The Innate Immune Toll-Like Receptor-2 modulates the Depressogenic and Anorexiolytic Neuroinflammatory Response in Obstructive Sleep Apnoea

Background The neurological mechanisms of the disease process of obstructive sleep apnea, the second most frequent sleep disorder, remain unclear whilst its links with several major neuropsychiatric disorders, such as depression, anxiety and even Alzheimer’s disorder, are increasingly recognised. A radical theory, that inflammation in the brain may underlie certain phenotypes of many of these disorders, has been proposed, and the microglial TLR2 system may serve as an important crossroad at the borderlands of several pathogenesis. This study undertook to investigate whether a neuroinflammatory response occurs under conditions of OSA, and whether it might be related to a modulated response due to TLR2 functionality in an established rodent model of OSA. Methods The effects of three weeks’ exposure to chronic intermittent hypoxia were monitored in mice with or without functional TLR2 (C57BL/6-Tyrc-Brd-Tg(Tlr2-luc/gfp)Kri/Gaj; TLR2−/−, C57BL/6-Tlr2tm1Kir), that were investigated by multimodal in vivo and ex vivo imaging, combining magnetic resonance and bioluminescence imaging and a variety of functional tests. Results An acute neuroinflammatory response was demonstrated following the three days in the basal forebrain of mice, and more chronically in other parts of the frontal cortex. Adaptive changes in specific neurocircuitry were demonstrated, with significant links to agitated (mal)adaptive behaviour under episodes of stress, and an increased ability to gain weight. Conclusions Our results suggest that microglial activation and an innate immune response might be the missing link underlying the pathogenesis of well known structural, psychologic and metabolic changes experienced by some patients with OSA.


Introduction
Obstructive sleep apnoea (OSA) is a major clinical problem due to its high prevalence and serious complications (1,2), including its links to anxiety disorders, depression(2, 3) and Alzheimer's disease (AD) (4)(5)(6)(7). OSA results from a mix of genetic, environmental, and lifestyle factors, with obesity and aging being the key risk factors. In patients with OSA, the upper airway narrows or collapses repeatedly during sleep, causing obstructive apnoeic events associated with intermittent hypoxia, recurrent arousals and increase in respiratory effort, leading to secondary sympathetic activation, oxidative stress and systemic inflammation. (8) To date, the neurological mechanisms of the disease process of OSA remain unclear. Their deciphering is of paramount importance, as it might aid the development of an effective neuroprotective therapeutic approach in AD, with which OSA appears to share a complex bidirectional link (4,9). Whilst several studies suggest that glial activation and associated inflammation play an important role in the pathogenesis of AD, depression (10,11) and several other psychiatric disorders (12), the occurrence of any neuroinflammatory process is yet to be demonstrated in OSA.
Toll-like receptors (TLRs) serve as important links between innate and adaptive immunity of the brain, and most recently the microglial TLR2 system has been demonstrated to play an important role in AD pathogenesis (13). Similarly, several recent pivotal studies suggested abnormalities in TLRs, including the TLR2 system, might be playing an important role in the pathophysiology of depression and suicidal behaviour (14), stress-induced neuroinflammation (15), elevated anxiety and social avoidance (16).
To investigate whether a neuroinflammatory response occurs under conditions of OSA, and to elucidate to what extent the combination of beneficial and harmful inflammatory events might be related to a modulated response due to TLR2 functionality, this study monitored the effects of three weeks' exposure to chronic intermittent hypoxia (IH) in an established rodent model of OSA (17) , (18). This mouse model has been shown to verifiably mimic the electroencephalographic arousals and significant hypoxaemia experienced by patients with OSA. Here its consequences were studied through time by multimodal in vivo and ex vivo imaging, combining magnetic resonance imaging (MRI) and bioluminescence imaging (BLI).

