Neuroprotective actions of Glucagon-like peptide-1 in differentiated human neuroprogenitor cells


  • Kalpana Velmurugan,

    1. Section of Endocrinology, Veterans Affairs Medical Center, Denver, Colorado, USA
    2. Department of Medicine, University of Colorado Denver, Aurora, Colorado, USA
    Search for more papers by this author
  • Ron Bouchard,

    1. Section of Endocrinology, Veterans Affairs Medical Center, Denver, Colorado, USA
    2. Department of Medicine, University of Colorado Denver, Aurora, Colorado, USA
    Search for more papers by this author
  • Gregory Mahaffey,

    1. Department of Medicine, University of Colorado Denver, Aurora, Colorado, USA
    Search for more papers by this author
  • Subbiah Pugazhenthi

    Corresponding author
    1. Department of Medicine, University of Colorado Denver, Aurora, Colorado, USA
    • Section of Endocrinology, Veterans Affairs Medical Center, Denver, Colorado, USA
    Search for more papers by this author

Address correspondence and reprints requests to Subbiah Pugazhenthi, Ph.D, Denver VA Medical Center and Department of Medicine University of Colorado Denver Mail Stop 8106, 12801 E 17th Ave Aurora, CO 80045, USA. E-mail:


Glucagon-like peptide-1 (GLP-1)-based therapies are currently available for the treatment of type 2 diabetes, based on their actions on pancreatic β cells. GLP-1 is also known to exert neuroprotective actions. To determine its mechanism of action, we developed a neuron-rich cell culture system by differentiating human neuroprogenitor cells in the presence of a combination of neurotrophins and retinoic acid. The neuronal nature of these cells was characterized by neurogenesis pathway-specific array. GLP-1 receptor expression was seen mainly in the neuronal population. Culture of neurons in the presence of Aβ oligomers resulted in the induction of apoptosis as shown by the activation of caspase-3 and caspase-6. Exendin-4, a long-acting analog of GLP-1, protected the neurons from apoptosis induced by Aβ oligomers. Exendin-4 stimulated cyclic AMP response element binding protein phosphorylation, a regulatory step in its activation. A transient transfection assay showed induction of a reporter linked to CRE site-containing human brain-derived neurotrophic factor promoter IV, by the growth factor through multiple signaling pathways. The anti-apoptotic action of exendin-4 was lost following down-regulation of cAMP response element binding protein. Withdrawal of neurotrophins resulted in the loss of neuronal phenotype of differentiated neuroprogenitor cells, which was prevented by incubation in the presence of exendin-4. Diabetes is a risk factor in the pathogenesis of Alzheimer's disease. Our findings suggest that GLP-1-based therapies can decrease the incidence of Alzheimer's disease among aging diabetic population.

Abbreviations used

Alzheimer's disease


advanced glycation end product


blood brain barrier


brain-derived neurotrophic factor


CREB binding protein

CNPase 2′

3′-Cyclic-nucleotide 3′-phosphodiesterase;


cyclic AMP response element binding protein




dibutyryl cyclic AMP


epidermal growth factor


fibroblast growth factor


Glial cell-derived neurotrophic factor


glial fibrillary acidic protein


insulin-like growth factor-1


nerve growth factor


neuroprogenitor cell

Alzheimer's disease (AD) which affects more than 35 million people worldwide is characterized by cognitive decline and progressive neurodegeneration. Hallmarks of AD pathology include accumulation of Aβ-containing plaques and neurofibrillary tangles, leading to synaptic dysfunction and neuronal loss (Querfurth and LaFerla 2010). The earliest event in AD pathogenesis is the abnormal processing of Aβ (Jack et al. 2010). A long asymptomatic phase is followed by neuronal dysfunction and then progressive decline in cognitive function (Jack et al. 2010). No therapeutic intervention capable of halting or reversing the disease process is currently available in spite of considerable advance in the understanding of AD pathology (Hardy 2006). Clinical trials with antioxidants, anti-inflammatory agents, cholesterol lowering agents, and drugs that target Aβ have all failed probably because of the inability to diagnose and initiate the treatment at an early stage (Palmer 2011). Effective treatment of this complex disease will require a combination of drugs that target multiple defects in this devastating disease. In this approach, it is essential to include a neuroprotective agent to repair neuronal injury along with drugs that can prevent the accumulation of neurotoxins including Aβ, free radicals, and inflammatory mediators in the AD brain. However, several neurotrophic growth factors showing protective actions in cultured neurons have limited ability to cross the blood–brain barrier (BBB). On the other hand, glucagon-like peptide-1 (GLP-1), a potential candidate with neuroprotective and anti-inflammatory actions, has the ability to cross BBB (Gengler et al. 2010).

Glucagon-like peptide-1, a hormone released by enteroendocrine L cells of the gut in response to food, increases glucose-dependent insulin secretion by β cells in the pancreas and decreases β cell apoptosis (Drucker 2006). GLP-1-based therapies are currently available in the treatment of type 2 diabetes to preserve β cell mass and function (Campbell and Miller 2009). Several studies have reported the actions of GLP-1 in other tissues especially in the brain [Reviewed in (Perry and Greig 2005; Holscher and Li 2010)]. Impairment of memory formation is observed in GLP-1R knock-out mice (Abbas et al. 2009). GLP-1 receptor (GLP-1R) activation has been shown to exert beneficial actions in Alzheimer's mouse models. For example, administration of liraglutide, a GLP-1 analog, to APP/PS1 mice for 8 weeks resulted in prevention of memory impairments in object recognition and water maze tasks (McClean et al. 2011). Significant decrease in β amyloid deposition is also observed in these mice.

