Amyloid Beta monomers regulate cyclic adenosine monophosphate response element binding protein functions by activating type‐1 insulin‐like growth factor receptors in neuronal cells

Summary Alzheimer's disease (AD) is a progressive neurodegenerative disorder associated with synaptic dysfunction, pathological accumulation of β‐amyloid (Aβ), and neuronal loss. The self‐association of Aβ monomers into soluble oligomers seems to be crucial for the development of neurotoxicity (J. Neurochem., 00, 2007 and 1172). Aβ oligomers have been suggested to compromise neuronal functions in AD by reducing the expression levels of the CREB target gene and brain‐derived neurotrophic factor (BDNF) (J. Neurosci., 27, 2007 and 2628; Neurobiol. Aging, 36, 2015 and 20406 Mol. Neurodegener., 6, 2011 and 60). We previously reported a broad neuroprotective activity of physiological Aβ monomers, involving the activation of type‐1 insulin‐like growth factor receptors (IGF‐IRs) (J. Neurosci., 29, 2009 and 10582, Front Cell Neurosci., 9, 2015 and 297). We now provide evidence that Aβ monomers, by activating the IGF‐IR‐stimulated PI3‐K/AKT pathway, induce the activation of CREB in neurons and sustain BDNF transcription and release.


Introduction
Alzheimer's disease (AD) is the most common form of dementia in the elderly. The onset of the disease starts a decade or more before the first signs of cognitive impairment appear. For this reason, a great attention is being paid to the early events occurring in the AD brain, such as synaptic dysfunctions (Masliah et al., 2001).
The cyclic adenosine monophosphate response element binding protein (CREB) mediates the transcription of BDNF. CREB is activated in neurons in response to trans-synaptic signaling and regulates the expression of genes that are essential for adaptive neuronal responses, as well as more complex neural functions, such as learning and memory (Alberini & Chen, 2012). CREB target genes include immediate-early genes, such as c-fos (Sheng et al., 1990), and others important for synaptic function, such as BDNF (Shieh et al., 1998) and neuronal nitric oxide synthase (nNOS) (Sasaki et al., 2000). CREB activation is induced through ser-133 phosphorylation mediated by a variety of kinases, including the PI3K-activated AKT (reviewed in Lonze & Ginty, 2002). The PI3K/AKT pathway, by inhibiting glycogen synthase kinase-3 beta (GSK3-b), could also release CREB-mediated transcription from the blockade operated by GSK3-b phosphorylation at ser-129 (Grimes & Jope, 2001;Zheng & Quirion, 2006).
Here we demonstrate that Ab monomers, both synthetic and natural, by activating PI3K/AKT pathway, were able to induce CREB activation in differentiated human neuroblastoma cells and in primary rat cortical neurons. This effect was not shared by pathogenic Ab oligomers. Accordingly, CREB activation is impaired in the brain of patients with AD and Tg2576 mice (Pugazhenthi et al., 2011;Bartolotti et al., 2016a,b), where it is associated with decreased levels of CREB-regulated BDNF release (Pugazhenthi et al., 2011). Contrary to what takes place in the disease, we found that CREB activation induced by Ab monomers was paralleled by an increased neuronal BDNF release.