Results
An acute two-site TLR2 response in mouse model of OSA Previous studies have shown that TLR2 regulates the hypoxic/ischaemic brain damage caused by stroke (4,19,21). To establish whether IH, that results from nocturnal apnoeic and hypopnoeic episodes in patients with OSA, provokes a similar inflammatory response in brain, we used an established mouse model of OSA (18,20) and transgenic mice that bore the dual reporter system luciferase/green fluorescent protein under transcriptional control of the murine TLR2-promoter. TLR2 induction/microglial activation and its spatial and temporal dynamics were then investigated longitudinally in real time using BLI, as previously described (19). The signals were analysed over a threeweek period following experimental (IH) and control (CTRL) protocol ( Figure 1A-D).
Using this relatively novel in vivo imaging approach, we demonstrated the inflammatory response with a marked chronic component. As shown in Figure 1A, the photon emission and TLR2 signals/microglial activation were significantly elevated, when compared to control baseline levels, throughout a three-week period interval following a daily IH experimental protocol. The quantitative analysis of photon emissions revealed that levels of TLR2-induction signals peaked after 24 hours (1.16x10 7 ±3.43x10 6 p/s/cm 2 /sr, n=18) reaching statistical significance following three days of the protocol (P=.018; baseline: 7.68x106±1.81x106 p/s/cm 2 /sr, n= 20 vs day 3:1.12x107±2.40x106,n=17) ( Figure 1A). A later smaller TLR2 peak was also noted following ten to twelve days of the experimental protocol. Thereafter, the total photon emission remained significantly elevated and relatively constant, up until the last analysed three-weeks timepoint (Figure1A,D).
Olfactory bulb microglia in mice receive and translate numerous inputs from the brain and the environment and likely serve as sensors and/or modulators of brain inflammation.
The subset of olfactory bulb microglial cells in mice was previously shown to continually express TLR2, enabling them to survey the environment in a 'primed' or alert state (19).

TLR2-transgene induction in the brain co-localises with microglia
To further characterize micro-regional differences of the inflammatory response recorded by BLI at the three key neuroanatomical sites -olfactory bulb and anterior olfactory nucleus (at baseline), septal nuclei (72 hours), and at later points, at the wider frontal cortical regions (see Figure 1) -we then explored cellular aetiology of TLR2-signal in TLR2CTRL and TLR2IHmice ( Figure 2). We were able to confirm that the endogenous TLR2-protein was indeed induced within these areas, and that the vast majority of cells expressing these receptors were microglia ( Figure 2E). (19). As shown in larger magnification photomicrograph in Figure 2F, the luciferase immunoreactivity co-localized with Iba1 immunostaining (microglial cells with amoeboid activated morphology) in the micro-region of septal nuclei, further suggesting the importance of this site early on in the inflammatory cascade.

TLR2 modulates the effects of chronic IH on structural brain changes
In order to investigate whether initially demonstrated neuroinflammatory response later on resulted in structural plastic changes, known to occur in patients with OSA(2), we utilised high resolution ex vivo magnetic resonance imaging (MRI). To fully verify involvement of the TLR2 system, mice with and without (TLR2 -/-) functional TLR2 gene were imaged after three weeks of IH or CTRL protocol (Figures1,2).
As shown in Figure 3, comparison of structural brain grey and white matter changes with  Table S3).

TLR2-induced neuroinflammatory plastic responses localised to brain derived neurotrophic factor (BDNF), neuroplastin and fibronectin-1 (FN1)-rich neurocircuitry
Together, our data provides further evidence for the role of TLR2 in modulation of the neuroinflammatory response, and suggested its spatio-temporal spread via synaptically connected enthorhinal and basal forebrain networks.
We further examined the molecular origins underlying neuromechanisms behind OSA-  Table S5). A modulatory role in observed structural changes was suggested for several major neuroplasticity genes, however, this association was evident only in mice with a functional TLR2-system (Table S4). The genes that were linked with TLR2-enabled structural network reorganisations were the brain derived neurotrophic factor (BDNF), neuroplastin, Calcium/calmodulin dependent protein kinase

