Altered ventilatory responses to hypercapnia-hypoxia challenges in a preclinical SUDEP model involve orexin neurons

Abstract

Failure to recover from repeated hypercapnia and hypoxemia (HH) challenges caused by severe GCS and postictal apneas may contribute to sudden unexpected death in epilepsy (SUDEP). Our previous studies found orexinergic dysfunction contributes to respiratory abnormalities in a preclinical model of SUDEP, Kcna1^−/− mice. Here, we developed two gas challenges consisting of repeated HH exposures and used whole body plethysmography to determine whether Kcna1^−/− mice have detrimental ventilatory responses. Kcna1^−/− mice exhibited an elevated ventilatory response to a mild repeated hypercapnia-hypoxia (HH) challenge compared to WT. Moreover, 71% of Kcna1^−/− mice failed to survive a severe repeated HH challenge, whereas all WT mice recovered. We next determined whether orexin was involved in these differences. Pretreating Kcna1^−/− mice with a dual orexin receptor antagonist rescued the ventilatory response during the mild challenge and all subjects survived the severe challenge. In ex vivo extracellular recordings in the lateral hypothalamus of coronal brain slices, we found reducing pH either inhibits or stimulates putative orexin neurons similar to other chemosensitive neurons; however, a significantly greater percentage of putative orexin neurons from Kcna1^−/− mice were stimulated and the magnitude of stimulation was increased resulting in augmentation of the calculated chemosensitivity index relative to WT. Collectively, our data suggest that increased chemosensitive activity of orexin neurons may be pathologic in the Kcna1^−/− mouse model of SUDEP, and contribute to elevated ventilatory responses. Our preclinical data suggest that those at high risk for SUDEP may be more sensitive to HH challenges, whether induced by seizures or other means; and the depth and length of the HH exposure could dictate the probability of survival.

1. Introduction

Sudden unexpected death in epilepsy (SUDEP) affects 1:1000 people with epilepsy each year (Hesdorffer et al., 2011; Thurman et al., 2014; Walczak et al., 2001). This risk increases to ~1:150 in patients with severe refractory generalized convulsive seizures (GCS). Clinically, approximately 40% of severe and focal seizures are accompanied by hypoventilation-induced drop in blood O₂ below 90% saturation levels and hypoxia (Azar et al., 2008; Bateman et al., 2008; Bruno et al., 2018; Carmenate et al., 2020; Lacuey et al., 2018; Seyal et al., 2010). GCS and post-convulsive apnea are associated with increased end tidal CO₂ levels and peri-ictal hypercapnia (Sainju et al., 2019; Vilella et al., 2019b, 2019a). The hypercapnic and hypoxemic (HH) blood gas instability can last several minutes following a severe seizure (Bruno et al., 2018; Farrell et al., 2016; Moseley et al., 2010). Clinical studies indicate that prolonged and severe post-ictal HH are associated with cardio-respiratory instability and contribute to increased SUDEP risk (Bruno et al., 2018; Moseley et al., 2010; Park et al., 2017; Seyal et al., 2011). Examination of data reported in the Mortality in Epilepsy Monitoring Unit Study (MORTEMUS) indicates that patients experienced intermittent HH (GCS and one or more apneas) prior to SUDEP with short recovery intervals (Ryvlin et al., 2013). Recovery from post-ictal repeated HH challenges that are separated by only brief recovery periods may be critical for survival.

Here, we developed a novel preclinical in vivo test with three brief HH challenges and recovery intervals, designed to mimic the intermittent HH prior to SUDEP. Using this test, we aimed to study the reliability of the HH ventilatory response. Three hypotheses were tested. Our first hypothesis was that the ventilatory response to a triple HH challenge is impaired in preclinical SUDEP. To test this hypothesis, we used Kcna1^−/− mice, which lack the Kv1.1 potassium channel and are a model of temporal lobe epilepsy and SUDEP (Glasscock et al., 2010; Moore et al., 2014; Simeone et al., 2016, 2013; Smart et al., 1998). Our previous studies have found that as Kcna1^−/− mice approach sudden death, they have increased GCS, dysregulated ventilatory responses to methacholine, a higher susceptibility to apneas and intermittent periods of hypoxemia, all indicative of blood gas instability (Dhaibar et al., 2019; Iyer et al., 2020, 2018; Roundtree et al., 2016; Simeone et al., 2018, 2016). Following the triple HH challenge, we found Kcna1^−/− mice had significant differences in several respiratory parameters indicating they struggled during the challenges. Indeed, most Kcna1^−/− mice could not survive a triple challenge when the severity of the HH was increased.

Recovering from severe or repeated HH is mediated by specialized central and peripheral chemoreceptors that detect blood gas instability and trigger adaptive respiration by increasing the output of key respiratory brainstem nuclei (Giannoni et al., 2017; Hodges and Richerson, 2010; Lazarenko et al., 2011; Oikawa et al., 2005; Richerson, 2004; Wong-Riley et al., 2013; Yang et al., 2003). There are several central chemosensors that detect HH either directly in response to changes in partial pressures of blood O₂ and CO₂ or indirectly in response to decreases in pH (acidosis) which is known to occur concurrently with HH (Azar et al., 2008; Borch et al., 1993; Siesjö et al., 1985; Ziemann et al., 2008). Orexin neurons are located in the lateral hypothalamus and their response to pH and CO₂ is critical for modulating respiration to maintain blood gas homeostasis, particularly after repeated challenges (Burdakov et al., 2013; Kuwaki, 2008; Li et al., 2016; Li and Nattie, 2010; Nakamura et al., 2007; Sakurai et al., 1998; Sunanaga et al., 2009; Williams et al., 2007). Orexin neurons project to multiple brain stem nuclei and are involved in regulating many functions including cardiorespiration and autonomic physiology (Burdakov et al., 2013; Ciriello et al., 2003; Date et al., 1999; de Oliveira et al., 2003; Lazarenko et al., 2011; Li and Nattie, 2014; Nattie and Li, 2012; Peyron et al., 1998; Sakurai et al., 1998; Shahid et al., 2011; Vicente et al., 2016; Williams et al., 2007; Young et al., 2005). Stimulation of orexin neurons leads to release of the neuropeptide and subsequent activation of orexin receptors type 1 or type 2 at several brainstem respiratory nuclei, including the retrotrapezoid nucleus, medullary raphe, and nucleus of the solitary tract, which then collectively mediate increased respiratory output (Burdakov et al., 2013; Ciriello et al., 2003; de Oliveira et al., 2003; Lazarenko et al., 2011; Nattie and Li, 2012; Shahid et al., 2011; Vicente et al., 2016; Williams et al., 2007; Young et al., 2005).

We previously reported that acutely blocking orexin receptors with a dual orexin receptor antagonist (DORA) reduces apneas and chronic intermittent hypoxemia, stabilizes respiration and heart rate, and postpones sudden death of Kcna1^−/− mice (Iyer et al., 2020, 2018; Roundtree et al., 2016; Simeone et al., 2018). Thus, our second hypothesis was that the orexinergic system contributes to the impaired HH ventilatory response of Kcna1^−/− mice. To test this hypothesis, mice were pretreated with a DORA and exposed to the triple HH challenge. The DORA normalized HH respiratory responses of Kcna1^−/− mice during the mild triple HH challenge and promoted survival in the more severe triple HH challenge.

