The neuroprotective effects of glucagon-like peptide 1 in Alzheimer’s and Parkinson’s disease: An in-depth review

1As a major peptide hormone and growth factor, glucagon-like peptide 1 (GLP-1) regulates several physiological processes in the body and brain. GLP-1 is produced in enteroendocrine cells of the lower gastrointestinal tract and continuously liberated at a low basal level. It is also produced in other cells and organs in the body such as the brain. Known as the incretin effect, GLP-1 and its synergetic sister hormone glucose-dependent insulinotropic polypeptide (GIP), are rapidly released by lower enteroendocrine L-cells (GLP-1) or upper enteroendocrine K-cells (GIP) in the gut following food intake and rising circulatory glucose levels to augment the glucose-stimulated release of insulin by pancreatic β cells. Other associated effects of GLP-1 include the inhibition of glucagon secretion by pancreatic α cells, reduced gastric emptying and intestinal transit, the suppression of appetite, enhanced satiety and further functions in other tissues (see;; ).

2In gastric L-cells, the bioactive forms of GLP-1, GLP7–37 (glycine-extended GLP-1), and GLP7–37 amide, are cleaved from its precursor (pre)proglucagon by prohormone convertase (PC)1 and PC3, which co-generates GLP-2 and oxyntomodulin. On the other hand, GIP is synthesized as a pro-peptide (pro-GIP) that is post-translationally processed to GIP in K-cells. Notably, proglucagon is cleaved in an alternative fashion by PC2 in pancreatic α cells to yield glucagon, but not GLP-1/2. Once GLP-1 is released, the peptide acts on GLP-1 receptors (GLP-1Rs) that are widely distributed in peripheral tissue, including the gut, stomach, pancreas, kidneys, heart, adipose cells, bones, and blood vessels.

3Importantly, while blood-borne GLP-1 readily crosses the blood brain barrier (BBB), incretins and their receptors are expressed in the central nervous system (CNS). A major source of preproglucagon and, thus, GLP-1, is found in PC1/3-expressing neurons in the caudal area of the medial nucleus of the solidary tract (NTS) and, partially, the area postrema, which both belong to the dorsal vagal complex (DVC). Preproglucagon-containing cell bodies were also identified in the granular cell layer of the olfactory bulb, dorsal and ventral medulla and lumbar sacral spinal cord. In this context, the activity of preproglucagon-expressing neurons in the NTS is stimulated by afferent vagal inputs that relay satiety signals from the periphery to the brain, such as gastric distention, the induction of peripheral GLP-1Rs or the release of the satiety-associated hormones cholecystokinin and leptin. Furthermore, NTS neurons form proglucagon and/or GLP-1-positive projections toward the olfactory bulb, various hypothalamic nuclei, the bed nucleus of the stria terminalis, lateral and medial septal nuclei, the amygdaloid complex, the septohippocampal region, nucleus accumbens and, more sparsely, medullary reticular formation, dorsal motor nucleus of the vagus and cortex. In contrast to the NTS-localized production and distribution of GLP-1, the GLP-1R is widely distributed in the CNS. In vivo, the enriched immunoreactivity or transcription of the GLP-1R was detected in neurons of the DVC, paraventricular and posterior thalamic nuclei, various hypothalamic regions, ventral, posterodorsal and interpeduncular tegmental areas, the periaqueductal gray and superior colliculus, while medium to low levels of GLP-1Rs were found in the posterior/caudal hippocampus, the hippocampal CA1, temporal/cerebral cortex, striatum, substantia nigra (SN) [which included tyrosine hydroxylase (SN)-expressing dopaminergic neurons], locus coeruleus, preoptic area, parabrachial nuclei, lateral septum, lateral habenula; zona incerta; substantia innominate, subfornical organ, interpenduncular nucleus, superior colliculus, ventral pallium; nucleus basalis of Meynert, central gray (especially medial), amygdala, spinal cord, the bed nuclei of the stria terminalis, the shell of the nucleus accumbens and the dorsal raphe nuclei. Besides neurons, astrocytes and microglia express GLP-1Rs (see for a general overview). Other studies using rodents or primates with immunofluorescent reporter GLP-1Rs verify the expression of GLP-1Rs by pyramidal (glutamergic), GABAergic and catecholaminergic neurons in feeding, energy homeostasis, fear, stress, motivation/reward, learning plus memory as well as Alzheimer’s disease (AD) and Parkinson’s Disease (PD)-associated key brain areas, such as the hippocampus, hypothalamus, cortex, striatum, amygdala, SNpc, or ventral tegmental area (VTA). Prominent in neurons, but not glial cells, GIP synthesis is more distributed in the CNS compared to GLP-1 and was identified in the hippocampal CA1-3, granule cell layer and neurogenic niche of the dentate gyrus, thalamus, cerebral cortex, cerebellum (Purkinje cells) and brainstem. GIP immunoreactivity was further observed in the olfactory bulb, basal ganglia (including the striatum, nucleus accumbens, ventral pallidum and SN), amygdala, lateral septal nucleus as well as multiple hypothalamic nuclei. Moreover, GIPRs are present in pyramidal cortical and hippocampal, DG granular and progenitor plus cerebellar neurons and further brain regions (see for more information about GIP).

4As Gqα-recruiting and G-protein coupled receptors (GPCR), the downstream signaling pathways of GLP-1Rs and GIPRs parallel those of insulin and lead to the induction of the neuroprotective phosphoinositide 3-Kinase (PI3K)/Akt (also known as protein kinase B)/mammalian target of rapamycin (mTor), cyclic adenosine monophosphate (cAMP)/protein kinase A (PKA)/cAMP-response element binding protein (CREB) as well as Raf/mitogen-activated protein kinase (MEK)1/2/extracellular signal-regulated kinase (ERK) pathways (for the GLP-1R pathways, see also Figure 1;; ).

5Given the role of GLP-1 in glycaemic control, the anti-apoptotic, growth-stimulating and insulin secretion-promoting effects on pancreatic β-cells, the ability to re-sensitize the cellular insulin-signaling, minor side effects, good tolerance by normoglycemic patients and the fact that GLP-1 does not desensitize, synthetic GLP-1R agonists have been on the market for decades to treat type 2 diabetes mellitus (T2DM) or, more recently, to improve weight loss. Due to renal clearance and the cleavage of GLP-1 by dipeptidyl peptidase-4, GLP-1 exhibits a short half-life time (t1/2) of 1–2 min. Thus, multiple proteolysis-resistant GLP-1 analogs, either short- or long-lasting, have been developed for the clinical use. This includes the structurally similar liraglutide (once daily, t1/2 = 13 h), semaglutide (once weekly, t1/2 = 165–185 h), dulaglutide (once weekly, t1/2 = 90 h), the more dissimilar Exendin-4 (Exenatide; twice daily, t1/2 = 2.4 h), which is derived from the saliva of Gila lizards and shows ∼ 53% sequence homology with GLP-1, as well as Lixisenatide (once daily, t1/2 = 3 h). Moreover, novel GLP-1R/GIPR dual agonists have been introduced for the clinical treatment of hyperglycaemia and T2DM, which show synergistic and superior metabolic effects compared to single incretin receptor agonists. In addition, we and others have reviewed the promising effects of GLP-1, GIP, and GLP-1R/GIPR dual agonists in animal and clinical studies for the treatment of various neurodegenerative diseases (see, for an overview).

6Insulin desensitization is found in the brains of patients with AD or PD, and GLP-1 receptor agonists can re-sensitize insulin signaling. Therefore, the use of incretins to normalize insulin signaling in the brain has been tested in animal models of disease and in patients with AD and PD (see section “Clinical trials show good protective effects in patients with AD or PD”). In the context of AD and PD, we have investigated the effects of five GLP-1R/GIPR dual agonists with various modifications, termed DA1-DA5. Of these, the cerebral uptake of DA4-JC4 and DA5-CH was enhanced by the addition of a cell-penetrating sequence (poly-lysine). Crucially, all forementioned GLP-1 analogs and GLP-1R/GIPR dual agonists cross the BBB in vivo, albeit at varying rates. In comparison, DA4-JC4 (t1/2 = 151.94 min) and, similarly, DA5-CH showed the highest BBB penetration rates, followed by medium BBB translocation of Exendin-4, lixisenatide and DA3-CH as well as poor brain influx of the lipidated peptides liraglutide, semaglutide, DA1-JC as well as the PEGylated DA2 and NLY01, a variant of Exendin-4 with prolonged t1/2. We confirmed that GLP-1R/GIPR dual agonists demonstrated superior neuroprotective effects compared to other synthetic GLP-1 analogs in AD and PD animal models, correlating with their greater ability to cross the BBB. A detailed review is found here.

7Oxyntomodulin is a natural dual GLP-1/glucagon receptor agonist. We and others tested a range of protease-resistant oxyntomodulin analogs in different animal models of AD, PD, or stroke. While they showed good protective effects, there was no improvement over single GLP-1 receptor agonists, suggesting that the glucagon receptor binding site does not contribute to the neuroprotective effects. We furthermore tested triple GLP-1/GIP/glucagon receptor agonists and did not find an improvement over dual GLP-1/GIP receptor agonists.

8Herein, given the success of GLP-1 analogs in phase II clinical trials (see section “Clinical trials show good protective effects in patients with AD or PD” for details), we will illustrate the manifold neuroprotective, cognition and motor-enhancing signaling pathways that are induced by the GLP-1R in detail and place them in the context with AD and PD.

9Extensive synapse loss, already apparent during the mild cognitive impairment (MCI) preclinical stage, is one of the earliest pathologic alterations in AD. Indeed, an exosome investigation revealed that the synaptic protein levels decreased years before clinical symptoms manifested and inversely correlated with cognition in AD and frontotemporal dementia patients, whereas Amyloid Beta (Aβ)1–42 or phospho-Thr181-Tau did not. The accumulation of amyloids in AD may harm synapses by various means, such as:

10The interaction of secreted (extracellular) Aβ with α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) on postsynaptic terminals, leading to AMPAR endocytosis, ubiquitination and degradation to impair long-term potentiation (LTP), elongate long-term depression (LTD), and elicit the loss of dendritic spines.

11The Aβ-induced intracellular Ca2+ amassment across N-methyl-D-aspartate receptors (NMDARs), voltage-dependent Ca2+ channel (VDCCs) and internal Ca2+ stores (see also section “GLP-1 mimetics suppress Ca2+ deregulation by amyloid beta and excitotoxicity”) that provokes the calcineurin-mediated phosphorylation of cytoskeleton components and spine loss as well as oxidative stress. The latter, in turn, further stimulates the activity of beta site amyloid precursor protein (APP) cleaving enzyme 1 (BACE1) to encourage the Aβ-generating cleavage of APP.

12The accumulation of hyperphosphorylated and microtubule-detached Tau at dendrites (as accelerated by oligomeric Aβ), which may interfere with kinesin/dynein-mediated transport, synaptic AMPAR insertion and, thus, postsynaptic currents.

13The interruption of the mitochondrial function by Aβ (elaborated in section “Oxidative stress and mitochondrial dysfunction”) and impaired mitochondrial transport to synapses due to Tau pathology, resulting in ATP-depletion and synaptic degeneration (see ).

14In this context, GLP-1R activation has repeatedly shown synaptoprotective properties. Various GLP-1R or GLP-1R/GIPR dual agonists prevented presynaptic synaptophysin and postsynaptic density protein 95 (PSD-95)/drebrin loss in the brains of mice injected with Aβ oligomers as well as multiple hippocampal and cortical regions of APP/presenilin-1 (PS1), 3 × Tg (APPswe, TauP301L, and PS1M146V) or 5xFAD (APPswe, APPI716V, APPV717I, PS11M146L/1L286V) AD rodents. This further includes synaptic protection in the i.c.v. streptozotocin (STZ)-microinjected AD rat model, the hippocampus, frontal cortex, and amygdala of Aβ-infused primates and the brains of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-injected PD animals. Indeed, in vivo ultra-structural investigations confirm that the administration of liraglutide or Exendin-4 partially or fully reversed the decline in synaptic vesicle numbers and postsynaptic areas, swollen membranes, enlarged synaptic clefts and more. In addition, GLP-1R activation preserved the levels of the dendrite marker microtubule-associated protein 2 (MAP2) as well as dendritic spine numbers in AD animal models.

15Mechanistically, as shown in cultured hippocampal neurons, liraglutide rescued the Aβ oligomer-driven downregulation of synaptic proteins and synapse densities in a GLP-1R and cAMP/PKA-dependent manner. In this context, PKA selectively phosphorylates AMPAR-GluR1 subunits at Ser845 to promote their synaptic incorporation and sustain LTP in the cornu ammonis 1 (CA1) (Figure 1;;;; ). Interestingly, cAMP/PKA-signaling, as induced by the GLP-1 analog, also weakened the interaction of applied oligomeric Aβ with neuronal synapses, similar to insulin. This mechanism might contribute to synaptoprotection. Besides Aβ, liraglutide maintained GAP43 expression, a growth cone component found at the tips of axons, in H2O2-challenged RCG-5 retinal ganglion cells. The GLP-1 mimetic reduced axonal degeneration by protecting axonal mitochondria from oxidative damage, while preventing the aberrant enhancement of autophagy/mitophagy.

16In addition to protecting synapses from amyloids and oxidative stress, GLP-1 directly supports neurite growth and synaptogenesis. In PC12, SH-SY5Y and adult sensory neurons, GLP-1R activation was shown to stimulate the cytoskeletal actin/tubulin polymerisation to elicit neurite multiplication, branching and outgrowth in vitro, comparable to NGF. A GLP-1 mimetic further promoted the expression of various synaptic proteins, including synapsin 1, synaptophysin and PSD-95, in SH-SY5Y cells and PSD-95 in the neocortex of adult mice. As we describe elsewhere (section “GLP-1R agonists promote neurogenesis”), these pro-neuritic and pro-synaptic effects are the result of (cAMP/PKA-supported) CREB activation and brain derived neurotrophic factor (BDNF) expression by GLP-1 mimetics.

17There is clear evidence that the GLP-1R modulates both pre- and postsynaptic plasticity. Early in vivo studies in healthy animals demonstrated that GLP-1R activation led to a spontaneous and rapid enhancement of hippocampal firing (1 s), followed by a steady decline in the neuronal activity. Given that a non-NMDA-glutamate receptor inhibitor blocked these effects, it is implied that GLP-1R agonists modulate glutamatergic signaling in the hippocampus.

18Generally, in pancreatic β cells and neurons, GLP-1 elicits insulin or neurotransmitter release, respectively, via (i) the cAMP/PKA-mediated opening of L-type VDCCs and Ca2+ influx, (ii) membrane depolarisation triggered by the PKA-induced inhibition of voltage-dependent potassium channels and possibly (iii) the PKA/Epac-facilitated priming of insulin or neurotransmitter-containing vesicles for their release upon Ca2+ accumulation (see; for an in depth review of the hippocampal synaptic regulation by GLP-1).

19Importantly, in a GLP-1R-dependent manner, long-term treatment of AD-like animals with incretin mimetics did not affect the baseline excitatory postsynaptic potentials, but facilitated the induction and maintenance of hippocampal LTP. These improvements were also observed in healthy, age-matched and GLP-1 analog-injected wilt-type (WT) animals, suggesting that GLP-1 enhances LTP even in the absence of neuropathological alterations. Similar LTP-enhancing effects were reported for GLP-1R/GIPR dual agonists. In contrast, the effects on paired pulse facilitation (PPF), which indicates presynaptic neurotransmitter release, were limited. PPF was not investigated or was not altered to begin within Aβ-based animal models, albeit two studies reported that GLP-1 analogies elevated PPF in APP/PS1 mice and in 18 month-old rodents. In turn, GLP-1R knockout mice displayed grave impairments in LTP induction and short interval PPF (25 ms).

20Moreover, GLP-19–36amide and Val8-GLP-1 were shown to normalize the Aβ-induced deficits in LTP induction, postsynaptic excitatory and inhibitory currents as well as changes in LTD in hippocampal slices. These LTP improvements were mediated by preventing Aβ-induced Ca2+-overload and calpain activation (Figure 1 and details in section “GLP-1 mimetics suppress Ca2+ deregulation by amyloid beta and excitotoxicity”) and by improving the mitochondrial function, oxidative stress and Akt/glycogen synthase kinase-3β (GSK-3β)-signaling (as expanded on in section “Oxidative stress and mitochondrial dysfunction”).

21Notably, the Aβ-associated induction of GSK-3β, as suppressed by GLP-1R/PI3K/Akt-signaling in the hippocampus, provokes the loss of GluR1-AMPARs as well as dendritic and postsynaptic degeneration. The administration of Exendin-4 did not alter glutamate receptor expression nor GluR2-AMPA and NR1-NMDA trafficking, yet promoted the synaptic insertion of GluR1 subunits of AMPA receptors and PSD-95 in the neocortex of both saline and i.c.v. Aβ oligomer-injected adult mice. Exendin-4 further upregulated CREB activity and subsequent BDNF synthesis. Furthermore, the synaptic GluR1-AMPA recruitment was mediated by CREB and independent of PI3K/Akt. In this context, BDNF co-induces GluR1 synthesis via the mTor pathway, modulates the synaptic AMPAR insertion in the hippocampus and improves memory formation, while Aβ suppresses the conversion of proBDNF to BDNF to interfere with AMPAR delivery (see also Figure 1; ).

22As such, the evidence suggests that GLP-1Rs mainly act on a postsynaptic level to potentiate the hippocampal LTP induction and sustain glutamatergic neurotransmission over the long-term by promoting synaptic AMPAR insertion. GLP-1R agonists accomplish the latter by suppressing GSK-3β via the PI3K/Akt pathway and through stimulating the CREB/BDNF pathway. Additionally, by engaging the cAMP/PKA pathway, GLP-1R agonists can restore impairments in glutamate release at presynaptic terminals in the hippocampal CA1 region. GLP-1R agonists have clear protective effects on synapses in animal models of AD and PD.

23In agreement with lessening synaptic injury and plasticity deficits, the treatment with GLP-1R or GLP-1R/GIPR dual agonists rescued cognitive decline, including deficits in spatial learning, recall and memory consolidation in the Morris Water Maze, short-term spatial working memory in the Y-Maze, hippocampus-mediated passive avoidance learning and memory retainment, cued and contextual fear learning and memory, active avoidance memory retention, as well as object recognition memory in various AD in vivo models and SAMP8 mice. A 4 week -long application of liraglutide further improved various cognitive measures in a small-scale pilot study that recruited adults with mood disorders. On the other hand, the spatial and object memory-enhancing effects of GLP-1 or its synthetic analogs were blocked by the co-administration of GLP-1R antagonists, while GLP-1R–/– mice exhibited deficiencies in spatial acquisition and recollection as well as associative and object discrimination memory. Notably, the vast majority of studies suggest that acute injections of GLP-1R or GLP-1R/GIPR dual agonists do not adversely affect anxiety, visual function, exploratory or locomotor behavior.

24Besides direct effects on learning and memory, Exendin-4 or GLP-1/GIP dual agonists further restored Aβ31–35-triggered distortions in the hippocampal circadian rhythm, including improvements in the circadian regulatory and CREB-modulating proteins Per1/2 and the synaptic re-modeler growth associated protein 43 (GAP-43). Moreover, GLP-1R activation enhanced the decreased theta band frequencies in response to the CA1 microinjection of STZ. Notably, theta rhythms are generated by projections of cholinergic and GABAergic neurons from the nucleus basalis toward the hippocampus, regulating the magnitude of evoked hippocampal action potentials. Indeed, the disturbance of theta rhythm was shown to impair hippocampal spatial learning, memory and movement.

25Due to its pivotal role in the neuronal growth, proliferation, differentiation, plasticity, neurogenesis and memory, the cAMP/PKA/CREB pathway is the most sought-after drug target for AD. The induction of the cAMP/PKA/pCREB pathway not only mediates plasticity and long-term memory formation in the hippocampus, but is also mandatory for neuronal survival. In the hippocampus, the CREB-associated expression of plasticity genes, including c-fos, activity-regulated cytoskeleton-associated protein and BDNF, drive declarative memory consolidation to convert short-term into long-term memory. Indeed, the genetic upregulation of CREB activity elevated long-lasting LTP and long-term memory consolidation in the hippocampal CA1 region, while the concomitant transcription of BNDF supports memory formation by CREB, while facilitating short-term memory.

26Unsurprisingly, cAMP/PKA/pCREB-signaling is downregulated in AD patients in key areas such as the hippocampus. The latter signaling pathway is impeded by Aβ-triggered cytosolic Ca2+ overload and the concomitant induction of the Ca2+-sensitive and CREB-proteolyzing enzyme calpain. Likewise, CREB induction was shown to be responsible for glucose transporter type 3 (GLUT3) expression, negatively associated with astrogliosis and impaired by oxidative stress.

27From a mechanistic standpoint, a study demonstrated that the GLP-1 mimetic-evoked improvements in associate memory retainment post training were dependent on GLP-1R/ERK-signaling in the murine hippocampus. Also promoted by GLP-1R activation, it was recently shown that the activation of ERK following cAMP/Epac-signaling was required to convert short-lasting and protein synthesis-independent LTP into protein translation-dependent forms of LTP to sustain plasticity. Indeed, ERK controls cognition by navigating protein synthesis, gene expression, dendritic spine remodeling, ion channel regulation and receptor insertion, such as those of AMPARs in synapses. As we elaborate in the respective sections, GLP-1 analogs further maintain the activity of the memory master regulator CREB (which lies downstream of ERK) by protecting from amyloid-provoked Ca2+ deregulation (section “GLP-1 mimetics suppress Ca2+ deregulation by amyloid beta and excitotoxicity”), oxidative stress (section “”Oxidative stress and mitochondrial dysfunction”) and cerebral insulin resistance (section “Insulin resistance and the neuronal energy metabolism”). Moreover, GLP-1R activation concertedly stimulates the PI3K/Akt, ERK, and cAMP/PKA pathways to enhance CREB activity and BDNF synthesis in hippocampal neurons in vitro and in vivo, even in non-pathological conditions (see section “Other growth factors”).

28Interestingly, the s.c. injection of liraglutide or DDP-4-inhibitors (which are mostly poorly BBB-penetrant, but block the degradation of peripheral GLP-1) enhanced the expression of GLP-1Rs in the hippocampus of WT, APP/PS1 or Aβ1–42 i.c.v injected mice. Since the adenovirus-mediated overexpression of GLP-1Rs by hippocampal neurons enhanced spatial, but not associative, learning in vivo, it can be hypothesized that the GLP-1-mediated upregulation of GLP-1Rs might be beneficial for memory acquisition. Interestingly, GLP-1R mRNA expression was found to be increased more than 10-fold in the SN of patients with PD compared to controls after treatment with a GLP-1R agonist.

