Acute hypoxia impairs posterior cerebral bioenergetics and memory in man

Abstract Hypoxia has the potential to impair cognitive function; however, it is still uncertain which cognitive domains are adversely affected. We examined the effects of acute hypoxia (∼7 h) on central executive (Go/No‐Go) and non‐executive (memory) tasks and the extent to which impairment was potentially related to regional cerebral blood flow and oxygen delivery (CDO2). Twelve male participants performed cognitive tasks following 0, 2, 4 and 6 h of passive exposure to both normoxia and hypoxia (12% O2), in a randomized block cross‐over single‐blinded design. Middle cerebral artery (MCA) and posterior cerebral artery (PCA) blood velocities and corresponding CDO2 were determined using bilateral transcranial Doppler ultrasound. In hypoxia, MCA DO2 was reduced during the Go/No‐Go task (P = 0.010 vs. normoxia, main effect), and PCA DO2 was attenuated during memorization (P = 0.005 vs. normoxia) and recall components (P = 0.002 vs. normoxia) in the memory task. The accuracy of the memory task was also impaired in hypoxia (P = 0.049 vs. normoxia). In contrast, hypoxia failed to alter reaction time (P = 0.19 vs. normoxia) or accuracy (P = 0.20 vs. normoxia) during the Go/No‐Go task, indicating that selective attention and response inhibition were preserved. Hypoxia did not affect cerebral blood flow or corresponding CDO2 responses to cognitive activity (P > 0.05 vs. normoxia). Collectively, these findings highlight the differential sensitivity of cognitive domains, with memory being selectively vulnerable in hypoxia.


INTRODUCTION
The human brain has evolved to be entirely dependent on O 2 to support the high rate of ATP formation to fuel the maintenance of ionic equilibria and uptake of neurotransmitters for synaptic transmission (Bailey, 2019a, b).In acute hypoxia, global cerebral blood flow (CBF) increases to maintain cerebral O 2 delivery (CDO 2 ) (Hoiland et al., 2016) and prevent metabolic compromise and/or structural damage to the neurovascular unit (Stacey et al., 2023).Nevertheless, extramitochondrial cellular processes are sensitive to small hypoxic insults, even when CDO 2 is sufficient to maintain bioenergetic function (Ainslie et al., 2016).Specifically, the synthesis of enzymes and related neurotransmitters is sensitive to O 2 (Kumar, 2011;Raichle & Hornbein, 2001), and neurotransmitter dysfunction might occur during even subtle reductions in the arterial partial pressure of O 2 (Ainslie et al., 2016).
Cognitive function is classified into central executive and nonexecutive domains (Cantelon & Giles, 2021).Central executive function refers to 'top-down' higher cognitive processes (e.g., inhibition, working memory, cognitive flexibility) (Diamond, 2013), whereas non-executive function includes attention, motor speed, information processing and memory (Cantelon & Giles, 2021).Currently, it is controversial which cognitive domains are most vulnerable to hypoxia (Ando et al., 2020;Petrassi et al., 2012), complicated, in part, by the intrinsic cognitive demands (Ando et al., 2020), considerable participant heterogeneity and experimental failure to account for basal habituation/learning effects (Marley et al., 2017).Conversely, middle temporal lobe-dependent cognitive function (e.g., memory) is known to be exquisitely vulnerable in hypoxia (Perosa et al., 2020), where structural damage to the hippocampus precedes memory impairment (Zola-Morgan et al., 1986), which is particularly evident in patients suffering from anoxic brain injury (Garcia-Molina et al., 2006).Furthermore, breath-hold diving training over several years can cause mild short-term memory impairments (Billaut et al., 2018) that might be related to intermittent repetitive structural damage to the neurovascular unit (Bailey et al., 2022).These findings suggest that memory is impaired in hypoxia.However, the underlying mechanisms and associated temporal kinetics remain to be established.
Despite an increase in global CBF in hypoxia, the response is heterogeneous and site specific (Binks et al., 2008;Lawley et al., 2017;Rossetti et al., 2021).Binks et al. (2008) indicated that greater blood flow is directed to phylogenetically older parts of the brain, which suggests that older brain regions might be more bioenergetically demanding to maintain functions in hypoxia compared with newer brain regions (e.g., frontal cortex and central executive function).
The hippocampus is phylogenetically old (Murray et al., 2018), and exposure to hypoxia (∼10 h) has been shown to reduce perfusion in the posterior cingulate and cuneal cortex, which are assumed to play a role in declarative and procedural memory (Lawley et al., 2017).

