A new physiologically realistic and clinically relevant model of sleep apnoea for investigating it’s effect as a comorbidity on neurodegenerative disease


 Sleep apnoea is a highly prevalent disease but often goes undetected and is associated with poor clinical prognoses when combined with many different disease states. However, most animal models of sleep apnoea (e.g., intermittent hypoxia) have recently been dispelled as physiologically unrealistic. Due to a lack of appropriate models, little is known about the causative link between sleep apnoea and it’s co-morbidities. To overcome these problems, we have created a realistic animal model of moderate sleep apnoea by reducing the excitability of the respiratory network. This has been achieved through controlled genetically-mediated lesions to the preBötzinger Complex (preBötC), the inspiratory oscillator. This novel model shows increases in sleep disordered breathing with alterations in breathing during wakefulness (decreased frequency and increased tidal volume) as observed clinically. The increase in apnoea episodes leads to a reduction in REM sleep, with all lost active sleep being spent in an awake state. The increase in hypoxic and hypercapnia insults leads to both systemic and neural inflammation. Alterations in neurophysiology, an inhibition of hippocampal long-term potentiation (LTP), reflect deficits in both long and short term spatial memory. This new physiologically relevant and clinically realistic model of sleep apnoea may be the key to understanding why sleep apnoea has such far reaching and often fatal effects on end organ function.


Introduction 27
Current estimates of the prevalence of moderate to severe (≥15 events.h -1 ) sleep apnoea (SA: >2 28 missed breaths during sleep) are ~4% of men and ~2% of women in the UK. However, recent large 29 cohort studies show that 91% of SA cases are undetected 1 , with an occurrence of ~50% of men and 30 ~25% of women over 40 2 . 31 During Rapid Eye Movement sleep (REM), individuals experience increased upper airway resistance 3 32 due to tonic motor inhibition lowering the muscle tone 4 . This leads to increased vulnerability to 33 repetitive pharyngeal collapse and a cessation in airflow even in the presence of respiratory effort, 34 otherwise known as obstructive sleep apnoea (OSA) 5 . During non-Rapid Eye Movement sleep (NREM), 35 a loss of descending inputs into the respiratory centres, means control of breathing is solely down to 36 metabolic factors, and breathing is more reliant on chemoreception 4 . Thus any defects in the 37 respiratory microcircuit will be exposed leading to a loss of drive to the respiratory muscles, termed 38 central sleep apnoea (CSA) 4 . Whilst it is believed that OSA is a REM related disordered with CSA is 39 linked to NREM, both CSA and OSA occur during either NREM or REM sleep 6,7 . In all cases, patients 40 experience intermittent hypoxic hypercapnia following cycling periods of apnoea and hyperpnoea 8 . 41 The blood/gas changes lead to oxidative imbalance facilitating the generation and build-up of reactive 42 oxygen species (ROS) leading to the development of systemic inflammatory-related biomarkers 43 associated with chronic inflammation 9 . Molecular markers of systemic inflammation are increased in 44 patients with SA 10 , leading to an increased risk of diseases with an underlying inflammatory 45 component, e.g., cardiovascular disease (CVD) 11 , diabetes 12 , and dementia 13 . Systemically, this can 46 lead to heart failure (HF) 14 , with mortality directly related to SA severity 15 and SA treatment improving 47 mortality 16 and morbidity 17 . As SA causes endothelial dysfunction through inflammation, and 48 endothelial dysfunction is capable of impairing cerebral blood perfusion 18 . In addition, an increase in 49 hypertension 19 and atherosclerosis 20 in SA patients leads to recurrent strokes 21 . Furthermore, the 50 hypercapnic hypoxia leads directly to oxidative stress and neuroinflammation 22,23 . All of these brain 51 insults individually or in combination lead to decreased hippocampal and grey matter volume 22,23 52 resulting in memory deficits, and cognitive decline 13,23 . 