Neuroprotective exendin-4 enhances hypothermia therapy in a model of hypoxic-ischaemic encephalopathy

Hypoxic-ischaemic encephalopathy (HIE) causes 25% of neonatal deaths worldwide. Rocha-Ferreira et al. demonstrate that a diabetes drug protects the neonatal brain in a model of acute HIE, and confirm that the required receptor is found in human perinatal brain tissue. Synergistic combination with clinical hypothermia enhances therapy, supporting potential translation.


Introduction
Hypoxic-ischaemic encephalopathy (HIE) is a serious complication of labour caused by reduced blood flow and oxygen supply to the neonatal brain. This can result in mortality for the infant or significant and lasting brain damage. HIE is a global problem with an estimated incidence of 1.5 per 1000 live births. Fifteen to 20% of HIE neonates die during the postnatal period, and an additional 25% develop irreversible and lifelong mental and physical disabilities including cerebral palsy (Shankaran, 2012). In 2010, HIE was associated with 2.4% of the total Global Burden of Disease (Lee et al., 2013).
The known temporal sequence of events in the brain following hypoxia-ischaemia in rodent models has defined a therapeutic window of opportunity of up to $8 h based on a time course of secondary energy failure and other measures of secondary brain injury (Blumberg et al., 1997;Gilland et al., 1998a, b). Early interventions have shown efficacy when initiated early and within this window (Nijboer et al., 2011). Reducing the body temperature (hypothermia) of human neonates within 6 h of hypoxiaischaemia onset and with duration of 72 h is the only clinically approved treatment (Edwards et al., 2010). Hypothermia is thought to protect neurons by reducing cerebral metabolic rate (Rosomoff and Holaday, 1954) and concentrations of glutamate and nitric oxide (Thoresen et al., 1997), inhibiting cerebral energy failure, preserving high-energy phosphates (Thoresen et al., 1995) and preventing inflammatory cascades (Inamasu et al., 2000). However, while hypothermia therapy is very promising, up to 55% of treated neonates cannot be saved (Gluckman et al., 2005). Therefore, there is a need to develop therapies that are either more effective than hypothermia, or can be used in combination to enhance its therapeutic efficacy.
Exendin-4 (also known as exenatide) is a small peptide drug approved by the Food and Drug Administration (FDA) in 2005 and European Medicines Agency (EMA) in 2006 for the treatment of type 2 diabetes mellitus. It is an analogue of the human glucagon-like peptide-1 (GLP-1) gut hormone peptide that plays a role in regulating blood sugar levels by enhancing insulin production from the pancreas. While GLP-1 has a half-life of $1.5 min (Deacon et al., 1996), exendin-4 can reach 60-90 min (Nielsen et al., 2004), making it of therapeutic value, and is administered twice daily after meals. The positive neurological effects of exendin-4 were first recognized by improvements in neuropathic aspects in type 2 diabetes mellitus patients under treatment (Grant et al., 2011). Exendin-4 also efficiently crosses the blood-brain barrier (Kastin and Akerstrom, 2003;Zanotto et al., 2017) and its cellular receptor (GLP1R) is found throughout the brain (Wei and Mojsov, 1995). Although there is no definitive unifying mechanism, in vitro and in vivo studies have suggested that exendin-4 has neuroprotective, neurotrophic (Perry et al., 2002), neurogenic (Bertilsson et al., 2008), antiinflammatory (Teramoto et al., 2011), anti-apoptotic (Tews et al., 2009) and mitoprotective (Fan et al., 2010;Kang et al., 2015) properties. These findings, together with the excellent safety profile of exendin-4 in patients with type 2 diabetes mellitus, has led to recent and ongoing clinical trials examining the neuroprotective properties of exendin-4 in patients with Parkinson's (NCT01174810; NCT01971242; Aviles-Olmos et al., 2013a; Athauda et al., 2017) and Alzheimer's disease (NCT01255163). Recently, the Parkinson's disease trial has reported that patients on exendin-4 show a statistically significant improvement in clinical motor and cognitive measures compared to control group. This effect persisted 12 weeks after ending the exendin-4 treatment (Athauda et al., 2017).
Given the need to develop effective treatments for neonatal HIE, the encouraging animal and clinical studies supporting the neuroprotective properties of exendin-4 make it an attractive therapeutic option. Little is known regarding the use of exendin-4 in perinatal animals, and there are no neonatal clinical studies. Therefore, we conducted a preclinical assessment of exendin-4 using an established mouse model of neonatal hypoxia-ischaemia causing widespread cerebral damage (Carlsson et al., 2012;Rocha-Ferreira et al., 2015). In this study, we confirmed the presence of GLP1R in the perinatal human and murine brains. We demonstrated significant dose-dependent neuroprotective and anti-inflammatory effect of exendin-4 treatment that is dose-dependent and the effects of which can be maintained even when administration is delayed posthypoxia-ischaemia. Furthermore, we conducted a toxicity study to examine the safety of high dose repeat administration of exendin-4 in perinatal mice and demonstrate safety and tolerance to the drug. Finally, we established its ability to be used synergistically with therapeutic hypothermia that enhances neuroprotection and ameliorates brain damage.