The neuroprotective actions of GLP-1 have been also demonstrated in cultured neurons. For example, human neuroblastoma cell lines over-expressing GLP-1R are protected from oxidative stress-induced cell death (Li et al. 2010). Glutamate-induced death of cultured rat hippocampal neurons is decreased by GLP-1 (Gilman et al. 2003). Thus, the dual actions of GLP-1 in β cells and in neurons have generated therapeutic interest because type 2 diabetes is a risk factor for AD (Irwin et al. 2010). However, there is limited information available on the molecular mechanism of GLP-1 actions, especially at the level of transcription factors that regulate gene expression patterns. Recently, we reported a neuron-rich cell culture model by differentiation of human neuroprogenitor cells (NPC) (Pugazhenthi et al. 2011b). In this study, we have improved the differentiation protocol and have used this model to demonstrate that GLP-1R activation leads to protection of these neurons exposed to Aβ oligomers through activation of cyclic AMP response element binding protein (CREB), a transcription factor needed for cognition and synaptic plasticity.

Materials and methods


Cell culture media and supplies were purchased from Gemini Bio Products, Inc. (Woodland, CA, USA) and Invitrogen-Life Technologies (Rockville, MD, USA). Neuroprogenitor cells derived from human fetal brain were obtained as cryopreserved neurospheres from an authorized vendor (Lonza, Inc., Walkersville, MD, USA) that isolates them from donated human fetal tissues (16–20 weeks gestational) after obtaining permission for their use in research applications by informed consent or legal authorization. Neurobasal medium, supplements for proliferation and differentiation of NPCs, epidermal growth factor (EGF), and fibroblast growth factor (FGF) were purchased from Stemcell Technologies (Vancouver, BC, Canada). Nerve growth factor (NGF) was from Invitrogen-Life Technologies. Poly-l-lysine, mouse laminin, 4′,6-diamidino-2-phenylindole (DAPI), and dibutyryl cAMP were purchased from Sigma Chemical Co. (St. Louis, MO, USA). Brain-derived neurotropic factor (BDNF) and antibodies directed against the active cleaved form of caspase-3 and caspase-6, phosphorylated form of CREB, Akt and Erk, CREB, synaptophysin, BDNF, MAP2a, glial fibrillary acidic protein (GFAP), 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNPase), and β-actin were obtained from Cell Signaling (Beverly, MA, USA). NRCAM antibody was from Santa Cruz Biotechnology (Santa Cruz, CA, USA). GLP-1R antibody was purchased from R & D Biosystems (Minneapolis, MN, USA). Anti-rabbit IgG and anti-mouse IgG conjugated to fluorescent probes (Cy3 or FITC) were from Jackson Immuno Research Laboratories (West Grove, PA, USA).

Expansion and differentiation of human neuroprogenitor cells

Human NPCs were cultured in suspension as neurospheres (Fig. 1a) in neurobasal medium along with proliferation supplements, EGF (20 ng/mL) and FGF (25 ng/mL). When neurospheres reached the size of 300–500 μm, they were transferred to 15-mL tubes and centrifuged at 150 g for 5 min. The supernatant was discarded, and the pellet in 0.2 mL medium was triturated 75 times to yield a single cell suspension and the culture was continued. After around 10 days of culture, the NPCs proliferated and the neurospheres reached the original size for the next passage. When cultured in vitro, NPCs can differentiate into neurons, astrocytes, or oligodendrocytes depending on the culture conditions. To obtain predominantly a neuronal population, 2000 (4-day-old) neurospheres were seeded per well in 6-well dishes, coated with poly-l-lysine (100 μg/mL) and mouse laminin (5 μg/mL) and cultured in the presence of differentiation supplements and 25 ng/mL FGF, 20 ng/mL NGF, 10 ng/mL BDNF, 100 μM dibutryl cAMP, and 1 μM retinoic acid. After 1 week, differentiation supplement (Stemcell Technologies) was replaced with B27 supplement (Invitrogen-Life Technologies).

Figure 1.

Human neuroprogenitor cells (NPC)-derived neuron-rich cell culture model. (a) Human NPCs derived from fetal brain were propagated as neurospheres in suspension in neurobasal medium containing proliferation supplements, epidermal growth factor and fibroblast growth factor-2. (b) Neurospheres were broken by trituration and cultured for 4 days before seeding them in dishes coated with poly-l-lysine and mouse laminin for differentiation. (c) NPCs started differentiating a day after culturing in the presence of nerve growth factor, brain-derived neurotrophic factor, dibutryl cAMP, and retinoic acid. Appearance of neuronal processes was seen after 4 days (d) and 2 weeks (e) of differentiation. Differentiated NPCs were fixed, immunostained for MAP2a (f), a neuronal marker, GFAP (g), an astrocyte marker and CNPase (h), an oligodendrocyte marker. Nuclei were stained blue with DAPI. Neurons constituted 80%–90% of the cell population. (i) Withdrawal of neurotrophins for 1 week leads to loss of neuronal phenotype. Decrease in the staining for MAP2a (neuron) and increase in the staining for glial fibrillary acidic protein (astrocyte) were observed. Representative images are presented.