Highlights
• What is the central question of this study?
We sought to examine the effects of acute hypoxia on central executive (selective attention and response inhibition) and non-executive (memory) performance and the extent to which impairments are potentially related to reductions in regional cerebral blood flow and oxygen delivery.
• What is the main finding and its importance?
Memory was impaired in acute hypoxia, and this was accompanied by a selective reduction in posterior cerebral artery oxygen delivery.
In contrast, selective attention and response inhibition remained well preserved.These findings suggest that memory is selectively vulnerable to hypoxia.
Furthermore, maintenance of appropriate hippocampal perfusion is crucial to preserving healthy memory (Johnson, 2023).
Given these knowledge gaps, the purpose of the present study was to examine how acute hypoxia affects central executive and memory functions in a single-blinded cross-over design.We hypothesized that (non-executive) memory would become progressively more impaired in hypoxia, specifically owing to a reduction in CDO 2 to the posterior circulation, in light of its aforementioned (enhanced) phylogenetic sensitivity.

Ethics
Ethical approval for this study was obtained from the Research Ethics Committee at the University of South Wales, UK (#201712BS01).This study conformed to the standards set by the latest revision of the Declaration of Helsinki, except for registration in a database, with verbal and written informed consent obtained from all participants.

Participants
Twelve recreationally active males (age, 23 ± 2 years; stature, 1.77 ± 0.07 m; mass, 79 ± 12 kg) were recruited from the University of South Wales via flyers and by word of mouth.All participants lived close to sea level (∼90 m) and had not been exposed to simulated or terrestrial high-altitude (>2500 m) in the previous 12 months.After a medical examination, they were confirmed to be healthy and free from any known cardiovascular, cerebrovascular or respiratory disease.
Furthermore, the participants were not taking any prescribed or overthe-counter medications or supplements.They were instructed to refrain from physical activity, caffeine and alcohol and to follow a lownitrate/nitrite diet for ≥24 h before the experiment (Bailey et al., 2017).
We confirmed that the participants followed these instructions before experimentation via interview.

Design
During the first visit, participants were familiarized with the Go/No-Go task until their reaction time (RT) was within 2SD from the mean.They were also familiarized with the memory task until they understood the task procedure adequately (see Cognitive function).
All participants completed two different experimental conditions in a normobaric environmental chamber (∼120 m 3 ) with the ambient temperature maintained at 21  et al., 2017).We adopted a block randomization approach using a computer-based random number generator.Subjects arrived at the laboratory (between 08.00 and 09.00 h) and were fitted with an indwelling cephalic venous cannula.They consumed a standardized meal (30 g of oats with 180 mL water) 30 min before the experimental trials and again at 2, 4 and 6 h to maximize compliance and avoid hunger/dehydration (Ogoh et al., 2021).
Cognitive function was assessed at baseline (0 h), 2, 4 and 6 h during each trial (Figure 1).Initially, participants completed baseline measurements of physiological and ventilatory variables.
Whole blood was obtained without stasis from an indwelling cannula located in a forearm antecubital vein.The first baseline measurement was performed after passive exposure to hypoxia for 10 min.The other baseline measurements (at 2, 4 and 6 h) were performed after consuming the standardized meal (Figure 1).After baseline measurements, participants performed the cognitive tasks.Before each cognitive task, cerebrovascular variables (baseline CBF velocity and cerebral oxygenation) were measured.This allowed us to assess haemodynamic and O 2 -delivery responses to cognitive activity.The cognitive tasks started with the memorization part of the memory task, followed by the Go/No-Go task.Then, the participants performed the recall part of the memory task (Figure 1).

Metabolic
Blood was collected into 1 mL syringes to determine Hb and into vacutainers containing EDTA-2Na for plasma samples.Each vacutainer was gently mixed and centrifuged at 600g (4 • C) for 10 min.Plasma supernatant was stored at −80 • C for glucose and lactate analysis using a Randox Daytona-plus Clinical Chemistry analyser (Prior et al., 2020).For glucose measurement, there was an analytical sensitivity of 0.02 mmol/L and a dynamic range of 0.02-250 mmol/L, with an intra-assay CV of <7.1% (Prior et al., 2020).For lactate measurement, there was an analytical sensitivity of 0.5 mmol/L and a dynamic range of 0.5−40 mmol/L, with an intra-assay CV of <1.5%.Insulin was measured using an Invitron Insulin ELISA kit, with an analytical sensitivity of 0.02 mU/L and a dynamic range of 0.02−250 mU/L, with an inter-assay CV of ≤ 7.1% (Prior et al., 2020).