29As stipulated by the amyloid hypotheses, the abnormal production, accumulation and aggregation of Aβ1–40 and Aβ1–42 monomers to oligomers and plaques, as a consequence of impaired Aβ degradation and clearance, has been thought to be a central pathological event in AD. Briefly, Aβ and the soluble fragment sAPPβ may be generated by the sequential cleavage of APP by BACE1 and γ-secretase (amyloidogenic pathway), whereas APP-processing by α-secretases [a disintegrin and metalloproteinase (ADAM) protein] and γ-secretase lead to the production of p3 and sAPPα (non-amyloidogenic and non-toxic pathway). As backed up by countless clinical failures, it is now clear that Aβ is a contributing, but not a leading, factor in the development of AD.

30Multiple studies confirm that GLP-1 analogs prevent neuronal death and the Aβ pathology in vivo. GLP-1 analog treatment prevented neuronal atrophy in the hippocampus and cortex of 3 × Tg mice and 5xFAD mice as well as the hippocampal CA1 region of rats following the i.c.v.-injection of STZ. A GLP-1 analog further rescued the cerebral, renal and splenic vasculature from Aβ-induced lesions and leakage in APP/PS1 mice, indicating vasoprotective effects. Moreover, various GLP-1R or GLP-1R dual agonists lowered the cerebral levels of APP in the cortex or across the brain, reduced soluble monomeric Aβ in the prefrontal cortex, hippocampus or centrally, decreased cortical or hippocampal Aβ oligomer pools and diminished the Aβ plaque load in the cortex, hippocampal CA1 area, dentate gyrus (DG)or globally in APP/PS1, 3 × Tg-AD, 5xFAD and i.c.v. STZ-injected rodents. Interestingly, Liraglutide preserved hippocampal pyramidal neurons in SAMP8 mice, an animal model sporadic AD that does not exhibit Aβ or Tau pathologies. This suggests that the GLP-1R-meditated neuroprotection in Aβ-based animal models presumably involves Aβ-dependent and Aβ-independent neuroprotective mechanisms.

31As illustrated in Figure 1, GLP-1R activation suppresses the cerebral Aβ pathology by various means. First, GLP-1 analogs enhanced the cortical and hippocampal expression of insulin degrading enzyme (IDE) in WT, APP/PS1 and 5xFAD mice. In this context, nephrilysin and IDE pose the main Aβ-degrading enzymes in the brain. The intraneuronal enzyme IDE preferably cleaves Aβ1–40 and Aβ1–42 monomers, but not Aβ oligomers or fibrils, implying that reductions in IDE increase the Aβ oligomer/monomer ratio. Indeed, IDE levels and concentrations were shown to decline in the hippocampus of MCI and AD patients, inversely correlating with the amount of Aβ1–42. As a possible link, in vivo studies suggest that the development of T2DM accelerates Aβ accumulation in the brain by reducing IDE and enhancing γ-secretase activity. Notably, as demonstrated in vitro, the stimulation of IR by insulin results in the PI3K/Akt-mediated upregulation of IDE as a negative feedback loop to prevent chronic IR-signaling. However, insulin-signaling and the induction of the PI3K/Akt pathway are impaired during AD. Indeed, reduced PI3K levels, as indicative of CNS insulin resistance, were observed in the brains of AD patients and in animals, correlating with lessened IDE and elevated Aβ quantities in vivo. In turn, GLP-1R agonists recover the expression of IDE by restoring Aβ-driven impairments in insulin sensitivity and boosting PI3K/Akt-signaling in the brain.

32Second, it was shown that Exendin-4 encouraged the membrane trafficking and cleavage activity of the APP-shedding enzyme ADAM10, the main α-secretase, in the neocortex of WT and oligomeric Aβ-injected mice.

33Third, GLP-1 mimetics interfere with the amyloidogenic turnover of APP by BACE1. The application of GLP-1 or exendin-4 decreased amyloidogenic APP processing in PC12 cells, as implied by the reduced intracellular and secreted sAPPβ levels. An in vitro study demonstrated that, by re-sensitizing the insulin pathway in neurons, liraglutide diminished the abnormally increased activity of BACE1, the key amyloidogenic secretase that generates Aβ from its precursor APP in concert with γ-secretase. This led to reduce Aβ/APP conversion ratios and Aβ plaque formation. Likewise, liraglutide or NLY01 attenuated BACE1 expression and APP turnover in the hippocampus of 5xFAD mice and blocked the neuronal upregulation of BACE1 in response to okadaic acid (OA)-triggered Tau hyperphosphorylation in vitro and in vivo. Notably, similar to the pro-inflammatory cytokine tumor necrosis factor alpha (TNF-α), GSK-3β is implicated in the activation of NF-κB, resulting in the increased expression of BACE1, APP cleavage, Aβ production, plaque formation, and impaired memory. In turn, the latter adverse changes could be prevented with the pharmacological inhibition of GSK-3β in APP23/PS45 mice. In this context, GLP-1R activation was shown to promote PI3K/Akt-signaling in neurons to inactivate the BACE1-upregulating GSK-3β in vitro and in vivo.

34Fourth, as another link to amyloidogenesis, GLP-1R induction normalizes the rate of N-glycosylation. A recent study showed that GLP-1 agonists preserved β-catenin levels through the Akt-mediated inhibition of GSK-3β. As dependent on the Akt/GSK-3β/β-catenin pathway, this resulted in the normalization of the aberrantly elevated N-acetylglucosaminyltransferase III (GnT-III) activity and concomitant increase in bisecting N-acetylglucosamine (GlcNAc) levels in GLP-1 mimetic and Aβ25–35-co-treated neurons in vitro and the hippocampus and cortex of APP/PS1 mice in vivo. While the mechanism is still elusive, it was shown that the nuclear translocation of β-catenin, as maintained by Wnt-pathway agonism that disassembles the “destruction complex” (GSK-3β, axin, diversin, and polyposis coli) or the pharmacological inhibition of the GSK-3β-mediated phosphorylation of β-catenin that targets it for proteasomal degradation, attenuated GnT-III expression, Aβ accumulation, plaque formation, gliosis and spatial memory deficits in APP/PS1 mice. In the context of AD, amyloidogenic proteins, including BACE1, APP, and Tau, were shown to be N-glycosylated by GnT-III. Interestingly, the GlcNAc-modification of BACE1, as stimulated by Aβ-triggered oxidative stress, hampered its lysosomal targeting and degradation. In addition, elevated GnT-III activity was shown to impair growth factor signaling, such as the blockade of nerve growth factor (NGF) receptor dimerization and epidermal growth factor (EGF) receptor phosphorylation or aberrantly increased EGF receptor internalization, which seemingly upregulates the induction of ERK. Given that AD patients displayed enhanced hippocampal GSK-3β activity, decreased β-catenin levels, heightened GnT-III expression and increased bisecting GlcNAc pools, treatment of Wnt pathway dysfunction, which might be achieved with GLP-1R activation, has been proposed as a therapeutic strategy for AD.

35Fifth, GLP-1 agonists block the induction of c-Jun N-terminal kinase (JNK) to prevent the generation of Aβ. In AD, JNK is induced by oxidative stress derived from various sources, for example in response to the Aβ or Tau pathology, neuroinflammation or mitochondrial dysfunction. Moreover, glial inflammation, as provoked by oligomeric Aβ and plaques, leads to the release of pro-inflammatory cytokines [i.e., interleukin (IL)-1β, IL-18, and TNF-α] that activate the corresponding pro-inflammatory cytokine receptors (PICRs) on neurons. Downstream inflammatory signaling by various modulators results in the mitogen-activated protein kinase kinase (MKK)-driven phosphorylation of p38 and JNK. Ultimately, JNK/p38 co-induce the activator protein 1 (AP-1)-mediated transcriptional upregulation of APP, while PICR stimulation separately provokes the NF-κB-driven expression of BACE1 in neurons. As an insulin receptor substrate 1 (IRS-1)-inactivating serine kinase, JNK, in cooperation with the inflammation/PICR-induced serine kinases inhibitor of κB–kinase β (IKKβ) and protein kinase R (PKR), also induces neuronal insulin resistance to exacerbate the production, accumulation and aggregation of Aβ (see Figure 1 and details in section “Insulin resistance and the neuronal energy metabolism”) as well as Tau hyperphosphorylation (section “GLP-1R mimetics suppress Tau hyperphosphorylation and aggregation during AD”). Lastly, Aβ1–42-driven endoplasmic reticulum (ER) stress is linked to the activation of JNK3, which was shown to augment APP phosphorylation by JNK3 at Thr668 to encourage amyloidogenic processing. Indeed, Aβ was shown to induce the JNK/TNF-α pathway in neurons to elicit the inactivation of IRS-1, CNS insulin resistance and memory decline, while Exendin-4 suppressed JNK, blocked the cerebral desensitization of insulin-signaling and restored spatial memory in APP/PS1 mice. Likewise, Liraglutide normalized the decreased ERK activity and the elevated levels of phosphorylated JNK1/2 in the brains of 3 × Tg triple transgenic mice. A study in rotenone-based PD models suggests that GLP-1R activation reduces oxidative stress in an Akt-dependent manner to prevent the detrimental activation of JNK. Moreover, as elaborated in section “Inflammation,” GLP-1 analogs resolve glial inflammation and cytokine production, which boost JNK activation via PICRs in neurons, in various contexts.

36It has been proposed that differences in the expression of insulin genes explain the vulnerability of distinct brain regions to the Aβ and Tau pathology, as exacerbated though the development of CNS insulin resistance during AD. For example, cell culture studies have demonstrated that insulin resistance in neurons creates an AD-like phenotype that exhibits attenuated IR/IRS-1/PI3K/Akt-signaling, enhanced activity of the Tau kinase GSK-3β activity due to the loss of Akt-signaling, deregulated ERK1/2, acetylcholinesterase and pro-inflammatory nuclear factor kappa B (NF-κB) activities, impaired plasma membrane recruitment of GLUT4 by insulin, the reduced expression of insulin-independent GLUT3, diminished glucose uptake, elevated BACE1 activity, increased Aβ production, secretion and plaque formation as well as, of course, heightened Tau hyperphosphorylation and aggregation. Notably, T2DM animal models show elevated Tau phosphorylation and inactivation of the insulin pathway (IR and Ser-IRS-1 phosphorylation) in the brain (summarized in ), while phospho-Ser-inhibited IRS-1 co-localized with neurofibrillary tangles (NFTs) in pyramidal neurons in the brains of AD patients.

37As repeatedly shown, GLP-1R and GLP-1R/GIPR dual agonists reversed Tau hyperphosphorylation in vivo. This involved a decrease in the cortical or hippocampal Thr181, Thr181/Ser396, Thr231, Ser199/202/Ser396, Ser199/202/Ser404, Ser202/Thr205, Thr212/Ser214, Ser396, Ser199/202/Ser396, and Ser199/202/Ser214/Ser396-phosphorylated Tau levels as well as hyperphosphorylated Tau neurofilament pools, as accumulating in the cell bodies and neurites of neurons, in APP/PS1 mice, 3 × Tg mice, Aβ1–42-microinjected rodents, i.c.v. STZ-injected rats, a 3 × Tg AD/T2DM crossover animal model, OA-microinjected rodents or hTauP301L mice. Liraglutide further diminished the numbers of neurons containing Ser202 and Ser212/214-phosphorylated Tau as well as oligomeric Tau and fragmented Tau in the DG and amygdala of primates. Moreover, GLP-1R induction reduced the numbers of Tyr231- pTau-positive neurons and the formation of Tau inclusions in the hippocampal tissue of Tau APP/PS1 and 3 × Tg mice. Notably, the injection of GLP-1 mimetics lowered the elevated total Tau levels in some AD animal studies, but had no effect on soluble Tau in others. In support of their Tau pathology-ameliorating properties, liraglutide reduced clasping behavior and massively extended the survival of hTauP301L mice, whereby the GLP-1 analog-mediated reduction in phospho-Tau levels correlated with life span improvements.

38Mechanistically, the latter AD-like rodents displayed reduced PI3K and Akt, but enhanced activity of the Tau kinase GSK-3β in the hippocampus, as indicative of desensitized insulin signaling. On the other hand, the application of incretin analogs re-induced the PI3K/Akt pathway in a GLP-1R-mediated manner and increased the expression of GLP-1 and GLP-1R. This led to the Akt-driven Ser9-phosphorylation and inactivation of GSK-3β, while the levels of activated (Tyr216-phosphorylated) GSK-3β, as stimulated by Aβ, were reduced. Indeed, liraglutide re-invigorated the insulin-evoked phosphorylation and activation of IRs, IRS-1, Akt as well as the inhibitory Ser9-phosphorylation of GSK-3β by Akt in insulin-resistant SH-SY5Y cells. The GLP-1-elicited normalization of insulin-signaling resulted in lessened Ser396 phospho-Tau pools, albeit it did not normalize the cellular glucose uptake. Notably, the effects of GLP-1 mimetics in AD models are mixed, which might be related to the treatment protocol. For example, GLP-19–36amide and Exendin-4 enhanced the hippocampal Akt and reduced GSK-3β activity in APP/PS1 mice, while liraglutide elevated the inactivated phopsho-Ser9-GSK-3β levels in the cortex or hippocampus of WT, 3 × Tg and 5xFAD animals. However, liraglutide failed to affect the induction of GSK-3β in other studies. Also, while Exendin-4 normalized the hippocampal expression levels of total GSK-3β in the i.c.v. STZ-injected rat AD/T2DM model, liraglutide did not reduce the elevated GSK-3β pools in Aβ-based AD animals. Notably, GLP-1 mimetics do not affect the cerebral Tau protein phosphatase 2A (PP2A) levels in vitro or in vivo. However, the administration of liraglutide prevented neuronal apoptosis and cognitive deficits in response to the PP2A-inhibitor OA in vitro and in vivo. Synoptically, the presented evidence suggests that GLP-1 agonists do not affect Tau phosphatase activity and predominantly suppress Tau hyperphosphorylation by enhancing insulin-sensitivity and the Akt-driven inhibition of GSK-3β during AD (Figure 1).

39Another indirect mechanism that prevents Tau hyperphosphorylation is the suppression of the p38 and JNK pathways as well as neuroinflammation by GLP-1R agonists (pathways depicted in Figure 1). It was discovered that increased JNK and p38 activities co-localize with NFT-bearing, but not apoptotic, neurons in tissue samples of patients with various Tauopathies, including AD. For example, the pro-inflammatory cytokine IL-1β, as released by Aβ or lipopolysaccharide (LPS)-stimulated microglia, induces the activation of p38 to drive Tau phosphorylation and synapse loss. Similarly, IL-18 elevated the expression of the Tau serine/threonine kinases GSK-3β, cyclin-dependent kinase 5 (Cdk5) as well as its regulatory subunit p35. Notably, along with the Tau phosphorylation at Thr212 and Ser214 by PKA, p38δ also appears to phosphorylate Tau at Thr217 to create the AT100 epitope that is commonly detected with Tau antibodies. Of note, p35 may be truncated to produce the more proteolytically resistant p25, which seems to accumulate in the brains of AD patients and constitutively activates Cdk5. It is thought that the Ca2+-sensitive calpain, as induced by Aβ-triggered Ca2+-overload, cleaves p35 into p25. Other negative events include the calpain-mediated cleavage of GSK-3β to increase its activity, Tau fragmentation [which may be neurotoxic and/or impair oxidative phosphorylation (OXPHOS)], the joint downregulation of dynamin 1, a presynaptic vesicle recycler, with Aβ, the proteolysis of synaptic proteins [i.e., PSD-95, NMDAR subunits NR1/2A/2B, metabotropic glutamate receptor subtype 1 (mGluR1)] and cleavage of PKA to interrupt CREB activity. On the other hand, studies in various disease models indicate that JNK is selectively stimulated by Aβ as well as proinflammatory cytokine exposure, but not Tau, in neurons in vivo and post-mortem hippocampal brain tissue of AD patients. The JNK isoforms (JNK1/2/3) act as Tau kinases, phosphorylating Tau at Ser199, Thr212 and Ser422 to accelerate Tau aggregation and microtubule depolarisation. Importantly, inflammatory cytokine/PICR-signaling in neurons and the concomitant activation of the IRS-1-inactivating serine kinases JNK, IKKβ, and PKR lead to cerebral insulin resistance in AD which, in turn, promotes Tau hyperphosphorylation by enhancing the activity of the Tau kinase GSK-3β. JNK further drives apoptosis by phosphorylating and activating the apoptosis-drivers BIM [a Bax and Bak activator and B-cell lymphoma 2 (Bcl-2)/myeloid cell leukemia 1 (Mcl-1) inhibitor] and Bcl-2-modifying factor (BMF) as well as suppressing the activity of their anti-apoptotic counterparts, i.e., Bcl-2. Aberrant JNK activation, due to the phosphorylation of PSD-95, also disrupts the recruitment of AMARs and NMDARs, triggers synapse (PSD-95 and drebrin) and spine degeneration and induces LTP impairments. In the aforementioned context, GLP-1 analogs were shown to inhibit the Tau pathology-associated activities of p38 in the hippocampus of APP/PS1 mice and JNK in multiple Aβ-induced AD animal models. Additionally, GLP-1R activation on glial cells is immunosuppressive (section “Inflammation”), thus forestalling the cytokine/PICR-induced activation of p38 and JNK, the upregulation of GSK-3β, Cdk5 and p35 as well as, of course, the desensitization of insulin-signaling in the brain (as expanded on in section “Insulin resistance and the neuronal energy metabolism”).

40Amongst other toxic implications, soluble oligomeric Aβ provokes neuronal death by stimulating Ca2+-overload in AD. Aβ may drive Ca2+ accumulation by forming pores in the plasma membrane. Furthermore, Aβ triggers reactive oxygen species (ROS) production due to the reaction with iron/copper ions or by impairing the mitochondrial electron transport chain (ETC), leading to plasma membrane lipid peroxidation, the subsequent inhibition of Ca2+ ATPases, membrane depolarisation and Aβ/ROS-induced NMDAR and L-type VDCC opening and excessive Ca2+-influx. The Aβ-induced Ca2+-influx across NMDARs seems to evoke a compensatory transcriptional downregulation of ryanodine receptor (RyR) subunits in the ER membrane, while, simultaneously, stimulating ER Ca2+ efflux into the cytosol across RyRs and inositol 1,4,5-trisphosphate (IP3R) receptors. This resulted in oxidative stress, mitochondrial fragmentation, mitochondrial membrane permeabilization and intrinsic apoptosis in vitro. Indeed, a recent study discovered that AD patients displayed upregulated efflux and downregulated influx Ca2+ transporter gene expression in mitochondria, suggesting that these are adaptive responses to maintain the mitochondrial function in response to the pathological cytosolic Ca2+ amassment in AD. Notably, Aβ-induced Ca2+ overload impairs the synaptic function by inducing calpain and the degradation of dynamin 1, which is implicated in synaptic vesicle recycling.

41In vitro studies confirm that GLP-1 treatment prevents pathological Ca2+ overload in neurons. GLP-1 and Exendin-4 dose-dependently rescued from Aβ25–35 or Aβ1–42-evoked apoptosis in primary hippocampal neurons and SH-SY5Y cells. The latter neuroprotection involved GLP-1R-induced cAMP accumulation, Akt and MEK1/2 activation. Importantly, Aβ suppressed the activation of the Akt/MEK1/2-signaling pathway and triggered Ca2+ release from the ER stores, whereas GLP-1R activation, as dependent on MEK1/2, prevented Ca2+ deregulation. Moreover, GLP-1 blocked the transcriptional increase in p53 and Bax that was provoked by Aβ. In this context, the pro-apoptotic protein Bax is upregulated upon cell stress by p53-mediated gene expression, while the p53/Bax pathway is downregulated in response to the activation of Akt. Besides shielding against Aβ-mediated ER Ca2+ efflux, GLP-1 prevented excitotoxic death by blocking glutamate-elicited Ca2+ influx through VDCCs and membrane depolarization in cultured hippocampal neurons. Notably, while GLP-1R activation maintains the neuronal Ca2+ homeostasis and prevents excessive L-VDCC activity and Ca2+ instream following K+-ion overload, Aβ or glutamate-associated stress, GLP-1-mediated cAMP/PKA-signaling evokes a rapid, but transient (2 min), elevation of intracellular Ca2+ through the opening of L-type VDCCs and ionotropic glutamate receptor channels (AMPA, NMDA, and kainite receptors) in hippocampal neurons. The short-term Ca2+ accumulation that is triggered by GLP-1 may evoke the spontaneous presynaptic release of glutamate in the hippocampus (see section “GLP-1 enhances hippocampal synaptic plasticity”).

42The above studies suggest that the neuroprotective effects of GLP-1 against Ca2+ overload are mediated by CREB. Previous studies have shown that CREB induction in response to metabolic stress and increased Ca2+ influx across NMDA receptors mediates survival under hypoxic and excitotoxic conditions by heightening the expression of activity-regulated inhibitor of death survival genes, such as bdnf, atf3, btg2, gadd45β/γ or Bcl-2. Indeed, the overexpression of CREB improves the neuronal tolerance of apoptotic stimuli, while non-functional CREB accelerates death. Notably, CREB is inactivated and degraded in hippocampal neurons following NMDA treatment, suggesting that glutamate overload drives apoptosis by impairing CREB activity. Likewise, oxidative stress (H2O2) impairs CREB induction by growth factors and dose-dependently downregulates CREB expression, thus enhancing the neuronal vulnerability to apoptosis. Moreover, Tau accumulation appears to interfere with CREB and drives synapse and memory impairments by stimulating the activity of the Ca2+-induced phosphatase calcineurin that dephosphorylates both CREB and calcium/calmodulin-dependent protein kinase IV. Given that MEK1/2 inhibitors suppressed the neuroprotective effects of lixisenatide against Aβ, it is implied that GLP-1 agonists achieve neuronal survival through the cAMP/exchange protein activated by cAMP (EPAC)/Raf/MEK1/2 and PI3K/Akt/Raf/MEK1/2 pathways. MEK1/2 subsequently induces ERK1/2, the latter activates p90RSK (also known as MAPKAP-K1) and p90RSK activates CREB through phosphorylation at Ser133. cAMP/PKA-signaling following GLP-1R induction is further involved in neuroprotection, with PKA regulating L-type VDCC activity and inducing the Ser133-phosphorylation of CREB (see also Figures 1, 2;;;; ). As a side note, while GLP-1 agonists stimulate short-term (2 min) Ca2+ influx via the opening of L-type VDCCs, studies in β-cells indicate that the local activation of L-type VDCCs, but not ER Ca2+ efflux or changes in the intracellular Ca2+ levels per se, are necessary for GLP-1R ligands to sustain ERK activation. Lastly, given that the amyloid intracellular C-terminal domain that is produced following APP/Aβ-cleavage interferes with Ca2+ gene transcription, the reduction in APP expression and processing by GLP-1 mimetics might contribute to the maintenance of the Ca2+ homeostasis in neurons.

