard is also the finding that, during protracted mild cold exposure, the shivering intensity appears to be preserved through a modification in fuel selection (i.e. carbohydrate and lipid oxidation decreases and increases, respectively) (Haman et al., 2016).
The purpose of the present study therefore was to examine whether, and to what extent, hypoxia would modulate thermoregulatory effector responses during continual exposures to moderate cold. To address this question, thermal, cardiovascular and perceptual responses were monitored during 2 h cold water immersions repeated three times serially in a 10 h period, while subjects were breathing, in normobaria, either normoxia or hypoxia. Although it might be argued that sustained exposure to cold air would have increased the ecological validity of the study, we utilized 20 • C water immersion (i.e.

New Findings
• What is the central question of this study?
In male lowlanders, does hypoxia modulate thermoregulatory effector responses during repeated whole-body cold stress encountered in a single day?
• What is the main finding and its importance?
A ∼10 h sustained exposure to hypoxia appears to mediate a gradual upregulation of endogenous heat production, preventing the progressive hypothermic response prompted by serial cold stimuli.
Also, hypoxia progressively degrades mood, and compounds the perceived thermal discomfort, and sensations of fatigue and coldness. a moderate-intensity cold stimulus) to elicit, in a safe manner and in a relatively short period of time, sufficient degrees of deep-body cooling to stimulate moderate levels of shivering .
We hypothesized that the ∼10 h hypoxic stressor would attenuate the cold-induced thermogenesis and peripheral vasoconstriction, and thus aggravate the progressive reduction in body core temperature prompted by the repeated cold stimuli.

Ethical approval
The experimental protocol was approved by the Human Ethics Committee of Stockholm (2018Stockholm ( /1433, and conformed to the standards set by the Declaration of Helsinki, except for registration in a database. Subjects were informed in detail about the experimental procedures before giving their written consent to participate, and were aware that they could terminate their participation at any time.

Subjects
Based on the mean effect sizes for changes in body core temperature from a previous study using a similar experimental protocol , a minimum sample size of 10 individuals was required to determine significant differences (α = 0.05, β = 0.8; G*Power 3.1 software, Heinrich-Heine-Universität, Dusseldorf, Germany; Faul, Erdfelder, Lang, & Buchner, 2007 (Eyolfson, Tikuisis, Xu, Weseen, & Giesbrecht, 2001). Subjects were nonsmokers, had no history of any cold injury, and were not taking any medication.
None of them were regularly exposed to cold water, or had sojourned at altitude ≥500 m during the month preceding the experiments.
Subjects were instructed to: (i) abstain from alcohol and strenuous exercise for 24 h prior to each test session, (ii) refrain from caffeine during the testing day, and (iii) maintain their habitual sleep (≥7 h) and eating routines the day before the test sessions.

Study design
The study was performed between November and February in a

Main test sessions
During the test sessions, the environmental conditions in the laboratory were kept constant: the mean (standard deviation; SD) temperature, relative humidity and barometric pressure were 27.1 (0.3) • C, 27 (6)% and 764 (12)

Respiratory measurements
Throughout the test sessions, subjects breathed through a low resistance two-way respiratory valve (model 2, 700 T-Shape, Hans Rudolph, Shawnee, KS, USA). The inspiratory side of the valve was connected via respiratory corrugated tubing to a bag filled with the premixed humidified breathing gas. Inspired and expired gases were sampled continuously, from either side of the respiratory valve.

Arterial pressures, heart rate and capillary oxyhaemoglobin saturation
During the baseline and CWIs, beat-to-beat systolic (SAP), diastolic (DAP) and mean (MAP) arterial pressures were measured continuously using a volume-clamp technique (Finometer, Finapres Medical Systems BV, Amsterdam, the Netherlands), with the pressure cuff placed around the middle phalanx of the left middle finger, and with the reference pressure transducer positioned at the level of the heart.
The Finometer-derived values were verified intermittently by electrosphygmomanometry (Omron, M6, Kyoto, Japan). Heart rate (HR) was derived from the arterial pressure curves as the inverse of the inter-beat interval. Capillary oxyhaemoglobin saturation (S pO 2 ) was monitored at 5 min intervals with an earlobe pulse oximeter (Radical-7, Masimo, Irvine, CA, USA).