Cognitive function
The cognitive task comprised central executive function (Go/No-Go) and a non-executive (memory) task (Akagi et al., 2019;Saito et al., 2021;Washio et al., 2021).The Go/No-Go task evaluates selective attention and response inhibition (Ando et al., 2013), and the memory task requires non-executive recall memory (Akagi et al., 2019;Lefferts et al., 2016a).The cognitive tasks were programmed and controlled using the software Presentation (v.19; NeuroBehavioral Systems, Berkeley, CA, USA).The participants viewed a computer screen placed ∼80 cm away from them in a supine position.The order of the cognitive tasks is shown in Figure 1.In the memory task, 30 words were presented to the participants for memorization and later recall from memory.
The list contained 30 concrete English words that were displayed for 1 s each.We used different word sets for each measurement in the memory task to avoid learning effects.The participants then completed the Go/No-Go task before the recall part of the memory task.In the recall part, 60 words (including 30 distracters) were presented, at a rate of one word every 2 s.In the Go/No-Go task, each trial started with a blank screen for 2.5 s, followed by a green square presentation at the centre of the computer screen for 1 s, which served as a preparatory stimulus.Then, one of four squares (red, blue, yellow and purple) was presented for 1 s.Half of the participants responded to red and blue squares as Go signals, and the others responded to yellow and purple squares as Go signals.In the Go trial, the participants were instructed to press a mouse button with the right index finger as quickly as possible.For the No-Go trial, the participants were instructed to withhold their response.The Go/No-Go task consisted of a block of 60 trials with equal probability.In the recall part of the memory task, the participants were instructed to press the mouse button if the word had been presented during the memorization.
Performance of the Go/No-Go task was assessed by RT and accuracy (as a percentage).Omissions of a response in a Go-trial or incorrect response in a No-Go trial were regarded as an error.Memory function was assessed by accuracy (as a percentage).All accuracy calculations were determined as the number of correct responses/total number of trials (×100).

Data sampling
Beat-by-beat data were sampled continuously at 1 kHz using an analog-to-digital converter (Powerlab/16SP ML795; ADInstruments, Colorado Springs, CO, USA) stored on a personal computer for offline analysis (LabChart v.7.2.2, ADInstruments, Colorado Springs, CO, USA).All files were given a coded number (not named) by an investigator blinded to the study.We also time-aligned the mean arterial blood pressure (MAP) and TCD channels, given the time delay (1.07 s) associated with MAP signal processing when using the Finometer PRO (Bailey et al., 2013).Baseline physiological and ventilatory variables were averaged over 1 min.Cerebrovascular variables (i.e., MCAv, PCAv, oxy-Hb, deoxy-Hb and TSI) were averaged over 30 s at baseline.Cerebrovascular variables were also averaged during the cognitive tasks.

Statistical analysis
Data were analysed using SPSS v.29.0 (SPSS, Chicago, IL, USA).or median (interquartile range), and the significance level was set at P < 0.05 for all two-tailed tests.

Basal responses
Baseline physiological and metabolic variables are summarized in
Cerebral oxygenation responses to cognitive activity are illustrated in Table 3.We observed that oxy-Hb and TSI changes in response to cognitive tasks were different between the memorization part, the Go/No-Go task and the recall part (P < 0.001, main effects).In contrast, deoxy-Hb change was not different (P = 0.435, main effect).

Variable
(P = 0.425, main effect).Figure 4b shows the accuracy of the memory task.Hypoxia impaired the accuracy of the memory task (P = 0.049 vs. normoxia, main effect), whereas time had no effect (P = 0.180, main effect).

DISCUSSION
The present study has identified three findings that have integrated translational relevance, albeit selectively constrained to male participants only.First, acute hypoxia was consistently associated with a reduction in both MCA (anterior) and PCA (posterior) DO 2 .
Second, memory was generally impaired in hypoxia, and this was accompanied by a selective reduction in PCA DO 2 .Third, in stark contrast, central executive function remained preserved in hypoxia despite a comparable reduction in CDO 2 .Collectively, these findings suggest that a local reduction in perfusion and CDO 2 to the posterior circulation selectively impair memory, whereas selective attention and response inhibition remain well preserved in hypoxia.