43Therefore, GLP-1 agonists protect from excessive Ca2+ accumulation and mitochondrial dysfunction by preventing the Aβ-driven Ca2+ release from the ER as well as aberrant Ca2+ influx across NMDARs/VDCCs in response to glutamate overload. Moreover, GLP-1R activation stimulates various key modulators (PI3K/Akt, cAMP/PKA and MEK1/2/ERK1/2/p90RSK) that jointly induce CREB-mediated survival gene expression.

44ER stress is a common pathological feature in most neurodegenerative diseases, including AD and PD. In the latter diseases, the accumulation of abnormally folded proteins, including protease-resisting Aβ, Tau and α-synuclein oligomers, may exceed the buffering capacity of cytosolic chaperones and the proteasomal degradation system to exacerbate amyloid amassment and trigger ER stress. To prevent damage, ER stress initiates the unfolded protein response (UPR), leading to the transcriptional upregulation of protein-re-folding ER chaperones and the general repression of protein translation. However, while initially protective, chronic UPR activation results in proteasomal dysfunction, oxidative stress, intracellular Ca2+ overload, mitochondrial damage and, ultimately, apoptosis.

45ER stress and the UPR involve the activation of three major pathways that both mediate cell survival by inducing autophagy, but also trigger apoptosis under prolonged cells stress. First, the PERK/eIF2a/activating transcription factor 4 (ATF4) pathway results in eukaryotic translation initiation factor 2a (eIF2a) phosphorylation, inhibition of eIF2a-dependent protein translation as well as the ATF4-conveyed expression of autophagy genes, but also the transcription of pro-apoptotic agents, such as CAAT/enhancer-binding protein (C/EBP) homologous protein (CHOP). Second, the accumulation of unfolded proteins is detected by the ER sensor binding immunoglobulin protein (BiP, also known as GRP78), resulting in inositol-requiring enzyme 1α (IRE1α)-associated signaling cascades. This leads to the autophagy-enabling and IRE1α-elicited degradation of XBP-1, the IRE1α-mediated induction of apoptosis signal-regulating kinase 1/JNK-signaling and the JNK-mediated dissociation of Bcl-2/Beclin-1 complexes to initiate autophagy. On the other hand, IRE1α may enhance pro-apoptotic stimuli, for example by inducing bcl-2 interacting mediator of cell death (BIM) and upregulating CHOP levels. Third, the activation of ATF6 enhances XBP-1-associated autophagy, but also induces CHOP. As a key driver of UPR-related cell death, CHOP evokes (i) the decreased transcription of anti-apoptotic Bcl-2, which is coupled to enhanced ROS-production by mitochondria and concomitant glutathione (GSH) depletion; (ii) the direct enhancement of pro-apoptotic BIM and PUMA synthesis, which stimulates Bax/Bak to induce pore formation and pro-apoptotic mitochondrial cytochrome c (CytC) release; (iii) the upregulation of the Akt-inhibitor tribbles-related protein 3 to enable forkhead box protein (FOXO)3a-driven PUMA induction; (iV) the increased expression of death receptor 5 and, thus, caspase 8-mediated apoptosis and (V) the activation of oxidoreductase 1α (Ero1α) to drive the IP3R-mediated liberation of ER Ca2+ stores, cytosolic Ca2+ amassment and the induction of the Ca2+-sensitive calpain/caspase 12 apoptosis pathway. Notably, Aβ-triggered cytosolic Ca2+ accumulation, as also induced by oxidative stress-induced mitochondrial injury, result in the Ca2+-induced swelling of mitochondria, membrane potential collapse, mitochondrial dysfunction and exacerbated ROS-production. Mitochondrial dysfunction, in turn, stimulates CHOP expression and apoptosis via the mitochondrial pathway [Bax/Bak-driven membrane-permeabilization, CytC release, apoptotic protease-activating factor-1 (APAF-1) and caspase 3 activation, as depicted in Figure 2;;; ]. Moreover, the chronically enhanced GSK-3β activity that is found in neurons during AD facilitates ER stress by upregulating CHOP synthesis, whereas the inhibition of GSK-3β was shown to prevent ER stress-induced apoptosis.

46Notably, the UPR response and expression of ER chaperons, for instance BiP and phospho-eIF2α, were shown to be elevated in APP/Aβ-based cell and animal models as well as the temporal cortex and hippocampus of AD patients. As an unfolded protein sensor, BiP initially has a useful function and was reported to detect Aβ1–42 and interact with APP to suppress the generation and secretion of Aβ by β/γ-secretase. On the other hand, phospho-eIF2α was demonstrated to aggravate BACE1 expression, Aβ generation and plaque formation in neurons. Increased phospho-eIF2α levels were also found in the brains of AD patients, seemingly triggered by insulin resistance-associated impairments in the neuronal ATP production and glucose metabolism. The increase in phospho-eIF2α pools further correlated with elevated BACE1 and plaque levels in vivo and in post-mortem brain tissue of AD patients.

47Animal experiments have confirmed that GLP-1R activation guards against amyloid-triggered ER stress and apoptosis. In the APP/PS1 animal model, the injection of a GLP-1R/GIPR dual agonist restored the reduced cerebral Akt and (inactivated) Ser9-phospho-GSK-3β levels to the levels of control mice, without affecting ERK1/2. Indeed, GLP-1R agonists are potent stimulators of the neuroprotective modulator Akt that directly suppresses the abnormally increased GSK-3β activity in the brain of AD animal models. Moreover, the synthetic dual incretin supressed the Aβ-triggered upregulation of the unfolded protein-sensor BiP, CHOP as well as caspase 12 in APP/PS1 mice, implying that ER stress was resolved. Liraglutide further downregulated the levels of BiP and phospho-eIF2α in the hippocampus of APP/PS1 mice.

48GLP-1 further prevents Ca2+-associated ER stress. As we describe in section “GLP-1 mimetics suppress Ca2+ deregulation by amyloid beta and excitotoxicity,” GLP-1 protects from Aβ or excitotoxicity-driven aberrant Ca2+ efflux from ER stores and Ca2+ instream though NMDARs/L-VDCCs. Cellular experiments with thapsigargin, which triggers excessive ER Ca2+-release, give further insight into the signaling mechanisms linked to ER stress. Following thapsigargin treatment, liraglutide suppressed the aberrantly upregulated expression of the ER stress indicator BiP [which lies upstream of IRE1α, PERK, and ATF6], activating transcription factor 6 and CHOP, while restoring the impaired activation of IRE1α and ATF6 in SH-SY5Y cells. Mechanistically, liraglutide induced Akt, normalized the lowered (Akt-induced) inhibitory phosphorylation of GSK-3β and pro-apoptotic Bcl-2-associated death promoter (Bad), recovered the reduced anti-apoptotic Bcl-2, p53 and signal transducer and activator of transcription (STAT)3 activities and diminished the basal expression of the apoptosis-eliciting protein BH3 interacting-domain death agonist (Bid). The impeded phosphorylation of ERK1/2 could not be restored, however. The GLP-1 mimetic further normalized the enhanced protein levels of the protein quality-control chaperones Ero1α and lessened calnexin pools, which indicates that the Ero1α/IP3R-driven, aberrant Ca2+ secretion from the ER and calnexin deprivation-associated protein misfolding and proteasome overload were ameliorated. The excessive activity of the stress survival transcription factor NF-E2-related factor (NRF-2), which is provoked by ER and oxidative stress, the amassment of misfolded or aggregated proteins or mitochondrial ROS generation, was also quenched by liraglutide. As such, liraglutide rescued cell death, improved the cellular proliferation, blocked the apoptosis-inducer caspase 12 and the subsequent degradation of poly ADP ribose polymerase in thapsigargin-assaulted neuroblastoma cells.

49Notably, despite the fact that GLP-1R activation suppresses ER stress, Exendin-4 enhanced the expression of the ER stress-associated transcription factor ATF4 to protect H2O2-subjected neuroblastoma cells from death. In this context, the PERK/eIF2a/ATF4 axis is required for the adaption and resistance of cells toward oxidative stress through the ATF-4 mediated transcription of autophagy and redox genes, such as GSH. However, ATF4 also controls the synthesis of pro-apoptotic factors, including CHOP, to initiate cell death upon persistent ER stress. Studies in pancreatic β-cells give further insight. As expected, Exendin-4 enhanced the β-cell survival by resolving ER stress in obese mice and isolated rat β cells. Intriguingly, while the GLP-1R agonist did not affect PERK, the incretin analog further potentiated the expression of ATF4/CHOP, yet diminished the inhibitory phosphorylation of the translation repressor eIF2a upstream of ATF4, in thapsigargin or tunicamycin-treated INS-1 and MIN6 cells. GLP-1 further upregulated the XBP-1-controlled expression of the chaperone DnaJ Heat Shock Protein Family (DnaJb9), interestingly without affecting the levels or induction of the ER stress mediator XBP-1, and augmented growth arrest and DNA damage-inducible protein (GADD34) levels following ER stress. The GLP-1R-evoked increase in ATF4/CHOP levels and the dephosphorylation of eIF2a were cAMP/PKA-dependent. Notably, PKA phosphorylates both human inhibitor 1 (I-1) and type 1 protein serine/threonine phosphatase (PP1), leading to the association of I-1 and PP1 with GADD34, the nuclear translocation of the latter protein complex and the PP1-mediated dephosphorylation of eIF2a to re-evoke protein translation. Furthermore, PKA mediates the direct phosphorylation of ATF-4 and the survival transcription factor CREB.

50Therefore, GLP-1R agonists modulate the cellular fate in favor of survival during ER stress. While GLP-1R activation resolves ER stress, blocks apoptosis and re-elicits protein synthesis via the cAMP/PKA/GADD34-PP1/eIF2a axis, GLP-1R-mediated PKA-signaling seems to selectively potentiate redox gene expression via the ER stress-associated transcription factor ATF4. That aside, GLP-1R agonists stimulate chaperone expression (calnexin and DnaJb9) following Ca2+ deregulation, which likely contributes to protein folding and forestalls proteasome dysfunction. The latter chaperone upregulation might be related to general increases in protein translation through the PI3K/Akt/mTor pathway by incretins.

51In their ramified or resting state, microglia monitor the CNS for the presence of damage-associated molecular patterns (DAMPs; including neuronal, glial, endothelial, or oligodendrocyte-derived material) or pathogen-associated molecular patterns (PAMPs; i.e., LPS), scavenge and phagocytose foreign particles, regulate the synaptic architecture and growth, support neurons and more. Similar to microglia, astrocytes undertake various important functions in the brain, such as the provision of the energy substrate lactate to neurons (“lactate shuttle”) or clearing synapses from excessive excitotoxic glutamate.

52Microglia and astrocytes can adopt a pro-inflammatory M1 profile in the presence of oxidative stress, amyloids, infections, injury or during the aging process, as characterized by the upregulation of reduced ionized calcium-binding adapter molecule 1 (IBA-1) (microgliosis) or glial fibrillary acidic protein (GFAP) (astrogliosis). The accumulating DAMPs, including neuron-derived Aβ, Tau, α-synuclein, myelin sheath fragments from damaged axons, neuron-specific enolase, advanced glycation end products (AGEs) and more, as well as PAMPs (i.e., the toll-like receptor 4 (TLR4) ligand LPS result in the activation of TLRs, inflammatory p38 and JNK-signaling, the induction of the master transcription factor NF-κB, the subsequent transcription and release of pro-inflammatory cytokines, such as IL-1β, IL-6, IL-12, interferon γ (IFN-γ) or TNF-α, and chemokines as well as the generation of ROS and inducible nitric oxide synthase (iNOS)-derived NO during AD and PD. In turn, microglia and astrocytes may adopt an anti-inflammatory M2 state to resolve such inflammatory conditions and enhance tissue healing, for example involving the production of IL-4, IL-5, IL-10, IL-13, transforming growth factor β1 (TGF-β) or the neurotrophins BDNF, NGF, and others.

53In AD, oligomeric Aβ was shown to engage with cluster of differentiation (CD)14/TLR2/TLR4 on microglia to induce p38-conveyed ROS-production and the activation of NF-κB, while the binding of Aβ to CD36 led to the heterodimerisation of TLR4/TLR6 and the NLR family pyrin domain containing 3 (NLRP3) inflammasome-evoked generation of mature IL-1β. Aβ fragments further synergised with IFN-γ to exacerbate reactive nitrogen species (RNS) formation and the microglial TNF-α expression. Similarly, aggregated forms of Tau drive pro-inflammatory cytokine (IL-6) and NO production by microglia, while in PD, neuron-derived, fibrillated α-synuclein was shown to activate TLR2. Aβ and α-synuclein were further shown to trigger inflammatory responses by astroglia, involving, but not limited to, the upregulation of microglia-attractant chemokines (such as monocyte chemoattractant protein-1 or regulated upon activation, normal T Cell expressed and presumably secreted), pro-inflammatory cytokines (IL-1α/IL-1β, IL-6, TNF-α), NF-κB, TLR2, matrix metalloproteinase (MMP)3/9 and ROS/nitrate, with phagocytosed α-synuclein accumulating intracellularly in astrocytes.

54Importantly, in vivo studies in PD models have demonstrated that the abnormally increased secretion of TNF-α and IFN-γ perpetuates microglial and astroglial neuroinflammation by sustaining the TNF-α/janus kinase/STAT- and IFN-γ/MEK/ERK-mediated activation of NF-κB. Ultimately, chronic neuroinflammation in AD and PD evokes the TNFα and IL-1β-mediated permeabilization of the BBB, immune cell infiltration in to the CNS, mitochondrial and axonal defects, synaptic damage, insulin resistance in the brain as well as cytokine/chemokine/ROS/NO-provoked microglial, astroglial and neuronal dysfunction and death.

55Various studies confirm that GLP-1 analogs show potent anti-inflammatory effects. Liraglutide, lixisenatide, semaglutide, exendin-4 or NLY01 [a pegylated and rather poorly BBB-penetrant version of Exendin-4] reduced IBA-1 or Mac-1 immunoreactivity, indicative of microgliosis, in the hippocampus, cortex and global brain tissue of aged WT, APP/PS1, 5xFAD, 3 × Tg-AD or sporadic STZ AD mice. Likewise, microgliosis was attenuated by GLP-1R agonists the SN pars compacta (SNpc), striatum or cortex of MPTP or α-synuclein-injected mice. In addition, GLP-1 mimetic treatment diminished astrogliosis in the hippocampus and cortex of various AD animal models plus the hippocampus of old WT mice, while lessening astroglial GFAP-immunoreactivity in the SNpc, striatum or cortex of MPTP-, 6-hydroxydopamine (6-OHDA) or α-synuclein-injected PD rodent models as well as hA53T α-synuclein transgenic mice. In agreement with suppressing glial activity, Exendin-4, NLY01 or liraglutide reduced the expression of NF-κB and the subsequent synthesis of the pro-inflammatory cytokines IL-1α, IL-1β, IL-6, TNF-α, IFN-γ, and complement component 1q (C1q) in the hippocampus and nigrostriatal regions of various AD and PD animal models. Moreover, the expression of anti-inflammatory IL-10 was partially rescued by NLY01 in MPTP-treated animals.

56Similar to GLP-1 mimetics, although superior in their effect, GLP-1R/GIPR receptor dual agonists suppressed microgliosis, astrogliosis, the expression of TLR4 and NF-κB, the production of pro-inflammatory cytokines, such as IL-1β, IL-6, and TNF-α, and the restoration of anti-inflammatory TGF-β1 and IL-10 pools in the hippocampus, cortex of elderly WT and AD-like animals or the SNpc and striatum of PD rodent models.

57Studies have shown that isolated neurons, microglia and astrocytes express GLP-1R. While the GLP-1R is widely present on the dendrites of neurons, such as pyramidal neurons in the hippocampal CA1-3 region and neocortex, the granule cell layer of the DG or Purkinje neurons in the cerebellum, the GLP-1R is only sparsely expressed by astroglia and microglia. However, in vivo studies indicate that GLP-1Rs are conditionally upregulated by glial cells following cerebral injury, as further enhanced by GLP-1 analog treatment, seemingly to quench inflammatory responses. Interestingly, aged mice showed an age-associated decline in the levels of GLP-1 and its precursor proglucagon, but not GLP-1Rs, in the medial prefrontal cortex, further accompanied by heightened numbers of microglia in this brain area, stronger co-localization of microglia with GLP-1Rs and impeded spatial learning. This suggests that attenuated GLP-1 production and, thus, anti-inflammatory GLP-1R activation are linked to cognitive impairments during aging. Notably, cortical and hippocampal proglucagon/GLP-1 was shown to be predominantly derived from microglia and induced by cAMP/PKA-signaling. Given that only primary amoeboid microglia, but not ramified (resting) microglia or astrocytes, showed GLP-1 immunoreactivity in vitro, it is further implied that the endogenous GLP-1 expression is selectively upregulated by activated microglia to reverse inflammatory responses in the brain. However, conditions of insulin resistance during obesity and the chronically increased presence of pro-inflammatory fatty acids were demonstrated to interfere with the microglial proglucagon synthesis and GLP-1 secretion.

58GLP-1R agonists were shown to directly suppress microglial inflammation, while indirectly forestalling microglia-evoked astroglial activation, to protect neurons from amyloid toxicity. Interestingly, albeit astrocytes were unaffected and neurons showed reduced expression, the treatment of mice with α-synuclein preformed fibrils selectively doubled the mRNA levels of GLP-1R in microglia. Given that GLP-1R levels were 10-fold elevated and predominantly co-localized with IBA-1/transmembrane protein 119 (TMEM119)-positive cells in the SNpc of PD patients, it is implied that GLP-1R is upregulated by pathology-induced microglia during PD. As such, it has been suggested that the predominant site of action of GLP-1 mimetics is mostly microglial. Notably, the application of α-synuclein fibrils induced the transition into the reactive M1 state and the secretion of IL-1α, IL-1ß, IL-6, C1q, TNFα, and leptin into the culture medium by primary microglia. Of these, IL-1α, C1q, TNFα, were reported to convert astrocytes into the pro-inflammatory M1 (A1) state, while antibodies for these cytokines blocked the shift of astrocytes into the A1 state and the subsequent death of co-cultured cortical and dopaminergic neurons. Indeed, when astrocytes were cultured in α-synuclein-containing microglia conditioned medium (MWM), the resulting astrocyte conditioned medium (ACM) was more toxic to primary cortical neurons than MWM. Furthermore, when NLY01 was given to microglia, which led to the reduced secretion of all earlier mentioned pro-inflammatory cytokines, the ACM-driven cell death in primary cortical neurons was prevented. Therefore, given that the direct application of NLY01 to cortical primary neurons failed to rescue them from death by ACM, it was concluded that GLP-1R agonists act on reactive microglia to reduce the secretion of pro-inflammatory cytokines. This, in turn, blocks the conversion of astrocytes into the reactive A1 state, thus forestalling the inflammation-evoked death of neurons. These anti-inflammatory and neuroprotective effects of NLY01, including reduced microglial and astroglial immunoreactivity and IL-1a/b, IL-6, TNFα, and C1qa synthesis, were replicated in α-synuclein-injected rodents and hA53T α-synuclein transgenic mice. As further confirmed, the GLP-1 mimetic was incapable of rescuing neuronal degeneration when GLP-1R was depleted in microglia. Additionally, NLY01 did not block α-synuclein uptake or the inflammatory TLR2/α-synuclein-interaction in microglia, but, instead, suppressed the activation of NF-κB through GLP-1R-signaling. Comparable to the SNpc of PD patients and the results of the latter study, post-mortem investigations revealed heightened GLP-1R mRNA levels in the hippocampus of AD patients and in vivo, likely as a countermeasure to reduce chronic inflammation, whereby GLP-1R was primarily exhibited by IBA-1-positive microglia. As observed for PD cell and animal models, NLY01 suppressed the Aβ1–42-evoked stimulation of hippocampal microglia, the release of multiple pro-inflammatory cytokines in a GLP-1R-mediated manner, forestalled microglia-driven astrocyte activation and the induction of various genes associated with astrogliosis, blocked inflammation/astrocyte-evoked neuronal death in vitro/in vivo and rescued learning and memory in 5xFAD and 3 × Tg-AD rodents.

59As further evidence for an anti-inflammatory and M2-encouraging role, liraglutide prevented irradiation-induced microglial and astroglial immunoreactivity in the cortex and hippocampus, inhibiting the secretion of IL-1β, IL-6, IL-12, and NO. However, when microglia were stimulated with LPS, the effects of Exendin-4 were mixed, either failing to reduce cytokine expression or mimicking the effect of anti-inflammatory polyphenols, preventing TNF-α release by microglia. Nonetheless, liraglutide suppressed micro-/astrogliosis and shifted reactive M1 microglia toward the anti-inflammatory and ramified M2 phenotype in vivo, which dampened palmitate-induced IL-6 and TNF-α synthesis in the DG and hippocampal CA1. Indeed, GLP-1R activation on primary microglia led to the expression of M2-associated markers, including IL-4 (cAMP/PKA/CREB-dependent) and Arg 1 and CD206 (alternative Gs-cAMP/PKA/p38β/CREB-mediated). Notably, GLP-1R agonism selectively induces the non-inflammatory p38β isoform and β-endorphin expression, but does not stimulate the pro-inflammatory isoform p38α, in microglia.

