Memory-enhancing effects of GEBR-32a, a new PDE4D inhibitor holding promise for the treatment of Alzheimer’s disease

Memory loss characterizes several neurodegenerative disorders, including Alzheimer’s disease (AD). Inhibition of type 4 phosphodiesterase (PDE4) and elevation of cyclic adenosine monophosphate (cAMP) has emerged as a promising therapeutic approach to treat cognitive deficits. However, PDE4 exists in several isoforms and pan inhibitors cannot be used in humans due to severe emesis. Here, we present GEBR-32a, a new PDE4D full inhibitor that has been characterized both in vitro and in vivo using biochemical, electrophysiological and behavioural analyses. GEBR-32a efficiently enhances cAMP in neuronal cultures and hippocampal slices. In vivo pharmacokinetic analysis shows that GEBR-32a is rapidly distributed within the central nervous system with a very favourable brain/blood ratio. Specific behavioural tests (object location and Y-maze continuous alternation tasks) demonstrate that this PDE4D inhibitor is able to enhance memory in AD transgenic mice and concomitantly rescues their hippocampal long-term potentiation deficit. Of great relevance, our preliminary toxicological analysis indicates that GEBR-32a is not cytotoxic and genotoxic, and does not seem to possess emetic-like side effects. In conclusion, GEBR-32a could represent a very promising cognitive-enhancing drug with a great potential for the treatment of Alzheimer’s disease.

the phosphorylation of the chromatin-bound histone HA2.X (a marker of DNA damage) were analysed to assess cyto-and genotoxicity, respectively. As shown in Fig. 3a,b, 24 hours of exposure to a single high concentration (100 μ M) of GEBR-32a did not produce any significant effect in both assays.
The emetic-like effects of GEBR-32a were investigated using the xylazine/ketamine-induced anaesthesia test, which indicated that the administration of GEBR-32a (0.003-3 mg/kg) did not significantly influence the duration of anaesthesia in adult mice (Fig. 3c). Under the very same experimental conditions, in our previous study the pan PDE4 inhibitor rolipram significantly shortened anaesthesia time already at 0.03 and 0.3 mg/kg 26 (Fig. 3c).
As for the pharmacokinetic, GEBR-32a (10 mg/kg) was rapidly absorbed and distributed to the brain (T max = 20 min) and also rapidly eliminated with a half-life of approximately 1 h ( Fig. 3d and Table 3). The brain to plasma AUC 0-t ratio was 2.71, indicating a favourable brain penetration of our PDE4D inhibitor.
Effects of GEBR-32a in the object location task (OLT) in wild type and Tg2576 mice. Figure 4a reports the performance of adult WT mice in the OLT, following administration of vehicle or of GEBR-32a Scientific RepoRts | 7:46320 | DOI: 10.1038/srep46320 (0.0003-0.01 mg/kg) with an inter-trial interval of 24 hours. With this long-time interval, vehicle-treated mice did not remember the disposition of the objects they had visited in the learning trial T1 and, therefore, they spent almost the same time in exploring the two objects during the test trial T2. Thus, their d2 index was not different from zero. On the contrary, when mice were treated with GEBR-32a at the dose of 0.003 mg/kg, 3 hours after the learning trial T1, they did recognize that one object had been moved from its original position and, during the test trial T2, they explored it for much more time than the object that had not been moved. Therefore, their d2 index was different from zero. Figure 4b summarizes the results in the OLT, following administration of GEBR-32a (0.001-0.3 mg/kg), but with an inter-trial interval of 1 hour. With this short-time interval, control aged WT animals (vehicle) with intact memory were able to remember the position of the two objects and, therefore, during the test trial T2 they explored the moved object more than the unmoved one, thus showing a d2 index > 0. On the contrary, age-matched vehicle-treated Tg2576 mice did not. Administration of GEBR-32a to aged WT animals did not increase their normal discrimination capabilities, whereas Tg2576 mice showed a significant improvement in memory performance, the d2 index being higher than zero at the doses of 0.03 and 0.3 mg/kg.