Genetic or pharmacological blockade of the orexinergic system has been associated with reduced hypercapnic chemoreflexes (Kuwaki, 2008; Li et al., 2016; Li and Nattie, 2010; Nakamura et al., 2007). Therefore, we tested a third hypothesis that orexin activity of Kcna1^−/− mice is more sensitive to pH changes (i.e., more chemosensitive) than WT. We conducted ex vivo extracellular electrophysiology experiments using a multielectrode array to record spontaneous activity of putative orexin neurons in the lateral hypothalamic network. Our results indicate that the chemosensitive activity of putative orexin neurons is increased in Kcna1^−/− mice.

2. Methods

2.1. Animals

Kcna1^+/− mice on a C3HeB/FeJ background were bred in-house to generate Kcna1^+/+ (wild type, WT) and Kcna1^−/− genotypes. Genotypes were determined by Transnetyx Inc. (Cordova, Tennessee) as previously described (Roundtree et al., 2016). Mice were reared in a quiet, temperature-controlled room on a 12 h light/dark cycle and had access to food and water ad libitum. Data did not differ between sexes; thus, males and females were combined. All Kcna1^−/− mice on the C3HeB/FeJ background strain succumb to sudden death following a seizure-related event (George et al., 2023; Simeone et al., 2016) and mortality in our colony occurs with a mean age of approximately postnatal day 50 (Iyer et al., 2020, 2018; Simeone et al., 2018, 2016). Approximately 7-week old mice were used, an age by which Kcna1^−/− mice experience multiple spontaneous seizures a day (ranging from mild to severe tonic clonic seizures) and are considered high-risk for sudden death (Iyer et al., 2020, 2018; Roundtree et al., 2016; Simeone et al., 2018, 2016). All experiments conformed to NIH guidelines in accordance with the United States Public Health Service’s Policy on Humane Care and Use of Laboratory Animals and were approved by Creighton University’s Institutional Animal Care and Use Committee.

2.2. SUDEP Common Data Elements (CDEs)

The SUDEP CDEs specify details pertaining to methods and outcomes to improve translational relevance and reproducibility (currently available at https://cureepilepsy.org/research-resources). Here, we used CDEs for (i) Core and Death-Related Information, (ii) Physiologic Measures, and (iii) Ex Vivo Electrophysiology.

2.3. Acute in vivo HH challenges

WT and Kcna1^−/− mice were gently placed in the whole-body plethysmography chamber (SQIREQ USA Inc., Plattsburgh, NY) and allowed to habituate. Prior to each experiment, the whole-body plethysmography sensor was calibrated with an empty chamber and fixed gas inflow and outflow rates. Respiratory signals of the animal were based on barometric changes inside the chamber. Mice were freely moving during the experiment and seizures were video monitored during the tests. A seizure was not induced prior to the tests to decouple the HH challenges from seizures to determine whether repeated HH challenges alone are sufficient to detect ventilatory differences in high-risk SUDEP. Following habituation, a stable 15-min baseline reading was obtained under normoxia and normocapnia conditions (21% O₂ balanced with N₂). Cohorts were exposed to either a ‘mild’ triple HH challenge or a ‘severe’ triple HH challenge using the three-gas mass flow controller which allows for precise regulation and rapid administration of gas concentrations on a short time scale (SCIREQ Scientific Respiratory Equipment Inc., an Emka Technologies Co., Montreal, QC, Canada). The mild triple challenge consisted of three 2-min HH challenges of 6% O₂ + 9% CO₂ with brief 3-min normoxia recovery intervals. The severe triple challenge consisted of a 3-min 6% O₂ + 3% CO₂ exposure, followed by 3-min of 6% O₂ + 9% CO₂, then a longer 5-min 6% O₂ + 9% CO₂ exposure with 3-min recovery intervals. Exchange of gases occurred over 1-min flush periods (as illustrated in Fig. 1A and 1H). Blood pH at baseline and immediately following the severe triple challenge was determined using microfiber-optic pH sensors (PreSens Precision Sensing, Regensburg, Germany).

Fig. 1.

In vivo, orexin contributes to the enhanced ventilatory response and high mortality rate of Kcna1^−/− mice following repeated HH challenges. (A) Top panel: A schematic of the mild triple HH challenge. Each 2-min challenge is in black. The 1-min periods of gas exchange are indicated with hatched bars. Lower panel: Raw traces depicting the reduction in WT breathing frequency and increase in minute ventilation during each challenge (mean ± SEM). During each challenge the peak (or zenith) of the response was averaged. Differences among groups were analyzed. (B) Breathing frequency was reduced during each challenge to similar degrees among the three groups (p = 0.16). Data are expressed as percent change from baseline. (C-G) Responses to the three challenges differed among groups: (C) tidal volume (F (2, 16) = 4.46, p < 0.05), (D) minute ventilation (F (2, 16) = 5.01, p < 0.05), (E) duty cycle (F (2, 16) = 5.86, p < 0.05), (F) peak expiratory flow (F (2, 16) = 6.39, p < 0.01), and (G) expiratory duration (F (2, 16) = 6.39, p < 0.01). N = 8 per group. (H) Left panel: A schematic of the severe triple HH challenge with each challenge in black. The 1-min periods of gas exchange indicated with hatched bars. Right panel: Bar graph depicts WT mice fully recovered (n = 7/7), a majority of Kcna1^−/− mice failed to survive (n = 5/7) and DORA pretreatment increased survival of Kcna1^−/− mice (n = 3/3) (Data are expressed as percent survival, X² = 186.0, p < 0.001). Base, baseline; C1, challenge 1; C2, challenge 2; C3, challenge 3; I_D, inspiratory duration. Data are expressed as the mean ± SEM and were analyzed using a mixed model ANOVA with repeated measure. Šidák’s multiple comparisons post hoc test distinguished whether Kcna1^−/− mice differed from WT controls (blue *p < 0.05, **p < 0.01) or Kcna1^−/− DORA mice (green # p < 0.05, ## p < 0.01). Dunnett’s multiple comparisons post hoc test identified whether the response to the last challenge differed from the first challenge within each group; differences only occurred in the Kcna1^−/− cohort (purple + p < 0.05, ++ p < 0.01).

2.4. In vivo signal analyses

Ventilatory waveforms were recorded and measured using IOX software (EMKA Technologies). Parameters including frequency, minute ventilation, tidal volume, peak expiratory flow, and expiratory duration were measured. Parameters were averaged in 10-s bins. Raw respiratory traces were manually vetted. The 10-s segments with noise artifact created by sensor interference due to large movements were identified with video-verification and the corresponding segment was removed prior to analyses. Following baseline stabilization, the average was determined during the last 5 min. During the challenges, once the response peaked, the average was determined. Overall, few segments contained noise artifact and required removal. At baseline few, if any, 10-s segments (0–4 segments) were removed from each subject. During the challenges, it was rare for a segment to contain noise artifact and only four segments total were excluded (including bins from all subjects during all challenges). Data acquired during the 1-min gas flush-in periods, and flush-out periods were not included.

2.5. Drugs

The DORA, TCS1102 (Tocris Biosciences), was dissolved in 100% DMSO (Sigma, St. Louis, Missouri). Mice were pretreated with 100 mg/kg (i.p.; 30–45 μl total volume) 30 min prior to the experiment, a timepoint at which a previous study found brain levels are 2370 nM, plasma levels are 3500 nM, and CSF levels are 43 nM (Bergman et al., 2008). TCS1102 has K_i values of 3 nM for orexin-1 receptor and 0.2 nM for orexin-2 receptor. The clearance rate is 3.7 ml/min/kg (Bergman et al., 2008).