60Generally, in vivo studies imply that reactive microglia and astrocytes are induced by and gather around Aβ plaques, co-localizing with markers of insulin resistance in neurons. In turn, GLP-1R agonist treatment was repeatedly shown to diminish the total soluble Aβ, Aβ oligomer or cerebral plaque burden in AD-like rodents (summarized in section “GLP-1R agonists are neuroprotective and prevent amyloid beta accumulation in vivo”), reduce α-synuclein accumulation in PD animal models and re-sensitize the neuronal insulin sensitivity in vitro and in the brains of AD and PD in vivo models (see also section “Insulin resistance and the neuronal energy metabolism”). This suggests that GLP-1R agonists forestall neuroinflammation by preventing insulin resistance and the mutually exacerbated amyloid pathology.

61Moreover, besides quenching microgliosis and cytokine production, Exendin-4 downregulated the expression of MMP3 in the SNpc of MPTP-treated mice. MMP3 represents a DAMP that is released by apoptotic neurons, similar to α-synuclein. In vitro studies suggest that cell stress-induced apoptosis enhances the expression of the pro-form of MMP3 in dopaminergic neurons, which is cleaved into MMP3 following pro-apoptotic JNK-signaling by serine proteases. MMP3 further accelerates α-synuclein fragmentation and aggregation in dopaminergic neurons, co-localizing with Lewy bodies in the brains of PD patients. Importantly, MMP3 that is liberated by dying neurons stimulates ERK/NF-κB-signaling, pro-inflammatory cytokine (IL-1β, IL-6, TNFα) and nicotinamide adenine dinucleotide phosphate oxidase-mediated superoxide production by nearby microglia. Indeed, MPTP-induced microglial inflammation, ROS, BBB damage, cerebral immune cell infiltration and dopamine neuron death in the nigrostriatal region were largely attenuated or fully prevented in MMP3–/– mice. In the context of AD, the expression of MMP3 was also upregulated by the PI3K/Akt pathway in microglia in response to Aβ1–42, while MMP3 was shown to correlate with total and phospho-Tau levels in the CSF of AD patients. Notably, MMP3 cleavage attenuates Tau aggregation, whereas MMP9, as activated by MMP3, promotes Tau oligomerization.

62Additional evidence indicates that the GLP-1R inhibits inflammatory cascades indirectly. A recent study showed that the application of Aβ25–35 resulted in NF-κB and NLRP3-mediated inflammatory cytokine expression, while the inhibition of TLR4 prevented the pathological microglial transition from M2 to M1 in vitro and in the APP/PS1 mouse model. Indeed, TLR4 plays a critical pathological role, given that a TLR4 polymorphism was shown to attenuate the risk of late-onset AD 2.7-fold. In this context, the GLP-1R/GIPR dual agonist DA5-CH downregulated TLR4 expression in the MPTP mouse model, while MPTP was shown to upregulate TLR4 levels by inducing the transcription factor AP-1 in vivo (at least in astrocytes). Furthermore, as a positive feedback loop, TLR4/NF-κB-signaling (i.e., as stimulated by Aβ) potentiate the transcription of TLR4 under inflammatory conditions. Synoptically, by suppressing oxidative stress, mitochondrial dysfunction and apoptosis in neurons, the release of death-associated DAMPs that would otherwise stimulate TLRs are reduced (see section “Oxidative stress and mitochondrial dysfunction” and Figure 2). Moreover, the activation of the GLP-1R receptor quenches oxidative stress in glial cells (see section “GLP-1 exerts direct anti-inflammatory and cytoprotective effects on astrocytes” for the cytoprotective effects in astrocytes), thus forestalling the stress-provoked induction of NF-κB and AP-1, the concomitant expression of TLR4/pro-inflammatory cytokines and the inflammatory switch from M2 to M1 in glial cells.

63Notably, inflammasome activation and the associated synthesis of NLRP3, caspase 1 and IL-1β are triggered by the Aβ/TLR4/NF-κB pathway in microglia. Interestingly, even though liraglutide did not affect the interaction of α-synuclein with TLR2 on microglia, the activation of the GLP-1R suppressed NF-κB activation in response to α-synuclein and MPTP-driven oxidative stress in neurons in vivo. Notably, GLP-1R activation also suppressed TNFα- or LPS-driven cell death, ROS and NO production, while upregulating antioxidant gene expression and protein levels of glutathione peroxidase 1 (GPx1) and superoxide dismutase (SOD)1 in a cAMP/PKA-, but not PI3K/Akt-, dependent manner in various microglial cell lines. Such studies strengthen the concept that GLP-1 mimetics exert indirect anti-inflammatory effects on microglia and astrocytes by protecting glial cells (and neurons) from oxidative and amyloid-driven injury.

64Multiple studies confirm beneficial consequences of GLP-1R activation on astrocytes. The application of GLP-1 induced the cAMP/CREB pathway in astrocytes to inhibit the LPS-provoked transcription and release of IL-1β, with trends for reduced IL-6 and iNOS, in a cAMP-dependent manner. Indeed, pro-inflammatory cytokine production, including IL-1ß and TNFα, was downregulated by and dependent on CREB activation in primary cortical astrocytes. Moreover, GLP-1 mimetics show clear protective effects. Liraglutide, by activating GLP-1R and recovering cAMP/PKA/CREB signaling, dose-dependently reducde ROS production, blocked pro-apoptotic caspase 3 cleavage, raised cell viability and blocked IL-1ß and TNFα secretion by cultured rat cortical astrocytes following AGE treatment. In this context, AGEs were shown to trigger oxidative stress and a reduction in anti-oxidative modulators in astrocytes in vitro, as marked by diminished GSH and SOD and elevated malondialdehyde (MDA), monoamine oxidase B and NO levels, leading to pro-inflammatory cytokine production. Additionally, liraglutide activated the cAMP/PKA pathway in astrocytes, which, in a PKA-dependent manner, reverted the Aβ1–42-induced downregulation of the mitochondrial fusion enhancers mitofusin-2 (Mfn2), optic atrophy 1 (OPA1) and the fission-inhibiting phosphorylation of dynamin-1-like protein 1 (DLP1, also known as DRP1) at Ser637 by PKA. The GLP-1R agonist further prevented Aβ1–42-driven mitochondrial fragmentation, normalized the collapsed mitochondrial membrane potential, rescued proton leakage, enhanced ATP production, reduced ROS generation and improved astroglial survival in vitro. Notably, the mitochondrial fusion/fission-ameliorating properties of GLP-1 mimetics in astrocytes are comparable to those in Aβ-assaulted neurons (see section “GLP-1R agonists suppress amyloid beta and GSK-3β-driven mitochondrial damage in AD” and Figure 2).

65Interestingly, in contrast to Aβ-induced mitochondrial fragmentation, the opposite might be the case in PD. Astrocytes and neurons isolated from the SNpc of PD patients showed an early decrease in DLP-1 activity, while another study identified an over twofold reduction of DLP-1 in mitochondria-enriched fractions of the SNpc of PD patients. In this context, albeit not accounting for any possible defects in mitochondrial fusion proteins, the deletion of astroglial DLP-1 led to mitochondrial elongation, fusion and increased motility, intracellular Ca2+ amassment and impaired glutamate removal, which resulted in excitotoxic injury of co-cultured dopaminergic neurons in vitro. Therefore, it would be interesting to assess the effects of GLP-1R activation on fusion/fission modulators in astrocytes and neurons in future PD studies.

66Besides cAMP/PKA-induced cellular protection, GLP-1-triggered PI3K/Akt-signaling prevented Aβ1–42-provoked increases in oxygen consumption, ATP and ROS production as well as a detrimental switch from aerobic glycolysis to excessive OXPHOS in astrocytes in vitro and in vivo. Notably, in contrast to OXPHOS-utilizing neurons, astrocytes prefer aerobic glycolysis (which is coupled to lactate production) as main bioenergetic pathway. On the other hand, astrocytes compensate with OXPHOS in response to bioenergetic and mitochondrial dysfunction (inverse Warburg effect).

67Synoptically, by stimulating the cAMP/PKA/CREB pathway in astrocytes, GLP-1R agonists elicit direct anti-inflammatory effects. Additionally, GLP-1 induces cytoprotective cAMP/PKA- and PI3K/Akt-signaling that forestalls pro-inflammatory responses by astrocytes in response to oxidative stress, mitochondrial dysfunction, bioenergetic deficits and/or pro-apoptotic signaling.

68Interestingly, by improving the astroglial energy metabolism and survival, GLP-1 analogs enhance the neuronal supply with lactate, BDNF and GSH, while possibly preventing excitotoxic damage. As implied in the previous section, GLP-1 prevented the Aβ1–42-induced mitochondrial dysfunction and the compensatory switch from the lactate-producing aerobic glycolysis, the main bioenergetic pathway in astrocytes, to the oxidative stress-associated OXPHOS in cultured astrocytes. Indeed, in a PI3K/Akt-dependent manner, GLP-1 prevented the downregulation of glycolytic enzymes [pyruvate kinase (PKM), PKM2, hexokinase 2 (HK2), pyruvate dehydrogenase kinase 2 (PDK2)], the transcription factor hypoxia-inducible factor 1 alpha (HIF-1α), lactate and NAD+ levels as well as the NAD+/NADH ratio in Aβ1–42-assaulted astrocytes. Similarly, 5xFAD mice showed elevated ROS generation, diminished redox capabilities (GSH), impaired ATP production, the alteration of copious bioenergetic genes, the reduced synthesis of glycolytic enzymes and excessive OXPHOS, as characterized by the abnormally heightened phosphorylation of the OXPHOS-inducer pyruvate dehydrogenase (PDH), and reduced activity of the PDH-modifier PDK. The latter restricts the PDH-mediated conversion of pyruvate into Acetyl-CoA for OXPHOS in mitochondria and stimulates the conversion of pyruvate into lactate by lactate-dehydrogenase (LDH) in the cytoplasm instead. All of the latter bioenergetic impairments could be reversed with liraglutide injections in vivo which, as observed in vitro, led to the stimulation of Akt. In this context, Akt stimulates the mTOR/HIF–1α pathway, with HIF–1α controlling the expression of the forementioned glycolytic enzymes as well as PDK and LDH-A (an isoform enriched in astrocytes and weakly expressed in neurons), thus favoring the processing of glucose into pyruvate and lactate. As a side note, astrocytes generally express higher levels of PDK2/4 and LDH than neurons, explaining the preference of astrocytes for aerobic glycolysis. Indeed, although failing to affect brain glucose metabolism in 3 × Tg AD mice, the administration of Exendin-4 led to elevated LDH activities (likely in astrocytes), pyruvate/lactate conversion and lactate levels in the brains of PS1K-KI mice, as accompanied by enhanced short- and long-term spatial memory.

69Interestingly, Aβ-stressed astrocytes decreased the viability of co-cultured primary neurons, whereas astroglia GLP-1 treatment not only prevented neuronal death, but also enhanced the axonal and dendritic outgrowth and synapse formation. In turn, the latter favorable effects of GLP-1 on neurons were inhibited by blocking astroglia glycolysis with 2-DG. Liraglutide further restored the secrteion of BDNF by Aβ1–42-stressed primary astrocytes in a PKA-dependent manner. Importantly, the GLP-1 mimetic only forestalled the astrocyte-induced death of co-cultured primary neurons and improved the neuronal numbers of neuritic interactions, somal neurites, secondary branches and total axon plus neurite length when Aβ1–42-assaulted astrocytes were treated with liraglutide in the absence of a PKA inhibitor. Thus, in the presence of amyloid pathology, GLP-1 analogs seem to stimulate the general energy turnover, the cAMP/PKA/CREB/BDNF pathway and BDNF secretion by astrocytes to evoke BDNF-driven synaptogenesis and neurite outgrowth in neurons.

70Moreover, the astroglia and NRF-2-mediated enzyme expression for the production and release of the antioxidant GSH, whose levels were restored by liraglutide in the cortex of 5xFAD mice, are crucial for the shielding of nearby neurons from oxidative damage.

71Therefore, GLP-1 protects astrocytes from amyloid toxicity and enhances the supportive function of astrocytes toward neurons. This involves the PI3K/Akt/mTOR/HIF–1α-dependent downregulation of OXPHOS-triggered oxidative stress, the re-invigoration of glycolysis and the neuronal lactate shuttle, the cAMP/PKA/CREB-driven synthesis and liberation of pro-synaptic and neuroprotective BDNF as well as the provision of anti-oxidative GSH by astrocytes.

72Notably, Aβ disturbs the homeostasis between astrocytes and neurons by impairing the astroglia expression of glutamate transporters excitatory amino acid transporter 1 (EEAT1) and EAAT2, which impedes the clearance of glutamate from the synaptic cleft, and glutathione synthase (GS), an enzyme that converts glutamate into glutamine. This leads to the accumulation of excitotoxic glutamate in the synaptic space and the reduced neuronal supply with glutamine, which is a crucial bioenergetic and neurotransmitter substrate for neurons. Intriguingly, while Aβ1–42 diminished glutamate uptake, EEAT2 and GS levels, GLP-1 could restore the astroglia expression of GS in an in vitro study. This indicates that the pro-metabolic and neuroprotective effects of GLP-1R agonists on astrocytes may protect neurons from excitotoxicity.

73Notably, GLP-1R activation on neurons directly prevents excitotoxicity (section “GLP-1 mimetics suppress Ca2+ deregulation by amyloid beta and excitotoxicity”). Indeed, the use of GLP-1 analogs rescued from the kainite- or ibotenic acid-driven excitotoxic apoptosis of hippocampal and basal forebrain cholinergic neurons in vivo. An in vitro study indicated that the neuroprotective effects of GLP-1 were linked to the blockade of glutamate and VDCC currents, the associated Ca2+ instream and membrane depolarisation in response to exogenous neuronal glutamate overload. For other excitotoxicity-suppressing effects of GLP-1 in the context of epilepsy, see also.

74In PD, mitochondrial dysfunction may arise as a consequence of gene mutations that exacerbate oxidative stress, impair the mitochondrial function and impede ATP generation, such as DJ1, leucine-rich-repeat kinase 2 (LRRK2) or the mitophagy-associated modulators PTEN-induced kinase 1 (PINK1) and Parkin. Other hypothesized reasons for mitochondrial damage include ROS-triggered mtDNA damage, environmental toxins that interfere with OXPHOS or non-mutant α-synuclein accumulation, which seems to enhance ER-mitochondria interactions and, hence, cause excessive Ca2+ transfer to mitochondria. It is thought that SNpc-located dopaminergic neurons selectively degenerate during PD due to their high energy demands, a concomitant increase in ROS production, the synthesis of autoreactive catecholamines (dopamine), poor anti-oxidant and Ca2+-buffering capabilities as well as weakly or non-myelinated axons. In this context, PD models rely on the use of complex I inhibitors, such as MPTP, rotenone or 6-OHDA, that are selectively taken up by dopaminergic neurons and interfere with the mitochondrial ATP production by the ETC, while elevating ROS generation. Given that an estimated ∼30% of dopaminergic neurons in the SNpc, ≤ ∼60% of the external SNpc projections, especially toward the striatum, and ≤ ∼70% of the dopamine supply of the dorsal striatum are lost when the characteristic motor symptoms occur in PD, it is evident that pharmacological interventions must ensue early.

75There is abundant in vivo evidence that GLP-1R activation protects nigrostriatal neurons and replenishes the production of dopamine in PD. Various synthetic incretin analogs, such as exendin-4, liraglutide, lixisenatide, semaglutide, (Val8)GLP-1-Glu-PAL, NLY01, or GLP-1R/GIPR dual agonists, were shown to prevent the atrophy of dopaminergic neurons in the SN and striatum and preserve dopaminergic fibers in PD animal models. Moreover, these GLP1R and GLP1R/GIPR dual agonists restored the nigral expression of TH, the rate-limiting enzyme of dopamine-synthesis, dopamine transporter, monoamine transporter 2 and vesicular monoamine transporter 2 (VMAT2), while improving the production of the dopamine precursor l-3,4-dihydroxyphenylalanine (L-DOPA), dopamine, the pools of other dopamine metabolites [dihydroxyphenylacetic acid (DOPAC), homovanillic acid (HVA), and 3-methoxytyramine (3MT)], dopamine turnover as well as norepinephrine and the serotonin breakdown product 5-hydroxyindoleacetic acid (5-HIAA) in the striatum or basal ganglia of MPTP, 6-OHDA, rotenone, α-synuclein pre-formed fibril, or LPS-induced mouse models of PD. Interestingly, Exendin-4 massively increased TH expression (60%) even in the absence of 6-OHDA in vitro, implying that GLP-1 not merely protects TH-expressing dopaminergic neurons, but also improves dopamine synthesis on a transcriptional level. GLP-1R or GLP-1R/GIPR dual agonists further decreased the production of the lipid peroxidation product and oxidative stress marker 4-hydroxynonenal (4-HNE), raised the synthesis of the neurotrophins glial cell line-derived neurotrophic factor (GDNF) and BDNF (more details in section “Other growth factors”) and prevented the accumulation of α-synuclein, phospho-α-synuclein and insoluble α-synuclein in TH-positive striatal and midbrain dopaminergic neurons in vivo.

76Mechanistically, GLP-1R activation enhanced the phosphorylation of Akt, induced CREB, heightened the synthesis of anti-apoptotic Bcl-2, reduced the levels of pro-apoptotic Bax and CytC, normalized the lowered Bcl-2/Bax ratio and diminished the levels of the apoptosis-effector caspase 3 in the nigrostriatal brain region of PD rodent models. We will investigate the underlying neuroprotective pathways in section “Oxidative stress and mitochondrial dysfunction.”

77In agreement with their neuroprotective effects, GLP1 or GLP1/GIP dual modulators preserved the motor function of the PD animals, including better balance and motor coordination in the rotarod or pole test, improved grip strength, lowered motor activity in the swimming test, lessened gait and postural abnormalities in the footprint or gait test, heightened spontaneous locomotor behavior in the open field test decreased rigidity in the catalepsy trial, normalized grooming behavior and rearing, improved sensory motor function in the cylinder test as well as accelerated functional recovery, as indicated in the apomorphine/amphetamine tests. The latter animal studies showed that GLP-1/GIPR dual agonists had superior effects compared to GLP-1R analogs.

78Strikingly, as a proof of concept, Exendin-4 preserved the motor abilities of PD patients in a randomized and double-blind phase II clinical trial, wherein the beneficial effects were still visible after a 12 week washout phase. Exosome analysis indicated that these PD patients showed elevated Ser-phosphorylated IRS-1 levels and impaired insulin-signaling in the brain. In turn, the application of Exendin-4 improved the impeded insulin sensitivity, as implied by enhanced Akt and phospho-mTor levels in treated PD patients. Indeed, the motor improvements in these PD patients were positively associated with the levels of mTor and phopho-mTor levels. Given the rising acknowledgment of insulin resistance as an early pathological key step in AD and PD, we have devoted a stand-alone section to the insulin-re-sensitizing mechanisms of GLP-1R agonists (see section “Insulin resistance and the neuronal energy metabolism” and Figure 1).

79Mitochondria may produce ATP by several pathways. While it has yet to be proven that neurons utilize lipid β-oxidation, they preferably metabolize glucose via glycolysis to create pyruvate. The latter is subsequently funneled into the tricarboxylic acid (TCA) cycle to generate a few units of ATP and, importantly, reduce nicotinamide adenine dinucleotide (NAD+) to NADH. In turn, NADH participates in OXPHOS that is executed by complex I–V of the ETC. Briefly, electrons (e–) won from NADH at complex I are moved toward complex IV within the inner-mitochondrial membrane, resulting in the reaction of e– and H+ to H2O at complex IV. Simultaneously, H+ ions are pumped outwards across complex I, III, and IV from the inner matrix to the intermembrane space. This establishes a gradient, leading to the re-flux of H+ ions into the inner matrix through complex V (ATP synthase) to produce ATP. Importantly, e– may leak at complex I and III, but also IV. ∼1–2% of the consumed oxygen reacts with these escaping e– to evoke the generation of the O2–. The latter ROS may be converted by SOD1 or SOD2 into H2O2, followed by detoxification into H2O by mitochondrial GPx or cytosolic catalase. As such, ironically, mitochondria pose the greatest cellular source of oxidative stress.

80As death signals accumulate in response to mitochondrial defects, including elevated ROS production, bioenergetic deficiencies, loss of ATP production and Ca2+ deregulation, pro-apoptotic Bcl-2 proteins are induced. Generally, anti-apoptotic members of the Bcl-2 family, for example Bcl-2 or Mcl-1, sequester their pro-apoptotic counterparts, including Bad, Bax, Bak, Bid, Bik, and others, on the outer mitochondrial membrane (but also ER) to manage the cellular survival. However, cellular and mitochondrial dysfunction stimulate pro-apoptotic Bcl-2 family members, mitochondrial pore formation by Bax and Bak, permeabilization of the outer mitochondrial membrane, the release of pro-apoptotic factors, especially CytC, and the sequential activation of APAF-1, caspase 9 and caspase 3 to drive intrinsic apoptosis. Notably, as an alternative, cells may be subject to extrinsic apoptosis across caspase 8 and caspase 3, as driven by ligand binding to death receptors on the cellular membrane, including Fas/Fas ligand (FasL) and tumor necrosis factor receptor-1/TNF-α.

81Ample evidence shows that GLP-1R or GLP-1R/GIPR dual activators guard against external oxidative stress, while preventing internal ROS production by injured mitochondria. In vitro, GLP-1, liraglutide, Exendin-4 and GLP-1R/GIPR dual agonists shielded PC12 neuronal, RGC-5 retinal and SH-SY5Y cells from H2O2-induced oxidative damage or Aβ-driven oxidative stress in a GLP-1R-dependent manner. Moreover, incretin hormones prevented H2O2-provoked mitochondrial injury and consequential ROS production by the mitochondria. Indeed, TEM images confirmed that Liraglutide prevented mitochondrial swelling and the disintegration of cristae upon H2O2 exposure in RCG-5 cells. In PD-specific contexts, with GLP-1R/GIPR dual agonists showing the best effects, GLP-1R induction by various incretin mimetics rescued SH-SY5Y cells and primary hippocampal, cortical or dopaminergic neurons from 6-OHDA or rotenone-provoked mitochondrial complex I dysfunction, the associated intercellular ROS accumulation and apoptosis.

82As illustrated in Figure 2, a recent study in GLP-1-treated HT22 hippocampal cells indicated that the neuroprotective effects of incretin hormones on H2O2-induced oxidative stress and mitochondrial dysfunction, glutamate overload, tunicamycin or thapsigargin-triggered ER stress or Aβ1–42-provoked neuronal death involve the activation of Akt and ERK1/2. More specifically, GLP-1R induction blocks caspase-dependent apoptosis through the mitochondrial pathway by modulating Bcl-2 family members. As supported by various studies, the neuroprotection from H2O2- or mitotoxin-induced oxidative stress necessitated the GLP-1R-induced activation of PI3K/Akt and PKA, resulting in attenuated levels of each member of the pro-apoptotic Bax/CytC/Caspase 3 pathway, decreased Bad phosphorylation, increased anti-apoptotic Bcl-2 expression and phosphorylation, elevated Bcl-2/Bax ratios and raised TH synthesis.