Immunofluorescence staining of neurons

Differentiated neurons were fixed in 4% paraformaldehyde for 30 min, washed in phosphate-buffered saline (PBS), treated with permeabilization buffer (5% bovine serum albumin and 0.2% Triton X PBS) for 60 min, and incubated with primary antibodies MAP2a (1 : 200), GFAP (1 : 1000), CNPase (1 : 250), GLP-1R (1 : 150), phospho CREB (1 : 500), phospho Akt (1 : 500), phospho Erk (1 : 500), and active caspase-3 (1 : 200) overnight at 4°C in a humidified chamber. After washing in PBS, cells were incubated with anti-rabbit-IgG or anti-mouse-IgG linked to cy3 or FITC and DAPI (2 μg/mL; nuclear stain) at 25°C for 90 min. Dishes were washed in PBS, mounted in mounting medium, and examined by digital deconvolution fluorescence microscopy using a Zeiss Axioplan 2 microscope (Carl Zeiss MicroImaging, Thornwood, NY, USA) fitted with Cooke SensiCamQE high performance CCD.

Human neurogenesis array

Total RNA was isolated from neurospheres and differentiated neurons using Versagene RNA isolation kit (Fisher Scientific, Pittsburgh, PA, USA). RNA samples following DNase treatment were converted to cDNA. The experimental cocktail was prepared by adding 1278 μL of the RT2 qPCR SYBR Green master mix, and 1173 μL H20 to 102 μL of the diluted cDNA and 25 μL of this cocktail was added to each well of the 96-well PCR array plate (SABiosciences, Frederick, MD, USA) containing primers for the 84 genes in human neurogenesis pathway (Table S1), five housekeeping control genes and three RNA, and PCR quality controls. Real-time PCR was performed with an ABI Prism 7700 sequence detector (Applied Biosystems, Foster City, CA, USA). The thermal cycling conditions were one cycle of 10 min at 95°C followed by 40 cycles of 15 s at 95°C and of 1 min at 60°C. After amplification, real-time data acquisition and analysis were performed through the Data Analysis Web Portal (SA Biosciences). Data analysis is based by the delta–delta Ct method with normalization of the raw data to glyceraldehyde 3-phosphate dehydrogenase as described.

Preparation of Aβ oligomers

The lyophilized Aβ peptide (1–42) was dissolved in hexafluoropropanol and incubated at 25°C for 1 h to monomerize Aβ. After evaporation of the solvent under vacuum, the peptide was stored at −20°C. Aβ was redissolved in dimethylsulfoxide to 5 mM, bath-sonicated, diluted to 100 μM with cold PBS containing 0.05% sodium dodecyl sulfate, and vortexed (30 s) (Ryan et al. 2010). After initial aggregation at 4°C for 24 h, Aβ oligomers were diluted to 50 μg/mL and incubated at 4°C for 2 weeks. Before use, the sample was centrifuged at 12 000 g for 10 min at 4°C and the supernatant was used. Aβ oligomer preparations were checked by western blot analysis using 6E10 antibody (Sigma). We were able to detect Aβ bands at the molecular weight range of 36–60 kDa, the species known to have neurotoxic properties.

Western blot analysis

Differentiated neurons following treatment were lysed using mammalian protein extraction reagent (M-PER, Pierce, Rockford, IL, USA) supplemented with phosphatase inhibitors (20 mM sodium fluoride, 1 mM sodium orthovanadate, and 500 nM okadaic acid) and protease inhibitor cocktail (Sigma). After centrifugation of cell lysates at 20 800 g for 15 min, the protein content of the supernatant was measured. Diluted samples containing equal amounts of protein were mixed with 2X Laemmli sample buffer, resolved on 12% sodium dodecyl sulfate–polyacrylamide gels, and transferred to polyvinylidene difluoride membranes. Blots were blocked and exposed to primary antibodies in TBST containing 5% BSA overnight at 4°C. After washing, the blots were exposed to alkaline phosphatase-conjugated anti-rabbit IgG or anti-mouse IgG and developed with CDP-Star reagent (New England Biolabs, Beverly, MA, USA). Band intensity was quantitated using Fluor-S MultiImager and Quantity One software from Bio-Rad (Hercules, CA, USA) and corrected for the levels of β actin.

Promoter assay by transient transfection

Neuroprogenitor cells were transduced with lentiviral CRE-luciferase for stable over-expression of the reporter and then differentiated into neurons. Alternatively, NPCs were differentiated in 12-well dishes and then transient transfection was performed using LipofectAMINE 2000 reagent (Invitrogen-Life Technologies) by the procedure described previously (Velmurugan et al. 2012). A luciferase reporter gene linked BDNF IV promoter (provided by Dr. Tonis Timmusk, Tallin University of Technology, Estonia) was used along with a constitutively active renilla luciferase (pRL-TK-luc) to correct for transfection efficiency. Following exposure to exendin-4, rolipram, dibutyryl cyclic AMP (DBC), and inhibitors of signaling pathways for 18 h, neurons were processed for the assay of luciferases using dual luciferase assay kit (Promega, Madison, WI, USA). The ratio of activities of two luciferases was taken as BDNF promoter activity.