Skin blood flow
Local

Arterial blood flow
During each baseline phase, and at minutes 30, 60, 90 and 120 of the CWIs, flow in the right radial, ulnar and brachial arteries was

Capillary glucose
During each baseline phase and at minute 95 of the CWIs, capillary blood was sampled from the right finger to measure glucose concentration. The skin was punctured with a lancet (Accu-Check, Scoftclix, Basel, Switzerland), and the second drop of blood was placed on a strip and immediately analysed with a portable analyser (Accu-Chek, Aviva, Roche, Manheim, Germany).

Perceptual measurements
During each baseline, and at minutes 1, 30, 60, 90 and 120 of the CWIs, subjects were asked to provide ratings of whole-body thermal sensation (from 1, cold, to 7, hot) and thermal comfort (from 1, comfortable, to 4, very uncomfortable). At the same time intervals, the affective valence was also assessed by means of the feeling scale (from −5, very bad, to +5, very good).
Approximately 10 min before each baseline phase, and at minute 100 of the CWIs, subjects were requested to complete the following questionnaires, based on how they felt at that particular moment: (i) the Profile of Mood States-Short Form (Shacham, 1983), which is a 37-item self-evaluation questionnaire of six subscales: tension, depression, anger, vigour, fatigue and confusion. The description of subjects' feelings was provided based on a five-point scale with anchors from 0, not at all, to 4, extremely. (ii) the 2018 Lake Louise Score, which is a self-assessment questionnaire of acute mountain sickness.
The Lake Louise Score evaluates the severity of headache, nausea, dizziness and fatigue; each item is rated with a score of 0, no symptoms, to 3, severe symptoms. The presence of acute mountain sickness was defined by a value ≥3, including headache. Both questionnaires were presented in hardcopy format, and were explained to the subjects by the same investigator prior to each test session. Subjects replied to the questions within ∼3-5 min.

Data and statistical analyses
Only data collected during the baseline and CWI phases were analysed.
Baseline values were calculated as averages of the final 10 min of the 20 min baseline phase. All physiological data obtained during CWI were reduced to 60 s averages. Due to the inter-and intra-individual variability in the duration of the CWIs, data were expressed as a function of the absolute time completed by all subjects, including the corresponding final value obtained in each CWI. For selected variables, the averages of the last 60 min of each CWI (CWI L-60 ) were also evaluated.

Thermal responses
The mean time series for ΔT rec and T sk are depicted in Figure 1.
Baseline T rec did not differ between trials (Normoxia: CWI A , 37.0 In normoxia, the baseline T sk was lower in CWI B and CWI C than in CWI A (P ≤ 0.001), and in CWI C than in CWI B (P = 0.02). In hypoxia, the baseline T sk was lower in CWI B and CWI C than in CWI A (P ≤ 0.001).
No differences were noted between the normoxic and hypoxic base- . Data in all trials were significantly different from baseline (P ≤ 0.001). Significant difference # between CWI A and CWI B , † between CWI A and CWI C , § between the normoxic and hypoxic CWI B , and * between the normoxic and hypoxic CWI C