83Indeed, PI3K/Akt and ERK1/2, as induced by growth hormones and neurotrophins such as insulin, GLP-1 or BDNF, are well-known survival pathways that are deregulated in neurodegenerative diseases. Neuroprotective PI3K/Akt-signaling activation enhances mTor/ribosomal protein S6 kinase beta-1 (S6K)-mediated protein translation and cell proliferation, which involves the Akt-mediated inactivation of the translation repressor eukaryotic translation initiation factor 4 (eIF4E)-binding protein (4E-BP). This general elevation in protein translation is likely responsible for the increase in TH expression (60%) that has been observed in GLP-1 mimetic-treated primary mesencephalic cell cultures (which are enriched in dopaminergic neurons) in vitro. Akt further activates Raf/MEK1/2 to reinforce ERK activity and phosphorylates CREB to drive the expression of the anti-apoptotic effectors Bcl-2 and Mcl-1. Furthermore, Akt phosphorylates and inhibits (i) pro-apoptotic Bad, a counter-regulator of the pro-survival mediator Bcl-2, (ii) caspase 9, which is the upstream activator of caspase 3, (iii) GSK-3β, a major apoptosis pathway in neurons and (iV) death-associated FOXOs that upregulate the apoptosis-inducing FasL, whilst (V) suppressing the pro-apoptotic JNK/p53/Bax axis. On the other hand, while chronic ERK1/2-signaling alone, similar to its serine/threonine protein kinase family members JNK and p38, may exert pro-apoptotic effects in some instances, for example during PD, ERK1/2 induces CREB-driven plasticity and survival gene expression. Notably, ERK further phosphorylates GSK-3β at Thr43. This primes GSK-3β for the inactivating Ser9 phosphorylation by p90RSK, followed by the upregulation of β-catenin.

84In the context of oxidative stress, the neuroprotective effects of GLP-1 involved the upregulation of the transcription factor ATF4, which mediates the expression of anti-oxidative and redox genes, for example GSH. Other studies suggest that GLP-1 agonists enhance the levels and activity of ATF4 through cAMP/PKA-signaling to prevent oxidative injury, as observed in response to cytosolic Ca2+ overload or ER stress. A study in H2O2-stressed human umbilical vein endothelial cells reported that Exendin-4, by activating the GLP-1R/cAMP/PKA pathway, evoked the CREB-mediated transcription of the antioxidant defense genes heme oxygenase 1 (HO-1) and NAD(P)H Quinone Dehydrogenase 1 (NQO-1). Studies in methyl glyoxal-injured PC12 cells further indicate that GLP-1 ameliorates the cellular redox balance by stimulating the PI3K/Akt/mTOR-dependent transcriptional upregulation of glutamate-cysteine ligase catalytic subunit (GCLc), the rate-limiting enzyme for GSH synthesis. Indeed, in vitro studies in GLP-1R-expressing mesenchymal stem cells support that GLP-1 analogs, as dependent on PI3K/Akt, dose-dependently shield against H2O2-driven loss of mitochondrial membrane potential, thus forestalling ROS production by defective mitochondria, intrinsic apoptosis across the mitochondrial pathway (increased Bcl-2 and reduced Bax, caspase 9/3 induction and CytC release) and upregulate the expression of SOD and the anti-oxidant GSH, while preventing ROS-associated lipid peroxidation and MDA formation.

85Neuroprotection in PD further involves the suppression of JNK by incretins. In this context, elevated levels of pJNK have been detected in the SN of PD rodent models and patients. Moreover, MPTP-induced dopamine neurons death necessitated JNK2/3/COX2-signaling, which seemingly enhances neurotoxicity by eliciting the COX2-induced oxidation of dopamine to create highly redox-reactive dopamine-quinones. In PD, JNK drives neuronal death across the mitochondrial intrinsic (Bax/CytC/caspase 3) and extrinsic (c-Jun/AP-1/FasL and more) apoptosis pathways. JNK is activated in response to various forms of stress in neurons, including oxidative stress, as reinforced by enhanced ROS-production from dysfunctional mitochondria or complex I inhibition by PD-toxins, and the induction of neuronal PICRs by the glia-derived pro-inflammatory cytokines IL-1 and IL-18. This suggests that GLP-1 agonists suppress the pro-apoptotic activation of JNK in dopaminergic neurons by enhancing anti-oxidative mechanisms (as elucidated below), preventing ROS-generation due to mitochondrial dysfunction and quenching microglial and astroglia inflammation during PD (more insight in section “Inflammation”). Indeed, GLP-1R-signaling prevented JNK activation following rotenone treatment in vitro/in vivo, whereas the pharmacological enhancement of JNK interfered with the GLP-1R-driven Akt phosphorylation and exacerbated the induction of the CytC/Bax/Caspase 3 cell death axis.

86Synoptically, in light of oxidative stress and summarized in Figure 2, GLP-1 agonists engage PI3K/Akt to suppress effectors of mitochondrial apoptosis pathway and, indirectly, pro-apoptotic JNK-signaling, induce the cAMP/PKA/ATF4 pathway to stimulate anti-oxidative gene transcription (i.e., GSH) and activate ERK1/2 and CREB for the transcription of survival and antioxidant defense genes (such as HO-1 and NQO-1). The latter anti-oxidative properties following GLP1R activation, in turn, forestall mitochondrial injury through oxidative stress and concomitant ROS-production as well as deficits in ATP production.

87The deposition of metal ions in the brain might contribute to oxidative stress in AD and PD. A major question is what triggers iron accumulation in the first place. In this context, chronic neuroinflammation may be an initiating trigger during AD and PD. Generally, iron is present in a bound form, as sequestered by both intracellular and macrophage-secreted ferritin in the blood plasma, as well as a free form (Fe2+), also known as labile iron pool (LIP). Free iron co-mediates essential cellular processes, such as mitochondrial respiration or the synthesis of nucleic acids. Iron is an essential co-factor for pathogens. As such, under inflammatory conditions, the release of pro-inflammatory cytokines alters the expression of iron-associated genes in the mononuclear phagocyte system to sequester iron in macrophages (including microglia). For instance, ROS, IL-1β and, in particular, IL-6, stimulate the STAT3-mediated expression and secretion of hepcidin antimicrobial peptide (HAMP), which is secreted, binds to and initiates the proteasomal degradation of the iron exporter ferroportin 1 (FPN1) in in surrounding cells. TNF-α and IFN-γ further elevate the transcription of the ion-importer transferrin receptor 1 (TFR1) by macrophages. While IL-1β and IL-6 further promote ferritin expression in macrophages to induce the storage of iron, ferritin, in both macrophages and non-immune cells, will eventually be saturated and the LIP increases. Notably, HAMP was not only shown to downregulate FPN1, but also the iron-internalizing TFR1 and divalent metal transporter 1 in cultured astrocytes, brain microvascular endothelial cells and neurons. Indeed, HAMP overexpression or treatment, in the absence of pro-inflammatory cytokines, prevented both the uptake and release of transferrin-sequestered or free iron in vitro and iron translocation across the BBB in vivo. As such, the main function of HAMP is to limit bidirectional iron transport in cells. Arguably, considering the presence of chronic neuroinflammation during the aging process as well as in AD and PD, HAMP, in combination with the production of other pro-inflammatory cytokines, may trigger cerebral iron accumulation. Therefore, it can be hypothesized that the anti-inflammatory effects of GLP-1 analogs (section “Inflammation”) might forestall the age- and neurodegeneration-associated iron deposition in the brain.

88Indeed, both MCI patients and APOE4 carriers showed higher cortical iron load that was positively correlated with increased Aβ plaque load. Another study found correlations between the abnormally elevated levels of iron, senile plaques and Tau inclusions, with iron sequestered in plaques and microglia, in frontal and mid-cortical layers of AD patients. Furthermore, iron and Aβ co-pathology was associated with reduced neuropsychological test scores in individuals with MCI or AD, while iron deposition in the frontal lobe was correlated with cognitive decline in AD. Therefore, given that iron chelators, for example deferrioxamine, have seen some success for the treatment of AD and PD animal models and patients (see i.e.,; ), the metal ion hypothesis of AD has been proposed.

89As summarized elsewhere, iron interacts with iron regulatory elements in APP mRNA to enhance the translation of APP, blocks furin expression to enhance BACE1 activity, interacts with γ-secretase to elevate Aβ generation, binds to Tau ad activates CDK5 and GSK-3β to induce Tau hyperphosphorylation and aggregation into NFTs. Of note, Aβ1–42 was demonstrated to bind, create and concentrate Fe3O4 within plaques, also known as iron oxide or magnetite. The latter has also been detected in polluted air and linked to dementia. However, while external magnetite was hypothesized to enter the brain via the olfactory bulb, it is not clear whether this is relevant and contributes to the magnetite that accumulates in the AD brain tissue. Interestingly, while Aβ monomers were shown to bind Fe3 in the region between Ser8 to Gly25, they could only do so when Fe3 was stabilized by an iron-chelator. In this context, Aβ1–40 and Aβ1–42 were shown to capture Fe3 from ferritin, whereby Aβ-bound Fe3 is readily converted into ROS-generating Fe2+. Other studies added that Aβ-Fe2+ interactions promote beta-sheet conformation and the self-polymerisation of Aβ monomers. Moreover, increased ratios of the Fe2+-oxidizing and detoxifying H-ferritin over the L-ferritin isoform were observed in the frontal cortex of AD and the caudate and putamen of PD patients. Thus, there appears to be a dynamic interaction between iron, the attempt to sequester iron by ferritin and the capture of ferritin-incorporated Fe3 by Aβ that triggers oxidative stress and plaque formation.

90Besides other adverse mechanisms (see ), Fe2+ is known to produce ROS in the presence of H2O2, resulting in the generation of the neurotoxic membrane lipid peroxidation product 4-HNE that is prominent in the brains of AD, but also PD, patients. Ultimately, intracellular iron overload leads to ferroptosis, a recently discovered form of cell death that is characterized by excessive iron-mediated ROS generation, lipid peroxidation and mitochondrial damage. Furthermore, ferroptosis was shown to be promoted by the downregulation of the anti-oxidative enzyme phospholipid glutathione peroxidase 4 (GPx4).

91There is evidence that GLP-1R activation prevents ferroptosis and oxidative stress. A recent study showed that iron deposition in hippocampal tissue as well as across the caudate nucleus, SN and putamen of T2DM patients have been correlated with cognitive decline. In this context, the treatment db/db diabetic rodents with liraglutide, besides various synaptoprotective, neuroprotective and spatial memory-enhancing effects, rescued neurons and their mitochondria from ferroptosis, as indicated by the normalization of the increased serum and hippocampal CA1, CA3, and DG iron levels. The latter iron-decreasing effects of the GLP-1 analog involved raising the downregulated expression of the iron-storing H-ferritin and mitoferritin, lifting the decreased synthesis of the iron-exporting FPN1 and attenuating the elevated levels of the iron-uptake protein TFR1. In addition, the GLP-1R agonist recovered the decreased expression of the ferroptosis-inhibitor GPx4, downregulated that of the lipid membrane re-modeling and ferroptosis-encouraging acyl-CoA synthetase long-chain family member 4 and reverted the decreased transcription of SLC7A11, a cysteine-translocating protein that is necessary for GSH synthesis. In agreement with the reduction in iron, liraglutide reduced the circulatory and hippocampal MDA and ROS generation, while increasing the levels of SOD2 and GPx, in db/db mice. Indeed, the anti-oxidative capabilities of GLP-1 (see section “GLP-1R-signalling protects from external oxidative stress, ROS production and the mitochondrial apoptosis pathway” and Figure 2) protected primary hippocampal neurons from Fe2+ or Aβ-triggered oxidative cell damage.

92Notably, as an autoreactive catecholamine neurotransmitter and similar to H2O2, intra- and extraneuronal dopamine may generate massive amounts of ROS (1O2, O2–, and H2O2) by reacting with metal ions such as copper and iron, resulting in severe DNA oxidation and damage in dopaminergic neurons. Generally, there is an age-dependent accumulation of iron in the basal ganglia, including the SN, putamen and globus pallidus. Beyond this age-related increase, although copper levels seem to be deceased, PD patients exhibit abnormally heightened iron deposition, lowered anti-oxidative GSH levels and decreases in the iron-chelator ferritin especially in the SN, but also related brain regions. It has been argued that the reduction in copper levels impairs the activity of iron-removing ferroxidases, thus augmenting the nigral amassment of iron, ROS production and tissue injury during PD. Indeed, the SN-microinjection of iron triggers parkinsonism in animals. Notably, post-mortem investigations showed that PD patients lose VMAT2 expression in the putamen, caudate and nucleus accumbens. VMAT2, whose expression is lost during PD, packages dopamine into presynaptic vesicles. This suggests that dopaminergic neurons are more vulnerable to the auto-oxidation of dopamine and concomitant DNA injury during PD. On the other hand, Exendin-4 was shown to preserve VMAT2-positive dopaminergic neurons in the SN of 6-OHDA-injected animals, proposing that GLP-1 prevents the cytosolic accumulation of redox-reactive dopamine species (Figure 2).

93Interestingly, GLP-1 analogs not only suppress oxidative stress to protect DNA integrity, but also elicit DNA repair mechanisms. Liraglutide and GLP-1R/GIPR dual agonists rescued cultured neurons from DNA fragmentation by H2O2 or AGE-associated ROS production. Similarly, in vivo, the TUNEL assay indicated that GLP-1R agonist treatment reduced apoptosis-associated DNA fragmentation in the hippocampus or SN following the microinjection of kainate or MPTP, respectively. Moreover, a study using menadione, a mitochondrial ETC inhibitor that enhances ROS production, showed that GLP-1R agonism in cortical neurons suppressed oxidative stress and evoked the CREB-mediated expression of apurinic/apyrimidinic endonuclease 1 (APE1), a member of the base excision DNA repair pathway, in an Akt (but not MEK1/2)-dependent manner (see Figure 2). Notably, APE1 expression was enhanced by GLP-1 or GLP-1 mimetics even in the absence of any stressors.

94In this context, the transcription of APE1 was shown to be reduced in the entorhinal cortex and mononuclear blood cells of AD patients, while the nuclear translocation of APE1 in the cerebral cortex was elevated in another study, as provoked by oxidative DNA damage in response to defective mitochondrial respiration and increased ROS production. Likewise, by driving mitochondrial dysfunction and ROS/RNS-generation, Aβ25–35 attenuated APE1 levels in vitro. In PD, genetic APE1 variants may be risk factors that accelerate the degeneration of dopamine neurons, whereas the overexpression of APE1 counteracted the MPTP-induced ROS amassment and apoptosis of PC12 cells. The knockdown of APE1 accomplished the opposite result. APE1 further appears to be the only upregulated BER enzyme in response to glutamate-associated oxidative stress as well as BDNF, while GLP-1-inducing agents were shown to raise the expression of BDNF in WT, AD and PD-like animals (see section “GLP-1 mimetics stimulate BDNF synthesis in neurons and glia”). As such, GLP-1R activation ameliorates oxidative stress-induced DNA damage by stimulating the CREB/APE1 axis in neurons in AD and PD.

95Indeed, mitochondrial biogenesis, which is a process in which mitochondria adapt to greater energetic demands by growing in size and numbers, is disrupted during AD and PD. In this context, a key effector of the mitochondrial biogenesis is peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α). For instance, the decreased expression of PGC-1α has been detected in the SNpc of PD patients, while AD patients displayed diminished CytC oxidase/complex IV levels, whose expression is under control of PGC-1α, in their post-mortem brain tissue and isolated platelets. A decline in PGC-1α and NRF-1/2 levels has also been observed in the hippocampus of APPswe/PS1dE9 mice. PGC-1α is strongly expressed in energy-demanding tissues, including dopaminergic neurons in the brain. Besides regulating the mitochondrial biogenesis, given that PGC-1α further co-induces other mitochondrial effectors [NRF-1/2 and transcription factor A, mitochondrial (TFAM)], modulates anti-oxidant gene expression (SOD, catalase, GSH), prevents mitochondrial dysfunction and blocks α-synuclein oligomerisation and apoptosis, PGC-1α plays a key role especially in PD.

96As confirmed in SH-SY5Y cells stressed with the PD toxin 6-OHDA, the application of a GLP-1R/GIPR dual agonist reverted the downregulation of PGC-1α and NRF-1. Similarly, Liraglutide blocked H2O2-provoked mitochondrial ROS production, membrane potential loss and structural damage, while potentiating the expression of PGC-1α to raise the number of mitochondria in RGC-5 cells. These biogenic effects were replicated in vivo, where a GLP-1R/GIPR dual agonist normalized the quantities and volume of neuronal mitochondria in the brains of 3 × Tg (APP/PS1/Tau) AD mice.

97Notably, in contrast to Akt/mTor-activating hormones that are associated with growth and nutrient abundance, such as insulin or GLP-1, conditions of caloric restriction and the release of Ghrelin induce the nutrient sensor AMPK to allow neurons to adapt to the lack of bioenergetic substrates (glucose). This involves the AMPK-mediated shut down of Akt/mTor-driven cell proliferation, growth and protein translation, the stimulation of PGC-1α/NRF-1/2-induced mitochondrial gene transcription and a switch toward lipid β-oxidation (see for details).

98Nonetheless, insulin resistance was found to impair the expression of PGC-1α and NRF-1 in T2DM patients. There seems to be a reciprocal relationship, with impeded PGC-1α activity triggering insulin resistance. In turn, IR knockout experiments and the use of insulin-re-sensitizing agents confirm that the resolution of insulin resistance restores the impaired activity of PGC-1α, energy metabolism and mitochondrial biogenesis. In this context, CREB/CREB-regulated transcription coactivator 2 (CRTC2)-co-signaling drives the expression of PGC-1α. It appears that insulin resistance heightens the activity of salt inducible kinases (SIKs) that phosphorylate CRTCs, leading to their sequestration by 14-3-3 proteins in the cytoplasm. Hence, this diminishes the nuclear translocation of CTRCs and the expression of PGC-1α under insulin-resistant conditions. GLP-1, in turn, not only improves insulin sensitivity, but also induces cAMP/PKA-signaling to inhibit SIKs and directly activate CREB to restore the CREB/CRTC2-driven transcription of PGC-1α. Indeed, in the sporadic UCD-T2DM rat model, the development of peripheral insulin resistance was associated with the hippocampal desensitization of the insulin pathway, resulting in enhanced lipid peroxidation (4-HNE) levels that were inversely correlated with those of PGC-1α. Liraglutide reversed these abnormal changes in the brain and further restored the levels of TFAM and other metabolic energy markers. Since cerebral insulin resistance is an early pathological event during AD and PD, while GLP-1 mimetics were shown to re-sensitize insulin-signaling in the brain (section “Insulin resistance and the neuronal energy metabolism”), this represents another mechanism that preserves the mitochondrial function during neurodegenerative diseases.

99Similar to PD, mitochondrial dysfunction is an early pathologic event in AD and promoted by Aβ. Indeed, mitochondrial abnormalities have been well documented in the brains of AD patients and reviewed elsewhere (see ). Aβ, which may enter mitochondria across cyclophilin D-containing mitochondrial permeability transition pores (miPTPs) or import by the translocase of the outer membrane (TOM) complex, contributes to mitochondrial injury by (i) the accumulation of APP/Aβ in the inner-mitochondrial membrane and import channels, which impairs oxidative phosphorylation (OXPHOS) and ATP generation, disrupts the mitochondrial membrane potential (MiMP) and increases ROS generation, (ii) the ROS-enhancing interaction of Aβ with Aβ-binding alcohol dehydrogenase (AβAD) within mitochondria and (iii) the binding of Aβ to cyclophilin D that evokes the opening of miPTPs, leading to, as further accelerated by oxidative stress or cytosolic Ca2+ overload, oxidative stress-associated mtDNA, lipid and protein damage, Ca2+ amassment in the mitochondrial matrix, swelling, depolarisation of the mitochondrial membrane potential, mitochondrial leakage and neuronal death. Aβ was further shown to impair the activities of complex I (in conjunction with Tau), complex III and, especially, complex IV of the ETC as well as the TCA enzymes α-ketoglutarate dehydrogenase (α-KGDH) and PDH. In turn, an impaired ETC heightens APP/Aβ processing. Other harmful events include that Aβ interferes with the anterograde axonal transport of mitochondria toward synapses, thus accelerating synaptic degeneration, and drives mitochondrial depletion and fragmentation due to the Aβ-induced deregulation of fusion and fission proteins. Many of the detrimental effects of Aβ are linked to the activation of GSK-3β, as shown in Figure 2 and explained below.

100Besides protecting from oxidative stress-associated mitochondrial damage, as discussed in the previous sections, GLP-1R induction shields mitochondria against Aβ and the associated harmful activation of GSK-3β. An ex vivo study demonstrated that the administration of GLP-1 protected mitochondria from Aβ1–42, as indicated by lessened mitochondrial ROS generation, in hippocampal slices. In vivo, GLP-1 further rescued the abnormally lowered Akt and heightened GSK-3β activities in the hippocampus of APP/PS1 rodents. Notably, the latter normalizing effects on Akt and GSK-3β, as related to the re-sensitization of the insulin pathway in the brain, were replicated by several GLP-1R or GLP-1R/GIPR dual agonists in different AD cell and animal models.

101The pathologic induction of GSK-3β is thought to be a key event in AD. The activation of GSK-3β, as driven by the Aβ-provoked impairment of insulin-signaling and the IR/PI3K/Akt pathway that suppress GSK-3β, negatively affects the mitochondrial biogenesis, motility, bioenergetics, integrity and mitochondria-associated apoptosis pathways. There are four mechanisms with which Aβ-associated GSK-3β-signaling injures mitochondria and all of them are resisted with incretin hormone treatment (Figure 2).