Study approval
All UK mice experiments were approved by the Ethics Committee of the University College London and were carried out by licensed personnel (PPL PCC436823) in concordance with the UK Home Office Guidelines [Animals (Scientific procedures) Act, 1986]. All mice experiments undertaken in Sweden conformed to the Swedish Board of Agriculture and were approved by the Gothenburg Animal Ethics Committee (61-2014 and 01-2016). CD1 mice (Charles River) were bred in house with a 12-h light/dark cycle and had free access to food and water. Breeding cages contained igloo housing with free access to an exercise plate. Once weekly, enrichment food in the form of dry nuts and fruits was sprinkled onto the cages to allow foraging. All animal experiments included male and female littermates randomly allocated to the different experimental groups to reduce bias and followed the ARRIVE guidelines (Kilkenny et al., 2011

Blood analysis
Postnatal Day 7 (P7) mice underwent four high doses of exendin-4 (0.5 mg/g per dose, Enzo) via intraperitoneal administration (5 mg/g, 12 h apart). Naïve or saline-injected mice acted as controls (n = 6 per group). Twelve hours after the last injection, blood samples taken via cardiac puncture and collected in EDTA-coated tube. The analysis was performed by MRC Harwell Clinical Pathology laboratory (Mary Lyon Centre, UK). Different parameters were obtained: total white blood cells, neutrophils, lymphocytes, monocytes, eosinophils and basophils counts, haematocrit, platelets, red blood cells, haemoglobin and mean corpuscular volume. Blood glucose levels (mmol/l) were measured using a blood glucose monitor (CodeFree, SD Biosensor) in naïve controls and mice following a single exendin-4 high dose administration (0.5 mg/g). Blood samples were collected via cardiac puncture at 0.5 h, 1 h, 2 h, and 4 h post-exendin-4 injection.

Cyclic AMP assay
Brains of P7 naïve mice were collected at 2 h, 4 h, 8 h and 12 h following a single intraperitoneal injection of exendin-4 (0.5 mg/g). Saline-treated controls were collected 2 h post-injection (n = 4 per group). All samples were processed using cAMP direct immunoassay kit as per manufacturer's instructions (Abcam).

Hypothermia treatment
Within 10 min post-hypoxia-ischaemia, P10 mice underwent a single high dose of either exendin-4 (0.5 mg/g) or saline and were placed in individual compartments within a hypothermia (33 C) or normothermia (36 C) chamber for 5 h. One probe in each chamber monitored environmental temperature and one animal in each chamber was randomly selected to be used as a temperature-monitoring sentinel to measure core temperature using a rectal probe (T21, 0.41 mm diameter; Physitemp Instruments). The use of a rectal probe is in agreement with previous hypothermia studies in mice and rats demonstrating a good correlation between core body temperature measured this way and brain temperature in small rodent pups (Thoresen et al., 1996b). The alternative use of telemetry is promising but more expensive and technically challenging. Room temperature was also recorded. Computer Software Daisy lab 10.0 (Physitemp Instruments) was used to monitor the temperature, and values from each probe were recorded every 5 min ( Supplementary Fig. 1). It was decided prior to the experiment that sentinel mice (n = 8 normothermia, n = 8 hypothermia) were excluded from the final analysis, as restraint-associated stress from the rectal probe has shown a neuroprotective effect (Thoresen et al., 1996a).

Tissue preparation
Highest level of widespread neuronal caspase-3 expression within the acute injury phase occurs 48 h post-hypoxia-ischaemia (Johnston et al., 2011). Therefore, this time point was chosen for early evaluation of neuropathological markers in the P7 study. P7 mice were perfused with 4% paraformaldehyde (PFA) 48 h post-hypoxia-ischaemia, brains transferred to 30% sucrose and snap frozen. P10 mice were perfused in the same manner 7 days after hypoxia-ischaemia, a time point when the secondary phase of brain injury is completed (Gilland et al., 1998a). Coronal brain sections (40-mm thickness) starting from the point of fusion of the corpus callosum were collected onto TBSAF wells. Every 10th section, in a total of five sections (400 mm apart) per brain underwent histochemistry, immunohistochemistry or immunofluorescence. Whole post-mortem preterm neonatal brains were fixed in 4% formalin for 7 weeks before anatomical positions from the frontal, occipital, and parietal lobes were selected and processed using a Leica tissue processor. The blocks were sectioned at 6-mm thickness using a microtome, mounted onto SuperFrost TM Plus slides and allowed to dry for 4 days.