Memory
We observed impairments in memory during hypoxia, which is in line with previous studies (Lefferts et al., 2016b;Wang et al., 2013).In the memory task, participants were instructed to remember words during memorization, which required temporary storage until recall was performed.Memories initially require rapid synaptic plasticity within the hippocampus for formation and are gradually consolidated in neocortical networks (Kitamura et al., 2017;Kumaran et al., 2016).
Therefore, the hippocampus is likely to play a crucial role in memory in the present study.
Rodent studies indicate that the hippocampus is more vulnerable to hypoxia-mediated oxidative stress and subsequent hypoxic-ischaemic injury compared with the cortex (Hota et al., 2007;Maiti et al., 2006).
Accordingly, the number of damaged cells in the CA3 region of the hippocampus has been shown to increase after exposure to 6400 m for 4 days (Shukitt-Hale et al., 1996).Furthermore, hippocampal vasculature (i.e., lower capillary density and red blood cell velocity) The ΔOxy-Hb (a) and ΔTSI (b) during memorization in the memory task, the Go/No-Go task and the recall part in the memory task.
Values are the mean ± SD or median (interquartile range).## P < 0.01 vs. memorization and Go/No-Go task.Abbreviations: oxy-Hb, oxyhaemoglobin; TSI, tissue saturation index.Numbers represent the time (in hours) after the exposure to hypoxia/normoxia.might constrain cerebrovascular substrate (O 2 /glucose) delivery and account enhancing vulnerability to hypoxia (Shaw et al., 2021).
In the present study, hypoxia reduced CDO 2 , which coincided with the observed impairment in memory.The hippocampal blood supply is generally provided by the collateral branches of the PCA and the anterior choroidal artery (Perosa et al., 2020;Spallazzi et al., 2019), and therefore, the present results indirectly, albeit intuitively, suggest that O 2 delivery to the hippocampus might be linked to cognitive dysfunction in hypoxia.In addition, we observed that Δoxy-Hb and ΔTSI were selectively elevated during recall compared with memorization and during the Go/No-Go task.Recall is involved in the activation of multiple regions, including the prefrontal cortex (Scalici et al., 2017), which is supported by our observations of greater increases in oxy-Hb and TSI.These findings suggest that, in conjunction with the hippocampus, the prefrontal cortex might also be associated with memory recall.Hence, reductions in CDO 2 in hypoxia might attenuate the activations in the prefrontal cortex during recall, contributing, at least in part, to the observed impairments in memory.

Central executive function
We demonstrated that RT in the Go/No-Go task was unaffected by hypoxia despite a clear reduction in anterior (MCA) DO 2 .The MCA supplies the majority of the lateral surface of the hemisphere, including the prefrontal, motor, parietal and temporal cortices (Berman et al., 1984).These findings suggest that central executive function, as assessed in the present study, was either resistant to a reduction in MCA DO 2 or, alternatively, that a threshold impairment in bioenergetic function needs to be surpassed before cognitive impairment ensues.
However, it should be noted that the type of cognitive task has interpretive implications (Ando et al., 2020).The Go/No-Go task requires selective attention and also control of motor inhibition (Guarino et al., 2018).Given the high accuracy of the performance and the absence of an impairment in hypoxia, this might be associated with relatively low task difficulty.Future studies using multiple central executive tasks (e.g., Stroop task) are encouraged to establish a wider understanding of the effect of hypoxia on central executive function.

Cerebral bioenergetics
The brain is a high-flow organ that depends on acute increases in regional blood flow and DO 2 to support increases in neural activity and synaptic transmission (Attwell et al., 2010;Bailey, 2019a, b;Iadecola, 2004).Neurovascular coupling (NVC) refers to the mechanism that links neural activity to consequent increases in local CBF (Hendrikx et al., 2019).In the present study, cognitive activity failed to effect adequate increases in posterior perfusion and consequent CDO 2 during the memory task in hypoxia.A recent rodent study demonstrated that NVC is weaker in the hippocampus compared with the visual cortex because of differences in vascular properties (Shaw et al., 2021).Thus, the present results would underscore the importance of basal CDO 2 in the hippocampus for the maintenance/preservation of memory in hypoxia.
Equally, previous studies have documented that hypocapnia attenuates NVC (Caldwell et al., 2021;Szabo et al., 2011).In the present study, however, hypocapnia did not affect haemodynamic and CDO 2 responses to cognitive activity.In these previous studies (Caldwell et al., 2021;Szabo et al., 2011), NVC was evaluated using visual stimulation.Visual stimulation generally evokes a robust NVC response that increases perfusion of the visual cortex substantially (Phillips et al., 2016), and the degree of NVC is likely to be greater (e.g., 9 ± 4 cm/s, peak absolute increase in PCAv during visual stimulation in the study by Caldwell et al., 2021) than that incurred in the present study.Hence, the absence of hypocapnic effects on haemodynamic and CDO 2 responses to cognitive activity is probably attributable to the differences in the tasks (i.e., visual stimulation vs. cognitive task).
Blood glucose is the primary energy source for the brain (Gold, 1995).In the present study, hypoxia did not affect glucose and insulin concentrations, and as anticipated, blood lactate was mildly elevated in hypoxia.Hashimoto et al. (2018) have suggested that the increase in lactate production from extracerebral tissue is likely to support brain function.However, these increases were minimal and cognitive function did not improve.Thus, we can assume that changes in these substrates contribute little (if at all) to cognitive function.