Induction of neuronal apoptosis and caspase-3 assay

Neuronal apoptosis induced by chronic exposure to Aβ oligomers was examined by dual immunofluorescent staining for MAP2a, a neuronal marker and the active cleaved form of caspase-3. In addition, caspase-3 assay was performed using a kit from Sigma (Sarkar et al. 2007). The cells were lysed with the lysis buffer provided and 8000 g supernatant fractions were used for the assay. The substrate for the assay was p-nitroanilide labeled DEVD. The released chromophore was read at 405 nM in a microplate reader.

Statistical analysis

All statistical analyses were performed by one-way anova with Dunnett's multiple comparison tests.


Proliferation and differentiation of human NPCs

Human NPCs were grown in suspension as neurospheres under proliferative conditions (Fig. 1a). When neurospheres were broken into single cell suspension and cultured in the presence of EGF and FGF-2, they grew to form new neurospheres. The neurospheres can be expanded for upto 10–12 passages, which provide abundant supply of human NPCs and reduce the use of human fetal tissue. These multipotent cells can be also frozen and revived to generate viable cultures. We generally use stocks from multiple batches for each set of experiments. To obtain neuron-rich cultures, we had recently reported differentiation with a new combination of NGF, BDNF, DBC, and retinoic acid (Pugazhenthi et al. 2011b). In the previous report, neurospheres were triturated to a single cell suspension and then seeded in coated dishes. We have now modified the procedure by seeding 4-day-old neurospheres (Fig. 1b) to ensure sufficient density during differentiation because neuronal processes need to establish contacts (Fig. 1c–e). In addition, differentiation supplement was replaced by B27 supplement, 3–4 days after seeding.

Characterization of differentiated neurons

To determine the percent of neurons and glial cells following differentiation of NPCs for 2 weeks, we performed immunofluorescent staining for markers of neurons (MAP2a), astrocytes (GFAP), and oligodendrocytes (CNPase). We observed 80%–90% of the differentiated NPCs to be neurons (Fig. 1f). GFAP-positive astrocytes (Fig. 1g) and CNPase-positive oligodendrocytes (Fig. 1h) were also detected. Counting of neurons and astrocytes was performed in 20 random fields for four independent batches of differentiated NPCs. Following withdrawal of growth factors and retinoic acid, the percent of neurons decreased yielding predominantly a glial phenotype in 1 week (Fig. 1i). In addition, we performed a neurogenesis pathway-specific gene expression array of 84 genes (listed in Table S1) with neurons and neurospheres. We observed up-regulation a panel of genes after differentiation for 2 weeks. Genes showing more than 2-fold increase are listed in Table 1. Majority of them were markers of neuronal phenotype. Up-regulated genes included growth factors (EGF, BDNF, GDNF), transcription factors (Pax6, MEF2C and STAT3), regulators of cell cycle (Inhibin β-A and CDK5 regulatory subunit), and regulators of synaptogenesis (acetylcholinesterase and NRCAM). Increases (3.5-fold) in the levels of the transcription factor STAT3, which plays a role in GFAP expression, was observed as a result of generation of astrocytes.

Table 1. Genes up-regulated in differentiated human neuroprogenitor cells (Fold induction; Mean ± SE)
  1. Human NPCs were differentiated into a neuron-rich culture for 2 weeks. RNA was isolated from undifferentiated NPCs and from neurons. A neurogenesis pathway-specific array was performed. Data analysis was performed by the delta–delta Ct method between the two groups. Values represent mean fold induction in neurons over NPCs from four independent experiments.

Acetelcholinesterase6.8 ± 1.5
BDNF5.0 ± 0.7
CDK5 regulatory subunit5.5 ± 0.6
Ephrin-B13.5 ± 0.3
EGF3.8 ± 0.3
GDNF8.1 ± 2.7
GRIN19.5 ± 2.9
HDAC45.5 ± 0.6
Inhibin beta A5.2 ± 0.7
MEF2C4.3 ± 1.8
Neuronal pentraxin I4.2 ± 0.4
NRCAM4.8 ± 1.4
Neuregulin 15.3 ± 1.1
PARD6B8.3 ± 0.9
PAX65.8 ± 0.6
STAT33.5 ± 0.3

NPC-derived neurons express GLP-1R

Although the main target of GLP-1 action is pancreatic β-cells where it enhances glucose-stimulated insulin secretion, the receptor for this growth factor is known to be expressed in other tissues, especially in the brain. GLP-1 is known to modulate food intake through mesolimbic GLP-1R (Dickson et al. 2012). To determine if NPC-derived neurons express GLP-1R, dual immunofluorescent staining for GLP-1R and markers of neurons and glial cells was performed. GLP-1R staining was observed mainly in neurons (Fig. 2a), but not in GFAP-positive astrocytes (Fig. 2b) and CNPase-positive oligodendrocytes (Fig. 2c). As expected, GLP-1 was present in insulin-stained β-cells of human islets (Fig. 2d). The presence of GLP-1R in human islets and differentiated neurons was further confirmed by western blot analysis (Fig. 2e). These findings suggest that GLP-1-based therapies can target β cells and neurons in aging diabetic population.