TUNEL staining
Murine brain sections were incubated in 3% hydrogen peroxide/methanol (15 min) followed by 2 h incubation at 37 C with terminal deoxytransferase and deoxyuridine triphosphate solution (Roche). The reaction was stopped by incubation in terminal deoxynucleotidyl transferase dUTP nick-end labelling (TUNEL) stop solution for 10 min. Slides were incubated in ABC solution (1 h, room temperature) and visualized using DAB enhanced with cobalt nickel in the presence of hydrogen peroxide. The reaction was stopped, and sections covered as described above.

Data analysis
Infarct area was measured using ImageJ software (NIH). The intact Nissl + areas of the isocortex, pyriform cortex, hippocampus, striatum, thalamus and external capsule brain regions were delineated bilaterally and the ipsilateral regions were subtracted from the contralateral regions and expressed as percentage tissue loss. This method assumes that the contralateral hemisphere represents maximal intact area (100%); however, there are instances where because of potential symmetrical differences between hemispheres, the intact ipsilateral hemisphere is larger in area size and the resulting measurement shows a negative value. TUNEL + cells were quantified under Â20 magnification in three separate fields per assessed brain region. Microglial activation was assessed using semi-quantitative score as previously described (Rocha-Ferreira et al., 2015), with a scale of: 0 (no activation, ramified microglia) to 4 (widespread amoeboid microglia). GFAP, ICAM1 and CD68 immunoreactivity was measured by quantitative thresholding image analysis using Image Pro Premier software (Media Cybernetics) as previously described (Rocha-Ferreira et al., 2016). In brief, three nonoverlapping RGB images from the assessed brain regions were captured using a live video camera (Nikon, DS-Fi1) mounted onto a Nikon eclipse E600 microscope at Â40 magnification for both GFAP and ICAM1 stainings. Similarly, 30 nonoverlapping RBG images from brain, heart, spleen, liver, lung, pancreas and kidney were captured for CD68 measurements. The threshold setting was kept constant for all acquired images. MAP-2 staining was used for quantification of volumetric tissue loss from five brain levels, starting from level 1 (L1, fusion of corpus callosum) and continuing until level 5 (L5, late hippocampus), as well as total volumetric tissue loss (Stridh et al., 2013). Macroscopic score of tissue loss was performed using a scale comprising of 0, no visible injury; 1, 25%; 2, 50%; and 3, 75% hemispheric injury/loss.

Statistical analysis
Experimental cohorts consisted of 14-25 mice, based on power calculations as described by Dupont and Plummer (1990). The levels of significance are 5% with 80% power as a minimum. The noise and effect size were estimated through our previous experience and publications using this mouse model (Wang et al., 2010;Carlsson et al., 2011) and calculations were performed using publicly available software, PS: Power and Sample Size Calculation v3.1.2, 2014 (http://biostat.mc. vanderbilt.edu/wiki/Main/PowerSampleSize) (Dupont and Plummer, 1998). Data were analysed using the GraphPad Prism v6.0. All assessments were performed blindly. Average AE standard error of the mean (SEM) was recorded for all data (Supplementary Table 1) and was first analysed with the Kolmogorov-Smirnov normality test. As the data did not follow Gaussian distribution, the Kruskal-Wallis non-parametric test was applied, followed by Dunn's test (Supplementary Table 2). *P 5 0.05, **P 5 0.01, ***P 5 0.001 and ****P 5 0.0001.

Data availability
All raw data are available from the corresponding author on request.