Cardiorespiratory responseṡ
M did not vary across the baseline phases, and was enhanced (P ≤ 0.001) in all CWIs (Figure 2a). In normoxia,Ṁ was greater during the initial part of CWI C than during the corresponding part of CWI A (P = 0.02; Figure 2a); yet the difference was blunted during CWI L-60 (P = 0.89; Figure 2b). The cold-induced elevation inṀ was unaltered by hypoxia during CWI A and CWI B (∼15% from the hypoxic CWI A ; P = 0.45), but was augmented during CWI C (∼25% from the hypoxic CWI A ; P ≤ 0.01; Figure 2a). The %Shiv peak was similar in the normoxic CWIs, but was increased by hypoxia in CWI B and CWI C (P ≤ 0.01; Table 1). TheṀ sensitivity to ΔT rec was not modified by the repeated normoxic CWIs (P ≥ 0.80; Figure 3a), whereas it was gradually enhanced by the repeated hypoxic CWIs, and particularly prominent in the hypoxic CWI C (P ≤ 0.05; Figure 3a). In eight out of eleven subjects, the shivering threshold was shifted towards a greater ΔT rec in the hypoxic than the normoxic CWI A (P = 0.13, d = 0.62). In comparison with the respective CWI A , the shivering onset (Figure 3b) was unaltered by the normoxic CWI C (P = 0.07, d = 0.61), and by the hypoxic CWI C (P = 0.10, d = 0.50). In the CWI C , however, the ΔT rec threshold for shivering was lower in hypoxia than in normoxia (P = 0.03, . V E and V T were consistently higher in hypoxia than in normoxia (P ≤ 0.01; Table 1). In hypoxia,V E was enhanced gradually over the repeated CWIs; it was higher in CWI B than in CWI A (P < 0.001), and in CWI C than in CWI A and CWI B (P < 0.01). In hypoxia, f R was greater in CWI C than in CWI A (P = 0.001; Table 1). P ETCO 2 was always lower in hypoxia than in normoxia (P < 0.001), and was lower in the hypoxic CWI B and CWI C than the hypoxic CWI A (P = 0.001; Table 1). RER dropped in all CWIs (P < 0.001); regardless of the breathing condition, it was lower in CWI B and CWI C than in CWI A (P < 0.01; Table 1).
In both breathing conditions, the cold-induced elevation in MAP was blunted in CWI B and CWI C (P ≤ 0.05). HR was consistently higher in hypoxia than in normoxia (P < 0.001). In hypoxia, HR was higher in  (3) Significant difference # between CWI A and CWI B , † between CWI A and CWI C , ‡ between CWI B and CWI C , and * between normoxic and hypoxic trial (P ≤ 0.05). CWI L-60 , the last 60 min of each cold-water immersion; DAP, diastolic arterial pressures; EE, energy expenditure; f R , respiratory frequency; P ETCO 2 , partial pressure of end-tidal carbon dioxide; RER, respiratory exchange ratio; SAP, systolic arterial pressures;V CO 2 carbon dioxide production; V T , tidal volume;V E , expired ventilation;V O 2 , oxygen uptake; %Shiv peak , estimated peak shivering metabolic rate.
CWI B than in CWI A , and in CWI C than in CWI A and CWI B (P < 0.01; Figure 4b).
The baseline finger CVC was lower in CWI B and CWI C , regardless of the breathing condition (P = 0.01; Figure 5a). In normoxia, the coldinduced reduction in finger CVC was greater in CWI B and CWI C than in CWI A (P ≤ 0.05). Finger CVC did not differ between the three hypoxic CWIs. At CWI L-60 , finger CVC was greater in the hypoxic CWI B and CWI C than in the normoxic CWI B and CWI C , respectively (P ≤ 0.02). Neither hypoxia nor the repeated CWIs affected forearm CVC ( Figure 5b).
The mean values of radial, ulnar and brachial arterial flow are presented in Table 2. During the normoxic baseline phases, the flow was lower in the radial and ulnar arteries in CWI B and CWI C than in CWI A (P ≤ 0.01); a similar tendency was observed in the brachial artery, but the difference was not statistically significant (P = 0.10). During the hypoxic baseline phases, the flow in the three arteries was lower in CWI B and CWI C than in CWI A (P < 0.001). The ulnar artery flow was higher in the hypoxic than the normoxic CWI A baseline (P = 0.05). In all three arteries, the cold-induced drop in flow was similar across all CWIs.