Figure 2.

Neuronal expression of glucagon-like peptide (GLP)-1R. Differentiated neuroprogenitor cells were dual immunostained with cy3 (red) for GLP-1 receptor (GLP-1R) and with FITC (green) for MAP2a/neuron (a), GFAP/astrocyte (b), or CNPase/oligodendrocyte (c). GLP-1R was detected predominantly in the neuronal population, but not in glial cells. (d) GLP-1R was also detected in insulin-stained β-cells of human islets. (e) Western blot analysis of GLP-1R was performed with human fetal lung fibroblasts (HFL), human islets (HI), and human differentiated neurons (HDN). The blots were reprobed for β actin. GLP-1R band was strongly detected in human islets and neurons, but not in fibroblasts.

Exendin-4 protects neurons from Aβ-induced apoptosis

Aβ accumulation in Alzheimer's brain is known to play an important role in neuronal loss by the pathway of apoptosis (Roth 2001). Therefore, we examined the effects of Aβ oligomers in cultured NPC-derived neurons. Chronic exposure of neurons to Aβ resulted in the activation of caspase-3, a marker for apoptosis, as shown by immunofluorescent staining for the active cleaved form of caspase-3 (Fig. 3a). Aβ oligomers caused modest visible neuronal injury at 2 μM concentration. Although not dramatic in the presented image, overall, some decrease in neuronal density was observed. Pre-incubation of neurons with exendin-4 resulted in significantly reduced activation of caspase-3. Western blot analysis showed increases in the active form of caspase-3 and caspase-6 in Aβ oligomer-treated neurons. Caspase-6 plays an important role during the execution phase of apoptosis. Aβ also decreased the active phosphorylated form of CREB, an anti-apoptotic transcription factor (Fig. 3b). Quantitation of the bands after correction for β actin levels showed 186% and 157% increases in the activation of caspase-3 and caspase-6, respectively, and 55% decrease in CREB phosphorylation in Aβ-treated neurons. Exendin-4 decreased the activation of caspases significantly (p < 0.01) and restored CREB activation to normal levels.

Figure 3.

Aβ oligomers induce neuronal apoptosis. neuroprogenitor cells cultured in 6-well dishes were differentiated for 2 weeks. Neurons were exposed to 2 μM Aβ oligomers in the absence and presence of exendin-4 (200 ng/mL) for 6 days. (a) Treated neurons were immunostained for active form of caspase-3 (red) and MAP2a (green). Representative Images are presented. (b) Another set of treated neurons were processed for the western blot analysis of active cleaved form of caspase-3 and caspase-6, phospho cyclic AMP response element binding protein (PCREB) and CREB. The blots were reprobed for β actin. Aβ-induced neuronal apoptosis was significantly reduced by pre-treatment with exendin-4. Decrease in CREB phosphorylation by Aβ oligomers (Aβo) was restored to normal levels by exendin-4 (Ex). Values are Mean ± SE of four independent experiments. **p < 0.001 compared with control; *p < 0.01 versus Aβ oligomer-treated neurons.

Exendin-4 stimulates CREB phosphorylation at serine 133

To determine the mechanism of anti-apoptotic actions of exendin-4, we examined the signaling pathways leading to activation of CREB, a neuroprotective transcription factor which is also needed for synaptic function, cognition, and memory formation. After binding to CRE, CREB is phosphorylated at serine 133, a step needed for its activation. We have reported previously that insulin-like growth factor-1 activates CREB through multiple signaling pathways (Pugazhenthi et al. 1999a,b, 2000). To determine if exendin-4 stimulates those pathways, we examined the phosphorylated form of CREB kinases. We observed increases in the levels of phospho Akt (Fig. 4a) and phospho Erk (Fig. 4b) in MAP2a-positive neurons. Exendin-4 also increased the phosphorylation of CREB in neurons (Fig. 4c). Pre-incubation of neurons with H89, an inhibitor of cAMP-activated protein kinase A, resulted in significant decreases but not complete loss of exendin-4-mediated CREB phosphorylation (Fig. 4d). These findings suggest that cAMP-independent pathways are involved in GLP-1-mediated CREB phosphorylation.

Figure 4.

Cyclic AMP response element binding protein (CREB) phosphorylation by glucagon-like peptide-1R activation. Differentiated neurons were exposed to exendin-4 for 20 min. The treated neurons were fixed and immunostained for MAP2a with FITC (green) and phosphorylated form of Akt (a), Erk (b), and phospho CREB (PCREB; c) with cy3 (Red). (d) Neurons were exposed to exendin-4 for 20 min in the absence and presence of 20 μM of H89, an inhibitor of protein kinase A, followed by immunostaining for MAP2a (green) and phospho CREB (Red). Representative images from four independent experiments are presented.