53 Current animal models of SA used to study these pathways and associations are severely flawed: 54 Chronic intermittent hypoxia, replicates the hypoxia observed during SA, but increased compensatory 55 breathing leads to hypocapnia, the opposite of the clinical condition. Airway occlusion replicates the 56 blood gases changes that occur during apnoea, but the frequency and duration of the apnoeas do not 57 replicate the clinical presentation. In both of these models the simulated "apnoeas" continue after 58 arousal or occur during wakefulness, which initiates the defence response which is physiologically 59 unrealistic, and will contribute to any of the related comorbidities. Here we have developed a 60 physiologically realistic model of SA, that only displays apnoeas during natural sleeping patterns. 61 Furthermore, our model displays both systemic and neural inflammation and exhibits cognitive 62 deficits, making it clinically relevant. This novel physiologically realistic and clinically relevant model of 63 moderate sleep apnoea can be utilised for investigating the effect of SA on associated conditions such 64 as neurodegenerative disease 65

Rats with sleep apnoea have neurophysiological changes resulting in reduced LTP
basal synaptic transmission appears unaffected by apnoea. To investigate long term potentiation (LTP) 129 we used theta burst stimulation (TBS, see methods). SA rats showed decreased LTP following theta 130 burst stimulation. Both shams (mean difference -0.6; std error 0.1, DF = 26; p = 0.00001) and SA rats 131 (mean difference -0.5; std error 0.1, DF = 26; p = 0.00003) showed a significant short term potentiation 132 immediately following TBS (Fig 4C,D). However, the short term potentiation was not converted into 133 long term potentiation in SA rats as the change in slope was reduced after  Fig. 6A-B). This was not confounded by anxiety or locomotor issues as rats 151 covered the same distance (Sham: 15 ± 5 m, n = 7 vs DTA: 17 ± 5 m, n = 9; p = 0.4; not shown). 152 Rats with sleep apnoea have deficits in long term memory 153 Whilst the SA rats displayed impaired short term memory, we next investigated whether they also 154 showed deficits in the conversation of short to long term memory.  (Fig. 7), with an 800% increase in IRF1 following IFNγ induced 194 M1 polarisation of macrophanges 38 . IFR1 induces transcription of caspase 1, which cleaves pro-  to produce the active IL-1β protein 39 (Fig. 7); therefore modulation of IL-1β is via post transcriptional 196 modification rather than increased transcription, hence IL-1β protein is increased even though 197 paradoxically it's mRNA is diminished. Caspase 1 further aids IL-1 β release, by cleaving Gasdermin D 198 (GSDMD) to form the pore responsible for IL-1β release 39 (Fig. 7). Once released, IL-1β induces the 199 secretion of MIP2 and MIP1β from paranchemyal cells 40 . In parallel, IFNy induces both MIP1α and 200 MIP1β release from macrophanges 41 , likely in a caspase 1 dependent manner 42 (Fig 7). 201 Interestingly, whilst IFNy activation leads to IL-1β and MIP secretion, it cannot by itself induce release 202 of IL-6 41 , RANTES 43 , or MCP-1 44 . Given that IL-6 is responsible for inducing CRP release from the liver 45 , 203 it is therefore not surprising that plasma levels of CRP also remained unchanged. In conjunction, whilst 204 IFNγ activated macrophages release TNFα 46 , this is inhibited by hypercapnia 47 , hence TNFα would not 205 be expected to be elevated in response to moderate sleep apnoea. Finally, cleavage of pro-IL-1α to IL-206 1α, occurs through a calpain dependent, caspase independent, pathway 39 . Given calpain is inhibited 207 by oxidative stress 48 , whilst caspase 1 is activated by it 49,50 , would provide a basis for why IL-1α and IL-208 1β are differentially regulated during sleep apnoea. 