Perceptual responses
The mean time series for thermal sensation, thermal comfort and affective valence are presented in Figure 6. Regardless of the breathing condition, subjects felt colder and thermally more uncomfortable during the initial part of CWI B and CWI C than the corresponding part of CWI A (P < 0.05). Yet hypoxia aggravated the sensation of coldness (P = 0.05) and thermal discomfort (P = 0.04) at the end of CWI C . Overall, the rates of affective valence were lower in the CWI B and CWI C baseline than in the CWI A baseline (P ≤ 0.04). In normoxia, subjects felt less pleasant during the first 30 min of CWI B and CWI C than during the corresponding part in CWI A , and at the end of CWI C (P ≤ 0.04). In hypoxia, the feeling of displeasure was compounded by the last two CWIs, especially by the CWI C (P ≤ 0.05).
The mean values of the Profile of Mood States-Short Form subscales are summarized in Table 3. In both breathing conditions, vigour was impaired, and fatigue was enhanced during CWI B and CWI C (P ≤ 0.05); hypoxia aggravated the sensation of fatigue (P ≤ 0.05). Moreover, in hypoxia, the self-reported tension and confusion were higher in CWI B and CWI C than in CWI A (P ≤ 0.05), and the perceived depression was enhanced in CWI C compared to in CWI A and CWI B (P = 0.05).

DISCUSSION
The study sought to determine, in male lowlanders, the impact of hypoxia on thermoregulatory functions during prolonged intermittent cold stress. We therefore employed a repeated-measures design, wherein subjects were exposed, for a ∼10 h period, to either normoxia or normobaric hypoxia, while a 2 h fixed moderate-intensity cold stimulus (i.e. passive immersion in 20 • C water) was applied sequentially at 2 h intervals. To eliminate any potential confounding influences on thermoeffector activity, the time of the day during which the serial immersions were performed was the same in the two test sessions, and subjects' pre-immersion T rec , body posture, as well as energy and fluid intakes were similar across all trials.
(a) (b) F I G U R E 3 Mean (95% CI) and individual values of the metabolic heat production (Ṁ) sensitivity (a) and shivering thresholds (b) obtained during the three successive cold-water (20 • C) immersions (CWI A : 1st trial, CWI B : 2nd trial, CWI C : 3rd trial) performed in normoxia and hypoxia. Number of subjects that exhibited a shivering response in normoxia: CWI A , 10; CWI B , 9; CWI C , 9; and in hypoxia: CWI A , 10; CWI B , 11; CWI C , 9.Ṁ sensitivity was calculated for the last 60 min of each immersion (CWI L-60 ). ΔT rec : changes in rectal temperature relative to baseline values. Significant difference † between CWI A and CWI C , and * between the normoxic and hypoxic CWI C The present results support previous evidence ) that, in normoxia, repeated whole-body cold provocations subjected in a single day may increase susceptibility to develop hypothermia. Also, in line with other human studies (Cipriano & Goldman, 1975;Johnston et al., 1996;Keramidas et al., 2019;Robinson & Haymes, 1990), acute systemic hypoxia (i.e. in CWI A ) aggravated the cold-induced drop in body core temperature. However, and contrary to our hypothesis, the ∼10 h sustained exposure to hypoxia mediated a gradual upregulation of endogenous heat production, which appeared to prevent the progressive hypothermic response prompted by the serial cold stimuli. Moreover, the hypoxic stressor progressively degraded mood, and compounded the perceived thermal discomfort, and sensations of fatigue and coldness.