2.6. Ex vivo slice preparation

Mice were anesthetized with isoflurane, decapitated, brains removed and prepared as we have previously described (Simeone et al., 2013, 2014a,b; Warren et al., 2018). Briefly, each brain was immediately placed into an artificial cerebral spinal fluid (aCSF) ice-cold slurry continuously bubbled with 95% O₂/5% CO₂ containing (in mM): 206 sucrose, 2.8 KCl, 8 MgSO₄, 26 NaHCO₃, 1.25 NaH₂PO₄, and 10 glucose, (pH 7.4, 294 mOsm). Coronal slices (400 μm) were prepared on a Leica VT1200 vibratome and transferred to a holding chamber containing the recording aCSF solution in a water bath at 34 °C for at least one hour before recording. The slices were then transferred to room temperature for the experiments.

2.7. Ex vivo multielectrode electrophysiology recording

The lateral hypothalamus was positioned over the 64-electrode grid on a MED64 probe (Alpha Med Systems, Inc., Osaka, Japan). The 1 mm² array has 50 μm² electrodes arranged in an 8 × 8 grid with a 150 μm interpolar distance. Slices were perfused at a rate of 1 ml · min⁻¹ with inline prewarmed (~33.0 °C) aCSF bubbled continuously with carbogen. Solutions were maintained at near-interface levels. Slices were washed with either pH 7.4 aCSF, pH 6.7 aCSF or 7.1 aCSF. Once solutions were switched it took 40 s for the new solution to perfuse into the multielectrode dish and another 80 s for complete bath exchange (as illustrated in Fig. 3A). The pH 7.4 aCSF solution was prepared as follows (in mM): 124 NaCl, 2.8 KCl, 2.4 CaCl₂, 2.5 MgSO₄, 26 NaHCO₃, 1.25 NaH₂PO₄, and 10 glucose (pH 7.4, ~305 mOsm). The pH 6.7 aCSF solution was (in mM): 124 NaCl, 2.8 KCl, 2.4 CaCl₂, 2.5 MgSO₄, 10.4 or 13 NaHCO₃ (for pH 6.7 and 7.1 respectively), 1.25 NaH₂PO₄, 10 glucose and NaCl adjusted to maintain osmolality of ~305 mOsm and pH. Extracellular recordings of multiunit activity were acquired at a 20 kHz sampling rate with a bandwidth of 10 Hz – 10 kHz with Mobius acquisition software (Alpha Med Systems, Inc., Osaka, Japan). To examine the responses to pH in an intact network, and thus better mimic the in vivo situation, synaptic activity was not altered.

Fig. 3.

Ex vivo, effect of sustained exposure to acidosis on putative orexin neuron firing rates. (A) Representative pH levels sampled every ten seconds during the five-minute wash in of pH 6.7 aCSF. (B) Wavemarks depicting the firing rate of an orexin neuron that was stimulated (O_Δ↑) and one that was inhibited (O_Δ↓) by low pH aCSF. Horizontal calibration bar: 2 s. (C) Proportion of O_Δ↑ neurons (empty bars), O_Δ↓ neurons (filled bars), and non-responsive neurons (grey) from WT slices (blue, n = 29 putative orexin neurons) and Kcna1^−/− slices (purple, n = 37 putative orexin neurons). (D) Whether a neuron was inhibited or stimulated (y axis) correlated with its original firing rate at baseline (x axis). (E) The magnitude of change differed between genotypes (F (1, 57) = 5.32, p < 0.05). The firing rates of putative O_Δ↑ increased by a greater magnitude in Knca1^−/− slices (p < 0.05), whereas the magnitude of change for putative O_Δ↓ neurons did not differ between groups. (F) Histogram of the Chemosensitivity Index (CI) for putative orexin neurons from WT (blue) and Kcna1^−/− (purple) slices. A neuron that is not chemosensitive has a CI of ~100. The CI of an inhibited neuron is <100, and of a stimulated neuron is >100. Data from experiments in which the pH was reduced either to 6.7 or to 7.1 (35 WT and 51 Kcna1^−/− putative orexin neurons) (G) Bar graph depicting the average CIs for O_Δ↓ and O_Δ↑ in WT (blue solid) and Kcna1−/− (purple checkered) slices (n = 29 WT putative orexin neurons and 37 Kcna1^−/− putative orexin neurons); *p < 0.05, **p < 0.01.

2.8. Ex vivo low pH challenges

Slices were perfused with pH 7.4 aCSF until a steady baseline was established. Baseline activity was recorded for at least five minutes. Slices were perfused with either pH 6.7 aCSF or pH 7.1 aCSF for five minutes before returning to pH 7.4 aCSF. Another cohort of slices was exposed to a triple challenge during which pH 6.7 aCSF was perfused three times (one minute each), with pH 7.4 aCSF recovery intervals (one minute each). The lateral hypothalamus contains multiple types of neurons including orexin neurons, melanin-concentrating hormone magnocellular neurons and various types of interneurons (Bonnavion et al., 2016; Stuber and Wise, 2016). Our extracellular multielectrode array recording technique does not afford the ability to definitively confirm the anatomical identity of the recorded neurons. However, unique characteristics of orexin neurons allow for identification during extracellular electrophysiological experiments. Specifically, orexin neurons are principal cells, that spontaneously fire action potentials, and are the only principal neurons in the region that express the inhibitory adenosine type 1 receptor (A1R) (Bonnavion et al., 2016; Dergacheva et al., 2016; Liu and Gao, 2007; Stuber and Wise, 2016; Thakkar et al., 2008; Warren et al., 2018). Thus, at the end of the experiment, the A1R antagonist 2-[4-(2,3,6,7-Tetrahydro-2,6-dioxo-1,3-dipropyl-1H-purin-8-yl) phenoxy]-acetic acid (XCC,10 μM made with 10 mM stock in DMSO, Tocris) was applied. Post-hoc signal analysis (described below) clustered waveforms to individual neurons and identified principal neurons; thus, spontaneously firing principal neurons were identified as putative orexin neurons if their firing rate increased following application of XCC as we have described (Warren et al., 2018). Even though a previous study reported that all orexin neurons respond to adenosine inhibition via the A1R (Liu and Gao, 2007), an immunohistochemical study found only 30% of orexin-positive cells were also A1R-positive (Thakkar et al., 2008); thus, it is possible that our methods have not identified all putative orexin neurons.

2.9. Ex vivo signal analyses

Signal analyses were conducted as we have previously described. (Simeone et al., 2013, 2014b; Warren et al., 2018) Traces were imported into Spike2 (v7) (Cambridge, England) and filtered with three IIR notch filters for machine noise and harmonics (60, 180, and 300 Hz) followed by a Finite Impulse Filter (FIR) 300–3000 Hz band pass filter. The standard deviation of the root mean square (RMS) of the noise of the trace was multiplied by four and served as the threshold for spike detection. Principal component analysis was used to separate single units (action potentials) into clusters representing individual neurons. Waveforms of individual neurons were classified as principal cells or interneurons based on spike asymmetry (a / b) versus spike width (Csicsvari et al., 1998; Henze et al., 2002; Quyen et al., 2008; Simeone et al., 2013; Warren et al., 2018).

