Hypoxia is a condition of reduced oxygen supply to brain tissue that can occur in a variety of circumstances, from low atmospheric oxygen to reduced blood flow. It is distinguished from anoxia, a more adverse condition of no oxygen supply, which is not considered in this study. Low oxygen levels can have severe adverse effects on affected tissue, including enhanced production of reactive oxygen species (ROS), impaired ATP production and loss of calcium homoeostasis (Kristian and Siesjo, 1998), all of which can lead to tissue damage and cell death. Neuronal tissue has a very high energy consumption, which makes it particularly vulnerable to hypoxic insult (Hossmann, 1999), for example following a stroke or respiratory arrest. This vulnerability is not limited to the immediate and drastic brain damage often associated with sustained reduction of oxygen supply to the brain. It can manifest in more subtle ways, notably increased risk of developing neurodegenerative disorders (Sun et al., 2006; Guglielmotto et al., 2009; Bhatia et al., 2017) as well as aggravate pathological processes in patients with existing neurodegenerative disease and animal models of neurodegenerative disease (Zhang et al., 2019). Compounding the link between hypoxia and neurodegenerative disorders is also a growing body of evidence that genetic traits which impair proper function of various hypoxia-responsive elements can increase the risk of developing neurodegenerative disorders, including Parkinson’s disease (PD) and Amyotrophic lateral sclerosis (ALS) (Greenway et al., 2006; van Es et al., 2011; Zou et al., 2012).

Hypoxia causes neurotoxic effects which may be the consequence of different mechanisms, both activity-dependent and activity-independent. Changes in neuronal excitability are widely reported in both dissociated neuronal cell cultures and brain slice preparations, albeit with contradictory findings. For example, some studies have found that hypoxia leads to neuronal hyperexcitability in Schaffer collaterals of rat hippocampal slices (Schiff and Somjen, 1985) and CA1 neurons in rat hippocampal slices (Godukhin et al., 2002), while other studies have reported decreased excitability in CA1 neurons in mouse hippocampal slices (Gu and Haddad, 2001) and dissociated mouse hippocampal neurons (Gavello et al., 2012). There may also be cell type-specific vulnerabilities which contribute to this deviance in findings, such as the reported vulnerability of inhibitory neurons (Turovsky et al., 2013). The influx of calcium into hypoxic neurons has been widely implicated in the neurotoxic process (Kristian and Siesjo, 1998), and NMDA-receptors appear to be an important avenue of Ca²⁺ influx due to their proximity to mitochondria, which act as intracellular calcium stores (Starkov et al., 2004). Blocking NMDA-receptors has been shown to reduce the formation of ROS (Lafon-Cazal et al., 1993). Activation of GABA-A receptors has also been shown to ameliorate oxidative damage during hypoxia due to Cl^– influx in rodents (Zhao et al., 2003; Kumar and Goyal, 2008), and inhibition of activity of hippocampal neurons with leptin also leads to improved outcomes following hypoxia (Gavello et al., 2012). Together these results suggest an activity-dependent mechanism for neuronal damage following hypoxia.

Although neuronal tissue is susceptible to hypoxia in general, there are notable differences in how different neuronal populations respond to hypoxia, and some populations are clearly more vulnerable than others. However, most of the research on neuronal susceptibility examines the vulnerable hippocampal neurons, while other vulnerable groups such as motor neurons are less frequently studied. However, not only are motor neurons particularly susceptible to hypoxia (Watzlawik et al., 2015), but repeated exposure to hypoxic conditions also appears to greatly increase the risk of developing motor neuron disorders like ALS (Vanacore et al., 2010). Motor neurons are highly energy demanding, and findings indicate that motor neurons with ALS pathology exhibit reduced mitochondrial respiration and oxidative ATP production, with a shift toward increased glycolysis (Hor et al., 2021) which can contribute to overall energy deficits which contribute to disease progression (Dupuis et al., 2011). This mirrors the effects of hypoxia, where mitochondrial activation of hypoxia inducible factor 1α (HIF-1α) in response to reduced concentration of oxygen leads to inhibition of oxidative phosphorylation and increased production of ATP through glycolysis, resulting in an overall energy deficit compared to normoxic conditions (Solaini et al., 2010). This makes the study of the responses of motor neurons to hypoxic challenge also relevant to studies of ALS, both through shared pathways and as a potential environmental contributor to the disease.

Studies of neuronal activity are underutilised in the study of hypoxia exposure, especially with regard to the effects on broader network dynamics. This study uses in vitro motor neuron networks on multielectrode arrays (MEAs) to examine responses to hypoxic insult in a longitudinal perspective. It considers the effects of network activity during the episode in maintaining healthy network function following hypoxia.

By the time the motor neuron networks in this study were mature at 49 DIV they exhibited stable activity in terms of firing rate, a high degree of burst propensity and stable levels of synchrony as seen up to 54 DIV in Figure 3, which are in line with observations for functionally mature networks (Pasquale et al., 2008). These networks also expressed markers of mature neuron cytoskeletons and relevant receptors including GABAb-1 receptors, shown in Figure 2C, as well as specific markers for motor neurons, shown in Figures 2A,B.