Cold-defence effector responses to repeated cold stress
The normoxic CWI C augmented the cold-induced fall in T rec , a finding that seems to be in accord with a previous work using an experimental design similar to the present . In that study , the hypothermic response induced by the repeated CWIs was attributed to a centrally mediated alteration in the recruitment of shivering-engaged muscles, given the delay in shivering onset. In the present study, however, a modest elevation inṀ was manifest during the initial part (i.e. until the 70th minute) of the normoxic CWI C (Figure 2a). On the basis of a previous work (Castellani, Young, Kain, & Sawka, 1999), it is highly unlikely that such shift in , and heart rate (HR; b) obtained during the baseline, the total (CWI Total ), and the last 60 min (CWI L-60 ) period of the three successive cold-water (20 • C) immersions (CWI A : 1st trial, CWI B : 2nd trial, CWI C : 3rd trial) performed in normoxia and hypoxia. Data in all trials were significantly different from baseline (P ≤ 0.02). Significant difference # between CWI A and CWI B , † between CWI A and CWI C , ‡ between CWI B and CWI C , and * between normoxia and hypoxia cold-inducedṀ was related to its circadian variation. Although the origin of the increasedṀ cannot be identified, we speculate that it might represent the enhanced recruitment of non-shivering thermogenic processes, involving mainly the skeletal muscle (e.g. the mitochondrial uncoupling, the calcium cycling), and perhaps to a minute extent brown adipose tissue thermogenesis (see Haman & Blondin, 2017;van Marken Lichtenbelt & Schrauwen, 2011;Wijers, Schrauwen, Saris, & van Marken Lichtenbelt, 2008). It might also reflect an overall increase in tonic motor unit activity evoking heat-producing isometric contractions (i.e. thermoregulatory muscle tone), which typically precede the onset of the overt tremorlike movements of shivering (Burton & Bronk, 1937;Lømo, Eken, Bekkestad Rein, & Nja, 2019;Meigal et al., 2003). Regardless of its source, this slight elevation in endogenous heat production, was insufficient to defend the body core temperature.
Notably, during the later portion of CWI C , the thermal drive for metabolic heat generation appeared to be blunted: the lower T rec failed to augment thermogenesis. A similar metabolic desensitization has been described during prolonged continuous cold-water immersion (Tikuisis, 2003;Tikuisis et al., 2002).
Judging by the absolute values obtained during the immersions, neither T sk (Figure 1b) nor the blood flow in the non-immersed cutaneous ( Figure 5) and muscle ( In addition, the MAP response to cold stimulus was blunted during the repeated CWIs (Figure 4a). Attenuation of the pressor response has also been induced by sustained periods of exertional fatigue (e.g. F I G U R E 5 Mean (95% CI) and individual values of cutaneous vascular conductance (CVC) of the left index finger (a) and forearm (b) during the baseline, the total (CWI Total ), and the last 60 min (CWI L-60 ) period of the three successive cold-water (20 • C) immersions (CWI A : 1st trial, CWI B : 2nd trial, CWI C : 3rd trial) performed in normoxia and hypoxia. Data in all trials were significantly different from baseline (P ≤ 0.01). Significant difference # between CWI A and CWI B , † between CWI A and CWI C , and * between normoxia and hypoxia. PU, perfusion unit after military sustained operations; Keramidas, Gadefors, Nilsson, & Eiken, 2018a;Young et al., 1998), as well as by short-and long-term regimens of cold adaptation (Keramidas, Kolegard, & Eiken, 2018b;Makinen et al., 2008;O'Brien et al., 2000;Tipton et al., 2013), and has been attributed primarily to a peripheral adrenergic desensitization (Opstad, 1990). Collectively, and regardless of the underlying thermoregulatory mechanisms, current results further indicate that, in normoxia, perturbations of thermal homeostasis imposed by repetitive whole-body exposures to moderate cold may enhance susceptibility to hypothermia.
Acute hypoxia aggravated the reduction in deep body temperature during CWI A , a result that conforms to those from previous investigations (Cipriano & Goldman, 1975;Gautier, Bonora, Schultz, & Remmers, 1987;Johnston et al., 1996;Keramidas et al., 2019;Robinson & Haymes, 1990). The novel finding of this study was that the hypoxic impact on cold thermoregulation appeared to be time dependent. The sustained exposure to hypoxia obviated the progressive hypothermic response prompted by the serial cold stimuli; given that, contrary to in normoxia, T rec did not differ between the hypoxic CWI A and CWI C (Figure 1a). Interestingly, unlike the transient hypometabolism commonly observed in hypoxic animals (Dzal & Milsom, 2019;Tattersall & Milsom, 2009), the ∼10 h hypoxic stressor augmented, in a gradual manner, the cold-induced thermogenesis. This seemingly paradoxical increase in metabolic thermosensitivity to the fixed 20 • C water stimulus was independent of the thermal inputs from the body core (T rec was similar between the hypoxic trials) and the shell (T sk was slightly elevated during CWI C ; see next paragraph). The F I G U R E 6 Mean (95% CI) values of thermal sensation, thermal comfort and affective valence during the three successive cold-water (20 • C) immersions (CWI A : 1st trial, CWI B : 2nd trial, CWI C : 3rd trial) performed in normoxia and hypoxia. Data are expressed as a function of the immersion time completed by all subjects, including the final value obtained in each CWI. B: baseline. Data in all trials were significantly different from baseline (P ≤ 0.001). Significant difference # between CWI A and CWI B , † between CWI A and CWI C , ‡ between CWI B and CWI C , § between the normoxic and hypoxic CWI B , and * between the normoxic and hypoxic CWI C during sustained hypoxic exposure (Kanstrup et al., 1999;Saito et al., 1988), although MAP was not enhanced in this study, might have exerted an influence on the thermoregulatory centres, modifying the integrative control function, and subsequently sensitizing the heat-producing thermoeffector. Moreover, it has been shown that, in anesthetized cats, the hypoxic stressor directly increases the activity of thermosensitive neurons, including the cold-responsive neurons, in the preoptic area of the anterior hypothalamus (Tamaki & Nakayama, 1987). Also, the metabolic cost associated with the hypoxia-evoked hyperpnoea and tachycardia might have contributed, at least to some extent, to the enhancedṀ. That hypoxia stimulated the thermogenic response by exciting other inputs to the hypothalamus appears less plausible. For instance, the hypoxia-evoked stimulation of peripheral chemoreceptors (Mott, 1963), and the development of hypocapnia (Gautier et al., 1987) would have been expected to suppress shivering. Likewise, baroreceptor stimulation may augment shivering in anaesthetized rabbits (Ishii & Ishii, 1960), whereas a reduction in MAP response was noted in the present hypoxic CWI B and CWI C .