Results
Ab(1-42) monomers induce CREB phosphorylation in primary cortical neurons and in differentiated neuroblastoma cells To determine whether Ab monomers (mAb) were able to promote the activation of CREB, we assessed the expression levels of p(ser133)-CREB in mature cultures of pure cortical neurons. A 30-min exposure was sufficient for mAb to induce p(ser133)-CREB expression in a dosedependent manner (0.01-1 lM) (Fig. 1A,B), as assessed by Western blot analysis. Consistent with the ability of mAb to activate the PI3K pathway by engaging IGF-IRs (Giuffrida et al., 2009(Giuffrida et al., , 2015, the selective inhibitor of IGF-IR, picropodophyllin (PPP, 500 nM), and the PI3K inhibitor, LY294002 (10 lM), reduced the phosphorylation of both p(ser473)-AKT (Fig. 1C,E) and p(ser-133)CREB (Fig. 1D,F), which were induced by 100 nM mAb. Unlike mAb, Ab oligomers (oAb) reduced the basal expression of p(ser-133)CREB (Fig. 1G,H) at a concentration as low as 100 nM, leaving unaffected basal levels of p(ser473)-AKT up to a 2 lM concentration (Fig. 1I,L). Appropriate concentrations of oAb for signaling studies were selected based on parallel assessment of neuronal viability by MTT assay. Both concentrations of oAb were devoid of toxicity at 24h, and toxic to a different extent after 48 hours exposure (% neuronal survival: controls = 100 AE 10.33; oAb 100 nM = 82.3 AE 5.03; oAb 2 lM = 67.93 AE 3.52).
We also assessed CREB phosphorylation in the differentiated neuroblastoma SH-SY5Y, exposed for 30 min to 100 nM of either mAb or oAb. Changes in CREB phosphorylation were assessed by p(ser-133)-CREB immunofluorescent staining of cell nuclei, which were counterstained with Hoechst dye 33342 (1 lg lL À1 ). Consistent with its transcriptional activity, p(ser-133)-CREB was largely imaged within the cell nuclei following exposure to mAb (100 nM). The same nuclear pattern was observed by treating the cells with forskolin (30 lM), which was used as a positive control for imaging CREB activation, but not with synthetic Ab oligomers (oAb, 100 nM) ( Fig. 2A,B).

mAb induces CREB DNA binding activity in cortical neurons
To assess whether mAb activates CREB to induce the transcription of target genes, whole neuronal lysates were used to assay CREB binding to immobilized CRE (cAMP response elements) consensus sequences. Consistent with its ability to induce p(ser-133)-CREB, 100 nM mAb for 30 min promoted a twofold increase in the spectrophotometric signal detecting the activated CREB/CRE transcription factor complex (Fig 3). A similar increase was observed with 30 lM forskolin, used as positive control for the experiment, whereas 100 nM oAb failed to induce the transactivating activity of CREB (Fig. 3).
Native forms of Ab from H4-APPswe cells induce pCREB(ser-133) in primary cortical neurons We wondered whether native forms of Ab were able to replicate the effects of synthetic Ab monomers. To answer this question, we used the H4-APPswe cell line, engineered to over-produce Ab1-40 (Ab40) and Ab1-42 (Ab42) by the introduction of the double Swedish mutation (K595N/M596L) in the amyloid precursor protein (APP) gene. Ab levels reached in the conditioned medium (CM) of this stable transfected neuroglioma are low, likely insufficient to induce significant aggregation of the peptide (Lomakin et al., 1997;Harper et al., 1999). We collected a 6h and 24h conditioned medium from H4-APPswe cultures, incubated in presence or absence of bor c-secretase inhibitors, and Ab levels were measured by ELISA (Fig. 4A). Data showed that, after 24h, Ab concentration in the conditioned medium reached the range of few hundred picomolars, and it was sensitive to the presence of the secretase inhibitors (Fig. 4A). The 6E10 monoclonal antibody, raised against the 1-16 fragment of Ab, failed to reveal amyloid bands other than the 4kD monomer in the neuroglioma medium even when collected after 48h (Fig. 4B). Hence, we used the 24h conditioned medium to treat cortical neurons, and we evaluated the activation of CREB after 30-min exposure. Unconditioned medium was used as control. As expected, we observed an increased expression of both p(ser473)-AKT and p(ser-133)CREB after exposing neurons to H4-APPswe CM, which was prevented in the presence of either bor c-secretase inhibitor-treated CM ( Fig. 4C-F). Interestingly, the exogenous addition of synthetic mAb to the secretase inhibitor-treated CM was able to rescue the levels of both p(ser473)-AKT and p(ser-133)-CREB ( Fig. 4C-F). mAb promotes the transcription and release of the CREB target gene, BDNF, in primary cortical neurons Mature cortical neurons were treated with 100 nM mAb or 100 nM oAb for 24h, and the transcript for BDNF was quantified by real-time PCR. Once more, 30 lM forskolin was used as a positive control. mAb induced a discreet and consistent increase in BDNF mRNA levels across the experiments (Fig. 5A), and the heightened BDNF transcription was accompanied by an increase in neuronal BDNF release, as assessed by ELISA (Fig. 5B). As in the case of p(ser-133)-CREB expression, the IGF-IR inhibitor, PPP, prevented the activity of mAb, which was instead mimicked by IGF1 (5 ng mL À1 ) (Fig. 5B). Unlike mAb, oAb did not affect BDNF transcription (Fig. 5A) and release (Fig. 5B).