Results
GLP1R is expressed in the human and murine neonatal brain Firstly, we sought to confirm the presence of GLP1R both in the human and murine brains. Confocal imaging using antibodies against GLP1R and neural cell-specific markers showed GLP1R co-localization predominantly in neurons (NeuN) in all assessed developmental ages: adult (10 weeks), P10 and P7 ( Fig. 2A). At 10 weeks of age, co-localization of GLP1R was (A) Immunofluorecence and scanning confocal microscopy studies demonstrating GLP1R is expressed in neurons (NeuN) in the mouse brain at 10 weeks (adult), P10 and P7, (B) with only small colocalization with astrocytes (GFAP) at 10 weeks of age, and (C) no co-localization with microglia (CD68) cells in any of the different developmental ages. The first row of micrographs (A-C) show negative staining for GLP1R antibody. (D) Immunofluorescence studies of post-mortem human preterm brain tissue shows co-localization of GLP1R expression in neurons (NeuN) in the frontal lobe and hippocampus Ammon's horn, with the first row of micrographs for each human brain region containing negative staining for GLP1R antibody. A median filter was applied to the images to reduce noise. (E) Relative quantification of GLP1R expression by quantitative PCR in different brain regions at 10 weeks, (F) P10 and (G) P7 (n = 6 per age group). CBL = cerebellum; CTX = cortex; HIP = hippocampus; HPT = hypothalamus; MDL = medulla; OB = olfactory bulbs; THL = thalamus. Scale bar = 20 mm in C; 10 mm in D.
observed mostly in neurons but also in astrocytes (GFAP). This astrocyte co-localization was not observed at P10 and P7 (Fig. 2B). No microglia (CD68) co-localization with GLP1R was observed in any of the different developmental ages (Fig. 2C). Post-mortem neonatal sections showed GLP1R expression in the frontal lobe and hippocampus Ammon's horn. Confocal imaging confirmed specific co-localization with neurons (NeuN) (Fig. 2D). Quantitative PCR analysis using primers against Glp1r and control Gapdh mRNA for normalization of data ( Supplementary Fig. 2) revealed GLP1R expression across the different brain regions examined in adult (10 weeks) and neonatal (P10 and P7) mice. Glp1r mRNA was particularly present in the cortex, cerebellum and olfactory bulb of adult (Fig. 2E), P10 (Fig. 2F) and P7 mice (Fig. 2G), and also in the hippocampus of P10 and P7 mice.

Exendin-4 reduces brain infarction in a dose-and time-dependent manner
To assess the efficacy of exendin-4 brain protection against neonatal hypoxia-ischaemia, tissue infarction was measured 48 h post-insult using different dosing and concentration regimen, and timing of intervention. Saline-treated hypoxia-ischaemia littermates served as controls (Fig. 3A). The Nissl measurement as a percentage of tissue loss showed overall profound and consistent injury in the ipsilateral hemisphere of saline-treated hypoxia-ischaemia controls (SAL, 50% AE 6.9%). A single high dose intraperitoneal administration of exendin-4 immediately after hypoxia-ischaemia significantly reduced tissue infarction (17% AE 8.6%, P = 0.0279). The same high dose given immediately after hypoxia-ischaemia and then every 12 h over a 48-h period provided added reduction of brain injury (2% AE 1.8%, P = 0.0214). However, dilution of exendin-4 (0.05 mg/g) given in the same four-dose regimen was not protective (37% AE 8%) (Fig. 3B). Delaying the first administration of exendin-4 by 2 h, in the same 12-h apart four high dose administrations still resulted in a significant protective effect (11.0% AE 6.9%, P = 0.0334), with no difference in protection observed between immediate and delayed four high dose regimens (Fig. 3C). Breakdown of Nissl measurements across individually assessed brain regions showed that immediate start of four high dose exendin-4 administrations had the highest regional protection: isocortex (P = 0.0222), pyriform cortex (P = 0.0629), hippocampus (P = 0.0005) and striatum (P = 0.0064). This is followed by the same four high dose administrations started 2 h after hypoxia-ischaemia: isocortex (P = 0.0334), striatum (P = 0.0069) and thalamus (P = 0.0406). Lastly, single high dose exendin-4 given immediately after hypoxia-ischaemia also reduced regional brain injury in pyriform cortex (P = 0.0013) and striatum (P = 0.0468) ( Supplementary Fig. 3). There were no differences between the different hypoxia-ischaemia-saline control groups, therefore, animals from all saline groups were randomly pooled. There were no differences between males and females in response to any of the treatments and the distribution of the sexes of mice per experiment are listed in Supplementary Table 3. Weight gain measurement was performed in all groups. Following hypoxia-ischaemia, a single high dose administration of exendin-4 did not cause weight loss; however, there was a reduced weight gain in comparison to saline controls, which reached significance at the 36 h (P = 0.0212) and 48 h (P = 0.0148) time points post-hypoxia-ischaemia. Both immediate and 2-h delay four high dose exendin-4 treatments resulted in initial non-significant weight loss when compared to baseline weight, with visible signs of recovery 48 h posthypoxia-ischaemia. In the exendin-4 (four doses group) mice weighed significantly less than the low-dose group at 12 h (P = 0.0150) and 24 h (P = 0.0096) time points, and significantly less than saline-injected mice at 24 h (P = 0.0074), 36 h (P = 0.0315) and 48 h (P = 0.0269) post-hypoxia-ischaemia (Fig. 3D). There was no exendin-4-mediated modulation of weight in the 2-h-delayed four-dose treatment (2H) and low-dose-treated groups. To assess how quickly exendin-4 was crossing the blood-brain barrier and potentially initiating a cellular response in the CNS, we measured alterations in cyclic AMP (cAMP) expression, a known second messenger of GLP1R signalling. Naïve animals were given one high dose of exendin-4 and cAMP was measured in the brain at 2 h, 4 h, 8 h and 12 h. Significantly higher total brain cAMP was measured at 2 h (P = 0.0021) and 4 h (P = 0.0281) when compared to saline-injected controls at 2 h ( Supplementary Fig. 4).