209 Hypoxia plays an important role in sterile information. Lowered O2 increases plasma IFNy 51 , and leads 210 to toll like receptor 4 (TLR4) 52 and MIP1a receptor 53 upregulation and activation. The increased IFNy 211 leads to IRF1 dependent inflammation which is enhance by hypoxia induced ROS stimulation of the 212 TLR4 pathway 54 . Furthermore, upregulation of MIP 55 and TLR4 56 will enhance caspase 1 activity 213 through NFKB activation, and amplifying cleavage pro-IL-1β to it's active form 49 (Fig 7). Hence IFNy-, 214 IRF1-, Caspase-, TLR4-, and MIP-all drive microgliosis 49,57 . Co-activation of TLR4 and the IFNy receptor 215 leads to significant neuronal dysfunction and cell death 58 . With neuronal cell death increased by 216 hypoxia induced caspase activity 50 . Therefore hypoxia, leads to IFNy induced microglial activation 217 which stimulates IRF1 to induce caspase activity to cleave pro-IL-1β, a pathway enhanced by activation 218 of TLR4 and MIP1α by ROS (Fig 7), leading to microgliosis, and neuronal damage. These effects on 219 neural inflammation may be why, IFNy is inversely correlated with cognitive ability 59 and grey matter 220 volume in the dorsal hippocampus 59 , and may therefore contribute to grey matter loss and mild 221 cognitive impairment (MCI) in SA patients 60 . 222 Microglial activation induces tau phosphorylation 61 in neurons, and leads to conversion of MCI to both 223 AD and VaD 62 . In addition to Tau phosphorylation sleep apnoea also contributes to Alzheimer's disease 224 (AD) through greater amyloid burden 63 . Sleep apnoea is independently associated with increased 225 dementia and MCI such as reduced memory and executive function 64,65 , with executive function being 226 more vulnerable than memory, or core intellectual and verbal abilities 66,67 . With treatment of sleep 227 apnoea being able to partial reverse this cognitive decline 68,69 . The loss of REM sleep, the activation of 228 microglia, and the direct effects of inflammation on hippocampal size and function, may be why sleep 229 apnoea has such profound effects on the brain, and the deficits of long and short term memory we 230 saw in our rats. 231 AAV-9: syn-DTA-GFP (AAV:syn-DTA) at a titre of 3.87x10 13 VP·ml -1 was aliquoted and stored at -80°C. 313 On the day of injection, viruses were removed and held at 4°C, loaded into graduated glass pipettes 314 (Drummond Scientific Company, Broomall, PA, USA), and placed into an electrode holder for pressure 315 injection. The AAV:syn-DTA used the synapsin promoter, transducing neurons with higher tropism for 316 the AAV 2/9 subtype, it did not transduce non-neuronal cells. 317

Viral transduction of preBötC neurons 318
Adult male Sprague Dawley rats (320-560 g) were anesthetized via intramuscular injection with 319 ketamine (100 mg·kg -1 ; Covetrus, Dumfries, UK) and medetomidine (250 µg·kg -1 ; Covetrus, Dumfries, 320 UK). Adequate anaesthesia was maintained with 0.5-2% Isofluorane (primal health care, India) in pure 321 oxygen (1 L·min -1 ) throughout the surgery as need. Rats received a presurgical subcutaneous injection 322 of atropine (120 µg·kg -1 ; Westward Pharmaceutical Co., Eatontown, NJ, USA) and meloxicam (2 mg·kg -323 1 ; Norbrook Inc., Lenexa, KS, USA). Rats were placed in a prone position into a digital stereotaxic 324 apparatus (Kopf Instruments, Tujunga, CA, USA) on a heating pad (TCAT 2-LV: Physitemp, Clifton, NJ, 325 USA) and body temperature was maintained at a minimum of 33°C via a thermocouple. The head was 326 angled so that the nose bar was -18mm below the intra-aural line. The injection arm was angled at 327 23°. Graduated glass pipettes containing the virus were placed stereotaxically into the preBötC (Fig  328  8A). The preBötC was defined as the area ventral to the semi-compact nucleus ambiguous 329 (coordinates: ±1.95 mm lateral and -0.2 mm rostral and +0.4 mm caudal, and -2.95 mm ventral from 330 the Obex; Fig 8A). The virus solution was pressure injected (~250-350 nL per side) bilaterally into all 4 331 sites. Pipettes were left in place for 3-5 minutes to prevent back flow of the virus solution up the 332 pipette track. Postoperatively, rats received sub-cutaneous injections of buprenorphine (100 µg·kg -1 ; 333 Reckitt Benckiser, Slough, UK) and atipamezole (1 mg·kg−1). For sham surgeries pipettes were 334 lowered 2mm dorsoventral from the obex, with the same mediolateral and rostral co-ordinates. Rats 335 were allowed 2 weeks for recovery and viral expression, with food and water ad libitum. 336 For preliminary experiments to determine the injection number and volume rats were injected with 337 ~300 nL of AAV:syn-DTA either at unilaterally at the aforementioned rostrocaudal coordinates or 338 bilaterally in the aforementioned rostral coordinates. 339