Discussion
Growing evidence exists that native Ab, which is produced during neuronal activity (Cirrito et al., 2005), may subserve physiological functions (reviewed in Giuffrida et al., 2010;Kepp, 2016). In particular, endogenous Ab seems to be relevant for the control of synaptic activity and memory formation (Puzzo et al., 2011;Piccini et al., 2012), via the activation of a-7-nAChRs (Calabrese, 2001). Based on this evidence, Teich et al. (2015) have suggested a central role for CREB and associated memory-related genes in the synaptic effects of native Ab.
Here, we provide the first evidence that Ab, in its monomeric form, induces the activation of CREB in neurons and promotes the transcription and the release of the CREB target gene, BDNF. Specifically, a 30min stimulation of primary cortical neurons with synthetic mAb (100 nM) resulted in phosphorylation of CREB at serine 133 (p(ser-133)CREB). This residue is a key site that must be phosphorylated for CREB to function as Amyloid-b  transcriptional regulator (Lonze & Ginty, 2002). Accordingly, mAbactivated neuronal CREB bound to CRE consensus sequences, and a nuclear localization of p(ser-133)-CREB was found in differentiated SH-SY5Y cells briefly exposed to mAb. Many routes converge into CREB phosphorylation (Lonze & Ginty, 2002), including the well-documented cAMP-PKA pathway. Indeed, forskolin, a drug that raises cAMP levels, was able to increase p(ser-133)-CREB in SH-SY5Y cells. However, at least under some circumstances, the PI3K/AKT pathway appears to control CREB phosphorylation in neurons (Zheng & Quirion, 2006). Consistent with our previous demonstration that mAb activates PI3K/AKT signaling by functioning as a positive allosteric modulator of IGF-IR (Giuffrida et al., 2009(Giuffrida et al., , 2015, both the specific IGF-IR inhibitor, PPP, and the PI3K inhibitor, LY294002, reduced strongly the activation of CREB promoted by mAb in cortical neurons. The evidence that native human Ab monomers, from the conditioned media of neuroglioma cells, induced CREB activation in neurons strengthened results obtained with synthetic Ab monomers. Ab peptides from different sources (i.e., cell-secreted, brain-derived or synthetic) are significantly different in terms of isoforms and effective concentrations (Reed et al., 2011).
Picomolar concentrations of a synthetic Ab solution, containing both monomers and oligomers, have been shown to mimic the memory enhancement effects of endogenous Ab (Puzzo et al., 2008). Instead, the solution used in this study contains a homogenous population of Ab monomers (Giuffrida et al., 2009), able to engage and activate IGF-IRs from recombinant cells at the lowest concentration of 10 nM (Giuffrida et al., 2015).
Although synthetic Ab monomers might oligomerize in the culture medium, spectroscopy techniques operating at single-molecule sensitivity levels have predicted that Ab stays monomeric up to at least 3 lM (Nag et al., 2011) (i.e., a concentration three times higher than the highest concentration of monomeric Ab used in this study). Moreover, the slow kinetic of Ab self-association in vitro (Kusumoto et al., 1998) is not consistent with the timing of CREB phosphorylation. Lastly, the contribution of Ab self-assemblies to the observed effect is unlike as concentrations of Ab oligomers as low as 100 nM were not able to promote CREB phosphorylation in neurons, but rather reduced it. However, Ab oligomers are highly heterogeneous and instable, ranging from low-n oligomers (dimers, trimers, tetramers) to higher order assemblies (MW > 50 kD) (Ferreira et al., 2015), as those used in the present study. As different oligomeric species might recruit distinct signaling mechanisms, we cannot predict the effect of very low concentrations of low-n oligomers, which have been shown to be relevant for proper synaptic function (Puzzo et al., 2011).