102First, GSK-3β opposes the insulin pathway to trigger reductions in the levels of PGC-1α, impairments in the mitochondrial biogenesis and mitochondrial fragmentation. PGC-1α is responsible for the transcription of key biogenesis genes, such as ETC enzymes (subunits of CytC oxidase/complex IV or the ATP5B subunit of ATP synthase/complex V), ANT1 (an ADP/ATP translocation channel), CytC (implicated in e– carriage and ATP synthesis by the ETC), NRF-1α (anti-oxidative transcription factor) and TFAM (mtDNA regulator and replication-inducer). The GSK-3β mediated degradation of PGC-1α seems to involve GSK-3β cleavage by Omi, which, in a GSK-3β-dependent manner, enhances the interaction of PGC-1α with SCF-Cdc4 E3 ligase to ubiquitinate and target PGC-1α for proteasomal degradation. On the other hand, GSK-3β (and p38) drive mitochondrial fragmentation by upregulating the fission-enhancers DLP1 and mitochondrial fission factor (MFF) levels and decreasing those of the fusion-modulator Mfn2. In this context, insulin-mediated PI3K/Akt-signaling regulates PGC-1α and mitochondrial biogenesis through the Akt-driven inhibition of GSK-3β, but also FOXO1 and PGC-1α. Furthermore, insulin was shown to augment the expression of important mitochondrial modulators, including Mfn2, TFAM or a subunit of CytC oxidase/complex IV. Of note, even though FOXO1 is responsible for the transcription of PGC-1α, FOXO1 may impede the activity of PGC-1α.

103There is evidence that GLP-1 mimetics restore the PGC-1α-driven mitochondrial biogenesis and rescue mitochondrial hyperfission in response to the activation of Aβ/GSK-3β and oxidative stress. A recent study showed that the application of GLP-1 agonists blocked the induction of GSK-3β and enhanced the transcription of PGC-1α, NRF-1, and TFAM to heighten the expression of the anti-oxidative enzymes catalase, SOD2 and GPx and prevent Tau hyperphosphorylation in AGE-injected animals. Moreover, this led to reduced mitochondrial ROS production, cristae damage and vacuole formation following AGE-treatment in vitro. In this context, besides enhancing mitochondrial biogenesis, PGC-1α further buffers the mitochondrial ROS production by navigating the expression of the aforementioned anti-oxidative enzymes. Interestingly, AGE-induced oxidative stress interrupted the interaction of PGC-1α with GSK-3β, whereas GLP-1 agonists restored protein binding. It is likely that the Akt-mediated (and inactivating) Ser9-phosphorylation of GSK-3β evokes this interaction with PGC-1α, although further studies are needed to decipher the downstream effect.

104Notably, the cAMP/PKA pathway protects from mitochondrial fragmentation though the PKA-driven phosphorylation and inhibition of the fission protein DLP-1 at Ser637. Both Liraglutide and Exendin-4 prevented the decrease in cAMP, phospho-PKA and phospho- Ser637-DLP-1 levels and restored those of the mitochondrial fusion inducers Mfn2 and OPA1 in the cortex of 5xFAD mice, which suppressed ROS overproduction, raised ATP generation and rescued cortical and hippocampal neuronal loss. The latter effects, including the cAMP/PKA-dependent inhibition of DLP-1, increase in fusion proteins as well as the protection from mitochondrial fragmentation, dysfunction and cell death were replicated in GLP-1 and Aβ1–42-co-treated cortical astrocytes in vitro. Indeed, the application of the GLP-1/GIP dual agonist DA4-JC enhanced the mitochondrial size and reduced mitochondria numbers in the hippocampus of 3 × Tg mice, confirming that GLP-1R activation protects from Aβ/GSK-3β-provoked mitochondrial fragmentation.

105To connect the improvements in PGC-1α and mitochondrial fission/fusion, the GLP-1R-mediated neuroprotection in AD animal models was shown to involve the PKA-mediated phosphorylation and activation of CREB. The latter transcription factor is jointly induced by the GLP-1R-mediated stimulation of the PI3K/Akt, Ras/Raf/MEK/ERK and cAMP/PKA pathways. Moreover, as dependent on CREB, Exendin-4 protected from caspase 3-driven apoptosis by the mitochondrial pathway in Aβ oligomer-stressed human neuroprogenitor cells. In the context of mitochondria, a study showed that the enhancement of cAMP/PKA-signaling restored phospho-CREB and PGC-1α levels as well as mitochondrial defects in APPswe M17 cells. Indeed, ATF2 and CREB jointly induce the transcription of the mitochondrial biogenesis-inducer PGC-1α, while PGC-1α enhances the transcription of Mfn2 and downregulates the expression of DLP-1 and phospho- Ser637-DLP-1 levels. Multiple studies further confirm that GLP-1R-induced cAMP/PKA-signaling is necessary for the protection from Aβ and mitochondrial damage in the brains of AD and T2DM rodents.

106Synoptically, GLP-1R agonists restore the neuronal insulin sensitivity and induce Akt to suppress the GSK-3β-mediated degradation of PGC-1α, oxidative stress and mitochondrial fragmentation following Aβ-exposure. Moreover, GLP-1R activation drives cAMP/PKA-signaling and the activation of CREB to normalize the expression of PGC-1α and fission/fusion-modulating proteins.

107Second, the induction of GSK-3β elicits the oxidative stress-promoted phosphorylation of the miPTP components cyclophilin D and voltage-dependent anion channel 2 to provoke excessive Ca2+-influx, swelling and mitochondrial membrane potential collapse. Indeed, the Akt pathway and inhibitory Ser9 phosphorylation of GSK-3β were shown to prevent miPTP opening and mitochondrial swelling in response to excessive ROS production. GSK-3β further counteracts Akt/NRF-2-mediated anti-oxidative gene transcription (such as HO-1 ), elicits the joint inactivation of the key TCA enzyme PDH with TPKI and inhibits mitochondrial complex I-IV and ATP synthesis. The impaired mitochondrial function, in turn, exacerbates ROS generation (see; Figure 2).

108In an Aβ1–42-induced AD rat model, Exendin-4 engaged PI3K/Akt-signaling to re-invigorate the impaired Akt phosphorylation, mitochondrial function, integrity, respiratory control ratio and ADP phosphorylation, while normalizing mitochondrial complex I, IV, and V activities (which were aberrantly enhanced by the acute Aβ challenge in this study, however). Similarly, Exendin-4 rescued the reduced complex I expression, ATP production and oxidative stress, as indicated by lowered SOD and increased MDA levels, in the hippocampus of 5xFAD rodents. Indeed, the mitochondrial enhancements by Exendin-4 correlated with spatial and working memory improvements in i.c.v. Aβ1–42-injected rodents. Besides Aβ, the latter GLP-1 mimetic blocked cell death, restored the mitochondrial biogenesis and the synthesis of mitochondrial ETC-related enzymes (complex I/NADH dehydrogenase 1, complex II/succinate dehydrogenase, complex III/cytochrome b 6, complex V/ATPase 6) via the PI3K/Akt pathway in INS-1E cells that were stressed with human islet amyloid polypeptide, which is a pancreatic amyloid.

109Transmission electron microscopy further confirmed that GLP-1 agonists protect from Aβ1–42-induced swelling, the loss of surface area and cristae as well as the deformation of mitochondria in hippocampal neurons in vivo. Importantly, mitochondrial cristae are sites for ETC complex formation. OPA1 remodels cristae by fusing parts of the inner-mitochondrial membrane, whereas Aβ predominantly accumulates in cristae and downregulates OPA1 levels. Notably, the insulin pathway regulates APP and Aβ trafficking across the trans-Golgi network to the plasma membrane through the MEK/ERK pathway, thus preventing the intraneuronal amassment of Aβ. As such, the re-sensitization of the insulin pathway in neurons by GLP-1 mimetics (section “Insulin resistance and the neuronal energy metabolism” and Figure 1) is likely to forestall Aβ accumulation and mitochondrial damage.

110Third, GSK-3β drives cell death by the mitochondrial apoptosis pathway through the direct inactivation of the anti-apoptotic Mcl-1 and activation of Bax, p53 and p21Cip1. Moreover, the activity of GSK-3β correlated with that of caspase 2/8, subsequent Bid cleavage and CytC liberation. On the other hand, the pharmacological inhibition of GSK-3β rescued caspase 3-induced apoptosis in SH-SY5Y cells following treatment with the mitochondrial complex I inhibitors MPTP and rotenone.

111As indicative of the suppression of the mitochondrial apoptosis pathway, the administration of GLP-1 analogs or GLP-1R/GIPR dual agonists enhanced the anti-apoptotic Bcl-2 levels in the hippocampus of WT, 3 × Tg, and 5xFAD rodents, normalized the increased Bax/Bcl-2 ratio in rats with hippocampal STZ-injections and lowered caspase 3 protein levels in the hippocampus of APP/PS1 rodents. In this context, the activation of the GLP-1R/PI3K/Akt leads to the Akt/CREB-conveyed transcriptional upregulation of Bcl-2 as well as the Akt-driven direct inactivation of GSK-3β and other pro-death effectors, such as the Bcl-2-sequestering Bad, pro-caspase 9 upstream of caspase 3, Bax and p53 (as part of the JNK/p53/Bax death axis) and FOXO transcription factors, for instance the FasL-synthesizing FOXO1 or the mTORC1-inhibitor FOXO3a.

112Fourth, as rescued by GLP-1 mimetics, GSK-3β is implicated in cholinergic dysfunction. As reversible with GSK-3β inhibitors, cellular stress elicits the detrimental mitochondrial translocation of GSK-3β to interfere with the mitochondrial biogenesis, dynamics, energy metabolism, mitochondrial membrane integrity, and survival pathways. As one of these stressors, Aβ1–42 provokes the activation of GSK-3β in mitochondria, which led to the GSK-3β-mediated inhibition of PDH and, hence, lowered levels of the acetylcholine-precursor acetyl-CoA in a study. In turn, Exendin-4 restored cholinergic dysfunction by engaging the GSK-3β-inactivating PI3K/Akt pathway, resulting in the normalization of the attenuated acetylcholine levels, lowered choline acetyltransferase and heightened acetylcholinesterase activities in the prefrontal cortex and hippocampus of i.c.v. Aβ1–42 infused rats.

113As master regulator of the cellular energy metabolism, the mTor pathway is responsible for the cellular growth and survival, protein translation, redox balance, and autophagy. Upon activation of growth factor receptors, including insulin and GLP-1, the induction of PI3K/Akt-signaling leads to the inactivation of the Rheb/mTORC1-repressor TSC1/2. Once activated, mTORC1 phosphorylates S6K and inactivates 4E-BP. This leads to increased protein translation, the mTORC1/4E-BP-specific induction of glycolytic genes via HIF1α, mTORC1/S6K/sterol regulatory element-binding protein (SREBP1/2)-mediated cell proliferation, lipid and nucleotide biosynthesis as well as the induction of the aerobic arm of the pentose phosphate pathway (PPP) that neurons utilize to generate NADPH for oxidative protection. Indeed, the induction of the Akt/mTor axis enhances the synthesis of glycolytic enzymes, for example HK2, PKM2, or LDH, as well as related effectors, such as GLUT1, and PPP enzymes, i.e., 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (PFKFB3). In addition, mTORC1, by inhibiting 4E-BP, enhances the translation of nuclear, mitochondria-associated mRNAs, for example those for TFAM that augment the replication of mtDNA and transcription of ATP synthase (mitochondrial complex 5) of the ETC. Lastly, the mTORC1/yin-yang 1 (YY1)-driven induction of PGC-1α is mandatory to sustain ATP production via oxidative phosphorylation (OXPHOS) and the mitochondrial biogenesis.

114An in vitro study indicated that the treatment with GLP-1 agonists upregulates the expression of pyruvate dehydrogenase even in unstressed SH-SY5Y cells, suggesting that GLP-1R-signaling directly supports the expression of glycolytic enzymes and, hence, the neuronal energy metabolism. Besides directly upregulating the activity of the mTor-inducer Akt in AD and PD animal models, GLP-1 re-invigorated the impaired IR/PI3K/Akt/mTor insulin pathway, as impeded by inflammation and Aβ, in neurons (Figure 1 and details in section “GLP-1R activation re-sensitises insulin signaling in the brain during AD and PD”). As a proof of concept, a phase II clinical trial in PD patients demonstrated that Exendin-4 enhanced insulin sensitivity, Akt and phospho-mTor levels in the brains of PD patients. Furthermore, liraglutide prevented a decline in the cerebral glucose metabolization and BBB glucose transfer rates in AD patients.

115While ignored in light of the amyloid cascade hypothesis in AD over the last decades, cerebral insulin resistance has now been acknowledged as an independent and early pathologic key event in both AD and PD. Multiple studies have demonstrated that AD patients exhibit pronounced deficits in the neuronal glucose metabolism and blood flow velocities in memory and cognition-processing brain regions, for example the temporoparietal and posterior cingulate cortex, correlating with the clinical progression of AD. Worsened blood flow and glucose hypometabolism, were also identified in the brains of PD patients which, interestingly, not only correlated to PD severity, but also the occurrence of dementia during PD. Moreover, both AD and PD patients showed the impaired expression of glycolytic and TCA-associated enzymes, while the formation of protein and lipid oxidation products was enhanced, symbolizing defects in the neuronal energy metabolism and elevated ROS-production in response to the bioenergetic dysfunction. The latter energetic impairments are believed to be an early consequence of insulin resistance in the brain, proposed to precede mitochondrial dysfunction and amyloid pathology. Importantly, it has been hypothesized that neuronal stress initially elicits glucose hypermetabolism in AD, followed by the development of cerebral insulin resistance and associated impairments in glucose metabolism that enforce a detrimental shift from glucose to alternative, but less efficient, energy substrates, for example ketone bodies and lactate. Indeed, markers of insulin resistance, such as increased quantities of Ser-phosphorylated and inhibited IRS-1 downstream of the IR, have been identified in the hippocampus and cortex of AD patients, correlating with plaque deposition, disease progression and memory decline. Likewise, PD patients displayed premature attenuations in IR receptor synthesis and insulin pathway inactivation in the SNpc, basal ganglia and other brain areas.

116Given that T2DM is a well-established AD and PD risk factor, cerebral insulin resistance is considered to be an important pathological event in neurodegenerative diseases. As a pivotal growth factor in the brain, the detrimental consequences of desensitized IR-signaling [IRS-1/PI3K/Akt/mTor and MAPK (ERK) pathways] result in synaptic, plasticity and memory deficits, accelerated plaque deposition, GSK-3β-mediated Tau hyperphosphorylation and aggregation, cellular growth, protein synthesis, mitochondrial and autophagy impairments, reduced protection from oxidative stress, diminished neuronal GLUT4 translocation (as regulated by the insulin pathway and inversely correlating with insulin resistance) and astroglia GLUT1 expression (which co-localizes with insulin resistance-associated Tau pathology), lowered cerebral glucose uptake, decreased blood flow due to the endothelial disturbance of insulin/NO-mediated vasoconstriction and, of course, bioenergetic defects due to the loss of glycolytic enzyme expression. Both AD patients and 3 × Tg AD mice displayed early peripheral glucose intolerance that was followed by attenuated PI3K/Akt-signaling, glycolytic flux, neuronal GLUT3 levels (but not astroglia GLUT1) and GLUT3 translocation to the plasma membrane as well as elevated glucose levels in the brain, suggesting a link between peripheral and cerebral insulin resistance. Notably, CREB regulates the expression of GLUT3, which is the predominant neuronal glucose transporter in the CNS. The cAMP/PKA/CREB pathway was shown to be impaired in the prefrontal cortex or hippocampus of AD patients, while the Aβ/Ca2+/calpain-mediated degradation of CREB, oxidative stress and astrogliosis reduce CREB levels, lead to CREB truncation and, as such, decrease the expression of GLUT3 during AD. Indeed, a study showed that the PKA-enhancer forskolin elevated the CREB-driven GLUT3 expression, glucose uptake and intracellular protein O-GlcNAcylation in SH-SY5Y cells.

117The latter studies suggest that the insulin resistance-associated reduction in the cerebrovascular blood flow and astroglia GLUT1 expression reduce the BBB translocation of glucose, while the neuronal impairment in the PI3K/Akt pathway disturbs the exocytosis of GLUT3 and promotes the extraneuronal accumulation of glucose in the brain. Moreover, defects in PI3K/Akt/mTor-signaling interfere with the global protein translation and glycolytic energy metabolism in neurons. On the other hand, as a secondary consequence of reduced growth factor signaling, pathology-associated stress, including Aβ-induced Ca2+ overload, oxidative stress and astrogliosis, disturb the neuronal CREB activity to reduce GLUT3 expression and exacerbate glucose amassment in the brain tissue.

118Moreover, the IR contributes to declarative memory consolidation. Mice haploinsufficient for an allele of the IR β-subunit displayed selective deficits in PI3K/Akt/mTor-dependent, translational and long-lasting forms of LTP, which led to impairments in the hippocampal consolidation of object recognition memory. Indeed, memory-enhancing effects were observed following intranasal treatment in healthy volunteers, older adults as well as early AD patients that presumably had not developed cerebral insulin resistance yet.

119Although not a focus of this review, there is plentiful evidence that GLP-1 agonists protect from cerebral and peripheral insulin resistance as well as T2DM-associated hippocampal damage in vivo (see ). First, besides T2DM-associated animals, GLP-1 mimetics were shown to ameliorate markers of peripheral insulin resistance, including abnormally raised plasma insulin, circulatory glucose and glycated hemoglobin A1c levels, which is a clinical marker for T2DM that reflects the plasma glucose levels over the last 3 months, in AD animal models. Importantly, the chronic application of GLP-1 mimetics or GLP-1R/GIPR agonists do not affect insulin sensitivity, circulatory or hippocampal insulin, blood glucagon and leptin as well as fasting nor brain glucose pools in animals without metabolic impairment, although the anorexic effects of GLP-1 may promote mild weight loss. Second, in T2DM animal models, GLP-1R activation further resolved the cerebral insulin resistance, leading to the reversal of PI3K/Akt-signaling impairments, GSK-3β overactivation and concomitant Tau hyperphosphorylation, oxidative stress (heightened MDA and diminished SOD activities), apoptosis mediator induction (decreased Bcl-2/Bax ratios and Bcl-2 expression, with elevated Bax and active caspase 3), interrupted autophagy [deregulated AMPK/mTor pathway with lessened microtubule-associated protein light chain 3 (LC3)-II-containing autophagosomes and Beclin-1 plus exaggerated p62 levels], synaptic injury, neuronal atrophy as well as learning and memory deficits.

120Importantly, GLP-1 agonists have shown insulin-re-sensitizing effects in the brains of AD animal models. For example, GLP-1 agonists prevented the loss of dendritic IRs in vitro and insulin receptors (IR) in the prefrontal cortex and/or hippocampus of rodents and non-human primates in vivo following exposure or infusion of Aβ-oligomers. Furthermore, neurons in the frontal cortex of APP/PS1 exhibited aberrantly distributed IRs, forming dense receptor bundles in the neuropil, as well as increased levels of IRS-1 with inhibitory Ser616-phosphorylation. The latter indications of cerebral insulin resistance were not observed in age-matched control mice and co-localized with plaques and plaque-attracted glial cells. Treatment with liraglutide, however, ameliorated the plaque burden, diminished the microglial and astroglia activation, reduced plaque-interaction with glial cells and, most importantly, decreased the levels of inactivated phospho-Ser616-IRS-1 as well as IR inclusions. Similarly, the GLP-1/GIP dual incretin DA-JC4 reverted the elevated phopsho-Ser1101-IRS-1 and diminished phospho-Akt levels in the cortex and hippocampus of the i.c.v. STZ-injected AD rodent model, while liraglutide elevated the levels of phosphorylated IRs in cortical areas of WT and 5xFAD mice.

121There are three mechanisms with which GLP-1R induction protects from cerebral insulin resistance, as illustrated in Figure 1. First, GLP-1R-signaling forestalls Aβ production, accumulation and secretion by neurons through the inhibition of BACE1, the upregulation of IDE and normalization of the impaired autophagy function that is responsible for amyloid removal (section “GLP-1R agonists are neuroprotective and prevent amyloid beta accumulation in vivo”).

122Second, previous studies indicate that Aβ-oligomers bind to hippocampal synapses, provoke Ca2+-influx, the activation of Ca2+-sensitive enzymes calcium calmodulin-dependent kinase II plus casein kinase II, and the weakening of insulin-signaling by eliciting the internalization of membrane IRs, in particular those on dendrites. These effects could be replicated with glutamate or K+-evoked neuronal depolarisation and Ca2+-influx, while the Aβ-mediated endocytosis of IRs could be prevented with NMDAR antagonists or a Ca2+-chelator. Furthermore, the APOE4 allele, but also a high fat diet, accelerate the Aβ1–42-enforced trapping of IRs in endosomes in hippocampal neurons, which impaired insulin-signaling, the mitochondrial function and glycolysis, AMPAR-GluR1 phosphorylation and synaptic insertion, postsynaptic plasticity and spatial memory. In turn, oxidative stress, the loss of IRs and synaptic spines as well as LTP impairments could be rescued with insulin or insulin-re-sensitizer treatment in an IR-dependent-manner. Synoptically, Aβ evokes IR internalization by stimulating aberrant Ca2+ accumulation in the cytosol, whereas insulin/IR-signaling seemingly reduces the synaptic binding of oligomeric Aβ by preserving the synaptic exposition of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs; steric hindrance) and through mediating the endocytosis of Aβ oligomers for the intracellular reduction to Aβ monomers and degradation via IDE.

123Importantly, GLP-1R agonists protect from the Aβ/Ca2+-induced IR defects in neurons. Notably, a cAMP-inducer rescued insulin-signaling in ApoE3xAPP mouse-derived hippocampal slices, while the induction of GPCRs, in a cAMP/PKA-dependent manner, was shown to prevent the K+-triggered excessive Ca2+ influx through L-VDCCs. In this context, GLP-1R-driven cAMP/PKA-signaling and CREB gene induction protect from the Aβ, K+ ion or excitotoxicity-enforced opening of NMDARs and VDCCs which, besides preventing the depletion of ER Ca2+ stores and Ca2+ overload (section “GLP-1 mimetics suppress Ca2+ deregulation by amyloid beta and excitotoxicity”). Notably, unlike IRs, GLP-1Rs were not downregulated by oligomeric Aβ in vitro/in vivo. Moreover, Exendin-4, in contrast to insulin, did not prevent the interaction of soluble Aβ aggregates with neurons, albeit liraglutide reduced the interaction of oligomeric Aβ with synapses in another study. This suggests that (i) the GLP-1 pathway does not seem to desensitize in response to the Aβ pathology or in general [which is also why GLP-1, but not GIP, analogs were chosen as clinical T2DM treatments], (ii) GLP-1 mimetics prevent IR endocytosis through the GLP-1R-mediated suppression of the Ca2+ overload that is provoked by Aβ oligomers and (iii) GLP-1 may prevent the harmful interaction of soluble and extracellular Aβ species with neurons, possibly by promoting insulin-signaling. As discussed elsewhere, GLP-1R stimulation also aids the removal of Aβ (section “GLP-1R agonists are neuroprotective and prevent amyloid beta accumulation in vivo”).