Effects of GEBR-32a in the Y-maze continuous alternation task in wild type and Tg2576 mice. The effects of GEBR-32a administration (0.0003-0.3 mg/kg) on Y-maze alternation performance in aged WT and Tg2576 mice are summarized in Fig. 5a. When treated with vehicle, both aged WT and transgenic mice did not show a performance significantly higher than chance level (50% alternations), indicating that under these test conditions they were not able to correctly remember the arms already visited. GEBR-32a significantly ameliorated the performance in control animals at all the doses tested (alternations always higher than 50%). Under the same conditions, GEBR-32a was not able to induce any improvement in Tg2576 mice. However, when the PDE4D  inhibitor was chronically administered (0.03 mg/kg for up to 23 days), also the transgenic animals showed alternations significantly higher than 50% (Fig. 5b), indicating amelioration of memory function.
Effect of GEBR-32a on hippocampal long-term potentiation in wild type and Tg2576 mice. As shown in Fig. 6, vehicle-treated Tg2576 mice displayed a significant LTP impairment in comparison with age-matched WT controls. In the latter, chronic GEBR-32a administration (0.03 mg/kg for 23 days) caused a clear, though not significant, enhancement of LTP in comparison with respective controls. Of great relevance, GEBR-32a was able to rescue the LTP deficit of Tg2576 mice to the level of vehicle-treated WT animals.

Discussion
In this study, we present GEBR-32a, a novel PDE4D inhibitor deriving from the lead optimization process of compound 8a, which was recently synthesized and characterized in our laboratories 30 .
A key strategy in modern drug design is to optimize absorption, distribution, metabolism and excretion (ADME) before a drug enters the clinical phase. In particular, drugs designed for neurodegenerative disorders need to readily cross the blood-brain barrier (BBB) in order to reach their targets. In pharmaceutical chemistry, introduction of fluorine atoms represents a common strategy to improve drug potency and optimize ADME parameters [31][32][33][34] . Actually, the number of fluorinated drugs approved by FDA and EMA has greatly increased over the last decade 35 . Fluorine substitution influences different aspects of drug-target interaction, as well as of ADME, by modulating basic or acid molecule pKa, structure conformation, hydrophobic interactions, lipophilicity and interaction with metabolizing enzymes [31][32][33][34][35] .
Our selectivity analysis, carried out on a panel of 20 PDE isoforms and variants, has confirmed that GEBR-32a is very active on PDE4D isoforms and its potency has been indeed improved in comparison to its parent compound 8a. For example, GEBR-32a showed an IC 50 of 2.43 μ M toward PDE4D3, whereas that of compound 8a was 7.60 μ M 30 . Fluorine introduction had even more dramatic effects on BBB permeability, the brain/plasma ratio being extremely improved from 0.76 for compound 8a to 2.71 for GEBR-32a. For comparison, this ratio for the well-known PDE4 inhibitors rolipram and roflumilast is 2 and 1, respectively 36,37 .
In addition, the presence of two fluorine atoms did not alter the ability of GEBR-32a to functionally inhibit PDE4D enzymes, as demonstrated by the significant elevation of intracellular cAMP levels observed in vitro, in both cultured cells and hippocampal slices (Fig. 2). Of note, 100 μ M GEBR-32a was able to cause a 12-fold increase of the forskolin-induced elevation of cAMP in cultured cells, while the same concentration of its parent compound 8a evoked only a 4-fold increase under the very same experimental conditions 30 , indicating a better cell membrane diffusion of the fluorinated derivative.
Although some natural occurring fluorine-containing molecules are highly toxic (e.g. sodium monofluoroacetate), it is now well known that introduction of fluorine into drugs does not necessarily confer toxicity and, on the contrary, may reduce the toxicity of parent compounds by preventing their metabolic bioactivation 31,35 . As a matter of fact, our preliminary toxicological screening demonstrated that GEBR-32a does not have either  Scientific RepoRts | 7:46320 | DOI: 10.1038/srep46320 cytotoxic or genotoxic effects, even when cells were exposed to a high concentration of the drug (100 μ M) for 24 hours (Fig. 3a,b).