Exendin-4-mediated induction of BDNF promoter

To determine exendin-4-mediated CREB activation at the functional level, we performed reporter assays. First, we stably over-expressed a luciferase reporter gene driven by multiple copies of CRE (4X) in NPCs by lentiviral transduction and then differentiated them to neurons. Exendin-4 induced the CRE-luciferase activity by 75% and 145% at a concentration of 100 ng/mL and 200 ng/mL, respectively (Fig. 5a). To determine the action of exendin-4 on a physiologically relevant promoter, we performed BDNF promoter assay by transient transfection. BDNF plays a significant role in synaptic plasticity and cognitive function. Human BDNF gene consists of 11 exons and nine functional promoters. Promoter IV, known to play an important role in BDNF gene transcription, contains a critical CRE site (Pruunsild et al. 2011). Cultured neurons were transfected with BDNF IV promoter linked to a luciferase reporter gene. Regulation of BDNF promoter by CREB was demonstrated by 57% decrease of promoter activity when cotransfected with a dominant negative mutant form of CREB (MCREB). Over-expression of wild-type CREB resulted in induction of the promoter by 226%. Treatment of transfected neurons with exendin-4, dibutyryl cAMP and rolipram, a PDE4 inhibitor that increases endogenous cAMP, resulted in 130%–275% increase in BDNF promoter activity (Fig. 5c). Aβ oligomers decreased the promoter activity by 58%, which was reversed by treatment with exendin-4 (Fig. 5d). Pre-incubation of transfected neurons with H89 (PKA inhibitor), U0126 (Erk inhibitor), and wortmannin (PI 3-kinase inhibitor) resulted in 50%, 36%, and 33% decreases in exendin-4-induced promoter activation, respectively (Fig. 5e). Thus, exendin-4-mediated CREB activation proceeds through multiple signaling pathways.

Figure 5.

Exendin-4 induces cyclic AMP response element binding protein (CREB)-responsive brain-derived neurotrophic factor (BDNF) promoter through multiple signaling pathways. (a) neuroprogenitor cells were transduced with lentiviral CRE-luciferase reporter and differentiated to neuron-rich cultures. Following withdrawal of growth factors, the neurons, stably expressing the reporter were cultured in the absence and presence of exendin-4 (Ex). (b–e) neuroprogenitor cells were differentiated in 12-well dishes and transfected with CREB-responsive human BDNF promoter IV using LipofectAMINE 2000 reagent. A constitutively active renilla luciferase was also included in the transfection to correct for transfection efficiency. (b) BDNF-luciferase was also cotransfected with the vector (pcDNA3) or plasmids encoding dominant negative mutant form of CREB (MCREB) or wild-type CREB (WT-CREB). (c–e) Transfected neurons were incubated in the presence of exendin-4 (Ex; 200 ng/mL), dibutyryl cAMP [(DBC); 750 μM], rolipram (Roli; 20 μM), Aβ oligomers (Aβo; 2 μM), protein kinase A inhibitor (H89; 20 μM), Mek inhibitor (U0126; 10 μM), or PI 3-kinase inhibitor (Wortmannin; 100 nM). The neurons were processed after 24 h for dual luciferase assay. The ratio of firefly and renilla luciferase activities was taken as the measure of BDNF promoter activity (b–e). Results are Mean ± SE of four independent experiments. #p < 0.05; *p < 0.01; **p < 0.001 compared with untreated or vector controls.

Anti-apoptotic effects of exendin-4 proceeds through CREB

To determine if the anti-apoptotic effects of exendin-4 require CREB, we over-expressed in neurons MCREB, a dominant negative mutant form of CREB in which the critical phosphorylation site is mutated (S133A) by adenoviral gene transfer. MCREB can bind to CRE site in a promoter but because of its inability to get phosphorylated, cannot bind to the coactivator CBP. Following transduction of neurons with adenoviral MCREB and exposure to exendin-4, there was a significant decrease in the levels of active phosphorylated form of CREB (Fig. 6a), suggesting that this mutant interferes with GLP-1-mediated CREB phosphorylation. There was also induction of apoptosis in neurons expressing MCREB as shown by increased immunostaining for the active cleaved form of caspase-3 (Fig. 6b). In addition, caspase-3 assay was performed with lysates from Aβ-treated neurons. Aβ oligomers increased caspase-3 activity by 145%. Exendin-4 decreased Aβ-mediated caspase-3 activation by 40% (p < 0.01). Transduction of neurons with adenoviral MCREB resulted in 130% increase in caspase-3 activation even in the absence of exposure to Aβ. Furthermore, the anti-apoptotic actions of exendin-4 were significantly lost following down-regulation of CREB. These observations suggest that CREB is needed for neuronal survival and to mediate the actions of GLP-1.

Figure 6.

The anti-apoptotic action of exendin-4 requires cyclic AMP response element binding protein (CREB). Differentiated neurons were transduced with adenoviral β galactosidase (AdV β Gal; control) or mutant form of CREB (MCREB). (a) After 2 days, the transduced neurons were exposed to exendin-4 (200 ng/mL) for 20 min, fixed and immunostained for MAP2a with FITC (green) and for phospho CREB with cy3 (red). Expression of the dominant negative mutant form of CREB (MCREB) resulted in significant decrease in exendin-4-mediated phosphorylation of CREB. (b) After 3 days of adenoviral transduction, the neurons were fixed and immunostained for MAP2a with FITC (green) and for the active form of caspase-3 with cy3 (red). Over-expression of CREB mutant resulted in induction of neuronal apoptosis. (c) Neurons transduced with indicated adenoviruses were exposed to 3 μM of Aβ oligomer (Aβo) in the absence or presence of exendin-4 (Ex) for 3 days and processed for the assay of caspase-3. Aβ treatment resulted in increases in the activities of caspase-3, which was significantly reduced by treatment with exendin-4 when control adenoviral β galactosidase was expressed, but not when transduced with adenoviral MCREB. Values are Mean ± SE of four independent experiments. *p < 0.01; **p < 0.001 versus untreated neurons transduced with adenoviral β galactosidase. #p < 0.01 compared with neurons exposed to Aβ oligomers in the absence of exendin-4.