Repeated high doses of exendin-4 is non-toxic in neonatal mice
The therapeutically optimal exendin-4 dose (0.5 mg/g) given every 12 h over a 48-h period in this study are substantially higher than those used clinically in diabetic patients. Therefore, we conducted an examination in naïve (no hypoxia-ischaemia) P7 mice for any toxicity or adverse effects. High dose exendin-4 did not alter blood glucose levels when compared to saline-treated or naïve controls (0 h, 0.5 h, 1 h, 2 h and 4 h time points) (Fig. 6A). The exendin-4 four high dose regimen in naïve animals resulted in reduced weight gain at 24 h (naïve versus EX4, P = 0.0552, SAL versus EX4, P = 0.0552) and 36 h (naïve versus EX4, P = 0.0558, SAL versus EX4, P = 0.05598 time points) (third and fourth injections), with partial recovery to baseline weight measurement (0 h time point) by 48 h after final injection (Fig. 6B). As exendin-4 has been shown to modulate peripheral immune cells (Yanay et al., 2015;He et al., 2017), blood tests were conducted after completion of the four high dosing regimen. Counts of various blood cell populations and biochemistry: total white blood cells (Fig. 6C), neutrophils (Fig. 6D), lymphocytes (Fig. 6E), monocytes (Fig. 6F), eosinophils (Fig. 6G), basophils (Fig. 6H), haematocrit (Fig. 6I), platelets (Fig. 6J), red blood cells (Fig. 6K), haemoglobin (Fig. 6L) and mean corpuscular volume (Fig. 6M) in mice treated with the high dose exendin-4 regimen were all normal with no significant differences to naïve unadministered mice and saline-injected mice. Given the systemic administration of exendin-4, in addition to the brain we also harvested the heart, spleen, liver, lung, pancreas and kidney for analysis. Haematoxylin and eosin staining did not reveal any obvious fibrosis or abnormalities in cellular or tissue architecture in mice receiving the high dose exendin-4 regimen when compared to naïve or saline-treated mice (Fig. 7A). Quantitative threshold image analysis using a macrophage-specific marker (CD68) did not show any significant inflammatory response in any of the organs examined from the exendin-4-administered mice compared to controls (Fig. 7B-H and Supplementary Fig. 5).