As inhibition of PDE4D activity has been consistently shown to ameliorate memory formation 26,28,30,38,39 , GEBR-32a was trialled in two different behavioural tests, the object location task (OLT) and the Y-maze alternation task. To verify that fluorine introduction had not influenced the pro-cognitive properties of the parent compound 8a in normal adult mice 30 , we tested GEBR-32a under the same experimental conditions. The OLT is used to assess episodic-like spatial memory by evaluating if mice recognize/remember that one of two objects, explored in the learning trial, has been moved to a different position in the test trial. In this case, vehicle-treated adult mice did not remember the spatial arrangement of the objects when tested in the OLT with an inter-trial delay of 24 hours, due to natural forgetting (Fig. 4a). On the contrary, when mice were treated with GEBR-32a, at a very low dose (0.003 mg/kg) three hours after the learning trial, a time point at which PDE4 inhibitors specifically improve memory consolidation 40 , they were able to remember that one object was not in the same position as in the learning trial and, consequently, spent much more time in exploring it in the test trial (d2 index > 0). Therefore, GEBR-32a is indeed able to improve consolidation of spatial long-term memory 3 h after learning.
However, the observation that a drug improves cognition in normal animals, although of interest for drug development, has limited relevance from a translational point of view. For this reason, we further analysed the memory-enhancing properties of GEBR-32a on aged Tg2576 mice that are widely used to model Alzheimer's dementia. When these mice are 9-12 months old, they show ß-amyloid plaque accumulation in brain tissue and synaptic and memory deficits 41 .
In these experiments, we evaluated the effects of GEBR-32a in the OLT with a 1-hour inter-trial interval, which allows normal animals to retain the information acquired during the learning trial. In this paradigm, in fact, vehicle-treated aged WT mice showed well-functioning short-term spatial memory (Fig. 4b), as they identified that one object had been moved from its original position and, therefore, they explored it for more time (d2 index > 0). For this reason, administration of GEBR-32a had no effects on their performance, as short-term  Table 3. Main pharmacokinetic parameters of GEBR-32a. GEBR-32a was administered subcutaneously to mice at the dose of 10 mg/kg. Blood samples and brain tissues were collected at 7 different time points and analysed for GEBR-32a content. The main pharmacokinetic parameters were calculated by non-compartmental analysis. C max : the maximum observed plasma/brain concentration; T max : the time corresponding to C max ; AUC 0-t : the area under the plasma/brain concentration versus time curve from time 0 to the last time point; t 1/2 : elimination half-life.
Scientific RepoRts | 7:46320 | DOI: 10.1038/srep46320 memory cannot be further improved probably due to ceiling effect. On the contrary, a memory impairment was evident in vehicle-treated Tg2576 mice, which were unable to distinguish the moved object and their d2 index was, in fact, not different from zero. Of great relevance, when these transgenic mice were acutely treated with GEBR-32a, they showed a relevant improvement of their short-term memory performance, exploring the moved object for more time during the test trial. The Y-maze continuous alternation task analyses spatial working memory by measuring the number and the order of entries of mice into the three different arms of the maze. In this case, a 50% alternation is considered chance level and reflects the absence of working memory performance. Indeed, adult WT mice had a performance compatible with intact working memory (alternations higher than 50%), which was not further improved by GEBR-32a administered 30 min before the trial (ceiling effect; data not shown). On the other hand, both aged WT and Tg2576 mice manifested memory impairment, since they scored alternations not different from chance level when treated with vehicle (Fig. 5a). When GEBR-32a was acutely administered 30 min before the test, aged WT mice had a significant memory improvement, whereas aged Tg2576 mice had not, suggesting that an acute increase of cAMP is not sufficient to ameliorate this type of cognitive deficit in this mouse model of Alzheimer's disease. However, if chronically treated with GEBR-32a, also Tg2576 mice performed significantly better than chance levels. These results demonstrate that the impaired working memory can be recovered by a chronic treatment with the selective PDE4D inhibitor.
In the attempt to correlate GEBR-32a memory-enhancing properties with cellular/molecular mechanisms, we investigated the effects of GEBR-32a on hippocampal long-term potentiation (LTP). Hippocampal LTP is a form of synaptic plasticity that is generally accepted to represent the molecular substrate of memory formation and is known to require a fine tuning of the cAMP/PKA/CREB pathway to function 24 . Moreover, a large body of evidence has shown that PDE4 inhibitors are able to rescue compromised LTP in different models of pathological conditions, including AD 42,43 .