Neuroinflammatory response: from effects on mood, cognition to effects on weight gain?
Finally, we wanted to assess if mice in our OSA model shared similar functional and behavioural changes to those frequently reported in patients with OSA, including their well-documented struggles with obesity, somnolence, fragile mood and memory and attention deficits (2,24). We also wanted to see if those changes were linked to neuroanatomical ROIs that were initially highlighted by our BL and MR imaging findings: frontal cortex(16), septal nuclei, ventral hippocampi (25) and PAG (26,27).
Weight: Firstly, the adaptive role for TLR2 system in weight gain was investigated. A newly proposed anorexigenic neural circuitry in rodents incorporates two ROIs implicated in the neuroinflammatory response in our study, the ventral hippocampus and lateral septal nucleus in the brain. suggesting functional TLR2-driven adaptive value of noted changes in mice exposed to IH. Similar structural differences were noted in control groups, although they were statistically remarkable only in the anorexigenic region of the right ventral hippocampi  (Figure 9). TLR2 -/deficient mice (TLR2 -/-CTRL), on the other hand, failed to gain further weight over the period of three weeks, and their weight remained constant throughout the protocol (Figure 9). Over a three-week period, a significant weight loss in both TLR2IH and TLR2 -/-IH mice was recorded, compared to their respective controls (Figure 9, Supplement). This reached a statistical significance after three days, and it remained significant until the end of the study (Figure 9; Tables   S7,8). However, whilst mice with functional TLR2 showed a steady state increase in weight following the initial loss, TLR2 -/mice were not able to regain initially lost weight ( Figure 9).
Mood and cognition: Next, the adaptive role for TLR2 system in neurocognitive changes was investigated. Using several behavioural tests (Table S6) known to assess and target psychomotor changes, affective and cognitive symptoms, we observed two primary findings, which were consistently replicated across several behavioural tests and their parameters (Table S6). Firstly, as expected, we demonstrated increased psychomotoric agitation and anxiety in all mice exposed to the IH protocol. This behaviour appeared independent of their TLR2-functionality. For example, during the tail suspension test (TST) the mice that were exposed to the experimental IH protocol spent significantly  Table S1). A significant aberrant connectivity between the two regions was also noted (P=.011;r=-.66).
In regard to other behavioural findings, only a statistical trend for more effective spatial navigation and cognitive processing was seen in mice with functional TLR2. For example, in Y-maze test the presence of TLR2 system appeared to partially rescue deficit in spatial acquisition following exposure to IH, otherwise recorded in TLR2 -/mice, as measured by the path efficiency (path efficiency%: TLR2IH 1.05±0.16; TLR2CTRL 1.09±0.20;TLR2 -/-IH 0.63±0.14;TLR2 -/-CTRL 1.04±0.20; P (TLR2IHvsTLR2 -/-IH) =.06). (see Table S6 for further details). Several other differential modulatory trends of genotype and phenotype on behavioural and cognitive parameters of mice are presented in Figure 10. Our data suggests an adaptive side to the neuroinflammatory response in regards to feeding (33) and or ability to gain weight. OSA has been associated with significant metabolic changes, diabetes and weight gain, thought to be at least partly modulated through its systemic inflammatory effects(4). It has been previously proposed that a significant interplay between central nervous inflammation and systemic inflammation might occur in affected patients, which then further modulates bidirectional links between metabolism, weight gain and neurological disorders (12). More specifically, our results could be taken to propose that activated TLR2-system enables continued feeding drive under conditions of repeated stress ( Figure 9). Presumably, similar drive could later cause maladaptive inability to control one's food intake and to lead to an increased link between our emotional states and eating, causing an excessive weight gain in patients with OSA. The feeding circuitry that controls emotional or cognitive aspects of food intake is still largely unknown. However, recently, Sweeney and Yang (2015) demonstrated an anorexigenic neural circuit originating from ventral hippocampus to lateral septal nuclei in the brain, revealing a potential therapeutic target for the treatment of anorexia or other appetite disorders (25). Interestingly, we were able to differentiate this network solely based on the functionality of the TLR2-system in our mice, and severity of changes appeared further exacerbated by the IH stress (also see Table S1 for further details). Whilst preliminary, these findings might go some way to suggest an aberrant feeding circuit that controls emotional and cognitive aspects of food intake in patients with OSA.