Discussion
There is an enormous unmet clinical need for effective interventions against neonatal HIE. We have conducted a Figure 8 Synergistic enhanced neuroprotection following combined exendin-4 and hypothermia treatment in the term hypoxia-ischaemia model. (A) Whole brain representative micrographs of normothermia + saline control (NT SAL, n = 24, 10 male and 14 female), single treatments normothermia + exendin-4 (NT EX4, n = 25, 12 male and 13 female) and hypothermia + saline (HT SAL, n = 25, 14 male and 11 female), as well as combined hypothermia and exendin-4 treatment (HT EX4, n = 25, 12 male and 13 female) 7 days after P10 hypoxic-ischaemic injury. (B) Macroscopic score evaluation. (C) Weight gain across the different groups. (D) Overall infarct volume and (E) through different injury levels, where level 1 indicates the most anterior level, as assessed using MAP-2-stained sections from the different treatments. Data presented as individual animals or as mean AE SEM and analysed using Kruskal-Wallis Dunn's test. *P 5 0.05, **P 5 0.01, ***P 5 0.001 and ****P 5 0.0001. Scale bar = 2 mm.
Exendin-4 reduces perinatal brain damage preclinical study of the therapeutic efficacy and safety of exendin-4 in a perinatal mouse model of HIE. Here we establish for the first time that exendin-4 treatment after neonatal HIE is highly neuroprotective. Furthermore, we demonstrate that exendin-4 can be used in synergy with hypothermia, the current clinical standard of care for HIE, to enhance its therapeutic efficacy.
In our late preterm results, intraperitoneal administration of 0.5 mg/g exendin-4 as a four-dose 12 h interval regimen started immediately after hypoxia-ischaemia is significantly neuroprotective. Delaying the start of this exendin-4 treatment regimen by 2 h also significantly protects the immature brain with no significant difference in efficacy when compared to immediate administration post-hypoxia-ischaemia. The significant therapeutic efficacy was also measured in other readouts of neuropathology including a reduction of TUNEL + cell death and microglia/macrophage, astrocytes and endothelial cells activation markers remained low in contrast to untreated hypoxia-ischaemia groups. Our 2 h delayed exendin-4 regimen results are of particular clinical relevance and may signify a potentially extended window of opportunity in which therapeutic exendin-4 could be administered. In the adult murine stroke model using transient middle cerebral artery occlusion, Teramoto et al. (2011) were able to retain reduced infarct volume only up to 1 h delayed administration of high dose exendin-4.
This study shows that exendin-4 alone or in combination with therapeutic hypothermia significantly protects the neonatal brain against term HIE. Hypothermia is the only standard treatment of care for term HIE in developed countries and is not used in preterm infants or in developing countries. The latter has a high association with infection, where hypothermia may have limited or even a detrimental effect (Robertson et al., 2008;Osredkar et al., 2014). Furthermore, hypothermia is only partially protective, and alternative or adjunct therapies are needed. Because of the variability in the efficacy of hypothermia in the neonatal rodent model of HIE, we administered a single high dose of exendin-4 in combination with hypothermia to allow clear dissection of a potential synergistic adjunct role between both treatments. Exendin-4 administration provided neuroprotection against tissue loss in the P10 term mouse comparable to the P7 single-dose results, demonstrating that the same dose is applicable for both developmental ages. Hypothermia alone also proved significantly neuroprotective; however, the addition of a single dose of exendin-4 treatment enhanced hypothermia protection both in the macroscopic injury score and regional infarct volume assessments.
Exendin-4 is an FDA and EMA approved drug used clinically for the treatment of type 2 diabetes mellitus (Furman, 2012). Epidemiological studies have established comorbidity links between type 2 diabetes mellitus and several adult onset neurodegenerative and cerebrovascular disorders, where patients with type 2 diabetes mellitus have increased risk of Alzheimer's disease (Peila et al., 2002), Parkinson's disease (Santiago and Potashkin, 2013) and stroke (Putaala et al., 2011). This suggests shared disease mechanisms, and inflammation, aberrant insulin and insulin-like growth factor 1 (IGF1) signalling and mitochondria dysfunction have shown to contribute to pathogenesis of these conditions (Kruyt et al., 2010;Donath and Shoelson, 2011;Clark et al., 2012;Aviles-Olmos et al., 2013b). Most neonatal studies rely on extensive research of adult experimental models, and despite biological differences, inflammation (Hagberg et al., 2015) and mitochondria dysfunction (Hagberg et al., 2014) are also major hallmarks of HIE. There are other reasonable links between hypoxia-ischaemia and the aforementioned adult disorders. Preclinical rat studies have shown neonatal hypoxia-ischaemia as a precursor to diabetes, metabolic syndrome and stroke (Mcpherson et al., 2009), and IGF1 treatment offers neuroprotection after hypoxia-ischaemia (Brywe et al., 2005). Therefore, exendin-4-mediated prevention of mitochondria damage and stimulation of mitochondria biogenesis (Fan et al., 2010), anti-inflammatory properties and sustained neuroprotection beyond the period of intervention (Athauda et al., 2017) make it an attractive therapeutic approach in the treatment of HIE. In fact, exendin-4 has already been used in neonatal animals to prevent adolescent and adult disorders without adverse effects. Transient neonatal preconditioning with exendin-4 in lamb (Gatford et al., 2013) and rat (Stoffers et al., 2003) intrauterine growth restriction models reduced both visceral fat accumulation, a risk factor for type 2 diabetes mellitus, and oxidative stress (Raab et al., 2009). Neonatal exendin-4 protected against juvenile and adult rat myocardial ischaemic injury and preconditioned mitochondria (Brown et al., 2010). Its administration in neonatal wild-type C57BL/6 mice prevented increased adult body weight and fat mass, and increased energy levels via GLP1R activation (Rozo et al., 2017).