Scientific RepoRts | 7:46320 | DOI: 10.1038/srep46320 To date, the serious emesis associated with non isoform-selective PDE4 inhibitors has impeded their clinical exploitation as promnesic drugs in those pathologies, such as AD, characterized by memory loss. Although rodents cannot vomit, it has been shown that preliminary screening for emetic-like potential of PDE4 inhibitors can be reliably carried out in these animals using the xylazine/ketamine-induced anaesthesia test 45 . It has been proposed that PDE4 inhibitors cause emesis by activating a sympathetic pathway through antagonism of presynaptic α 2 -adrenergic receptors, a hypothesis supported by the evidence that α 2 -adrenergic antagonists stimulate vomiting, while clonidine, a selective α 2 -receptor agonist, is able to prevent PDE4 inhibitor-induced emesis. In addition, α 2 -adrenergic antagonists, as well as PDE4 inhibitors, reduce the time of anaesthesia induced by a xylazine/ketamine mixture 45,46 . Using this test, we have recently confirmed that rolipram is able to reduce the anaesthesia time at the same dose improving memory in mice 26 . On the contrary, our initial hits 26,30 and our optimized lead GEBR-32a (Fig. 3c) are ineffective on anaesthesia time at doses 100-1000 times higher than the pro-cognitive ones, thus indicating that they are very likely devoid of emetic side-effects.
Therefore, the memory-improving effects of GEBR-32a are further dissociated from possible emetic-like effects, as compared to its parent compound 8a. A possible explanation could be that GEBR-32a (as 8a) inhibits PDE4D isoforms involved in memory processes per se. As for inhibition mechanisms, our relatively high IC 50 values reflect binding of GEBR-32a to the catalytic domain, yet interactions outside the catalytic domain cannot be ruled out, resulting in more or less selectivity for specific PDE4D isoforms. Clearly, more research is needed into this aspect, including co-crystallization of GEBR-32a and PDE4D isoforms to elucidate their interactions. Moreover, we need to know which specific PDE4D isoform to target for the treatment of memory decline in Alzheimer's disease 47 .
As for the molecular mechanisms underlying the pro-cognitive effects of GEBR-32a in AD mice, it can be hypothesised that our PDE4D inhibitor reverses the Aβ -mediated inhibition of the cAMP/PKA/CREB pathway, as it happens with rolipram 42,48 , thus leading to the rescue of hippocampal LTP and memory deficits. Moreover, it has been reported that PDE4 inhibition can also reduce neuroinflammation by dampening the production of the proinflammatory cytokines IL-1b and TNFα and by increasing the levels of TGFβ -1 and BDNF, which are known to play a key role in hippocampal synaptic plasticity and memory [49][50][51] .
In conclusion, GEBR-32a is a selective PDE4D inhibitor endowed with a very favourable toxicological and pharmacokinetic profile, and is able to improve spatial memory processes without undesired emetic-like side effects. Although further safety studies are needed, GEBR-32a is a promising therapeutic agent for the treatment of cognitive decline in AD and related dementias.
Yield: 56%. 1   All chemicals were purchased from Chiminord and Aldrich Chemical (Milan, Italy). Solvents were reagent grade. Unless otherwise reported, all commercial reagents were used without any further purification. Aluminium backed silica gel plates (Merck DC-Alufolien Kieselgel 60 F254, Darmstad, Germany), were used in thin-layer chromatography (TLC) for routine monitoring the course of reactions. Detection of spots was made by UV light. Merck silica gel, 230-400 mesh, was used for chromatography.
Melting points are not "corrected" and were measured with a Buchi M-560 instrument. IR spectra were recorded with a Perkin-Elmer 398 spectrophotometer. 1 H NMR spectra were recorded on a Varian Gemini 200 (200 MHz) instrument, chemical shifts are reported as δ (ppm) relative to tetramethylsilane (TMS) as internal standard; signals were characterized as s (singlet), d (doublet), t (triplet), q (quartet), m (multiplet), br s (broad signal); J in Hz. Elemental analyses were determined with an elemental analyzer EA 1110 (Fison-Instruments, Milan, Italy) and the purity of all synthesized compounds was > 95%.