Thermoperceptual responses to repeated cold stress
The thermal perception, which in turn initiates conscious thermobehavioural actions, is driven primarily by thermoafferent inputs from the body core and the shell (Schlader & Vargas, 2019). Although their relative contribution to the thermoperceptual output is still unclear, a hierarchical arrangement seems to exist, whereby internal body temperature constitutes the main determinant, especially for thermal (un)pleasantness (Attia, 1984;Mower, 1976). During the first 30 min of the normoxic CWI B and CWI C , subjects' ratings for coldness and thermal discomfort were amplified. The sensory and hedonic evaluation of the 20 • C water stimulus appeared to be dissociated from central and peripheral thermal cues; T rec and T sk were similar in the three normoxic trials. This transient enhancement in thermal discomfort was probably associated with subjects' psychological state prior to and during the initial part of the immersion. In particular, the preceding CWI(s) in the normoxic test session provoked a gradual shift towards a less pleasant generalized affective state, which in turn might have influenced the central integration of temperature information (Arnsten, 2015), enhancing thermoperceptual responsiveness (Auliciems, 1981;Barwood, Corbett, Tipton, Wagstaff, & Massey, 2017). It is noteworthy, however, that after the first 30 min of immersion, any inter-trial difference in thermoperception dissipated.
During the normoxic CWI C , the lower values in ΔT rec , and the somewhat greater generalized unpleasantness, appeared to be inadequate stimuli to invoke distinct temperature-related sensations.
Presumably, the serial cold-water immersions might have elicited a habituation effect (Golden & Tipton, 1988;O'Brien et al., 2000), which did emerge when the impact of psychological strain on thermoperception was attenuated.
The sensation of coldness and thermal displeasure, however, were exacerbated by the hypoxic CWI B and CWI C , a response that prevailed throughout the cold provocations. This thermoperceptual sensitization was independent of the magnitude of cold-induced drop in T rec , which did not vary between the trials, and in T sk , which was either similar, or even slightly diminished in the latter part of CWI C . The aggravated perceptual responses to cold can probably be ascribed to hypoxia per se, and its direct impact on neural circuits mediating thermal sensation and comfort (Craig, 2002); yet our study does not allow us to draw any firm conclusions on the underlying neural mechanisms. It is also plausible that the feelings of coldness and thermal discomfort were dictated by the negative affective state and the enhanced levels of perceived fatigue produced by hypoxia.
Lastly, the conscious perception of the augmented shivering in hypoxia could have provoked, independently, some degree of discomfort (Gagge, Stolwijk, & Hardy, 1967), which, conceivably, might have magnified the thermal discomfort (Schlader & Vargas, 2019). Whether the magnitude of thermoperceptual sensitization induced by hypoxia would eventually facilitate the decision-making to thermoregulate behaviourally remains to be settled.