The optimal exendin-4 dose concentration used in this study (0.5 mg/g) is significantly higher than in the treatment of type 2 diabetes mellitus (0.1 mg/kg) and in the ongoing Alzheimer's and Parkinson's disease trials. The doses used in this study are based on the findings by Teramoto et al. (2011), which required the same dose concentration or higher (10, 50 mg/100ml per mouse) to exert protection against transient middle cerebral artery occlusion (Teramoto et al., 2011). We observed that reducing the dose concentration by 10-fold to 0.05 mg/g resulted in abolition of the exendin-4 neuroprotective effect in the HIE model. This high dose requirement may reflect the acute and rapid nature of the injury. Hypoglycaemia is a serious risk for infants suffering from HIE, and blood glucose is continuously monitored in neonatal intensive care units (Tam et al., 2012). Therefore, we monitored blood glucose at different time points after high dose exendin-4 injection, and the known glucose lowering effects of exendin-4 was not observed when compared to saline-treated/basal level controls. This differs from many studies where exendin-4 was associated with substantial reduction in blood glucose levels. However, many of these preclinical exendin-4 studies use diabetic models, which have altered baseline glucose (Young et al., 1999). Clinically, exendin-4 is given to patients with type 2 diabetes mellitus as an adjunct to metformin, sulfonylurea or basal insulin (Inzucchi et al., 2015), and it is the exendin-4 addition to metformin and/or sulfonylurea that enables type 2 diabetes mellitus patients to achieve glycaemic control via improvement of b-cell function (Kendall et al., 2005). These results are compatible with exendin-4's positive effect on glucose homeostasis. Moreover, exendin-4 administration in healthy individuals reduces fasting glucose levels without reaching hypoglycaemic levels (Edwards et al., 2001). Our assessment of any adverse effects to this transient systemically administered high dose exendin-4 regimen did not reveal any toxicity through blood, macrophage-inflammatory or histological analyses. Our results indicate that the high dose regimen used in this study is safe and well-tolerated in mice. These findings are supported by a pharmacology and toxicity review of exenatide (Byetta Õ ) reporting that doses of 450 times the clinical dose in normal glycaemic monkeys produces no hypoglycaemia, neurological signs or pathology (https://www.accessdata.fda.gov/drugsatfda_ docs/nda/2009/021919s000pharmr.pdf). Chronic toxicity studies in monkeys administered very high doses of up to 1000 mg/kg over a 28-day period resulted in no mortality. Any risks are further mitigated as we have advocated high doses administered once or every 12 h over a limited 48 h period (for optimal therapy) when the pathological cascade associated with HIE is most significant. The lack of commercially available kits to measure exendin-4 makes comparison of plasma levels in this study with those in humans difficult. However, the same review shows that a 200 mg/kg subcutaneous administration in mice leads to a C max value of 318 507 pg/ml compared to a human that had received 10 mg/subject and a C max of 251 pg/ml.
In our study, high dose exendin-4 in post-hypoxic-ischaemic injury or naïve mice resulted in an initial weight drop, with a return to baseline level (0 h time-point equivalent) 48 h after start. However, a 10-fold lower dose of exendin-4 (0.05 mg/g) given four times had no effect on body weight when compared to controls, suggesting that this dose might be too low for exendin-4 to sufficiently activate GLP1R in the brain of neonates. Interestingly, P10 mice treated with one high dose exendin-4 used in conjunction with hypothermia showed no weight change in comparison to controls. This could signify that hypothermia may have a modulating effect on exendin-4-mediated activation of GLP1R, although not enough to diminish its neuroprotective effects in the HIE model.
Exendin-4 was administered intraperitoneally and several studies have shown that it can readily cross the bloodbrain barrier (Kastin and Akerstrom, 2003;Zanotto et al., 2017) and interact directly with GLP1R in neural cells. Our measurements of cAMP (a second messenger of GLP1R) in the brain at regular time intervals following a single intraperitoneal administration of exendin-4 to naïve P7 mice showed significantly elevated levels at the earliest 2 h time point. This suggests that exendin-4 rapidly crosses the blood-brain barrier and initiates a pharmacological response. GLP1R is expressed in the human (Wei and Mojsov, 1995), rat (Gö ke et al., 1995) and mouse (Hamilton and Hö lscher, 2009) brain. Similarly, we have shown ubiquitous expression of GLP1R throughout the naïve mouse brain irrespective of developmental age (P7, P10 and 10 weeks) and confirm for the first time in the human preterm neonatal brain. In the neonatal period (P7 and P10) GLP1R seemed to predominantly co-localize with neuronal cells, whereas GLP1R in adult mice also showed co-localization with astroglia. We believe that this is the first study to confirm GLP1R expression in the neonatal mouse brain, and its main co-localization with neurons at this developmental stage.