Extendin-4 sustains neuronal phenotype in the absence of neurotrophins

Glucagon-like peptide-1 is known to induce neurogenesis in the brain (Perry and Greig 2002; Hamilton et al. 2011). We had demonstrated in the initial characterization that following withdrawal of neurotrophins, differentiated neurons switched to a glial phenotype (Fig. 1i). We used this model to determine if GLP-1 can prevent this reversal. Following withdrawal of neurotrophins, significant decreases in the levels of BDNF and synaptic markers including NRCAM, synaptophysin, and MAP2a were observed as shown by western blot analysis (Fig. 7a). Quantitation of the bands revealed the decreases to be 40%–55%. Exendin-4 was able to prevent this change in neuronal phenotype as shown by restoration of the levels of neuronal markers. In addition, growth factor withdrawal resulted in significant increase in the level (150%; p < 0.001) of GFAP, a marker for astrocytes (Fig 7a and c). GLP-1R activation decreased GFAP levels by 45% (p < 0.01). However, when NPCs were differentiated in the presence of exendin-4 alone, neuronal differentiation was not as efficient as with the combination of neurotrophins, DBC, and retinoic acid (results not shown). Increased secretion (188–277%) of BDNF in the culture medium of exendin-4-treated neurons was observed (Fig. 7d).

Figure 7.

Exendin-4 maintains neuronal phenotype in the absence of neurotrophins. Human neuroprogenitor cells were differentiated in the presence of neurotrophins, dibutyryl cyclic AMP and retinoic acid for 2 weeks. Growth factors and dibutyryl cyclic AMP were withdrawn (GF/WD) by culturing neurons in neurobasal medium containing B27 supplements in the absence and presence of 200 ng/mL of exendin-4 (Ex) for another 10 days. (a) Western blot analysis for neuronal and astrocyte markers was performed. Representative images are presented. (b and c) The bands were quantitated by scanning and corrected for the levels of β actin. Sy: Synaptophysin Results are Mean ± SE of four independent experiments. *p < 0.01; **p < 0.001 versus neurons cultured in the presence of neurotrophins. (d) Differentiated neurons were cultured in the absence of neurotrophins (control) and in the presence of indicated concentrations of exendin-4 for 3 days. The medium was collected and brain-derived neurotrophic factor (BDNF) levels were determined using a Quantikine ELISA kit from R & D Systems. #p < 0.05; *p < 0.01 compared with control.


A neuroprotective agent that can cross BBB has therapeutic significance in the treatment of AD. To test such agents, we have developed a cell culture system by differentiating human NPCs into a predominantly neuronal population. These neurons express GLP-1 receptors and can be maintained in culture for up to 10 weeks. Chronic exposure of neurons to Aβ oligomers resulted in the induction of apoptosis as shown by activation of caspases. Exendin-4, a long-acting analog of GLP-1 induced BDNF promoter through multiple signaling pathways and protected the neurons from Aβ-induced apoptosis, an effect that was dependent on the activation CREB, a neuroprotective transcription factor.

Human NPC-derived neuron-rich culture is a very good model to study the mechanism of Aβ-induced neurodegeneration and to screen for neuroprotective agents. To further characterize these neurons, we performed neurogenesis pathway-specific gene expression profiling. Differentiation of NPCs with neurotrophins and retinoic acid was accompanied by coordinated regulation of gene expression patterns. The list of up-regulated genes included acetylcholinesterase, BDNF, CDK5R1, ephrin-B1, HDAC4, MEF2C, NRCAM, neuregulin 1, PARD6B, and PAX6 (Table 1). Previous studies have reported the functional significance of many of these genes in neurogenesis. For example, the regulatory subunit of CDK5 is essential for neurite outgrowth (Nikolic et al. 1996). Histone deacetylases (HDACs) play an important role in epigenetic control of gene expression. HDAC4, known to regulate neuronal survival is enriched in the brain (Chen and Cepko 2009). PAX6 is a transcription factor that increases the generation of neurons from NPCs (Kallur et al. 2008). Ephrin-B1, present in dendritic spines, plays a role in synaptic plasticity (Migani et al. 2009). Although the differentiated NPCs were predominantly neuronal in nature, astrocytes were also generated which was reflected in the up-regulation of the transcription factor STAT3 which is needed for the induction of GFAP, a marker for astrocytes. Deletion of STAT3 has been shown to promote neurogenesis and inhibit astrogliogenesis (Cao et al. 2010). Detailed follow-up studies on these critical genes can provide us the mechanism of neurogenesis in vivo and can lead us to developing strategies for enhancing neurogenesis.