Interestingly, the hypoxic networks showed no changes in activity during the episode itself compared to the normoxic controls. There were no discernible differences in firing rate, burst firing or synchrony in the uninhibited hypoxic networks compared to the pre-hypoxia baseline (Figures 5C,E,G). The same was predictably true for uninhibited normoxic networks (Figures 5B,D,F). Previous reports have indicated that exposure to hypoxia can lead to acute hyperexcitability leading to elevated firing rate in hypothalamic neurons (Shonis and Waldrop, 1995) while others have found that chemical hypoxia leads to reduced excitability (Englund et al., 2004), although investigations of ongoing activity during hypoxia rather than post-hypoxic recovery are rare, especially in motor neurons. Such changes in activity were not observable in the current motor neuron networks, and changes in activity following exposure to hypoxic challenge only became evident later on.

It should be noted that the astrocytes co-cultured in the motor neuron networks may provide neuroprotective effects to the motor neurons in this study, a widely reported phenomenon (Marina et al., 2016). However, since all experimental groups contained the same ratio of neurons to astrocytes, and evidence suggests that astrocytes are more resilient than neurons to hypoxic conditions (Alves et al., 2000), it seems unlikely that potential effects of astrocytes is a major contributing factor to the observed group differences.

This study demonstrated that mature networks of motor neurons exposed to hypoxic challenge exhibit aberrant fluctuations in network activity which were not evident in networks maintained in normoxic conditions, seen clearly in Figure 3. The fluctuations occurred over the course of 59–64 DIV, and involved changes from high to low firing rate, from normoxic levels of burst propensity to low levels, and from normoxic levels of network synchrony to elevated levels. This loss of homoeostasis and stable network function differs significantly from the normoxic networks, as seen in Figure 4, showing the detrimental effects of the hypoxic exposure on network function. Interestingly, this effect did not become apparent until 5–6 days following the hypoxic challenge, as seen in Figure 3. This may suggest that there are compensatory mechanisms or adaptive responses which can maintain function close to that of normoxic counterparts, but that these measures are only effective for a limited time.

Furthermore, the timing of the observed changes in activity follow a very characteristic pattern, with deviations from normoxic-like levels observed on alternate days which correspond to observations immediately prior to cell media replacement, meaning that more of the nutrients in the cell media were depleted compared to the observations where network activity was more similar to that of normoxic networks. It is possible that this is a result of an increased metabolic burden due to effects of the hypoxia and any homeostatic mechanisms which fail to maintain as nutrients are depleted. High metabolic requirements are seen across multiple populations of neurons vulnerable to neurodegenerative disease (Muddapu et al., 2020) and may contribute to their selective vulnerability. Similarly, during hypoxia, due to the uncoupling of oxidative phosphorylation, the production of ATP becomes less efficient, eventually leading to elevated oxidative stress (Jain et al., 2015; Snyder et al., 2017; Chen et al., 2018). The effects of hypoxia can thus be more adverse for neuronal populations with high energy requirements. The results from this study may therefore suggest that motor neuron networks exposed to hypoxia exhibit increased metabolic stress, which leads to loss of functional homoeostasis and could be a predisposing factor for further network degeneration.

Some of the changes in network activity appear to be related to activity during the hypoxic challenge, since inhibition of the network with GABA improves functional outcomes in terms of maintaining network burst propensity. While inhibition of the normoxic networks appeared to have no lasting effect, the hypoxic networks inhibited during the hypoxic challenge exhibited normoxic levels of burst propensity throughout the experiment (Figures 3B, 4E), despite exhibiting similar fluctuations as the uninhibited hypoxic networks in terms of the other activity parameters (Figures 3A,C, 4D,F). Network bursts are an emergent property of activity of networks of neurons and is considered to be more reliable in terms of signalling and can carry more information (Zeldenrust et al., 2018). The maintenance of this parameter in the silenced hypoxic networks indicates improved functional outcomes as a result of the silencing, even though there are still clear consequences of the hypoxic challenge. It is possible that there are activity-dependent mechanisms which occur during hypoxia which lead to a delayed loss of function, either due to plasticity-induced changes during the episode or due to metabolic stress incurred by maintaining activity during hypoxic challenge. Acute hypoxia like in this study has been reported to activate HIF-1 through mitochondria-dependent mechanisms, a function of which is to upregulate anaerobic respiration and inhibiting the oxygen dependent TCA cycle, which places the affected cells in an energy-deficient state (Solaini et al., 2010; Hernandez-Gerez et al., 2019). From the results of our study, it appears that the activity-dependent damage was due to maintenance of normal activity, indicated by the results in Figure 4, rather than due to changes in activity during the hypoxic challenge. Maintaining activity during such circumstances of less energy production can lead to increased production of ROS which have the potential to cause damage to the network. Elevated levels of oxidative stress markers have previously been reported in mouse models of ALS exposed to hypoxic conditions (Kim et al., 2013). It seems likely that inhibition the networks during the hypoxic challenge would reduce the energy requirements of the motor neurons, thus reducing the metabolic stress. Inhibiting the networks contributed to maintaining pre-hypoxic functionality in terms of network burst propensity, although the persistent fluctuations in firing rate and synchrony indicate that the hypoxic challenge had effects on the networks which were not dependent on activity during the hypoxic challenge itself.