We determined the degree of chemosensitivity using the chemosensitive index (CI) for each neuron as described by Wang et al. (Wang et al., 1998; Wang and Richerson, 2000, 1999). The CI was calculated as the mean firing rate (FR) induced by a change in pH of 0.2 units; thus, the CI provides a normalized value:

$$CI=100\%x{10}^{\left(\mathit{\text{log}}\left[\frac{F{R}_{\mathit{\text{pHacid}}}}{F{R}_{\mathit{\text{pH}}7.4}}\right]\right)\left(\frac{0.2}{\Delta pH}\right)},$$

where FR_pHacid = mean firing rate at the new acidic pH, FR_pH7.4 = mean firing rate in pH 7.4 aCSF and ΔpH = pH 7.4-pH of acidic aCSF. Accordingly, a neuron that did not change FR during the pH shift would have a CI of 100, whereas a neuron that is inhibited would have a CI of <100 and a stimulated neuron would be above 100; thus, the further away from 100, the more chemosensitive regardless of direction of response. Neurons were deemed chemosensitive if they had a CI > 105 or < 95.

2.10. Statistical analyses

Data were analyzed using Graphpad PRISM v10 and are presented as the mean ± standard error. Differences in in vivo ventilatory responses and ex vivo chemoresponses were determined with a mixed model ANOVA with repeated measures and Sidak’s post hoc test for multiple comparisons. Differences between groups were determined with a Mann-Whitney test to account for non-parametric variance when data did not conform to Gaussian distribution. Differences in survival were assessed with Fisher’s exact contingency analyses.

3. Results

3.1. Experiment #1: In vivo, Kcna1^−/− mice have elevated ventilatory responses to a ‘mild’ triple HH challenge

First, we needed to develop a test. The few instances of clinically recorded SUDEP indicate multiple potential HH challenges (i.e., GCS + multiple apneas) occurred prior to terminal apnea. There was a wide range in the number of challenges and their respective durations. We decided on administering three HH challenges that were each two minutes in duration, and importantly, the recovery intervals were brief. To further optimize the test, the severity of the HH challenges were adjusted until specific criteria were met: the three challenges had to be sensitive enough to detect differences between the epileptic and control groups, and the three challenges had to be mild enough for all subjects to fully recover and survive. These criteria were met when three challenges of 6% O₂ and 9% CO₂ were administered (Fig. 1A).

The reliability of the ventilatory response to the triple HH challenge was determined. During each challenge, the parameters were assessed in 10-s bins and the peak (or zenith) of the response was averaged and plotted. When compared with baseline, the breathing response changed during each challenge. The reduction in breathing frequency during each challenge (F (2,32) = 4.070, p < 0.05) did not differ among groups (F (2, 16) = 2.075, p = 0.16) (Fig. 1B). For both genotypes, this indicated that more time was spent inhaling and exhaling during each breathing cycle and suggested that the larger tidal volume (F (2.59, 41.41) = 120.9; p < 0.01) was the primary contributor to the elevated minute ventilation (F (2.24, 35.77) = 130.0, p < 0.01) (Fig. 1C and D). The increased duty cycle indicated more time was spent inspiring during each breathing cycle (F (1.15, 18.36) = 19.08, p < 0.01) (Fig. 1E). The greater peak expiratory flow (F (2.26, 36.07) = 150.2; p < 0.01) and expiratory duration (F (1.054, 16.87) = 15.05, p < 0.01) also supported more force and time dedicated to expunging CO₂ (Fig. 1F and G).

The magnitude of change for WT mice did not differ from the first to the third challenge indicating a reliable/predictable response. In contrast, the tidal volume (Fig. 1C, p < 0.05), minute ventilation (Fig. 1D, p < 0.05) and peak expiratory flow (Fig. 1F, p < 0.05) were greater in Kcna1^−/− mice during the first challenge, then diminished by the third challenge. While the expiratory duration during the first challenge was normal in Kcna1^−/− mice, subsequent durations were much shorter (Fig. 1G, p < 0.01).

3.2. Experiment #2: In vivo, most Kcna1^−/− mice succumb during a ‘severe’ triple HH challenge

To determine whether the altered ventilatory responses of Kcna1^−/− mice would be detrimental in extreme situations, a more severe in vivo test was developed to assess survival probability. Preliminary experiments varied the severity and duration of HH until the three challenges were calibrated to be mild enough that all control mice survived and severe enough that respiration degraded and failed in most Kcna1^−/− mice. The test which met these criteria used parameters of increasing severity: 6% O₂ + 3% CO₂ for three minutes, 6% O₂ + 9% CO₂ for three minutes, then 6% O₂ + 9% CO₂ for five minutes, with three-minute normoxia recovery intervals (Fig. 1H). A separate cohort of mice was used in this test. All WT mice successfully mounted the necessary responses and survived this triple challenge. In contrast, whereas respiratory parameters were similar for both genotypes during the 6% O₂ + 3% CO₂ challenge (data not shown), respiration degraded and failed in the majority of Kcna1^−/− mice during the prolonged exposure to 6% O₂ + 9% CO₂ with one and four mice expiring during the second and third challenges (2/7 survived), respectively (Fig. 1H), thus ventilatory responses were not analyzed. Together, these data indicate that the abnormal ventilatory responses of Kcna1^−/− mice may contribute to a high rate of mortality during challenges involving HH.

3.3. Experiment #3: In vivo, blockade of orexin receptors restores the ventilatory response of Kcna1^−/− mice to the mild triple HH challenge and promoted survival during a severe triple HH challenge

Previous data indicate that aberrant orexin signaling in Kcna1^−/− mice may contribute to respiratory and blood gas instability (Iyer et al., 2020). Pretreatment with a DORA thirty minutes prior to the mild triple HH challenge did not influence expiratory duration, but did stabilize tidal volume, minute ventilation, duty cycle and peak expiratory flow of Kcna1^−/− mice to levels resembling those of WT controls (Figs. 1C–G). In a separate cohort all Kcna1^−/− mice pretreated with the DORA survived the severe triple HH challenge (Fig. 1H). These data suggest that orexin signaling likely contributes to the volume and inspiration aspects of the Kcna1^−/− abnormal ventilatory response to a mild triple HH challenge. In addition, these data support the notion that aberrant orexin signaling in vivo is pathologic and contributes, in part, to mortality following more severe repeated HH challenges.

3.4. Experiment #4: In response to low pH, stimulated putative orexin neurons in Kcna1^−/− lateral hypothalamic slices are more chemosensitive

Orexin neurons are chemosensitive and modulate the activity of neurons in many brainstem respiratory nuclei, some of which are themselves chemosensitive (Burdakov et al., 2013; Ciriello et al., 2003; de Oliveira et al., 2003; Lazarenko et al., 2011; Nattie and Li, 2012; Shahid et al., 2011; Vicente et al., 2016; Williams et al., 2007; Young et al., 2005). Our in vivo experiments indicated that the orexinergic system is involved in the abnormal respiratory responses of Kcna1^−/− mice. To determine whether the chemosensitivity of orexin neurons of the Kcna1^−/− lateral hypothalamic network differed from WT, coronal slices of lateral hypothalamus were positioned on a multielectrode array (Figs. 2A–B) and spontaneous, extracellular multi-unit activity was recorded (Fig. 2C). Multi-unit activity reflects action potentials from one or more neurons. Wave forms detected by each electrode were first identified as individual neurons via principal component analyses and then sorted into principal cells or interneurons based on spike width and asymmetry (Fig. 2C) (Simeone et al., 2013). Orexin neurons are principal cells that spontaneously fire. Liu et al., previously demonstrated that blocking the inhibitory adenosine type 1 receptors (A1R) disinhibited only orexin neurons in the lateral hypothalamus (Liu and Gao, 2007); thus, at the end of each experiment, the A1R antagonist XCC was bath applied to identify putative orexin neurons (Fig. 2D). Overall, we identified 1344 neurons (WT: mice = 9, slices = 14, neurons = 633, principal cells = 95; Kcna1^−/−: mice = 9, slices = 17, neurons = 711, principal cells = 115). Of the principal cells, 117/210 increased firing in the presence of XCC and were classified as putative orexin neurons (WT = 51, Kcna1^−/− = 66). XCC did not alter the firing rates of interneurons.