Several reports have shown that caspase activation, a marker of apoptosis, is observed in neurons of the AD brain [reviewed in (Roth 2001)]. We have reported Aβ-induced apoptosis in rat hippocampal neurons and detection of markers of apoptosis in the AD post-mortem brain (Pugazhenthi et al. 2011a). In this study, we observed that activation of caspase-3 and caspase-6 by Aβ oligomers is significantly reduced by exendin-4 (Fig. 4). Caspase-6, in addition to its known role as an executioner caspase, is involved in the pathogenesis of several neurodegenerative diseases (Graham et al. 2011). Substrates of caspase-6 include nuclear transcription factors. It has been also shown that caspases activated in the AD brain play critical roles in Aβ deposition and Tau pathology in addition to their role in the loss of neurons. For example, caspase-cleaved tau undergoes conformational changes that facilitate filament formation, and Aβ localizes to cleaved tau (Rissman et al. 2004). Over-expression of bcl-2, an anti-apoptotic gene, in the triple transgenic Alzheimer's mice results in reduced β amyloid deposition, decreased Tau pathology and improvement in cognitive function (Rohn et al. 2008). Thus, neuronal apoptosis pathway is an important therapeutic target in the neurodegenerative diseases.

Although GLP-1 actions in cultured neurons and transgenic AD mouse models have been reported, its mechanism of action at the molecular level is not clearly understood. GLP-1 is known to activate G-protein coupled receptors leading to the activation of adenylyl cyclase resulting in the generation of cyclic adenosine monophosphate (cAMP). Protein kinase A, activated by cAMP and PI 3-kinase play major roles in the neuroprotective actions of GLP-1 in SH-SY5Y cells over-expressing GLP-1R (Li et al. 2010). Transcription factors provide a crucial link between signaling pathways and nuclear gene expression. Cyclic AMP response element binding protein (CREB) is a nuclear transcription factor that enhances cognition and memory formation (Frank and Greenberg 1994). In the human AD post-mortem brain, decreases in the phosphorylated form of CREB (Yamamoto-sasaki et al. 1999) and CREB-regulated BDNF levels (Phillips et al. 1991) have been reported. We have characterized in rat neuronal cells, the signaling pathways that mediate CREB activation by insulin-like growth factor-I (Pugazhenthi et al. 1999b, 2000). We have also determined the mechanism by which 4-hydroxynonenal, a lipid peroxidation product that causes neuronal injury in AD brain, interferes with CREB function (Pugazhenthi et al. 2003, 2006). In this study, we demonstrate that GLP-1 induces CREB-regulated BDNF promoter through multiple signaling pathways (Fig. 5e). GLP-1 was able to protect neurons from Aβ-induced apoptosis, but this action was lost in neurons transduced with a dominant negative mutant form of CREB (Fig. 6). We also demonstrate that following withdrawal of neurotrophins, the levels of synaptic markers are significantly decreased which is prevented by GLP-1R activation (Fig. 7). We have previously reported decrease in IFN-γ-stimulated STAT-1 protein levels in human islets exposed to exendin-4 and PDE inhibitors (Pugazhenthi et al. 2010). Thus, GLP-1 actions in multiple pathways are relevant to the treatment of AD.

The association between diabetes and AD was first reported by a large population-based Rotterdam study (Ott et al. 1999). Later studies also strongly suggested that type 2 diabetes is a risk factor for AD [(reviewed in (Akter et al. 2011)]. The link between diabetes and AD has been further supported by several studies in animal models. For example, diet-induced insulin resistance in Tg2576 mice, a transgenic mouse model for AD leads to increased Aβ production (Ho et al. 2004). Crossing of APP23 transgenic mice with diabetic mice (ob/ob and NSY mice) results in exacerbated cognitive dysfunction (Takeda et al. 2010). In addition, these cross-bred mice show an accelerated diabetic phenotype compared with diabetic mice. There are several pathways in diabetic brain that resemble early events in AD. (i) Insulin and Aβ peptide are degraded by insulin degrading enzyme which during hyperinsulinemia is sequestered away from Aβ. (ii) Insulin resistance leads to decreased activation of Akt which inhibits GSK3β, one of the kinases that phosphorylate Tau protein. Therefore, during insulin resistance, increased GSK3β activation will lead to hyperphosphorylation of Tau, an important component of neurofibrillary tangles. (iii) Accumulation of advanced glycation end products is an important cause of diabetic complications. Receptor for advanced glycation end products (RAGE), is also a putative receptor for Aβ. Increases in the number of Aβ dense plaques and abundance of RAGE is observed in post-mortem brain samples of AD with diabetes when compared with AD brain samples (Valente et al. 2010).

We have previously reported the anti-apoptotic actions of GLP-1 in β cells of human islets (Sarkar et al. 2007; Velmurugan et al. 2012). With the availability of a wide range of drugs for managing diabetes and its complications, the life expectancy of diabetes is steadily increasing (Ioacara et al. 2011). We will be facing a growing population of AD patients with diabetes in the future (Han and Li 2010). Therapies in diabetic patients with dual beneficial effects in the brain and in β cells will have the advantage of preserving the β cell mass and of delaying aging-associated brain pathology.


This work was carried out with the use of resources and facilities at Denver VA Medical Center. This study was supported by Merit Review grant NEUD-004-07F from the Veterans Administration (to S.P.). RT-PCR analysis was performed at the University of Colorado Cancer Center Core Facility. Digital deconvolution microscopy was carried out at the VA Microscopy Core Facility. Authors declare that there is no conflict of interest associated with this manuscript.