Acute mountain sickness and thermoregulatory function
Based on the Lake Louise Score values, significant acute mountain sickness occurred in seven subjects. Symptoms were accentuated during the second half of the test session (≥4 h in hypoxia), and mainly included headache and light-headedness; two subjects complained of nausea as well. It is highly unlikely that the latter symptom was ascribed to hypoxia-induced malabsorption of food, given that no evidence exists that this occurs at altitude lower than 5500 m (Kayser, Acheson, Decombaz, Fern, & Cerretelli, 1992;Mekjavic et al., 2016). The blood glucose availability was also maintained, despite the moderate calorie intake during the 11 h session. The impact of acute mountain sickness on human thermoregulation remains perplexing. Field studies (Maggiorini, Bartsch, & Oelz, 1997;Roggla, Moser, Wagner, & Roggla, 2000) have suggested that increases in body core temperature might be associated with the development of acute mountain sickness, whereas Loeppky et al. (2003) have demonstrated that, during a 12 h hypoxic confinement in a hypobaric chamber, a reduction in internal temperature was related to acute mountain sickness. In the present study, no correlation was detected between the Lake Louise Score values and the cold-induced ΔT rec (r s = 0.12; P = 0.59), orṀ (r s = 0.32; P = 0.14). Still, whether, or to what extent, acute mountain sicknessdepended perturbations in autonomic function might have contributed to the augmented metabolic and thermoperceptual responsiveness to cold is unclear.

Methodological limitations and delimitations
Considering that the magnitude of stress reactivity is determined by the intensity, duration and mode of (inter-)action of stressor ( An inter-individual variation in the magnitude and direction of the thermoeffector responses was also observed. Yet, given the small sample size and the relatively homogeneous group of subjects tested (as regards their age, sex, aerobic capacity and body morphology), the source of the detected variation cannot be identified (see Atkinson & Batterham, 2015). Further work is also required to determine whether a sex-specific influence of hypoxia on cold thermoregulation exists, provided that the thermoregulatory function might differ in women, due to morphological differences and/or the hormonal fluctuation during the menstrual cycle (Burse, 1979).
Subjects were sufficiently rewarmed prior to each CWI, as indicated by the baseline T rec values. Yet the rewarming treatments may have eliminated the circadian fluctuations in body internal temperature (Krauchi & Wirz-Justice, 1994), thus marginally attenuating the natural afternoon elevation in T rec . Also, whether the rate of body-core cooling was confounded by the slightly lower baseline T sk noted prior to the afternoon CWIs (CWI A vs. CWI B : ∼0.5 • C; CWI A vs. CWI C : ∼0.8 • C) is unclear. The changes in body core temperature were evaluated based on T rec , which is regarded a slow indicator and might be influenced by variations in rectal (mucosal) and abdominal blood flow (see Taylor, Tipton, & Kenny, 2014). In addition, the study would have benefited from the integration of a multimodal quantification of shivering, in particular by the use of surface electromyography (Arnold et al., 2020;Haman & Blondin, 2017).
Arguably, any changes in body fluid distribution and plasma volume typically evoked by CWI (Stocks et al., 2004) and sustained hypoxia (Sawka, Convertino, Eichner, Schnieder, & Young, 2000) might have influenced, at least partly, our findings; despite that the immersion depths as well as subjects' body postures did not vary between the trials. However, the similar drop in subjects' body weight observed at the end of each test session suggests that the magnitude of body fluid loss during a session was probably similar in hypoxia and normoxia.
Both in normoxia and hypoxia, the steep increase inṀ observed during the initial 10 min of CWIs was attributed to hyperventilation, which, along with tachycardia, describe the main components of the 'cold-shock response' (Tipton, 1989). In line with previous work (Barwood et al., 2017), the magnitude of hyperventilation tended to be blunted by the three repeated normoxic CWIs, especially by CWI C (changes from the baseline in the normoxic CWI A , 126 (46) To evaluate the efficiency of the blinding process, subjects were asked, in the middle of each test session, to report whether they believed they were breathing a normoxic or a hypoxic gas. Subjects replied 22 times, of which they were indecisive nine times, guessed incorrectly four times, and correctly nine times. It cannot be excluded, however, that the thermoperceptual responses might have been biased, at least to some extent, by subjects' awareness of the breathing condition (i.e. a placebo effect). Lastly, the use of a unipolar scale for thermal (dis)comfort might have dampened any variation between the breathing conditions, especially during the pre-immersion phase.