Fig. 2.

Multi-electrode array recording in the lateral hypothalamus. (A) The right ventral aspect of a representative coronal slice with the 64-electrode array positioned under the lateral hypothalamus (LH). The amygdala, 3rd ventricle (yellow vertical line), and mammillothalamic tract (m.t.) are indicated. (B) An illustrative overlay of a microelectrode array is drawn on lateral hypothalamic tissue to depict the size and distribution of the extracellular electrodes (white boxes) relative to orexin neurons (green immunoreactive neurons). (C) A raw trace of multi-unit activity detected by one electrode. (Lower) Representative waveforms following principal component analysis distinguishing a principal cell and an interneuron. (D) Wavemarks depicting firing frequency for a putative orexin neuron, a non-orexin principal cell, and an interneuron. The firing rates of putative orexin neurons increased following bath application of XCC. Horizontal calibration bar: 2 s.

Physiologically, exposure to hypoxia can induce acute metabolic acidosis due to lactic acid release (Borch et al., 1993). Prolonged GCS, as seen prior to SUDEP, can reduce brain pH as low as 6.8 through accumulation of lactic acid (metabolic acidosis) and CO₂ (respiratory acidosis) (Ziemann et al., 2008). In the current study, blood pH was decreased after the in vivo severe triple challenge by a mean of 0.2 ± 0.05 pH units with a range of 0.1–0.6 units. Brain/CSF pH closely follows blood pH (Akaishi et al., 2019), thus, following baseline recordings with aCSF pH 7.4, the effect of reducing the bath aCSF pH on putative orexin neuron activity was determined by perfusing aCSF pH 6.7 for five minutes. The pH of the solution perfused into the multielectrode array dish reached pH 6.8 ± 0.1 (Fig. 3A). The intent of these experiments was to determine effects of low pH exposure on the output of putative orexin neurons while the synaptic integrity of entire lateral hypothalamic network remained intact.

Nearly all putative orexin neurons, 93% of WT (27/29) and 92% of Kcna1^−/− (34/37) responded to the decrease in pH. Similar to chemosensitive neurons in other brain regions (Putnam et al., 2004), some were stimulated (O_Δ↑), and others were inhibited (O_Δ↓) (Fig. 3B). Reducing pH inhibited 72% of WT putative orexin neurons compared with 54% of Kcna1^−/− putative orexin neurons (p < 0.01, Fisher’s exact test) (Fig. 3C). The direction of change correlated with the baseline firing rate at pH 7.4 (i.e., O_Δ↑ tended to have slower firing rates in control conditions, whereas O_Δ↓ had faster firing rates; p < 0.05, Pearson’s correlation coefficient test) (Fig. 3D). Notably, the firing rate of Kcna1^−/− O_Δ↑ increased by a greater magnitude when compared to WT O_Δ↑ (WT 57 ± 22% v. Kcna1^−/− 292 ± 95%, p < 0.05) (Fig. 3E). There was no difference in the magnitude of inhibition (WT −55 ± 4% v. Kcna1^−/− −61 ± 5%, p = 0.9). Pooling all putative orexin neurons revealed that the overall effect of pH reduction on WT networks was inhibition, whereas Kcna1^−/− networks experienced stimulation (WT −29 ± 8% v. Kcna1^−/− 76 ± 45%, Mann Whitney test, p < 0.05). In a subsequent experiment, reducing the aCSF to a less severe pH of 7.1 had qualitatively similar, albeit more mild effects (proportion O_Δ↓: WT 67% v. Kcna1^−/− 36%; magnitude O_Δ↑: WT 17% v. Kcna1^−/− 181 ± 61%; magnitude O_Δ↓: WT −36 ± 11% v. Kcna1^−/− −55 ± 14%; n = 6 WT and 14 Kcna1^−/− neurons). Reduction of pH did not alter the firing rates of putative interneurons or putative non-orexin principal cells.

Next, we calculated a chemosensitivity index (CI) for each neuron as described by Wang et al. (1998) and Wang and Richerson (1999, 2000). The CI measure enables the comparison of relative chemosensitivities across reported experiments with varying pH shifts, animals, and brain regions (Putnam et al., 2004). A neuron that is not chemosensitive has a CI of 100, whereas a neuron that is inhibited has a CI of <100, and a stimulated neuron is above a CI of 100. Thus, the further away from 100, the more chemosensitive, irrespective of the direction of the response. Fig. 3F depicts a histogram of the CIs obtained from the experiments in which the pH was reduced (n = 35 WT and 51 Kcna1^−/− putative orexin neurons). The bin size was 10 and CIs of 100 ± 5 were considered insensitive to acidosis (9% of WT and 8% of Kcna1^−/− neurons, respectively). The histogram and the corresponding bar graph of averaged CIs indicated that inhibited WT and Kcna1^−/− putative orexin neurons had equivalent chemosensitivity, whereas stimulated Kcna1^−/− neurons were relatively more chemosensitive than WT (F (1, 75) = 62.26, p < 0.01) (Fig. 3G).

These results suggest that putative orexin neurons had stable responses, albeit in either direction, when exposure was sustained for at least 5 min. However, clinically, there were also short duration apneas with brief recovery intervals (Ryvlin et al., 2013), and our in vivo test which artificially modeled this sequence of events found significant ventilatory differences between WT and Kcna1^−/− mice. Next, we determined whether brief exposures to acidosis ex vivo would elicit responses similar to the 5-min exposure or reveal additional differences. To achieve brief exposures of approximately pH 7.0–7.1 and mimic that which occurs in vivo during brief hypercapnia/hypoxemia and apneas (Dulla et al., 2005; Kintner et al., 2000; Magnotta et al., 2012; Schears et al., 2005), an aCSF solution with a pH of 6.7 was perfused into the bath for one minute and then switched to pH 7.4 at which point the bath pH trough stabilized for 30 s and averaged 7.07 ± 0.01. Slices were exposed to three one-minute low-pH challenges (pH 6.7 aCSF), with one-minute recovery intervals (pH 7.4 aCSF) (Fig. 4A). Neuronal firing rates were determined during baseline, during each trough, and during each peak. Firing rates during each low-pH trough were compared with the firing rate during the prior peak (i.e., during baseline or a recovery). Putative orexin neurons that responded to at least one of these three brief challenges were included. Similar to the long exposure, the first brief exposure stimulated or inhibited putative orexin neurons, with inhibition accounting for 62% (10/16) and 47% (7/15) of WT and Kcna1^−/− putative orexin neurons, respectively (p < 0.05, Fisher’s exact test) (Fig. 4B). In contrast, the magnitude of stimulation and inhibition did not significantly differ between genotypes during the brief exposure (Fig. 4C).