The enzymatic profile of GEBR-32a has been carried out by Scottish Biochem (Glasgow, Scotland, UK), using recombinant human PDE4 enzymes expressed in baculovirus, as previously reported 25 . In a first assay, GEBR-32a was preliminary tested at a single concentration (10 μ M) in duplicate on a panel of 20 PDE isoforms and variants; then, IC 50 values were determined only for those PDEs whose inhibition by 10 μ M GEBR-32a was more than 50%. The compound was tested at five different concentrations (1 nM to 100 μ M) and the IC 50s were obtained by non-linear regression analysis of the concentration-inhibition curves (GraphPad Prism software).
Tg2576 and wild type (WT) mouse colonies were maintained in the animal facilities of the Columbia University. The correct genotype was identified by PCR on tail samples. Adult Sprague-Dawley rat colony was maintained in the animal facilities of the Department of Pharmacy, University of Genoa. Only male animals were used in the experiments.
For behaviour experiments, adult (3-4 months of age) WT, aged WT and Tg2576 mice (12-21 months of age) were housed individually on a reverse light/dark cycle (light from 19:00 to 07:00). Rats were housed on a regular light/dark cycle (light from 07:00 to 19:00). Animals were kept at constant temperature (22 ± 1 °C) and relative humidity (50%) with free access to food and water.
All Hippocampal slices (400 μ m thick) were obtained from adult Sprague Dawley rats using a McIlwain tissue chopper and pre-incubated in physiological medium 52 (7 slices/condition) at 37 °C for 15 min, first with different concentrations of GEBR-32a and, then, with 0.1 μ M forskolin for further 15 min. Finally, slices were left in hot medium (90 °C) for 15 min, sonicated and centrifuged at 10000 g for 15 min. The supernatant was used for cAMP determination, using the same kit as above, and the pellet for protein quantification.
Results from cells and slices are presented as fold increase and represent mean ± S.E.M. of the number of experiments reported in the figure legend. Normal distribution was evaluated by Kolmogorov-Smirnov test and then data have been analysed by one-way ANOVA followed by Newman-Keuls or Dunnett's tests, as appropriate.

Cytotoxicity and genotoxicity in cultured cells.
For the cytotoxicity and genotoxicity assays, HTLA cells were treated for 24 hours with GEBR-32a (100 μ M) at 37 °C. At the end of the incubation period, lactate-dehydrogenase release was measured in conditioned media using the Cytoxicity Detection Kit PLUS (Roche, Germany) according to manufacturer protocols.
To evaluate genotoxicity, cells were processed for total protein extraction as described previously 53 . Immunoblots were done according to standard methods, using the following antibodies: mouse monoclonal to gamma H2A.X (2F3, phospho S139) and rabbit polyclonal to Histone H2A.X (Abcam, UK); anti-mouse and anti-rabbit secondary antibodies coupled to horseradish peroxidase (GE Healthcare, UK). Proteins were visualized with an enzyme-linked chemiluminescence detection kit according to the manufacturer's instructions (GE Healthcare). Chemiluminescence was monitored by exposure to films and signals were analyzed under non-saturating condition with an image densitometer (Bio-Rad, CA, USA).

Pharmacokinetic analysis.
A total of 21 male BALB/c mice were used for each drug and three mice were used for each time point 30 . GEBR-32a was dissolved in DMSO, diluted in 0.5% methylcellulose to yield a final concentration of 1 mg/mL and administered subcutaneously at the dose of 10 mg/kg. Blood samples (approximately 250 μL) were collected via retro-orbital puncture into tubes containing sodium heparin at 10 min, 20 min, 40 min, 1 h, 2 h, 3 h, and 5 h post-injection. Plasma was separated by centrifugation (11000 rpm, 5 min, 4 °C) and stored at −70 °C before analysis. After blood harvest, mice were sacrificed by cervical dislocation and brains were excised, weighed, rinsed by cold saline and then frozen at −70 °C until analysis.