Temporal kinetics
Reductions in CBF in the brain regions associated with memory were found to be more pronounced after prolonged (10 h) compared with acute (2 h) hypoxic exposure (Lawley et al., 2017), which suggests that prolonged hypoxia might reduce regional cerebral metabolism in the brain regions related to memory.However, in contrast to our original hypothesis, we found that memory did not become progressively worse in hypoxia, despite changes in CDO 2 ; differences that might relate to the comparatively shorter exposure time used in the present study.

Experimental limitations
There are several major limitations that warrant critical consideration.
First, our findings are constrained to male participants only, which is unfortunate given that incidence rates of neurogenerative diseases, notably Alzheimer's disease, are higher in women compared with men (Beam et al., 2018).Future researchers are actively encouraged to include both men and women to help identify biological sex differences in responses to various stimuli that could be influential for brain health and brain ageing.Second, despite retrospective analyses revealing that we were adequately powered (1 − β = ≥ 0.80 at P < 0.05) to detect main effects for all primary outcomes, we recognize that our sample size is small and that this is likely to have contributed to the absence of hypoxic effects on central executive function.
Given that we were not in a position to include more participants owing to logistical/financial constraints, future research needs to consider larger sample sizes to draw firmer conclusions and make these findings more applicable to the general population.Third, MCAv and PCAv were measured using TCD, reflecting an indirect surrogate measure of regional CBF.In hypoxia, the net CBF response reflects a balance between hypoxic cerebral vasodilatation and hypocapnic vasoconstriction driven by hyperventilation-induced hypocapnia (Ogoh et al., 2014).Given that MCA diameter appears to increase in hypoxia (Wilson et al., 2011), we cannot rule out the possibility that we underestimated volumetric changes in hypoxia.In addition, hippocampal vascularization seems to be categorized into a mixed supply from both the PCA and the anterior choroidal artery (Perosa et al., 2020).
This implies that there are inter-individual differences in hippocampal blood flow supply.Hence, associations between cognitive function and regional cerebral perfusion/CDO 2 warrant further investigation.
Follow-up studies using advanced imaging techniques, such as arterial spin labelling, would be helpful to improve our understanding of the association between memory and regional CBF and substrate delivery in hypoxia.Finally, the extent to which hypobaria per se (and not hypoxia) impacts cognitive function remains to be established, despite preliminary evidence for comparable effects (Hohenauer et al., 2022).

CONCLUSIONS
The present study examined the effects of acute hypoxia (∼7 h) on central executive (Go/No-Go) and non-executive (memory) function.
In hypoxia, cerebral bioenergetic function taking the form of MCA and PCA DO 2 were consistently lower compared with normoxia.The reduction in PCA DO 2 was accompanied by impaired memory, whereas selective attention and response inhibition remained well preserved, despite comparable reductions in CDO 2 .Collectively, these findings, albeit selectively constrained to males, suggest that cognitive function, in particular memory, is selectively vulnerable in hypoxia.
Overview of the experimental protocol.Cognitive function was assessed at baseline (0 h), 2, 4 and 6 h, after measurements of cardiopulmonary and metabolic variables.Abbreviations: CBFv, cerebral blood flow velocity; HR, heart rate; MCA, middle cerebral artery; NIRS, near-infrared spectroscopy; PCA, posterior cerebral artery; P ET,CO 2 , end-tidal partial pressures of CO 2 ; P ET,O 2 , end-tidal partial pressures of O 2 ; S pO 2 , peripheral O 2 saturation; VE , minute ventilation.

Figure
Figure4aillustrates RT in the Go/No-Go task.Hypoxia did not affect RT (P = 0.189 vs. normoxia, main effect), and RT did not change as a function of time (P = 0.738, main effect).Accuracy of the Go/No-Go task remained high throughout the experiment (Table1) and was unaffected by hypoxia (P = 0.201 vs. normoxia, main effect) or time Abbreviations: Deoxy-Hb, deoxyhaemoglobin; Oxy-Hb, oxyhaemoglobin; TSI, tissue saturation index.