The role of CREB in AD pathology has been highlighted by the observation that pCREB is decreased in the nuclear fraction of postmortem prefrontal cortex (PFC) of individuals with AD, as compared to age-matched, cognitively normal controls (Bartolotti et al., 2016a,b). In the same study, the authors reported that pCREB expression in peripheral blood mononuclear cells paralleled pCREB expression in the PFC, proposing pCREB as a biomarker for cognitive function in AD. Whether or not peripheral pCREB levels could depend on circulating Ab monomers remain to be established. Noteworthy, skeletal muscle cells and insulinoma cells, which express functional IGF-IRs, respond to mAb stimulation (Giuffrida et al., 2015).
Consistent with the demonstration that mAb induced CREB activation in neurons, mAb was found to increase the neuronal transcription of BDNF. This increased transcription was associated with a rise in BDNF secretion. It remains to determine whether BDNF levels are dynamically regulated by the activity-dependent Ab release. Such a regulation would be in agreement with the evidence that in our hands mAb was able to increase BDNF transcription and pre-proBDNF synthesis (i.e., the precursor protein synthesis) as early as 6h after neuronal treatment (not shown).
Ab oligomers failed to induce the transactivating activity of CREB, and therefore, they did not affect BDNF transcription and release. The last two-ones were assessed at a time preceding frank neuronal loss, suggesting that a selective oAb-induced downregulation of BDNF levels was not among the earliest signs of neurodegeneration. Although oAb did not appear to modify CREB activity, it reduced the basal expression of pCREB. The discrepancy between CREB phosphorylation status and CREB transcriptional activity, observed in the case of Ab oligomers and not with Ab monomers, indicates that monomers and oligomers could not share the phosphorylating signaling required for CREB-regulated BDNF expression. Accordingly, following a brief exposure to oAb, we observed a decreased expression of pCREB, whereas basal expression levels of pAKT stayed unchanged. Hence, monomers and oligomers of Ab likely impinge on different CREB phosphorylation pathways. To this regard, Ab oligomers have been shown to bind the insulin receptor (Xie et al., 2002;Townsend et al., 2007) and to promote a state of insulin resistance through the ERK-dependent phosphorylation of insulin receptor substrate-1 (Zhang et al., 2015). Conversely, Ab monomers do not interact with the insulin receptor (Giuffrida et al., 2015) and have no effects on the ERK pathway (Giuffrida et al., 2009). It has been suggested that the impaired insulin signaling provoked by Ab oligomers enables the reentry of postmitotic neurons into the cell cycle (Norambuena et al., 2017), a seminal process in AD pathogenesis (Copani et al., 2006). Whether or not Ab monomers can oppose to this process, by sustaining IGF-1 signaling, remains to be determined. Present data suggest that, at least, monomers and oligomers exert distinct effects on CREB protein functions, and that a loss of functional monomers and a building-up of toxic oligomers could operate coordinately in determining the CREB dysregulation observed in AD (Pugazhenthi et al., 2011;Bartolotti et al., 2016a,b).