For blood analysis, 25 μL of plasma were transferred into a 1.5 mL Eppendorf tube and were added with 25 μL of methanol and 25 μL of internal standard (500 ng/mL voriconazole), followed by the addition of 100 μL methanol.
As for brains, 100 mg of tissue were placed into a plastic tube and added with 500 μL of methanol to facilitate homogenization, which was carried out using a Fluko F6/10 superfine homogenizer for approximately 1 min. Homogenized samples were then treated by ultrasound for 10 min and centrifuged at 11000 rpm for 5 min. A 25 μL aliquot of the homogenized samples was transferred into an Eppendorf tube and added with 20 μL of internal standard (500 ng/mL voriconazole) followed by the addition of 100 μL of methanol. After centrifugation at 11000 rpm for 5 min, a 5 μL aliquot (plasma or brain) was injected in a LC/MS/MS system (Shimadzu LC-20A HPLC system; AnaShiseido, Tokyo, Japan) coupled with a TSQ Quantum Vantage triple quadrupole mass spectrometer equipped with a HESI source (ThermoFisher, San Jose, CA, USA). Chromatographic conditions were the following: guard column SecurityGuard C 18 column (4 mm × 3.0 mm I.D., 5 μ m, Phenomenex, Torrance, CA, USA), analytical column SB C 18 (150 mm × 4.6 mm I.D., 5 μ m, Agilent, USA), buffers 0.1% formic acid in 10 mM ammonium acetate/0.1% formic acid in methanol (10:90), flow rate 0.6 mL/min. Mass spectrometric conditions were the following: source HESI, scan mode SRM, polarity positive, vaporizer temperature 420 °C, ion sweep gas pressure 1 bar, auxiliary gas pressure 5 bar, capillary temperature 320 °C. Calibration curves were prepared in heparinized blank mice plasma (3-3000 ng/mL) or blank brain homogenate (1-10000 ng/g). Concentrations were calculated using a weighted least-squares linear regression (W = 1/x 2 ). The linear regression equations of the calibration curves are presented in the Table 4. The main pharmacokinetic parameters were calculated by non-compartmental analysis using Phoenix 1.3 (Pharsight USA). AUC 0-t : the area under the plasma/brain concentration versus time curve from time 0 to the last measurable concentration was calculated by using the linear trapezoidal rule. C max : the maximum observed plasma concentration. T max : the time corresponding to C max . t 1/2 : the elimination half-life was calculated as 0.693/λ z . λ z was obtained by log-linear regression using the terminal points of the plasma concentration-time curve.
Behavioural analysis. Drug preparation for behavioural testing. For all behavioural analysis, GEBR-32a was dissolved in dimethylsulfoxide (DMSO) and stored at 4 °C, this stock solution was used for further dilutions in 0.5% methylcellulose. The vehicle condition received 0.5% methylcellulose with 0.005% DMSO, and all drug solutions had a fixed percentage of 0.5% methylcellulose and 0.005% DMSO.