Although there are conflicting reports on the immediate effects of hypoxia exposure on network activity, it is evident that the mechanisms involved, at least in part, increased oxidative damage through calcium dysregulation (Kristian and Siesjo, 1998; Starkov et al., 2004). In this context, there is a particular association with the activity of NMDA-receptors and findings suggest that blocking them can reduce the production of ROS during hypoxia (Lafon-Cazal et al., 1993) as well as ameliorate post-hypoxic hyperexcitability (Godukhin et al., 2002). When addressing hypoxia and considering interventions it is vital to know what pathways can be targetted in order to improve outcomes for patients, for example those who require assisted ventilation, which can be associated with acute hypoxic events (Keith and Pierson, 1996; Mahmood et al., 2013). This study suggests that functional outcomes can be preserved by targetting activity-dependent pathways.

The study of hypoxia and the mechanisms by which it damages neural networks bears a direct relevance to neurodegenerative diseases like ALS. The body of evidence, which shows that mutations in hypoxia-responsive elements like angiogenin (ANG) can increase the likelihood of developing ALS (Greenway et al., 2006; van Es et al., 2011; Zou et al., 2012), indicates that some mechanisms of healthy responses to hypoxia are also involved in the ongoing health of motor neurons which fail in ALS. Additionally, SOD1 genetic models of ALS exhibit altered responses to hypoxia, which may exacerbate their susceptibility to hypoxic stress (Cimini et al., 2014), and elevated levels of hypoxia responsive elements including ANG, apoptosis inducing factor and HIF-1α (Zhang et al., 2011) are observed in both ALS patients (Cronin et al., 2006; van Es et al., 2014; Morelli et al., 2020) and mouse models (Xu et al., 2011). Furthermore, exposure to repeated hypoxic episodes appears to increase the risk of developing ALS (Vanacore et al., 2010), and studies in mouse models of ALS have found that stabilising HIF-1α, in effect enhancing the hypoxic response mechanism, reduced degeneration of spinal motor neurons and myofibrils of the lower limbs, as well as increasing the lifespan of the mice (Nomura et al., 2019). Mouse models of ALS exposed to hypoxic conditions also exhibit accelerated disease progression compared to normoxic controls (Kim et al., 2013). These findings suggest a relationship between the mechanisms of resilience to hypoxia and the risk of developing ALS as well as the severity of the disease progression. The results of this study indicate that there are activity-dependent mechanisms which lead to network dysfunction following hypoxia. Not unexpectedly, studies of ALS show that activity-dependent processes, hyperexcitability and formation of ROS is directly linked with motor neuron damage (Kaus and Sareen, 2015; Verma et al., 2018; Marcuzzo et al., 2019). ALS has also been observed to alter cortical hub regions, i.e., highly interconnected structures which are vital for proper network function, early on in the disease process (Verstraete et al., 2011; Ma et al., 2015). Given that cortical hubs are known to be highly metabolically demanding (Bullmore and Sporns, 2012; Collin et al., 2014), it is natural to consider that they might be some of the first parts of the neural network to show deterioration in circumstances where there is reduced capacity to respond healthily to activity-dependent adversity, such as a reduction in functionality of the hypoxia-responsive machinery.

In conclusion, this study examined the responses of motor neuron networks to a hypoxic challenge and how the functionality of the networks develops over time, with and without inhibition of activity during the challenge. We found that although activity during the hypoxic challenge remained similar to both baseline measurements and to normoxic controls, hypoxic networks began to show clear deviations from the behaviour of normoxic networks, involving fluctuations in activity. Inhibition of network activity during the hypoxic challenge contributed to preserved bursting behaviour compared to uninhibited networks, an improved functional outcome. These responses could not have been captured without a longitudinal assessment of the network activity, which is rarely seen in the existing literature. Elucidating patterns of dysfunction following hypoxic challenge and the mechanisms involved has implications beyond the hypoxia itself, such as in diseases like ALS, where the mechanisms of degeneration appear to be at least in part shared with those observed in hypoxia. Assessing these mechanisms can thus contribute to a more precise investigations of these properties in a highly multifactorial neurodegenerative disorder.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

VF conceived of and designed the study, acquired the data, performed the statistical analysis, and drafted the manuscript. AS and IS participated in the design of the study and helped to revise the manuscript. All authors read and approved the final version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.