Concluding remarks
The results reported in this work strength the relevance of physiological Ab. Ab depletion is lethal to cultured neurons (Plant et al., 2003), and its blockade impairs learning and memory in adult mice (Morley et al., 2010;Puzzo et al., 2011Puzzo et al., , 2017. Low doses of synthetic Ab ensemble mimic the memory enhancement effects of endogenous Ab (Puzzo et al., 2008), likely via the regulation of transmitter release (Morley et al., 2010;Puzzo et al., 2011). Here we demonstrate that Ab monomers are specifically able to activate CREB, a converging point for mechanisms and pathways involved in memory formation (Teich  , 2015). By activating IGF-IRs and the ensuing PI3K pathway, Ab monomers regulate the expression levels of the CREB target gene, BDNF (Scheme 1), which is deeply involved in the regulation of cognitive functions (Budni et al., 2015). Our data suggest a new model whereby Ab monomers may preserve cognitive decline.

Experimental procedures Primary neuronal cultures
Animal care and experimentation were in accordance with institutional guidelines. Cultures of pure cortical neurons were obtained from rats at embryonic day 15 as previously described (Giuffrida et al., 2009). Cortical cells were dissected, mechanically dissociated, and seeded in Neural Q ™ Basal Medium supplemented with GS21 (Globalstem, Rockville, Maryland, Stati Uniti). Cortical cells were plated on 24-well plates precoated with 0.1 mg mL À1 poly-D-lysine and incubated at 37°C with 5% CO 2 in a humidified atmosphere. Cytosine arabinoside (1-b-D-arabinofuranosylcytosine, Ara-C) (3-10 lM) was added to the cultures 18 h after plating to avoid the proliferation of non-neuronal elements and was kept for 3 days before medium replacement.

Neuroglioma cell line, H4-APPswe
The neuroglioma H4-APPswe was cultured in Opti-MEM (Gibco, Thermofisher) supplemented with 10% fetal calf serum (Gibco, Thermofisher), penicillin and streptomycin (100 mg mL À1 each), and 2 mM L-glutamine at 37°C, 5% CO 2 . To obtain the conditioned medium from this cell line, cells were seeded on 24-well plates, and grown until confluence. Then, cells were washed twice with PBS buffer and medium was replaced with Neural Q basal Medium supplemented with penicillin and streptomycin (100 mg mL À1 each), 2 mM Lglutamine, and 2% GS21 (Globalstem). To allow medium enrichment of endogenous Ab, cells were maintained for 6 and 24 h, in the presence or in absence of b-secretase inhibitor (5 lM) or csecretase inhibitor (1 lM) (Calbiochem). Then, medium was collected and immediately assessed for its concentration of Ab. The remaining part was briefly stored at À80°C. The 24h conditioned medium was subsequently used to treat mature cortical neurons. When required, synthetic Ab monomers (100 nM) were added to rescue the lack of endogenous Ab (b-/c-secretase Inhibitor CM).

Ab peptides preparation
Ab 1-42 (HFIP-treated) was purchased from Bachem Distribution Services GmbH, Germany, dissolved at a 5 mM concentration in anhydrous dimethyl sulfoxide (DMSO) and stored at À20°C. At the time of its use,

Scheme 1 Schematic drawing of the interactions between Ab
(1-42) and IGF-I receptor signaling. Monomeric forms of Ab activate type-1 IGF receptors (IGF-IRs), resulting in CREB phosphorylation (via PI3K activation) and BDNF transcription. Dashed lines refer to a possible, yet not proven, PI3K-dependent pathway of CREB activation. a solution of 100 lM Ab in ice-cold DMEM F-12 was prepared and allowed to oligomerize overnight at 4°C according to the Lambert protocol (Lambert et al., 1998) and our previously described method (Giuffrida et al., 2009). Monomers and oligomers (>50 kD) were isolated from the peptide suspension by filtration through cutoff filters as previously described (Giuffrida et al., 2009).

MTT assay
Cortical cells were seeded in 24-well plates precoated with poly-d-lysine (0.1 mg mL À1 ) and incubated at 37°C with CO 2 (5%) under a humidified atmosphere; 18h after plating, cytosine b-d-arabinofuranoside ((3-10 lM)) was added to the cultures to avoid the proliferation of glial cells. Mature neuronal cultures between 6 and 8 days in vitro were treated with mAb 100 nM and two different concentrations of oAb (100 nM and 2 lM) for 24 and 48h.