124Third, while insulin exerts anti-inflammatory effects, inflammation is also the primary cause of insulin resistance in the brain (see Figure 1; ). For example, the i.c.v-injection of LPS caused memory impairments and desensitized insulin-signaling in the hippocampus of rats, as implied by Ser307-phosphorylated IRS-1 and decreased phospho-Akt levels, whereas insulin co-treatment rescued the latter. Mechanistically, similar to what happens in the periphery during T2DM, the TLR/NF-κB-mediated production and release of pro-inflammatory cytokines by stimulated microglia and astrocytes, including IL-1β, IL-6, IL-18, and TNF-α, result in PICR activation on neurons during AD and PD. Neuronal PICR induction leads to the MKK-mediated induction of JNK and p38, which both enhance the AP-1-driven transcription of APP, as well as the activation of IKKβ, resulting in IKKβ/NF-κB-conveyed inflammatory cytokine and BACE1 expression. NF-κB and AP1 may also be activated in response to oxidative stress. Importantly, besides enhancing the neuronal Aβ production, JNK and IKKβ directly elicit the insulin-desensitizing Ser-phosphorylation of IRS-1, whereas TANK-binding kinase 1 (TBK1), the upstream activator of IKKβ, was shown to trigger the insulin resistance-associated phosphorylation of IR at Ser994. Aβ was further shown to potentiate the TNF-α/PICR-dependent activation of PKR, another IRS-1 Ser-kinase in neurons, to drive IRS-1 phosphorylation and insulin resistance as well as ER stress-associated eIF2α induction and subsequent synaptic damage, LTP impairment and memory deficits in vivo. Indeed, the latter adverse effects following i.c.v. Aβ oligomer-infusion were absent in PKR–/– or TNFR1–/– rodents. The same study also suggested that the ER stress provoked in hippocampal neurons by oligomeric Aβ, as evident by enhanced XBP1s, BiP, and CHOP levels, contributed to insulin resistance. Indeed, ER stress, as suppressed by GLP-1 agonists (section “GLP-1 analogues counteract endoplasmic reticulum stress”), augments insulin pathway desensitization by potentiating inflammatory responses, such as the IRE1α-mediated activation of JNK, PERK, and ATF6 that jointly induce NF-κB-driven pro-inflammatory cytokine production.

125Importantly, incretin hormones shield from the negative effects of inflammation on insulin sensitivity. A study that used a mixture of Aβ-based AD models, including hippocampal primary neurons, murine brain slices and monkeys, demonstrated that Exendin-4, in a GLP-1R-dependent manner, diminished the levels of phospho-Ser312/ Ser616/Ser636 IRS-1, enhanced the activating Tyr465 phosphorylation of IRS-1 and restored (JNK-associated) axonal transport deficiencies as well as spatial learning and memory retainment. Aβ oligomers were shown to trigger IRS-1 phosphorylation and insulin resistance by stimulating JNK activity, whereas a GLP-1 analog protected from insulin pathway desensitization in a manner that was dependent on the suppression of JNK as well as the expression and release of TNF-α by hippocampal neurons. As elaborated in section “Inflammation,” GLP-1 mimetics further exert potent anti-inflammatory effects on microglia and astrocytes in both AD and PD models. Therefore, GLP-1 agonists prevent (Aβ-induced) pro-inflammatory cytokine production by neurons and glial cells, resulting in decreased neuronal PICR activation, reduced JNK/IKKβ/PKR-driven Ser-phosphorylation of IRS-1 and, hence, improved insulin pathway sensitivity.

126Besides AD, Exendin-4 and a GLP-1R/GIPR dual agonist reversed PD-related and 6-OHDA-induced insulin pathway impairments, including lessened Akt activity, CREB induction and elevated Ser312-phosphorylated IRS-1 levels in vitro plus lowered Ser129 pIRS-1/IRS-1 ratios in vivo. Moreover, Exendin-4 re-sensitized the Insulin/Akt/mTor pathway, as indicated by exosome analysis, in a phase II clinical trial with PD patients, leading to enhanced motor function (see section “Clinical trials show good protective effects in patients with AD or PD”).

127Notably, during T2DM, hyperglycaemia provokes increased AGE formation that, in turn, contributes to systemic oxidative stress, AGE/RAGE-induced inflammation and insulin resistance. Indeed, all of these harmful alterations during T2DM have been clinically, epidemiologically and pathologically connected to the accelerated progression of AD and PD. Post-mortem examinations have confirmed the enlarged presence of cytotoxic AGEs in neurons, astrocytes and microglia in the hippocampus and parahippocampal gyrus of AD patients as well as the neocortex of individuals with PD. Moreover, AGE-driven oxidative stress exacerbates Aβ expression, while AGE-driven amyloid glycation enhanced Aβ aggregation, α-synuclein crosslinking and Lewy body formation as well as Tau hyperphosphorylation by activating GSK-3β (summarized in ). Notably, AGEs showed a threefold higher binding affinity to ApoE4 compared to the ApoE3 allele, which might be a mechanism that accelerates plaque formation, the development of AD and cognitive decline in ApoE4 carriers.

128Both in vitro and in vivo evidence implies that a decline in glyceraldehyde-3-phosphate dehydrogenase (GAPDH), which is the sixth glycolytic enzyme that converts glyceraldehyde-3-phosphate to 1,3-diphosphoglycerate, enhances the build-up of methyl glyoxal (Figure 2). The latter is a glycating compound that reacts with other proteins and DNA to generate AGEs, as observed during T2DM and hyperglycaemia. Albeit increased GAPDH activity was reported in a small-scale post-mortem study, substantially decreased GAPDH activity and elevated quantities of ROS-inactivated and S-glutathionylated GAPDH were detected in the temporal cortex, inferior parietal lobule or isolated skin fibroblasts of individuals with AD. Likewise, nuclear GAPDH was identified in the SNpc of PD patients.

129As concluded elsewhere, GAPDH forms multimeric complexes with Aβ, while Aβ-associated and Aβ-independent oxidative stress leads to posttranslational disulphide bride formation, nuclear trapping and the enzymatic inactivation of GAPDH. Comparably, GAPDH-containing and insoluble paired helical filament Tau was found in the temporal cortex of AD and temporal lobe of Tauopathy patients. It is likely that the insulin resistance-associated downregulation of glycolytic enzymes, including GAPDH, and the associated worsening of the Aβ and Tau pathology play a major role in GAPDH dysfunction, subsequent glyceraldehyde-derived AGE formation, oxidative stress and inflammation.

130Indeed, the GAPDH/glyceraldehyde-derived AGE methyl glyoxal was demonstrated to be toxic in vitro and shown to build up in the cytoplasm of neurons, but not astrocytes, in hippocampal and parahippocampal brain regions of AD patients. Methyl glyoxal also reacts with dopamine to produce 1-acetyl-6,7-dihydroxy-1,2,3,4-tetrahydro-isoquinaline (ADTIQ) (see Figure 2). The latter not only imitates the effects of the PD-toxin MPTP, but ADTIQ was further heightened in hyperglycaemia-stressed cells and the brains of T2DM rats. Moreover, ADTIQ was found to amass in the putamen, SNpc, caudate nucleus and other brain areas of PD patients. Synoptically, methyl glyoxal depletes the cellular stores of the anti-oxidant GSH for its detoxification by the glyoxalase system, weakens the mitochondrial membrane potential, enhances the mitochondrial ROS and lessens ATP production, stimulates pro-inflammatory cytokine (IL-1β, TNF-α) expression, induces intracellular dopamine accumulation as well as impairs growth factor (BDNF) pathway signaling in neurons, leading to apoptosis.

131In vitro studies in high glucose or AGE-stressed neurons imply that GLP-1 or its analogs rescue diabetes-associated structural mitochondrial damage and membrane potential loss, ROS production, DNA oxidation, Aβ secretion, Tau hyperphosphorylation and neuronal death. Moreover, GLP-1R agonism protected from systemically administered AGEs/methyl glyoxal, leading to reduced hippocampal Tau hyperphosphorylation and apoptosis (caspase 3) induction, re-invigorated PGC-1α activity and mitochondrial biogenesis, enhanced anti-oxidative gene transcription as well as the rescue of synaptic injuries and spatial memory deficits in vivo.

132More specific for methyl glyoxal, liraglutide ameliorated the mitochondrial membrane potential and survival of SH-SY5Y cells by engaging Akt as well as MEK1/2 and its downstream target p90RSK, inhibiting the synthesis of pro-apoptotic Bax and Bik, blocking caspase 3 activity and restoring the decline in anti-apoptotic Mcl-1 levels. Furthermore, as shown in methyl glyoxal-stressed PC12 cells, GLP-1 prevents apoptosis by re-establishing the cellular redox balance (GSH vs. glutathione disulfide). This improvement in redox signaling was dependent on the PI3K/Akt/mTor-driven transcriptional upregulation of GCLc, which is the rate-limiting gene in GSH synthesis. Since cAMP and MEK inhibitors attenuated the anti-apoptotic effects of GLP-1 in these methyl glyoxal-treated PC12 cells, it is implied that the activation of the survival transcription factor CREB, as linked to cAMP/PKA- and MEK1/2 /ERK1/2/p90RSK-signaling, was involved (see also Figure 2). Indeed, as triggered with 2-iodo-4’-methoxychalcone, a recently discovered GLP-1R agonist, GLP-1R activation enhanced Akt, the inhibitory Ser9-GSK-3β and CREB phosphorylation, increased Bcl-2 expression via CREB, reduced the apoptosis-driving Bax, CytC and caspase 3/9 levels, upregulated BDNF and 75NTR/tyrosine receptor kinase B (TrkB) growth factor receptor synthesis, reduced ROS production and stimulated anti-oxidative defense mechanisms that included enhanced (GSK-3β-suppressed) NRF-2/HO-1 expression, SOD activity and GSH pools in methyl glyoxal-injured SH-SY5Y cells. CHA79 further prevented the decline in glyoxalase-1, which is involved in the detoxification of methyl glyoxal.

133Notably, in comparison to astrocytes, an in vitro study demonstrated that neurons are 6 times more vulnerable to methyl glyoxal due to the weaker expression of glyoxalase enzymes. In this context, astrocytes shield nearby neurons by internalizing methyl glyoxal, followed by the GSH-mediated reduction of the toxic carbonyl compound to hemithioacetal and subsequent processing to D-lactate via glyoxalase-1/2. Nonetheless, high methyl glyoxal concentrations eventually enforce AGE generation, glycolytic dysfunction and impaired glutamate uptake even in astrocytes, elevating the neuronal risk of exocytotic injuries and methyl glyoxal-induced anti-oxidant depletion. Given that GLP-1 agonists exert various pro-survival effects on astrocytes, such as enhancements in the expression of GSH and multiple neuro-supportive mechanisms (section “GLP-1 exerts direct anti-inflammatory and cytoprotective effects on astrocytes”), it is likely that GLP-1R activation in astrocytes ameliorates AGE/methyl glyoxal-induced neuronal damage.

134Autophagy describes an intracellular cellular clearance process that eliminates waste products, such as misfolded proteins or damaged organelles. There are 3 main forms of autophagy (illustrated in ) and, in the context of this review, we will specifically refer to the best characterized one: macroautophagy. Briefly, macroautophagy involves the engulfment of junk with autophagosomes that subsequently fuse with lysosomes for the enzymatic degradation of the captured material.

135However, as observed during the aging process and exacerbated in AD and PD, the gradual impairment of autophagy is thought to elicit the accumulation and aggregation of toxic waste products and amyloids, such as Aβ, Tau and α-synuclein. Importantly, post-mortem investigations have revealed that AD patients exhibit functional deficits in the lysosomal activity of proteolytic enzymes and the fusion of waste-containing autophagosomes with lysosomes. Thus, the latter impairments lead to the amassment of non-degraded junk, such as lipofuscin inclusions and amyloids, in autophagic vacuoles. Importantly, even though the rate of autophagy, as marked by reduced levels of the autophagy-initiating protein Beclin-1, is attenuated during an early stage in AD patients, therapeutic approaches to enhance the rate of autophagy only seem to work before degradation-resistant amyloid aggregates have formed. Similarly, PD patients exhibited a decrease in autophagy modulators, the accumulation of LC3-II-containing autophagosomes, as indicative of the impaired downstream fusion with lysosomes, the formation of intraneuronal α-synuclein deposits and the sequestration of proteins that regulate the maturation of autophagosomes (LC3, ULK1/2 etc.) by these α-synuclein aggregates in brain regions such as the SNpc. Indeed, it has been revealed that not the deceleration, but the functional impairment of autophagy, triggers neurodegeneration and an AD-like phenotype.

136As reviewed and further discussed in, there is clear evidence that GLP-1 agonists enhance the function of autophagy in various peripheral tissues, which mediates the survival of pancreatic β-cells during hyperglycaemia and T2DM. Similarly, treatment with GLP-1 agonists was shown to prevent T2DM-associated autophagy impairments in the brain in vivo.

137Our group demonstrated that the application of GLP-1 analogs enhances the transcription of autophagy inducers, including Atg3 and Atg7, and elevated LC3 levels, which implies an increase in autophagosome synthesis, even in non-stressed SH-SY5Y cells. The latter effects were PI3K-dependent, suggesting that the increase in autophagy proteins was driven by the GLP-1R/PI3K/Akt/mTORC1-mediated upregulation of protein translation.

138Generally, the mTor pathway regulates macroautophagy in response to the nutritional state. The lack of energy substrates elicits the release of hormones such as Ghrelin that induce the phosphorylation of the energy sensor AMPK, leading to TSC1/2 activation, the inactivation of mTORC1 drives autophagy by the ULK1/2 initiation complex. In turn, nutrient availability and growth signals, such as insulin or GLP-1, attenuate the induction of autophagy via the PI3K/Akt-mediated suppression of TSC1/2.

139In the context of macroautophagy, Beclin-1, along with vacuolar protein sorting 34 and VPS15, initiate the formation of autophagosomes. Furthermore, Atg4 cleaves immature LC3 at the C-terminus, followed by modification with phosphatidylethanolamine via Atg7 and Atg3 to promote the critical incorporation of LC3 as LC3-II into autophagosomes. While abnormally heightened P62/SQSTM1 levels indicate dysfunctional autophagy, it has been shown that polymeric p62 complexes are necessary for LC3 recruitment, autophagosome maturation and amyloid removal.

140Despite the fact that GLP-1 mimetics stimulate the PI3K/Akt/mTor pathway and, hence, reduce the rate of autophagy, they were shown to prevent autophagy dysfunction following ER stress and Aβ insult in AD animal models (Figure 1). ER stress, as induced by the accumulation of unfolded proteins and amyloids, oxidative stress or excessive Ca2+-amassment in the cytoplasm, impedes autophagy through two mechanisms. This involves the induction of the PERK/eIF2a arm that represses mTORC1-associated protein translation (including that of autophagy-mediators) as well as the activation of apoptosis modulators, i.e., caspase 3 and calpain, which cleave Atg proteins and Beclin-1 to drive suicide, if ER stress persists. In vivo studies indicate that the incretin dual agonist DA-CH3 not only raised Akt-signaling to resolve Aβ-triggered ER stress in the brains of APP/PS1 mice, but also normalized the decreased expression of Atg3, Atg7, Beclin-1 and the autophagosome-associated marker LC3. Furthermore, i.c.v. Aβ1–42-injected rodents developed lipofuscin inclusions in the hippocampus, which was prevented by liraglutide treatment, indicating that the autophagy function was improved. Similarly, liraglutide rescued the synthesis of Beclin-1 and LC3, with a trend for normalized Atg7, in response to thapsigargin-induced cytosolic Ca2+ overload and ER stress in SH-SY5Y cells. Notably, thapsigargin was demonstrated to terminate autophagy by impeding autophagosome synthesis and the fusion of autophagosomes with lysosomes. As such, the latter results support that GLP-1R activation prevents autophagy impairments in response to Aβ amassment, ER stress and Ca2+ deregulation in neurons.

141Autophagy-improving effects following GLP-1R activation were also observed in PD models. GLP-1 analogies and GLP-1R/GIPR dual agonists prevented the 6-OHDA-provoked downregulation of autophagy initiators, including Beclin-1 and p62, in neuroblastoma cells. Further in vivo studies in the MPTP and 6-OHDA mouse models showed that the expression of autophagy markers was impaired in nigrostriatal brain areas, whereas the treatment with liraglutide, semaglutide Exendin-4 or GLP-1R/GIPR dual agonists reversed the decline in Beclin-1, Atg7, p62 and LC3-positive autophagosomes. Strikingly, the Beclin-1 or p62 levels were even doubled in 6-OHDA/DA-JC1-co-injected animals compared to control mice. As indicative of improved autophagy function and the associated amyloid removal, GLP-1R or GLP-1R/GIPR dual agonists reduced α-synuclein accumulation and the deposition of insoluble α-synuclein/p-α-synuclein and ubiquitin inclusions in dopaminergic neurons in the brains of various PD in vivo models. Indeed, GLP-1 mimetic treatment extended the life span of α-synuclein hA53T Tg mice by 100 days.

142Mitophagy is a quality control process and a selective form of autophagy that is implicated in the degradation of dysfunctional and ROS-generating mitochondria. Indeed, the restoration of mitophagy, as deregulated in AD and PD, poses a therapeutic strategy that has been shown to prevent amyloid accumulation and aggregation, brain inflammation, neuronal loss and cognitive impairments. Classic Parkin-dependent mitophagy is concertedly executed by PINK1 and Parkin, while mutations in these genes lead to the degeneration of dopaminergic neurons and early onset PD. PINK1 senses mitochondrial damage, including membrane depolarisation and ROS leakage. Upon damage, PINK1 accumulates at the mitochondrial surface and attracts plus phosphorylates cytosolic or inner-mitochondrial Parkin. This results in the polyubiquitination of mitochondrial proteins, autophagy receptor recruitment and the formation of LC3-II-containing autophagosomes to engulf and digest the marked mitochondrion.

143As shown in 3 × Tg AD mice, a GLP-1R/GIPR dual agonist boosted mitophagy by raising the impaired hippocampal PINK1 and Parkin expression. Notably, the transcription of PINK1 and Parkin is regulated by NRF-1, while the expression of NRF-1 is under control of PGC-1α. GLP-1R agonists restored the expression of NRF-1 upon 6-OHDA-induced mitochondrial damage in neuroblastoma cells and in the brains of AGE-injected mice, while normalizing the decreased neuronal PGC-1α levels in response to H2O2, Aβ, or AGE-induced oxidative stress in vitro and in vivo in 3 × Tg AD or AGE-treated mice. Thus, GLP-1R activation prevents mitochondrial damage (section “Oxidative stress and mitochondrial dysfunction”) and the pathology-associated downregulation of transcription factors (PGC-1α/NRF-1; see Figure 2) that synthesize the mitophagy modulators PINK1 and Parkin.

144Intriguingly, while enhancing the impeded mitophagy in AD animals, it was reported that a GLP-1 analog quenched mitophagy-signaling in the nigrostriatal region of MPTP mice. Notably, excessive autophagy or mitophagy can promote cell death. It has been suggested that an imbalance between the mitochondrial degradation via mitophagy, as driven by PINK1, Parkin and BNIP3L/Nix, as well as mitochondrial multiplication through biogenesis, mainly induced by PGC-1α, is linked to the development of neurodegenerative diseases. Excessive mitophagy, yet impaired biogenesis, triggers apoptosis through the depletion of mitochondria and ATP, whereas aberrantly enhanced mitochondrial biogenesis in the absence of mitophagy encourages the accumulation of ROS-generating, damaged mitochondria. As indicated in H2O2-treated RGC-5 cells and confirmed with the autophagy-inducer rapamycin, Liraglutide rescued oxidative stress-induced cell death by suppressing autophagy and normalizing the heightened autophagosome formation (LC3-II/LC3-I conversion ratio) and Beclin-1 expression, while recovering the lowered p62 levels. Furthermore, Liraglutide reduced the mitochondrial injury, potentiated biogenesis (PGC-1α), yet blocked the H2O2-provoked expression of mitophagy markers [Parkin and BNIP3L/Nix, a Parkin-independent mitophagy inducer]. Comparably, Liraglutide rescued from apoptosis and mitochondrial defects in high glucose/hyperglycaemia-burdened retinal ganglion cells in vitro and in vivo, respectively, by attenuating autophagy, PINK1/Parkin expression and, hence, mitophagy. This implies that the anti-oxidative and mitoprotective effects of GLP-1 agonists (Figure 2) forestall the excessive need of repair mechanisms (autophagy and mitophagy) by protecting from oxidative damage in PD or T2DM/insulin resistance-associated hyperglycaemia. As such, GLP-1 mimetics restore the cellular homeostasis between autophagy/mitophagy and mitochondrial biogenesis.

145BDNF, which is regarded as the predominant and most widely expressed neurotrophin in the CNS, is a member of the neurotrophin family that encompasses NGF and neurotrophin 3/4. BDNF and its receptor TrkB play a major role in neuroprotection, dendritic branching, synapse formation, neurogenesis as well as plasticity, learning and memory. Interestingly, BDNF shares similar downstream signaling pathways with insulin, IGF-1 and GLP-1 that include the activation of the neuroprotective IRS-1/2/PI3K/Akt pathway that modifies survival modulators, such as Bcl-2 and Bad, and stimulates mTor-dependent protein translation as well as the Ras/Raf/MEK/ERK pathway, which results in CREB-mediated survival, growth, plasticity and de novo BDNF gene expression. Moreover, BDNF induces the PLC-γ/DAG/PKC pathway implicated in intracellular Ca2+ accumulation and the defense from pro-inflammatory (TLR) and pro-apoptotic GSK-3β/p38/caspase 3-signaling.