4.5
Practical perspectives -does hypoxia counteract 'shivering fatigue' or impede metabolic habituation to cold? Castellani et al. (1998) have postulated that, in normoxia, protracted intermittent cold stress leads to a forced reduction in body core temperature, possibly reflecting a condition of 'thermoregulatory fatigue' . On the basis of this premise, the present results might suggest that prolonged hypoxia prevents the development of 'shivering fatigue' , hence reducing the risk of accidental hypothermia. Such a notion must, however, be questioned. Although no direct assessment of fatigue was conducted in this study, a large body of evidence has documented that diminished systemic O 2 availability does in fact facilitate both central (i.e. a reduction in central motor drive) and peripheral (i.e. functional changes at or distal to the neuromuscular junction) fatigue (Goodall, Twomey, & Amann, 2014).
Rather, our findings may suggest that the repeated cold stress elicited a regulated hypothermic response (i.e. habituation), which ostensibly was disturbed by the hypoxic stressor. Assuming that the metabolic habituation to cold confers a survival advantage by preserving energy reserves and gross motor function (Carlson, Burns, Holmes, & Webb, 1953;Golden & Tipton, 1988), the excessive metabolic response instigated by hypoxia may provide an adaptive benefit in short-term (as observed herein), but might in fact be disadvantageous for long-term survival in a moderate cold environment. In addition, the concurrent activation of opposing thermoeffectors (i.e. enhancement of both heat-producing and heat-dissipating responses) observed in hypoxia could also be an energy-costly process over time. Altogether, these hypoxia-driven adjustments might describe a thermoregulatory 'allostatic overload' (McEwen, 2007;Ramsay & Woods, 2014); this concept, however, remains, hypothetical, and warrants further investigation. Finally, it is also likely that, in real field conditions, any thermoregulatory benefits obtained from the hypoxia-induced enhancement inṀ would be overridden by a more prominent acute mountain sickness, which is often prevalent during prolonged (i.e. ≥4-6 h) and moderate-to-severe (i.e. ≥2500 m) altitude exposure.
In conclusion, the present findings support previous evidence that, in male lowlanders, the cold-induced drop in deep body temperature may be aggravated by (i) repeated moderate cold stress encountered within a single day , and (ii) acute systemic hypoxia (Cipriano & Goldman, 1975;Johnston et al., 1996;Keramidas et al., 2019;Robinson & Haymes, 1990). A ∼10 h sustained exposure to hypoxia, however, appears to mediate metabolic and thermoperceptual sensitization to whole-body cold stress, and hence to prevent the progressive hypothermic response prompted by the serial cold stimuli. The nature and function of such a hypoxia-dependent adaptive response should be elucidated in future studies.