Fig. 4.

Ex vivo, effect of repeated, brief exposures to acidosis on putative orexin neuron firing rates. (A) The pH of the bath aCSF solution surrounding the tissue sampled every 10 s. Wash-in of pH 6.7 aCSF is indicated by the horizontal bars. Firing rates were analyzed during baseline and during each trough (pH 6.7 aCSF) and peak (pH 7.4 aCSF) indicated by the circles. (B) The proportion of O_Δ↑ neurons (empty bars) and O_Δ↓ neurons (filled bars) from WT slices (blue) and Kcna1^−/− slices (purple) during C1. (C) The magnitude of change during C1 did not differ between genotype (F (1, 27) = 1.60, p = 0.22). (D) Stimulation or inhibition during the second or third low-pH challenge (x axis) correlated with whether the neuron was able to recover to its original baseline firing rate (y axis). (E-F) Chemosensitivity Index for putative orexin neurons that were (E) inhibited or (F) stimulated during C1, C2 and C3. N = 16 WT and 15 Kcna1^−/− putative orexin neurons, data are expressed as mean ± SEM, * p < 0.05. C1, challenge 1; C2, challenge 2; C3, challenge 3.

Unexpectedly, there was no response consistency across subsequent low-pH challenges: i.e., O_Δ↑ were likely to transition to O_Δ↓ or not respond, and vice versa. Correlation analysis indicated that whether a neuron was stimulated or inhibited during the second or third low-pH challenge was related to whether the neuron recovered to its original baseline firing rate after the first or second low-pH challenge, respectively (p < 0.01, Pearson correlation coefficient test) (Fig. 4D). That is, if the recovery firing rate overshot the baseline firing rate, then the subsequent low-pH challenge would likely be inhibitory (or the neuron would fire slower than the previous challenge). In contrast, if the recovery firing rate was less than the baseline firing rate, the subsequent low-pH challenge would likely be stimulatory (or the neuron would fire faster than the previous challenge). This suggests that the direction of response of putative orexin neurons may have a dynamic relationship with extracellular and intracellular pH and the brevity of exposures contributed to the individual variability of responses.

Another unexpected finding was that regardless of the inconsistency of responses of individual neurons to brief challenges, the overall chemosensitivity index for inhibitory and stimulatory responses was similar between WT and Kcna1^−/− putative orexin neurons for the first two challenges but differed in the last challenge; thus, by the third challenge both the inhibited and stimulated neurons were more chemosensitive in Kcna1^−/− slices (Fig. 4E and F, respectively).

4. Discussion

The mechanisms contributing to a SUDEP event remain elusive; however, clinical, and preclinical evidence suggest that cardiorespiratory abnormalities may be involved (DeGiorgio et al., 2011; Dhaibar et al., 2019; Iyer et al., 2020; Moore et al., 2014; Ryvlin et al., 2013; Vilella et al., 2019b). Indeed, our previous studies using the Kcna1^−/− mouse model of SUDEP found significant, detrimental alterations to cardiorespiratory parameters that progressively worsened, were associated with SUDEP risk, and may involve a dysregulated orexinergic system (Iyer et al., 2020; Simeone et al., 2018). In the present study, we developed two in vivo gas tests to mimic blood-gas changes associated with GCS and apneas to determine the ventilatory response of Kcna1^−/− mice at high risk of SUDEP, and developed similar ex vivo tests to interrogate the chemoresponse of the orexinergic network. Specifically, there are eight novel findings reported in this study: (i) Kcna1^−/− mice have elevated ventilatory responses during repeated mild HH challenges when compared with WT controls, and (ii) succumb to more severe challenges. (iii) Blockade of orexin receptors stabilizes the Kcna1^−/− ventilatory response and (iv) improves the rate of survival from 29% to 100%. (v) A majority of putative orexin neurons respond to exposure to low pH. Similar to chemosensitive neurons in other brain regions, putative orexin neurons are either stimulated or inhibited. (vi) When compared with controls, a higher proportion of putative Kcna1^−/− orexin neurons are stimulated by low pH and these neurons are more chemosensitive. (vii) Unexpectedly, repeated, brief low-pH challenges can elicit a stimulatory and inhibitory response from the same putative orexin neuron. (viii) Repeated, brief low-pH challenges increase the chemosensitivity index of Kcna1^−/− putative orexin neurons.

4.1. The HH ventilatory response and SUDEP

Our previous studies demonstrated that Kcna1^−/− mice at high risk of SUDEP experience chronic intermittent oxygen desaturation associated with seizures and apneas-hypopneas, but also occurred in the absence of seizures/apneas (Iyer et al., 2020; Simeone et al., 2018). Moreover, Kcna1^−/− mice exhibited abnormal breathing patterns including hyperventilation-induced apnea which occurs in people with central apnea or respiratory failure and is attributed to a chemoresponsive instability (Munemoto et al., 2013; Pinsky, 2015; Xie et al., 1997, 1994). Collectively these data suggest Kcna1^−/− mice struggle to stabilize blood gases.

Thirty percent of GCS and consequent apneas are accompanied by severe peri-ictal hypoxemia and hypercapnia that persist anywhere from 60 s to 30 min (Bateman et al., 2008; Bruno et al., 2018; Farrell et al., 2016; George et al., 2023; Seyal et al., 2010). During a GCS patients often stop breathing due to tonic muscle paralysis (Azar et al., 2008; Carmenate et al., 2020; Devinsky, 2004). The resulting hypoxemia and hypercapnia trigger a chemoresponse that increases minute ventilation to expel excess CO₂. This hypercapnia/hypocapnia oscillation lessens as CO₂ normalizes and eupnea returns. If too much CO₂ is expelled it leads to respiratory alkalosis and hypocapnia-induced apneas to increase CO₂ again (this is clinically apparent as post-convulsive central apnea, PCCA). In the presence of severe metabolic acidosis, as happens with seizure-mediated production of lactic acid and hypercapnia, breathing becomes deep and labored/stertorous (also called Kussmaul breathing, a form of hyperventilation) for several minutes until blood gases stabilize and eupneic breathing returns (Azar et al., 2008; Carmenate et al., 2020). Clinical observations of SUDEP events recorded that the terminal generalized tonic-clonic seizure was followed by a brief period of ventilation, bouts of apnea and finally complete respiratory arrest (Ryvlin et al., 2013). This suggests an inability to stabilize blood gases during the SUDEP event. To determine whether the respiratory abnormalities of high risk Kcna1^−/− mice would be detrimental during an HH event, we developed two gas tests to mimic the sequence of HH prior to SUDEP. In our mild test, we found that the ventilatory response of both genotypes decreased breathing rate and increased minute ventilation, tidal volume, duty cycle, peak expiratory flow and expiratory duration. This slower and deeper ventilation with more forceful expiration is compensatory, similar to Kussmaul breathing, and would serve to expel CO₂ in an attempt to stabilize blood gases. The differences between genotypes were in the magnitude and the sustainability of the response. Kcna1^−/− mice had greater ventilatory responses to the initial challenge, which degraded with subsequent HH challenges, whereas the WT response remained stable for each challenge. In particular, the Kcna1^−/− expiratory duration was significantly decreased by the third challenge indicating reduced ability to expel CO₂ and possible blood gas instability. Nevertheless, all mice successfully recovered. This was not the case when the duration of each HH challenge was increased in our severe test. Kcna1^−/− mice struggled to maintain the compensatory breathing pattern and succumbed to respiratory arrest, whereas all WT mice recovered.