146Plasma and CNS investigations of the BDNF and TrkB levels in AD patients are inconclusive, showing elevated, lowered or no change in the levels of BDNF or TrkB. However, there seems to be a decline in the cortical levels of BDNF, more apparent in advanced stages of AD, that was also observed in Aβ-based AD rodent models (see ). It was reported that MCI and AD patients displayed massive reductions in the levels of proBDNF (21% / 30%) and BDNF (34% / 62%) in the parietal cortex that preceded acetylcholine dysfunction and correlated with cognitive impairments. Such BDNF reductions were seen in the prefrontal cortex, also inversely correlating with cognitive impairments in more advanced AD patients, as well as in Tau-bearing and non-tangle-affected neurons in the frontal cortex of AD patients. Interestingly, strong BDNF expression was discovered in dystrophic neurites that co-localized with plaques, while TrkB was selectively downregulated in Tau-loaded neurons. First, given that neuronal insulin resistance and Tau hyperphosphorylation mutually enhance each other, this suggests that the decrease in neurotrophin receptor expression was caused by impaired insulin-signaling. Second, in cortical neurons, APP-cleavage products, involving Aβ1–42, appear to disrupt the anterograde and retrograde trafficking of BDNF-containing vesicles, while sublethal concentrations of Aβ1–42 block BDNF-signaling (PI3K/Akt and Ras/Raf/MEK/ERK pathways) by interfering with the TrkB-mediated Tyr-phosphorylation of the docking proteins Shc (necessary for Grb2/SOS/Ras-signaling) and IRS-1 (responsible for PI3K and Grb2/SOS/Ras induction). Because Aβ interrupts BDNF-signaling, it is likely that early increases in BDNF/TrkB during AD reflect a compensatory response.

147BDNF is also expressed in motor areas, such as the basal ganglia or cerebellum, and is, hence, implicated in PD. In this context, BDNF/TrkB-signaling assists the survival of dopaminergic neurons in the SNpc and VTA as well as other nigrostriatal areas, while mediating the synthesis of dopamine D3 receptors and TH. Indeed, as summarized in, post-mortem investigations have confirmed a decline in BDNF levels in the SNpc, caudate nucleus and putamen, which was correlated with the magnitude of dopamine neuron death. In vitro studies further show that α-synuclein overexpression impairs ERK and PKC, enhances the (Akt-inhibited) GSK-3β activity and blocks the CREB-mediated transcription of BDNF.

148In vivo studies have demonstrated that GLP-1R, GLP-1R/GIPR, or GLP-1R/GIPR/glucagon receptor agonists enhanced the hippocampal or neocortical BDNF expression in WT mice, while restoring BDNF levels in the hippocampus of APP/PS1 mice, 3 × Tg and 5xFAD AD animal models as well as the SNpc and striatum of MPTP-treated rodents. Exendin-4 further recovered the downregulated CREB and BDNF synthesis in the hippocampus of a T2DM animal model. Lastly, GLP-1 stimulated the expression of BDNF, but also GDNF and NGF, in BV-2 and primary adult microglia in a PI3K and PKA-dependent fashion. Similarly, GLP-1R activation on astrocytes preserved the astroglial BDNF secretion following Aβ assault by inducing the cAMP/PKA/(CREB) pathway.

149Mechanistically, an in vitro study demonstrated that GLP-1R activation and downstream PI3K/Akt, ERK, and cAMP/PKA-signaling jointly elicit the phosphorylation of the transcription factor CREB to increase BDNF promotor activity and expression (Figure 2). Moreover, while the withdrawal of differentiation agents and neurotrophins from the culture medium led to the differentiation of human neuroprogenitor cells toward GFAP-positive astroglia, the application of Exendin-4 dose-dependently elevated BDNF secretion and preserved the neuronal phenotype, including the presence of neuritic and synaptic markers such as synaptophysin and MAP2a.

150Notably, oxidative stress and the formation of lipid peroxidation products was shown to reduce CREB expression, CREB phosphorylation and the CREB-mediated transcription of BDNF due to the induction of the JNK/c-Jun stress pathway and the competitive BDNF promotor binding by c-Jun. Given the anti-oxidative capacities of GLP-1 analogs (see Figure 2 and section “Oxidative stress and mitochondrial dysfunction”), this is another mechanism that preserves BDNF expression.

151Another trigger of BDNF impairments is inflammation. For example, Aβ1–42-provoked decreases in the cerebral BDNF levels could be rescued with the anti-inflammatory cytokine TGF-β1, whereas reduced BDNF transcription co-occurred with elevated IL-6 and TNF-α mRNA levels as well as reactive microglia numbers in the SNpc, striatum and hippocampus of PD and Lewy Body Disease patients. IL-1β was also shown to interrupt BDNF-signaling by eliciting the inactivating phosphorylation of IRS-1, a downstream mediator of both IR and TrkB. The release of inflammatory modulators by microglia and astrocytes, such as IL-1β, IL-6, and TNF-α, activate the corresponding PICR on neurons, resulting in the GSK-3β-driven inhibition of CREB-mediated BDNF expression, pro-apoptotic p38/caspase 3-signaling and the potentiation of NF-κB-conveyed inflammatory gene transcription. As such, it is implied that the anti-inflammatory effects following GLP-1R activation (section “Inflammation”) prevent inflammation-triggered impairments in the signaling pathway and the expression of BDNF.

152GDNF synthesis occurs in a few selected brain regions, including the dorsal and ventral striatum or SNpc, whereas GDNF family receptor α subtypes (GFRα1-4) are distributed across the CNS. GDNF shows high affinity for GFRα1, which further leads to the co-activation of the receptor Ret. While GFRαs are seemingly absent in the striatum, they are densely expressed by nigral dopaminergic neurons. Therefore, GDNF is considered to be the most crucial neurotrophic factor in the midbrain dopaminergic system. Mechanistically, GDNF/GFRα1/Ret-signaling leads to activation of the neuroprotective Gab1/2/PI3K/Akt and Grb2/SOS/Ras/Raf/MEK/ERK pathways to modulate the neuronal survival, differentiation, maintenance and organization. Presumably due to its cell-specific actions, GDNF is 5–10-fold more potent than BDNF in the rescue of nigrostriatal dopaminergic neurons, which has made GDNF the pursued therapeutic approach for PD. Indeed, PD patients have been shown to exhibit selective and grave reductions in the levels of GDNF in the SNpc (see;; ). Moreover, α-synuclein accumulation was shown to interfere with GDNF-signaling in nigral dopaminergic neurons by reducing the Nurr1-induced transcription of the GDNF receptor Ret. However, despite having high potential, the cerebral GDNF infusion in clinical PD trials failed to deliver benefits and led to severe side effects. It has been postulated that viral strategies to selectively increase GDNF expression in nigrostriatal regions, improved infusion protocols or the co-treatment of the Ret/GDNF-inhibiting α-synuclein pathology are necessary.

153Similar to BDNF, a range of studies has confirmed that single GLP-1 or dual-activating GLP-1/GIP incretin analogues stimulate the phosphorylation of Akt and CREB to rescue GDNF expression in the SNpc and striatum of MPTP-microinjected PD rodents. A recent study showed that both the expression of BDNF and GDNF in nigral dopaminergic neurons is mediated by phospho-CREB and dependent on the (Akt-associated) induction of mTORC1-driven protein translation (Figure 2; ). Given that α-synuclein intercepts Ret/GDNF-signaling in dopaminergic neurons, while GLP-1R activators were shown to prevent the accumulation of soluble, insoluble and phosphorylated α-synuclein species in PD in vivo models, GLP-1 mimetics might be a safer, easier administrable and more effective alternative to endogenous GDNF treatment.

154Albeit at a low rate, in rodents, neurogenesis occurs across adulthood and describes the birth, proliferation and differentiation of neural stem cells toward functional neurons. Neural stem cells are exclusively found in two specialized brain areas: the subgranular zone of the DG as well as the subventricular zone (SVZ). DG-derived neuroblasts may integrate as mature glutamatergic granule cells into the hippocampal formation to support memory, whilst SVC-originating neuronal precursor cells can migrate to the olfactory bulb. There is also evidence that DG immature neurons improve information resolution in the hippocampus, with higher numbers of newborn neurons facilitating pattern separation and declarative memory. Unsurprisingly, neurogenesis not only declines with age, but is impeded during AD and PD. Therefore, some therapeutic approaches aim to boost neurogenesis.

155Murine in vivo studies showed that GLP-1Rs are expressed by adult neural progenitor cells of the SVZ. Likewise, GLP-1R expression was observed in granular neurons and polymorphic cells of the DG, although the receptor levels on DG granule cells appear to significantly decline during early adulthood.

156As proof of concept, the injection of various GLP-1 or GLP-1/GIP mimetics enhanced the numbers of doublecortin (DCX)-displaying immature neurons in the SVZ, the nearby medial striatum or DG, increased neuroblast differentiation [bromodeoxyuridine (BrdU) and Ki67] and heightened their differentiation into adult neurons (NeuN) in the brains of WT, APP/PS1 or various T2DM animal models. Liraglutide failed to induce neurogenesis in the DG in APP/PS1 mice in another study, however. Given that GLP-1 mimetics improved the birth of neurons in WT mice as well, this suggests that neurogenesis was enhanced as a consequence of GLP-1R activation and not only through the improvement of the brain pathology. Besides in vivo studies, neurosphere cultures derived from the SVZ that were cultured on a liraglutide-disposing biodevice showed an increased birth of DCX-positive neurons as well as signs of differentiation. Similarly, GLP-1 and exendin-4 enhanced the neuronal proliferation, as assessed via BrdU and ATP levels, and the expression of the differentiation markers MAP2, β-III-tubulin and neuron-specific enolase in SVZ-derived neurospheres and the rat embryonic striatal ST14A stem cell line.

157The available evidence indicates that GLP-1 mimetics possess neurotrophic capabilities, enhancing the proliferation, differentiation and maturation of immature neurons in the DG and SVC. Explicitly, GLP-1 analogs were shown to induce PI3K/Akt and cAMP/PKA/CREB-signaling, resulting in a PI3K- and PKA-, but not ERK-, dependent and dose-related improvement in the viability and proliferation of SH-SY5Y neuroblastoma cells. Notably, in the APP/PS1 mouse model of AD, progenitor neuron proliferation in the DG (as measured by BrdU) begins to be impaired at an age of around 3 months, when oxidative stress and plaque formation start to incline, whilst the total quantities of DCX-positive immature neurons decreased from an age of 10 months. Moreover, GSK-3β, as inactivated by Akt and overactive during AD, led to growth cone collapse and neurite retraction in vitro as well as postsynaptic protein (PSD-95) reductions, GluR1 loss and dendritic shortening in immature DG neurons plus microgliosis and inflammatory cytokine expression in the DG in vivo. Given the pro-growth [PI3K/Akt/mTor pathway], anti-oxidative (section “Oxidative stress and mitochondrial dysfunction”), anti-amyloid (section “amyloid beta and Tau pathology in AD”) and GSK-3β-inhibiting (section “GLP-1R agonists suppress amyloid beta and GSK-3β-driven mitochondrial damage in AD”) effects of incretins, early GLP-1 treatment both stimulates growth and prevents proliferation deficits in immature neurons.

158Interestingly, GLP-1 preserved the neuronal phenotype of isolated human neuroprogenitor cells even after withdrawal of other differentiation-boosting agents (i.e., retinoic acid, EGF and FGF), albeit the neuronal differentiation was less efficient. As indicated in another study using PC12 cells, GLP-1 seems to potentiate the differentiation-enhancing effect of other neurotrophins (NGF in this case). Indeed, when GLP-1 was applied to SH-SY5Y cells following initial retinoic acid treatment, the incretin hormone attenuated vimentin levels (a marker of proliferating neural progenitor cells) and induced AMPA, NMDA, and dopamine receptor and synaptic protein expression, neurite outgrowth, the development of neuron-like morphologies and the ability to generate action potentials. Furthermore, amongst other effects, GLP-1R induction enhanced Na2+ and L/T-type VDCC currents, paralleling those of mature neurons.

159The use of pharmacological inhibitors demonstrated that the GLP-1-boosted growth, differentiation toward glutamatergic/dopaminergic neurons and survival were dependent on the PI3K/Akt/CREB axis and partially on MEK. Other studies indicate that GLP-1R-induced Ras/Raf/MEK/ERK-signaling, as confirmed with MEK inhibitor treatment of PC12 cells, drive GAP-43 expression as well as GAP-43-associated neurite outgrowth and branching. The neurite number and length-boosting effects of Exendin-4, as observed in cultured adult sensory neurons and SH-SY5Y cells, involved concomitant cytoskeletal adaptions, including an increase in F-actin and tubulin polymerisation as well as the phosphorylation and inactivation of the depolymerising factor cofilin. Therefore, GLP-1 induced PI3K/Akt-signaling supports the survival and proliferation of immature neurons, whereas the Ras/Raf/MEK/ERK pathway elicits neurite outgrowth.

160In addition, incretin hormones stimulate the differentiation of progenitor neurons across the CREB/BDNF axis. In vitro studies confirm that GLP-1R activation evokes CREB-phosphorylation via the PI3K/Akt, Ras/Raf/MEK/ERK and cAMP/PKA pathways, even under unstressed conditions, leading to the expression of BDNF and the differentiation toward a neuronal phenotype. It is widely accepted that the local, CREB-mediated expression of BDNF in the hippocampus (without any external BDNF sources from the blood) is mandatory for memory formation and neurogenesis. Indeed, CREB expression co-occurs with that of DCX and seems to terminate with the emergence of adult neuron markers in mature granule cells, such as NeuN. Moreover, CREB enhances neurite outgrowth, dendritic spine formation and functions as a pivotal survival factor in newborn neurons, responsible for the transcription of key modulators such as the anti-apoptotic protein Bcl-2 or the neurogenic factors BDNF and NGF. In this context, BDNF was shown to be expressed and secreted across dendrites by mature granule cells of the DG, followed by the induction of TrkB on GABAergic interneurons. The latter neurons subsequently promote the differentiation and maturation of neighboring neuroblasts as well as their DG integration as mature neurons. Importantly, GLP-1R or GLP-1R/GIPR dual agonists were shown to enhance CREB activity in various non-pathologic or pathologic contexts in vitro, in adult mice as well as AD, PD, and T2DM animal models in vivo. Moreover, GLP-1R activation led to the upregulation of BDNF in the hippocampus and other brain areas of WT and the aforementioned rodent models of neurodegeneration (section “GLP-1 mimetics stimulate BDNF synthesis in neurons and glia”).

161Besides neurons, glial cells contribute to neurogenesis through the hippocampal supply with BDNF. Similar to neurons, by inducing the PI3K/Akt and cAMP/PKA pathways, GLP-1 mimetics facilitate the glial delivery of neurotrophic factors by improving the microglial production and release of BDNF, GDNF, and NGF, while preventing the loss of the astroglial BDNF secretion in response to Aβ.

162Notably, neuroinflammation and the release of pro-inflammatory cytokines (i.e., TNF-α), but not amyloidosis per se, are thought to be the major cause of neurogenesis impairments. This discovery is supported by fact that the treatment with various anti-inflammatory agents rescued neurogenesis in AD and PD-like animals. It is likely that neuroinflammatory processes impair neurogenesis by interrupting BDNF-signaling and the transcription of BDNF. Given the potent direct and indirect anti-inflammatory effects of GLP-1 in glial cells and neurons (section “Inflammation”), it can be postulated that the administration of GLP-1 analogs maintains neurogenesis in light of age- and pathology-associated neuroinflammation.

163Based on the protective effects of GLP-1 receptor agonists on a wide range of physiological systems described above, several clinical trials have been conducted and have shown good effects. The results of four clinical trials in AD or PD have been conducted, testing the GLP-1 receptor agonists exendin-4 or liraglutide. These trials have shown clear and impressive effects that improve clinical pathologies of PD and AD. These results are a clear proof of concept that the use of such growth-factors are indeed effective and show meaningful effects.

164In a single-blinded, open-label pilot study that tested the drug exendin-4 (NCT01174810), 45 patients with moderate PD who were on L-Dopa-based treatment were randomized to either exendin-4 (Byetta) twice daily treatment for 12 months, or were controls without treatment. Drug treatment was for 12-month, followed by a 2-month “drug wash-out” period where no drugs were given to test if the drug effect is still visible. After 14 months, the result showed significant differences. A significant improvement was found in the Mattis dementia rating scale-2 (Mattis DRS-2) which tests the PD-associated cognitive impairments. Furthermore, a clear improvement in motor control was observed in the Movement Disorders Society Unified PD Rating Scale (MDS-UPDRS) part 3 in the drug group. The drug group did not deteriorate, while the control group showed steady and strong decline in all measures. The motor test battery MDS-UPDRS part 3 had been conducted by Neurologists blind to the treatment. Importantly, the benefits in both motor and cognitive abilities were still visible in follow-up tests 12 months after the last drug treatment. Again, the drug group did not deteriorate after 24 months in cognitive and motor assessments, while the control group had deteriorated severely. This demonstrates that the improvements in motor control and cognitive tests are not due to a placebo effect.

165Based on these encouraging results, a follow -up phase II clinical trial with a placebo control group had been run (NCT01971242). The once-weekly formulation of exendin-4 (Bydureon) was tested in PD patients for 60 weeks, accompanied by a placebo group. In this study, 60 patients with moderate PD were treated with exenatide (Bydureon; once weekly) or placebo for 48 weeks in addition to their dopaminergic replacement drugs. This was followed by a “wash-out” period of 12-weeks. Patients that had received the drug showed far better values in the part 3 of MDS-UPDRS motor tests than the placebo group following 48 weeks of exenatide therapy. As in the pilot study, the drug group did not deteriorate, while the placebo group did. After 12 weeks of no-treatment, the difference between groups was still visible. CSF analysis demonstrated that exendin-4 can cross into the brain, and furthermore showed that it was no longer present after wash-out. The improvements in MDS-UPDRS part 3 scores were still visible after “wash-out” at 60 weeks, demonstrating that the drug effect is not an simply an acute effect, satisfying the definition of disease-modification. In addition, a range of additional secondary measures was assessed, most of which showed an improvement.

166Exosomes are vesicles that are released from cells. Neuronal exosomes taken from blood samples from patients that took part in this clinical trial showed improvements in insulin signaling in the brain, similar to GLP-1 drug effects in diabetes. It was found that standard biomarkers of insulin signaling such as IRS-1 phosphorylation at tyrosine sites was improved in the brains of PD patients. This demonstrates that the preclinical research results translate into the clinic.

167A randomized, double-blind phase II clinical trial showed that liraglutide can improve PD pathology (NCT02953665). PD patients received once-daily subcutaneous injections of liraglutide for 52 weeks in addition to standard clinical medication to treat PD. The analysis included 37 active and 18 placebo subjects. At 54 weeks, non-motor symptom scores (NMSS) had improved in the liraglutide group and worsened in the placebo group, a 13.1 point adjusted mean difference (p = 0.07), just missing statistical significance. However, a significant improvement in MDS-UPDRS part-II scores in the treatment group had been observed (p = 0.001). This test battery evaluates everyday activities such as walking, talking, eating, getting dressed, getting out of a car, tremor, personal hygiene activities, and more. Furthermore, statistically significant improvement of Global MDS-UPDRS and PDQ-39 (quality of life) scores were found. The result shows that liraglutide improves PD pathology beyond the effects of L-Dopa.

168We have conducted a placebo- controlled double-blind phase II clinical trial testing liraglutide in over 200 MCI/AD patients for 1 year (ELAD study, NCT01843075). It analyzed the effects on cognition (ADASexec tests), brain volume changes as measured by MRI brain scans, and other parameters that are still being analyzed. It was found that liraglutide protected patients from cognitive impairment as measured in ADASexec tests (p < 0.01). In MRI brain scans, brain temporal lobe volumes shrank less in the drug group compared with the placebo group (p < 0.001), and the total gray matter cortical volume shrank less in patients, too (p = 0.002), indicating that neuronal loss has been reduced by the drug.

169Other clinical trials testing GLP-1 type drugs in PD patients are currently ongoing, testing the drugs semaglutide (NCT03659682), lixisenatide (NCT03439943), NLY01 (NCT04154072), or PT320 (NCT04269642) in PD patients, with additional phase 2 (NCT04305002) and phase 3 (NCT04232969) trial testing exenatide in PD patients, highlighting the growing interest and demand for discovery and development of GLP-1R agonists. Two phase III clinical trial testing semaglutide in AD patients are currently underway, too (NCT04777396 and NCT04777409).

170The first clinical trials in AD and PD show clear neuroprotective effects, with others reporting later this year. Phase 3 trials have started, demonstrating the confidence and interest of the industry in this novel approach. In addition, dual GLP-1/GIP receptor agonists are being developed as novel treatments for AD and PD. As current GLP-1 receptor agonists have been developed to treat diabetes, it is important to design drugs specifically for treating CNS diseases. Diabetes drugs have been developed to remain in the blood stream for a long time, ensuring protection from hyperglycaemia day and night. However, this means that the drugs do not enter the brain very well, reducing target engagement. There is a clear correlation between the neuroprotective potency of drugs and their ability to cross the BBB, as demonstrated by growth factors such as NGF, BDNF, and GDNF. While showing excellent neuroprotective effects in cell culture or when injected directly into the brain, they do not have any effects when injected peripherally. Hence, novel dual agonists have been developed to cross the BBB at an enhanced rate, showing superior effects in animal models compared with GLP-1 receptor agonists that have been developed to treat diabetes. Further improvements can be made by adding other receptor agonist binding sites to the peptide, with GLP-1 of GIP dual agonists only just starting this area of drug discovery. The future looks bright for this novel research area, and we may well be at the beginning of a paradigm shift in treating chronic neurodegenerative disorders such as AD and PD.

171The peptide hormone and growth factor GLP-1 plays key roles in cell growth, energy utilization, autophagy, protection from oxidative stress and in the control of chronic inflammation responses. Neurodegenerative disorders such as AD and PD are driven by chronic inflammation and impairments in growth factor signaling. Reversing this impairment by the use of GLP-1 mimetics has shown improvements in cell growth and repair, and in the re-activation of synaptic activity and energy utilization, while reducing the chronic inflammation response and removing misfolded protein build-up. Neuronal survival and functionality is rescued, and key pathological markers such as motor impairment, loss of the striatal-dopaminergic function and neuronal loss in PD, and loss of glutamatergic neurons and cognitive processes in AD are markedly improved in preclinical studies as well as in first clinical trials in AD and PD patients. These encouraging results demonstrate the big potential that the pharmacological treatment with GLP-1 mimetics holds.

172NR wrote the manuscript and made the figures. CH reviewed and revised the manuscript. Both authors contributed to the article and approved the submitted version.