Indirect pCREB immunofluorescence analysis
To confirm CREB phosphorylation and to prove its nuclear localization after Ab monomer treatment, SH-SY5Y were seeded in 35-mm glass bottom dishes and differentiated with RA as described in the cell culture section. To perform the experiment, cells were washed and stimulated for 30 min with monomeric Ab1-42 (100 nM), forskolin (30 lM), or oligomeric Ab1-42 (100 nM) in PBS buffer. Cells were then fixed in 2% formaldehyde and permeabilized using 0.1% Triton X-100. Unspecific binding was blocked by 30 min of incubation in 4% bovine serum albumin (BSA) in 0.1% Triton X-100-PBS. pCREB was detected by incubating overnight cells with rabbit anti-p (ser-133)CREB antibody (1:500, Cell signaling). After PBS washing, cells were exposed for 1 h at RT to the secondary antibody (anti-rabbit Texas Red conjugated). Hoechst (Molecular Probes, 1 lg lL À1 ) was used to stain nuclear DNA. Images were analyzed under a Leica DMI 6000B epifluorescence inverted microscope with Adaptive Focus Control. Sixty red (pCREB)/blue (Hoechst) ratio values were taken from twenty nuclear ROI (region of interest)/microscopic field, in 3 fields/dish.

Real-time PCR
Total RNA was extracted from primary cortical neurons using TRIzol reagent (Invitrogen, Thermofisher), following the manufacturer instructions. cDNA was synthesized from 1 lg of total RNA using High Capacity cDNA Reverse Transcription kit (Applied, Thermofisher) with the oligo (dT) primers. Quantitative real-time PCR for the expression of BDNF was performed using SYBR Green Master Mix with the following primers: BDNF-Forward 5 0 -TCAAGCTGG AAGCCTGAATGAA-3 0 , BDNF-reverse 5 0 -CCC AGT CAG GTA ACC ACT AAC AC-3 0 . GAPDH was used as endogenous controls for normalization (GAPDH-fw 5 0 -GAACA TCATCCCTGCATCCA-3 0 , GAPDH-rv 5 0 -CCAGTGAGCTTCCCGTTCA-3 0 ; Amplification reactions were performed on an ABI Prism 7500 (PE Applied Biosystems) according to the manufacturer's instructions.
Relative expression levels of the tested gene were determined using the 2 ÀDDCt .

CREB DNA binding assay
Primary neuronal cultures were treated with two different concentrations of mAb (10 and 100 nM) for 30 min in PBS. 100 nM Ab oligomers and 30 lM forskolin were used as negative and positive control, respectively. After treatments, neurons were harvested in RIPA lysis buffer (Alfa Aesar, GmbH Co KG) containing protease and phosphatase inhibitors cocktail mixes (both from Pierce-Thermofisher). Protein concentration was calculated using BCA protein assay kit (Pierce-Thermofisher), and an equal amount of proteins (15 lg) from each sample was tested. The DNA binding activity was measured by CREB(pSer133) Transcription factor Assay kit (Cayman CHEMICAL, Ann Arbor, MI, USA) in whole cell lysates, following the manufacturer's instruction.

Enzyme-linked immuno sorbent assay (ELISA)
Levels of Ab42 in conditioned medium of H4-APPswe were determined with the Human/Rat beta-Amyloid (42) ELISA Kit, High Sensitivity from Wako Chemicals USA, Inc. Medium-containing human recombinant Ab42, collected at 6 and 20 h incubation, was diluted with the provided sample buffer Amyloid-b monomers and CREB functions, S. Zimbone et al. immediately prior to the assay. Levels of BDNF in neuronal culture supernatants were measured with Rat BDNF ELISA kit Thermofisher.

Funding
This work has been supported by grant SIR RBS1148TJD from MIUR (Italian Ministry for Research and University) to Maria Laura Giuffrida. Stefania Zimbone has been supported by a SIR project postdoctoral contract.
Author contributions E.R., M.L.G., and A.C. were responsible for the design, acquisition, analysis, interpretation, preparation, and approval of manuscript draft. S.Z., M.L.G, I.M., F.G., and G.P. performed experiments, analyzed the data, made the figures, and approved the final version of the manuscript.