The precise mechanisms of action of exendin-4 neuroprotection are still not fully understood. Several studies have demonstrated a multitude of neuroprotective actions, with exendin-4-mediated PI3K surge in different areas of the CNS, leading to increase in phosphorylation of AKT via PI3K signalling pathway. This mechanism of action is thought to increase the anti-versus pro-apoptotic bcl-2 family protein balance (Brywe et al., 2005;Athauda and Foltynie, 2016), attenuate neuroinflammation and stabilize the blood-brain barrier post-transient middle cerebral artery occlusion in non- (Chen et al., 2016) and diabetic mice (Li et al., 2016). Intracellular cAMP levels are also raised in exendin-4-treated post-transient middle cerebral artery occlusion as a result of increased GLP1R expression (Teramoto et al., 2011;Kim et al., 2017). Exendin-4 administration also increases islet-brain 1, partially inhibits JNK and attenuates downstream COX-2 and prostaglandin E 2 after transient middle cerebral artery occlusion (Kim et al., 2017). Exendin-4 treatment has resulted in mitochondria biogenesis (Fan et al., 2010;Kang et al., 2015), protection from reactive oxygen species (Teramoto et al., 2011;Li et al., 2013), inflammatory inhibition (Teramoto et al., 2011), neurogenesis (Bertilsson et al., 2008), neurotrophic effects (Perry et al., 2002) and cell survival (Teramoto et al., 2011;Candeias et al., 2017). These neuroprotective effects are of extreme importance in offsetting the mechanisms of HIE injury.
The current study was designed to assess exendin-4 drug preclinical safety and efficacy as a standalone therapy, its capacity to augment therapeutic hypothermia neuroprotection, and not specifically the exendin-4 mechanisms of action. However, aspects of our study suggest that exendin-4 may act to maintain neuronal viability and modulate neuroinflammation. Microglia, astroglia and endothelial cells are a source of inflammatory response and reactive oxygen species after HIE (Dietrich, 2002;Hagberg et al., 2015;Rocha-Ferreira et al., 2016). In our study, four high dose treatments of exendin-4 started either immediately or 2 h post-hypoxia-ischaemia significantly reduced proliferation of microglia and its morphological change into phagocytic amoeboid microglia. GFAP expression in astrocytes also remained low and there was reduced ICAM1 immunoreactivity. ICAM1 upregulation is associated with diapedesis, transendothelial migration and further recruitment of peripheral inflammatory cells (Dietrich, 2002). In neonatal hypoxia-ischaemia, ICAM1 is also associated with alterations of the brain microvasculature and breakdown of the blood-brain barrier (Lai et al., 2017). Our ICAM1 results showed increased ICAM1immunoreactivity, particularly in the areas surrounding the tissue infarction observed in the hypoxia-ischaemia group, an effect not observed in the exendin-4-treated groups. This suggests that exendin-4 treatment prevents breakdown of the blood-brain barrier. This is in agreement with a study by Zanotto et al. (2017) where exendin-4 treatment reversed blood-brain barrier permeability and reduction of bloodbrain barrier-specific proteins in diabetes mellitus rats. The suggested beneficial effect of reduced activation of these cells occurs in parallel with significantly reduced tissue infarction assessed through Nissl staining and cell death (TUNEL assay) in exendin-4-treated mice. The neuroprotective effect of exendin-4 alone seemed moderately longlasting, as seen in the P10 single dose treatment part of the study. Our studies showed predominant neuronal expression of GLP1R. Therefore, it cannot be excluded that the reduced glial cell activation may be secondary to reduced neurodegeneration. Additionally, high dose exendin-4 administration resulted in significantly higher levels of brain cAMP, suggesting involvement of the cAMP signalling pathway, also proposed in other studies (Teramoto et al., 2011;Kim et al., 2017). Interestingly, metformin, another type 2 diabetes mellitus medication with diverse pharmacological activities, also ameliorates brain infarction in neonatal hypoxic-ischaemic rats (Fang et al., 2017).
This preclinical study in a mouse model of acute perinatal HIE is an essential first step to potentially advance the use of exendin-4 for clinical benefit. Using severe hypoxic-ischaemic injury in both the late preterm and term models has shown exendin-4's efficacy in both ages. Preterm treatment with exendin-4 alone is highly attractive as therapeutic hypothermia is not common practise at this age, and the successful 2-h delayed start of exendin-4 treatment offers a larger therapeutic window of opportunity within which to administer the drug. Exendin-4 term treatment combined with therapeutic hypothermia has provided enhanced protection as seen in the brain macroscopic score and regional volume measurement. Other proof of concept studies of exendin-4 have already been translated into early clinical trials for neurological diseases such as Alzheimer's disease (NCT01255163) and, in particular, Parkinson's disease for which initial reports are highly promising (NCT01174810; NCT01971242; Aviles-Olmos et al., 2013a; Athauda et al., 2017). This study and these early clinical trials support the continued investigation of exendin-4 for clinical translation as a treatment for HIE.