EEG and EMG placement 340
Following viral injections the head was levelled so the nose bar was set to 0 at the intraural line. rostral to the lamboidal suture in a position on the right of the sagittal suture to the most lateral extent 347 of the parietal bone (Fig 8b). EMGS were placed bilaterally into the trapezius muscles. Electrodes were 348 connected to custom built headmounts. Headmounts were adhered to the skull with superbond 349 dental cement (prestige dental, Bedworth, UK), with additional support from Vertex Orthoplast cold-350 curing orthodontic acrylic resin (prestige dental, Bedworth, UK) 351 Rats were placed into a custom-made 4.5 L plethysmography chamber (Fig 8C), with an airflow rate of 353 2 L·min -1 . Pressure transducer signals were amplified and filtered using the NeuroLog system 354 (Digitimer, Welwyn Garden City, UK) connected to a 1401 interface. All data were acquired on a 355 computer using Spike2 software (Cambridge Electronic Design, Cambridge, UK). In humans sleep-356 disordered breathing is defined as perturbed breathing for ≥10 secs, which is ~2 breaths. In our 357 experiments rats respiratory rate was 105 breaths·min -1 , or 2 breaths every 1.15 seconds. For our 358 analysis apnoeas were defined as an absence of breathing for ≥1.3 seconds, and hypopnoeas by a drop 359 in minute ventilation of ≥50% over the same time period. Apnoeas and hypopneas were scored by 360 hand, and the length of each event recorded. Respiratory parameters were measured during quiet 361 wakefulness at the beginning of the recording. Airflow measurements were used to calculate: tidal 362 volume (VT), signal trough at the end of expiration subtracted from the peak signal during inspiration, 363 converted to mL following calibration, and standardised to body weight); and respiratory frequency 364 (f) calculated as breaths per minute using peak of inspiration to peak of inspiration averaged over 1 365 minute. Minute ventilation (VE) was calculated as VT x f. 366

Sleep-Wake recordings 367
Delta waves and neocortical EEG activity were obtained by differential recordings from electrodes 3 368 and 4 (Fig. 7B). Theta rhythms were obtained by differential recordings from electrodes 1 and 2 369 positioned in the dorsal hippocampus and the frontal cortex respectively (Fig. 7B). EEGs were 370 bandpass filtered 0.5Hz-60Hz. EMGs were recorded by differential recordings through the trapezoid 371 wires (Fig. 7B) and bandpass filtered at 70Hz-560Hz. Videos were recorded to aid with sleep-wake 372 scoring. All data were acquired on a computer using Spike2 software (Cambridge Electronic Design, 373 Cambridge, UK). EEG and EMG signals were processed using the OSD4 script: EEGs were band filtered 374 with a power spectra of 2-4.5 Hz (delta) and 4.5-9 Hz (theta). EMGs were smoothed over 1 second. 375 The 3 hour long recordings were separated in 5 second epochs. Sleep-wake scoring epochs were 376 categorised into 4 predefine classes:1) WAKElow amplitude desynchronized EEG activity and high 377 EEG power; 2) NREM -high amplitude EEG delta waves and low EMG power; 3) REM -Continuous low 378 amplitude EEG theta activity leading to high theta:delta (T:D) ratio, and no EMG power, muscular 379 paralysis could be observed on the associated video; 4) DOUBT -Epochs that could not be defined 380 into the previous 3 categories. 381

Barnes Maze 382
The Barnes maze is a 122cm diameter circular maze with 20 x 90cm holes; 19 false holes and one 383 escape hole (Stoelting, Dublin, Ireland) (Fig 8E). The maze is one metre off the ground. Spatial cues of 384 different colour and shape are spaced evenly around the maze at regular intervals in plain sight of the 385 rat. The surface of the table is brightly lit (>1500 lux) to create an adverse environment. Rats 386 underwent a 17 day Barnes maze protocol: Days 1-3 (learning phase) the escape hole (randomly 387 assigned on day 1) contained an incentive (peanut butter); days 4-12 (acquisition phase) the incentive 388 is provided in the home cage after the test, allowing the rat to be rewarded for completing the test 389 but removing the olfactory stimulus from the maze; days 13-17 (reversal