Discussion
In conclusion, utilising the rodent model of OSA that verifiably replicates arousals and hypoxaemia in patients with OSA (17) and visualising in vivo TLR2-activation, we were able to demonstrate early activation of microglia in regions of basal forebrain, with later widespread frontal projections, suggesting a pivotal role for TLR2 in brain's response to OSA injury. Our results provide the first in vivo evidence of an OSA-induced inflammatory response, with septal nuclei, the major source of cholinergic input to the hippocampus, being highlighted as a region of particular vulnerability early on in the neuroinflammatory process ( Figure 1C). Moreover, we were also able to link the initial neuroinflammatory response to later hypotrophic and hypertrophic neuroanatomical changes, with potential primary molecular drivers also highlighted by our results. Through these findings, a fingerprint of a distinct OSA-effected neurocircuitry has emerged, with frontal regions and septal nuclei being suggested as an initial TLR2-dependant seed sites.
Our neuroimaging results are in broad agreement with previous findings in patients with OSA where compensatory mechanisms, activation of various homeostatic gene programmes and astroglial neurogenesis have all been proposed to underlie some of the initial adaptive and later maladaptive changes (36,37). In further agreement, our data also suggests that differential plastic response to OSA in the brain may depend on regional genes expressional profiles (38) For example our 'MR-gene' mapping findings support permissive and cohesive role for the TLR2-system in the interplay with BDNF, RSGRP1, fibronectin and neuroplastin-driven significant discrete and transformative neurophysiologic and behavioural changes.
Of the listed genes, the enhanced BDNF-levels have long been associated with increased plasticity and promotion of a growth-permissive environment, and its depletion with many neurological and psychiatric disorders (39). The concept of the interplay between BDNF and TLR2-system in our study might also explain proadaptive (preconditioning) protective effect on visuo-spatial memory in our TLR2IH mice, which, however, showed only significant trend. BDNF has been shown to enhance synaptic plasticity and neuron function in response to physical activity, learning and memory, and its baseline expression and activity-dependent upregulation in the hippocampus is believed to be under control of the medial septal nuclei, and important ROI in our study, with an important regulatory role in REM sleep ( Figure 2). Patients with OSA show a rapid decrease in serum and plasma BDNF levels during initiation of the treatment (with PAPdevice), likely reflecting enhanced neuronal demand for BDNF in this condition. (40) Similarly, OSA patients had increased TLR2-expressions on blood immune cells, which could be reversed with PAP treatment (41). TLR2-deficiency has been shown to impair neurogenesis previously (42) and more recently, the TLR2-receptor has been shown to enhance adult neurogenesis in the hippocampal DG after cerebral ischaemia (43). Of particular note to our findings, the loss of TLR2 has been recently shown to abolish repeated social defeat stress-induced social avoidance and anxiety in mice, and its deficiency mitigated stress-induced neuronal response attenuation, dendritic atrophy, and microglial activation in the medial prefrontal cortex (16). TLR2 has also been shown to modulate inflammatory response caused by cerebral ischemia and reperfusion via linking to endogenous ligands, such as fibronectin (44). Fibronectin, on the other hand, has been shown to act as a reparative molecule that promotes cellular growth and its levels are enhanced after brain injury in variety of disorders, including the AD (39,45). Finally, neuroplastin is known to play a role in synaptic plasticity (e.g. long-term potentiation), formation and a balance of the excitatory/inhibitory synapses (46), in shaping of brain's cortical thickness (47), and its involvement in early tissue response in hippocampi of AD patients has also been recently shown (48).  Representative images of control (A-C) and animals exposed to IH (D-F) at 72 hours. All animals were exposed to ex vivo bioluminescence imaging where TLR2 signal was confirmed in olfactory bulb and anterior olfactory nucleus in both groups, whilst additional signal was      Non-parametric statistics were performed using FSL randomize with 5000 permutations and threshold-free cluster enhancement.