To our knowledge this is the first report of HH-induced ventilatory responses in epileptic animals. Clinical and preclinical epilepsy studies investigating either the hypoxia ventilatory response (HVR) or hypercapnia ventilatory response (HCVR) have reported conflicting results: HVR, mediated by peripheral chemoreceptors in the carotid body, is increased and HCVR, mediated by central chemoreceptors in the hypothalamus and brainstem, is either increased or decreased (Sainju et al., 2021, 2019; Teran et al., 2023). One study found that Kcna1^−/− mice have increased HVR and normal HCVR, however, it is impossible to estimate the relationship to SUDEP risk at the time of the experiment because the study did not specify the background strain, seizure incidence, or age of their Kcna1^−/− mice (Kline, 2005). Ongoing studies will address this gap in knowledge. A recent MRI study of people with epilepsy found a hypercapnia challenge increased BOLD signals and connectivity between chemosensitive regions in the brain stem and lateral hypothalamus supporting the notion of increased HCVR chemosensitivity in epilepsy (Hampson et al., 2022). The discrepancies of the impact of epilepsy on HCVR may be due to differences in epilepsy/seizure type, but also may be due to differences in oxygen content during challenges. HCVR studies are often conducted with normoxia or varying degrees of hyperoxia (50–95% O₂) co-administration (Brugniaux et al., 2018; Chua et al., 1996; Ciarka et al., 2005; Hampson et al., 2022; Kline, 2005; Kouchi et al., 2022; Ott et al., 2019; Sainju et al., 2021, 2019; Teran et al., 2023). Hyperoxia is used to silence the effects of peripheral chemoreceptors which send excitatory signals to central chemosensory regions, such as the nucleus tractus solitarius, to increase ventilation. However, after a seizure and during an actual SUDEP event hypoxemia and hypercapnia occur concurrently presenting a more complicated ventilatory response as the peripheral and central chemoreceptors interact and influence the final response (Bruno et al., 2018; Farrell et al., 2016; Forster and Smith, 2010; George et al., 2023; Giannoni et al., 2017; Laffey et al., 2003, 2000; Moseley et al., 2010; Nattie and Li, 2012; Sainju et al., 2019; Vannucci et al., 2001; Vilella et al., 2019b, 2019a; Wong-Riley et al., 2013; Yang et al., 2003).

4.2. The role of orexin in the HH ventilatory response

The orexinergic system plays a significant role in the basal respiratory abnormalities of Kcna1^−/− mice (Iyer et al., 2020). Orexin neurons project to all respiratory nuclei including rostral ventrolateral medulla, medullary raphe, locus coeruleus and nucleus tractus solitarius (Date et al., 1999; Lazarenko et al., 2011; Marcus et al., 2001; Peyron et al., 1998; Young et al., 2005). These regions contain neurons that control the timing and air flow of inspiration and expiration. SUDEP risk in Kcna1^−/− mice is associated with an increase in the number of lateral hypothalamic neurons expressing orexin, and DORA treatment attenuates the basal respiratory abnormalities and increases longevity of Kcna1^−/− mice (Iyer et al., 2020; Roundtree et al., 2016). Orexin neurons are chemoresponsive and are considered important central contributors to HVR and HCVR. For example, Nakamura et al. found that HCVR was diminished in orexin-neuron-ablated mice and other studies report that pharmacological blockade of orexin receptors or genetic knock out of prepro-orexin attenuates HVR and/or HCVR (Kuwaki, 2008; Li et al., 2016; Li and Nattie, 2010; Nakamura et al., 2007). Moreover, intracerebroventricular injection of orexins A and B into prepro-orexin knockout mice partially restores the HCVR (Nakamura et al., 2007). In the present study we found that acute administration of a DORA attenuates the elevated ventilatory response of Kcna1^−/− mice to control levels during the mild HH test and promotes survival of all Kcna1^−/− mice during the severe HH test. Collectively, these data support the notion that the elevated ventilatory response is due to activity of central chemoreceptors and that the orexinergic system is hyperactive in the Kcna1^−/− mouse model of SUDEP.

There is evidence that chronic intermittent hypoxemia, similar to that seen in Kcna1^−/− mice, increases prepro-orexin mRNA expression in the hypothalamus of rats (Liu et al., 2014). Overexpression of orexin in Kcna1^−/− mice (Roundtree et al., 2016) may be a compensatory mechanism to increase the respiratory drive to stabilize blood gasses. In this light the elevated chemoresponse we found here and detrimental effects on sleep and seizures we previously described (Roundtree et al., 2016), are unintended side effects. Increased orexin has been found in patients with hypercapnic respiratory arrest and obstructive sleep apnea (Igarashi et al., 2003; Zhu et al., 2011); however, in epilepsy there are conflicting reports (the reader is referred to the review by Berteotti et al., 2023). It would be interesting to determine orexin levels in patients at high risk for SUDEP. We predict the results may be more consistent.

4.3. Chemosensitivity and orexin neurons

Recently Goyne and Bayliss proposed strict criteria for identifying types of cells as central chemoreceptors involved in HVR or HCVR: (1) Activation and inhibition of the candidate cell group have opposite effects on respiration; (2) inhibition of candidate cells blunts the respiratory response to CO₂; (3) cell activity in vivo tracks pH or PCO₂; (4) CO₂/H⁺ modulation of cell activity is a direct effect, at least in part; and (5) interfering with the specific molecular mechanism(s) by which a cell senses CO₂/H⁺ inhibits the normal hypercapnic ventilatory response (Gonye and Bayliss, 2023). Accordingly, hypoxia and hypercapnia significantly increase c-Fos expression in a subset of orexin neurons in in vivo animal studies, and the studies mentioned above demonstrate the effect of orexin receptor antagonists and orexin neuron ablation on HVR and HCVR (Spinieli et al., 2022; Sunanaga et al., 2009). An early ex vivo electrophysiological study using extracellular and whole-cell, patch-clamp techniques reported that hypoxia alone and hypercapnia alone either stimulated, inhibited or had no effect on neurons in the caudal hypothalamus (Dillon and Waldrop, 1992). Subsequent studies used transgenic mice expressing GFP in orexin neurons to specifically record from identified orexin neurons. One study found that hypercapnia and acidosis stimulate orexin neurons, whereas the other reported hypoxia and HH/acidosis were inhibitory (Dergacheva et al., 2016; Williams et al., 2007).