phase) the escape hole is 390 placed at 180°, this phase is to test the ability of the rats to unlearn a task. The task was completed 391 when the mouse entered the exit box. All runs were recorded using The spontaneous spatial novelty preference test was conducted using a radial arm maze (Stoelting, 401 Dublin, Ireland), adapted into a Y-maze. Each arm was 50 cm long, 10 cm wide, with 13-cm-high walls 402 (Fig 8D). All rats were placed into the same entry arm. For the first trial, one of the arms of the Y-maze 403 was blocked, therefore rats could either go left or right according to a pseudorandom sequence. Rats 404 could also can move in the central region. First trial lasted for 10 mins. After 1 hour, rats underwent a 405 second trial to assess short term memory. The test was repeated with access to all arms. All runs were 406 recorded using a camera system (Henelec Model 335 BWL SONY) attached to a computer for offline 407 analysis (Any-MAZE v4.96, Stoelting). Total distance, duration, and number of entries into each arm 408 were measured. An entry into an arm was defined as placement of 2 paws into that arm. 409 Immunology 410 Rats were anesthetized with Isofluorane (0.5-2%; primal health care, India) in room air (1 L·min -1 ). 411 Inferior vena cava blood samples were taken and placed on ice before being spun at 13K RPM at 4⁰C 412 for 10 mins and the plasma collected; samples were stored at -80⁰C. Plasma sample were run on 413 sandwich ELISA kits testing 8 inflammatory cytokines (TNF-α, IL-6, MCP-1, IFN-γ, Rantes, MIP, IL-1a, IL-414 1b; Signosis EA-1201) and noradrenaline (whole antibody; Cusabio CSB-E07021h). Optical densities 415 were converted to concentrations via standard curves (Signosis EA-1202). Due to the variability in the 416 ELISA kits, measurements that did not lie on the linear part of the curve were discarded and the 417 samples were run again. To further reduce the impact of the variability of the ELSIA kits, samples were 418 pseudorandomised to kits and run in duplicate on different kits. 419 Collection of brain tissue 420 Rats were humanely killed by overdose of Isofluorane. The brains were removed. The medulla and 1 421 cortical hemisphere were placed into 4% paraformaldehyde (PFA) for immunocyctochemistry. The 422 other cortical hemisphere was placed into ice cold saline for electrophysiological studies. 423 The as means ± SD. For the hippocampus: To identify activated microglia, we counted DAPI stained nuclei 443 that we encapsulated with Iba1, anywhere in the visual field. Cell counts provided in the text are 444 reported as means ± SD. 445 Extracellular recording of fEPSPs in hippocampal slices 446 The extracellular recording of synaptic transmission and plasticity was carried out using two recording 447 rigs to increase productivity. The data from the two rigs was not significantly different and so was 448 pooled. 449 Recording on Rig 1: A slice was transferred to the recording chamber submerged in aCSF (124 mM 450 NaCl, 3 mM KCl, 2 mM CaCl2, 26 mM NaHCO3, 1.25 mM NaH2PO4, 1 mM MgSO4, 10 mM d-glucose 451 saturated with 95% O2-5% CO2, pH 7.5) and perfused at 4-6 ml/min (32°C). The slice was placed on a 452 grid allowing perfusion above and below the tissue and all tubing (Tygon) was gas tight (to prevent 453 loss of oxygen). To record field excitatory postsynaptic potentials (fEPSPs), an aCSF-filled 454 microelectrode was placed on the surface of stratum radiatum in CA1 (Fig 8F). A bipolar concentric 455 stimulating electrode (FHC) controlled by an isolated pulse stimulator model 2100 (AM Systems, WA) 456 was used to evoke fEPSPs at the Schaffer collateral-commissural pathway ( Fig 8F). Field (f)EPSPs were 457 evoked every 30 s (0.03 Hz). Stimulus input/output curves for fEPSPs were generated using stimulus 458 strength of 1-5 V for all slices (stimuli duration 0.2 ms). The paired-pulse ratio was measured over an 459 interval range of 20 to 500 ms (ratio averaged for 5 repeats at each interval). For the synaptic plasticity 460 experiments, the stimulus strength was set to produce a fEPSP slope ~40 % of the maximum response. Recording on rig 2: The slice was placed on a grid allowing perfusion (6-7 ml/min at 32°C) above and 468 below the tissue and all tubing (Tygon) was gas tight (to prevent loss of oxygen) Field excitatory 469 postsynaptic potentials were recorded as outlined for rig 1. A bipolar concentric stimulating electrode 470 (FHC) controlled by a DS3 isolated current stimulator (Digitimer, UK) with field (f)EPSPs evoked every 471 45 s. Stimulus input/output curves for fEPSPs were generated using stimulus strength of 20-300 µA 472 (stimuli duration 0.2 ms), stimulus strength was set to produce a fEPSP slope ~40 % of the maximum 473 response. Long-term potentiation was induced by theta burst stimulation (as above for rig 1). Signals 474 were acquired using WinLTP software (Vs 2.3, WinLTP Ltd, Bristol, UK). 475

Experimental paradigm 476
For the transduction, the order of rats and whether they received viral injections or were shams were 477 pseudorandomised. Experimenters were blinded to the condition of the rats. Rats were acclimated to 478 the plethysmography chamber the 2 nd week post-surgery. Plethysmography began 3 weeks post 479 transduction and continued for 6 weeks. Rats then underwent 1 day of Y-maze, and 17 days of Barnes 480 maze, testing the order of which was randomised for each batch. Rats were then allowed 1 more week 481 before tissue collection and LTP experiments were performed. Experimenters were unblinded to the 482 condition of the animals once analysis had been performed ( fig 8G). 483 Preliminary experiments to determine the injection volume required to induce sleep apnoea were 484 performed. Rats were acclimated to the plethysmography chamber the 2 nd week post-surgery. 485 Plethysmography began 3 weeks post transduction and continued for 6 weeks. Rats were then allowed 486 1 more week before tissue collection. Experimenters were unblinded to the condition of the animals 487 once analysis had been performed. Given the lack of sleep apnoea induced in rats in the preliminary 488 experiments, dual injected rats were pooled into he sham operated rats for analysis. 489

Data analysis 490
Outliers were removed after Iglewicz and Hoaglin's robust test for multiple outliers with an outlier 491 criterion: Modified Z score ≥3.5 (https://contchart.com/outliers.aspx). Sharipo-Wilk tests were 492 performed on sham operated groups. were deemed non-Gaussian by a Sharipo-Wilk test for normality, and tested via a Kruskal-Wallis Test. 504 68% of data are contained within 1 SD, to make data comparable we calculated the data range from 505 15 th to the 85 th quartile, herein known as Range70. Data are expressed as median ± Range70, and 506 displayed graphically as median ± quartile 15-85. 507 Data for Barnes duration (P = 0.0), Barnes maze distance (P = 0.0), Barnes maze Escape hole failures 508 (P = 0.0), and LTP (P = 0.0002) were deemed non-Gaussian by a Sharipo-Wilk test for normality, and 509 tested via a repeated measures 2 way ANOVA with a Sidak correction. Data are expressed as mean ± 510 SEM. 511 Data for input-output curves (P = 0.00005), were deemed non-Gaussian by a Sharipo-Wilk test for 512 normality, and tested via a 2 way ANOVA with a Sidak correction. Data are expressed as mean ± SEM. 513  Input-out curves for SA and sham mice. Top) representative data from individual rats, and Bottom) 727 group data. There is no significant difference between the two. B) Paired pulse facilitation was also 728 unaffected. Top) representative data from individual rats, and Bottom) group data. There is no 729 significant difference between the two. C) Average recordings of the slope of EPSPs from 730 electrophysiological recordings show LTP is diminished in rats with SA (red), but not shams (black), 731

References
following theta burst stimulation (TBS). Top) representative data from individual rats, and Bottom) 732 group data. D) Analysis was performed on 5 minute epochs from LTP traces from DTA injected rats. 733 and sham rats. Data represented as mean ± SD. 734 both arms. B) Heat maps from representative rats. C) Rats with SA (red) take longer to complete the 739 maze and cover more distance on the maze than sham operated rats (Black). D) Heat maps from 740 individual rats. E) Cartoon displays search strategies used by sham and SA rats; SA rats are less likely 741 to use spatial clues to find the escape hole than sham controls. 742 743