Two potential mechanisms for the stimulatory effect of acidosis are TWIK acid sensing postassium leak (TASK) channels and voltage-insensitive proton-gated acid sensing cationic channels (ASICs). Both are expressed by orexin neurons and there is a fair amount of evidence for their involvment in the orexin contribution to HCVR (Burdakov et al., 2006; Song et al., 2012; Wang et al., 2021). Focal infusion of a TASK1/3 channel inhibitor into the lateral hypothalamus reduced phrenic nerve discharge at baseline and attenuated the inhibitory effects of alkaline pH on respiration (Wang et al., 2021). Similarly, focal acidification of the lateral hypothalamus increased phrenic nerve discharge and respiration, which was counteracted by either orexin neuron ablation or TASK1/3 inhibitors. ASIC1a may also contribute to orexin neuron chemosensitivity as ASIC1a inhibition abolished pH effects on phrenic nerve discharge (Song et al., 2012). In contrast, another study found that TASK1/3 knockout mice had orexin neurons that responded to pH changes ex vivo and concluded that these channels may not be essential for chemosensitivity (González et al., 2009). The mechanisms mediating the inhibitory effects of hypoxia and HH/acidosis remain vague, but may involve calcium-activated potassium channels, ATP-dependent potassium channels, a reduction of glutamatergic neurotransmission and/or internalization of TASK1/3 channels (Dergacheva et al., 2016; Feliciangeli et al., 2010; Inoue et al., 2020). Further studies regarding mechanisms, as well as addressing the lack of in vivo evidence for pH and PCO₂ directly influencing orexin neuron activity, are necessary before orexin neurons meet the criteria of Goyne and Bayliss and are considered direct chemoreceptors.

Regardless of whether the chemosensitivity of orexin neurons is direct (i.e., intrinsic) or indirect (i.e., influence of other chemosensitive excitatory, inhibitory, or glial cells), the present study aimed to determine the chemosensitivity of putative orexin neurons within an unfettered lateral hypothalamic network; therefore, synaptic neurotransmission was not blocked. A limitation of the extracellular multielectrode array recording technique is that we were required to rely on anatomical location, spontaneous firing, and sensitivity to an A1R antagonist to identify putative orexin neurons; however, the advantage of is that it enabled us to record many neurons simultaneously and capture the integrated chemosensitive output of the functioning lateral hypothalamic network.

In vivo hypoxemia alone initiates hypercapnia and acidosis, whereas HH deepens the effect on pH; therefore, we chose to use significant pH drops as perturbations in our ex vivo experiments. In contrast to the two patch clamp studies mentioned above in which all orexin neurons responded in the same, albeit opposing, manner (Dergacheva et al., 2016; Williams et al., 2007), we found lowering pH results in stimulation and inhibition of putative orexin neurons. This finding aligns with the mixed responses reported by a number of electrophysiological studies examining chemosensitive neurons in brainstem respiratory nuclei (Putnam et al., 2004). It is also consistent with the finding that <50% of orexin neurons co-label with cFos after a hypoxic or hypercapnic challenge (Spinieli et al., 2022; Sunanaga et al., 2009). Overall, more Kcna1^−/− putative orexin neurons were stimulated by acidosis (38% v. 21% WT) and the magnitude of stimulation was significantly greater than WT. Wang et al. developed a method to quantify chemosensitivity regardless of response directionality, degree of acidosis or skewed response magnitudes, which are commonly associated with initial firing rates (Wang et al., 1998; Wang and Richerson, 2000, 1999). This chemosensitivity index (CI) has been used to compare data from different brain regions, experimental preparations, species, and laboratories (Putnam et al., 2004). Here, the CI for WT putative orexin neurons was 115, similar to chemosensitive neurons in the locus coeruleus (CI = 112–135), but far lower than neurons in other regions such as the retrotrapezoid nucleus (CI = 330) (Putnam et al., 2004).

This difference in CI is logical given the function and projection profile of each region. The retrotrapezoid nucleus innervations remain localized to regions of the brainstem that contain the respiratory pattern generator, thus its main function is regulating respiration (Guyenet et al., 2012). In contrast, adrenergic neurons of the locus coeruleus and orexin neurons send projections throughout the brain and participate in diverse physiological functions (Nattie and Li, 2012; Nikolenko et al., 2024). In this regard, a lower chemosensitivity is advantageous so not to continually disrupt many physiological functions and reserve wide-spread mobilization for emergency situations. For example, orexin plays a prominent role in arousal from sleep, however, it is unnecessary to wake when breathing naturally slows during NREM sleep and there is a gradual buildup of CO₂ which facilitates transition into REM sleep (Madan and Jha, 2012); but, is necessary to wake up from potentially dangerous sleep apnea.

Kcna1^−/− putative orexin neurons have an increased CI of 164, higher than the nucleus tractus solitarius and on par with nuclei of the ventral medulla (Putnam et al., 2004). This suggests that Kcna1^−/− putative orexin neurons will have heightened chemoresponses that disrupt multiple functions. Accordingly, in the present study we demonstrated that Kcna1^−/− mice have an elevated chemoresponse and in previous studies we found Kcna1^−/− mice have abnormal cardiorespiration and severe sleep deficiency involving more sleep-wake transitions and diminished REM sleep (Iyer et al., 2020, 2018; Roundtree et al., 2016; Simeone et al., 2018; Warren et al., 2018). Currently, the reason for the increased chemosensitive activity of Kcna1^−/− putative orexin neurons is unknown. The simplest explanation is the lack of the Kv1.1alpha subunit renders the orexin neurons hyperexcitable as in the hippocampus (Simeone et al., 2013); however, expression of Kcna1 mRNA is minimal in the hypothalamus of WT mice (Allen Mouse Brain Atlas, https://mouse.brain-map.org/gene/show/16258, accessed April 26, 2024). Future studies will address the knowledge gap regarding the mechanisms underlying the changes in chemosensitivity.

The repeated apneas preceding SUDEP occur in quick succession. When we attempted to mimic this with three brief exposures to acidosis in our ex vivo preparation, we observed that the CI for putative orexin neurons of both genotypes increased dramatically from 110 to 160 for WT and 130 to 223 for Kcna1^−/−. Similarly, the number of orexin neurons co-labeled with cFos increases with repeated hypoxia compared to a single hypoxic event (Yamaguchi et al., 2015). This may represent a means to react quickly with a strong ventilatory response; although the Kcna1^−/− chemoresponse is assumed too strong and becomes detrimental.

An unexpected finding with this experiment was that the chemosensitive activity of putative orexin neurons readily switched from stimulation by acidosis in one brief exposure to inhibition in the next exposure. This is difficult to rectify, but a possible explanation is that TASK channels can undergo rapid, g-protein mediated internalization which would attenuate the cell’s response (Feliciangeli et al., 2010; Inoue et al., 2020). However, inhibited neurons also changed to stimulated neurons which would require re-insertion. It is unlikely that both could occur on less than a minute time-scale. During the longer 5 min acidosis exposure whether a neuron was inhibited or stimulated correlated with its firing rate (i.e., faster neurons were inhibited and slower neurons were stimulated). During the brief exposures, a similar relationship occurred between whether the firing rate returned to baseline upon reintroduction of pH 7.4 and the response to the subsequent acidosis. Most likely this was due to the brevity of the exposures causing the local pH to constantly fluctuate. Together, these data introduce the interesting concept that the directionality of the acidosis response is dependent on internal firing rate threshold of putative orexin neurons; a concept which requires further study. Regardless of the mechanism, quick, repeated acidosis increased chemosensitivity of putative orexin neurons.

5. Conclusion

Outside epilepsy, early mortality of patients with chronic heart failure correlates with increased chemosensitivity to hypoxia and/or hypercapnia (Giannoni et al., 2009). Overall, our data indicate that individuals at high risk for SUDEP could be exceptionally sensitive to hypoxia-hypercapnia challenges, whether induced by seizures or other means; and the depth and length of time of HH could dictate the probability of survival. The novel in vivo tests presented here need to further validatation in additional preclinical models but provide preclinical support for the development of mild HH clinical tests, which are needed to improve the accuracy and personalization of SUDEP risk.

Data availability

Data will be made available on request.