Object Location Task (OLT). The OLT has been derived from the Object Recognition Task (ORT) 54 . The apparatus consisted of a circular arena, 40 cm in diameter. The back half of the 40-cm high wall was made of grey polyvinyl chloride (PVC) and the front was made of transparent PVC. Fluorescent red tubes and a light bulb provided a constant illumination of about 20 lux on the floor of the apparatus. Two objects were placed in symmetrical positions at the mid-line between the black and transparent halves of the arena, about 5 cm away from wall. Four sets of 2 identical objects were used and these objects were presented to the animals in a balanced manner to avoid learning or place biases. The rodents were unable to displace the objects. Prior to the trials, mice were put in an empty cage for 4 minutes to increase arousal during testing. A test session comprised two 4-min trials. During the learning trial T1, the apparatus contained two identical objects placed on a horizontal line in the middle of the arena (object a1 and a2). The animals were always introduced into the apparatus with their nose towards the transparent wall segment (i.e. facing outwards to the front of the arena). Subsequently, the rodents were put back in their home cage for the inter-trial interval. After the retention interval, the animals were put back into the arena for the test trial T2, in which one object (a3) was in the same position as during T1 and the other (b) was moved to a different position to the front or back of the arena. The time spent exploring each object during T1 and T2 was recorded manually on a personal computer. Exploration was defined in the following manner: directing the nose to the object at a distance of no more than 2 cm and/or touching the object with the nose. Sitting on the object was not considered as exploratory behaviour. In order to avoid the presence of olfactory cues, the objects were thoroughly cleaned with a 70% ethanol solution before each trial. Prior to compound testing, animals were handled for 5 minutes on 3 consecutive days and allowed to explore the arena for 5 min. Subsequently, animals were accustomed to the complete OLT testing procedure including injections prior or after testing. In a first set of experiments, GEBR-32a (0.0003-0.01 mg/kg s.c.) was evaluated in adult WT mice with a 24 h inter-trial interval and was administered 3 h after the first trial (T1) of the OLT 40,55 . Afterwards, it was tested (0.001-0.3 mg/kg s.c.) in memory deficit models (aged WT and Tg2576 mice) with a 1 hour inter-trial interval and was administered 30 min before the first trial (T1) of the OLT. Of note, drug conditions were tested in semi-random order, so within one testing session, multiple treatment conditions were tested. The experimenter was blinded to the conditions that were being tested. Similar to the ORT, the readout parameters of the OLT are the time that rodents spent on exploring each object during T1 and T2. The exploration time (in seconds) of each object during T1 are presented  as 'a1' and 'a2' . The time spent in exploring the familiar and the moved objects in T2 are represented as 'a3' and 'b' , respectively. Using these data, the following parameters were calculated: e1 = a1 + a2; e2 = a3 + b; d2 = (b-a3)/ e2. The d2 index is a relative measure of discrimination that corrects for exploratory activity 56 . The d2 index can range from -1 to 1. A significant positive difference from zero indicates that the mice remembered the object locations from T1. Of note, mice require a minimum amount of exploration in order to show reliable memory performance 57 . Therefore, animals were removed from the analysis if they spent less than 9 s exploring the objects during T1 or T2. Normal distribution of data was evaluated by Kolmogorov-Smirnov and one-way ANOVA was performed to evaluate differences between the conditions. A one sample t-test was used to compare the d2 index of the conditions to zero (i.e. chance level).
Y-maze continuous alternation task. The apparatus was made of grey Plexiglas with three arms symmetrically placed together at a 120° angle. At the beginning of the trial, each mouse was placed in one of three arms (randomly divided and balanced) and was then allowed to freely explore the apparatus for 6 minutes. The number of entries into a different arm and the order was measured. An entry was only counted if all four paws of the animal were placed completely inside the arm. When a mouse visited all 3 arms consecutively, it made a triad. In between trials, the apparatus was cleaned with a 70% ethanol solution to avoid olfactory cues. In a first set of experiments, GEBR-32a (0.0003-0.3 mg/kg s.c.) was administered acutely to aged WT and Tg2576 mice 30 min before the trial. In a second set of experiments, GEBR-32a was administered chronically (23 days) at the dose of 0.003 mg/kg s.c.
In order to reduce the number of animals necessary for experimentation and to increase statistical power, two testing sessions were performed (at day 22 and day 23) and combined. Of note, there was no statistical difference between testing sessions (day effect; data not shown) and, therefore, both test sessions were pooled. To measure spatial working memory, the percentage of alternations was calculated by taking the number of triads and dividing it through the maximum possible alternations (total entries minus 2) and multiply this by 100. A score of 50% alternations is considered chance level and, therefore, a significant difference from 50% is indicative of functional working memory. Normal distribution of data was evaluated by Kolmogorov-Smirnov and one-way ANOVA was performed to evaluate differences between the conditions. For each condition, one-sample t-test was used for comparison versus the score of 50%.
Ketamine/xylazine induced α 2 -adrenoceptor-mediated anaesthesia test. To investigate possible emetic-like effect of GEBR-32a, mice were anesthetized with of 60 mg/kg ketamine and 10 mg/kg xylazine (i.p. injection, 1 ml/ kg). GEBR-32a (0.003-3 mg/kg) or vehicle were administered 15 minutes after the induction of anaesthesia and mice were put in dorsal position 26 . The outcome measurement was the total time the animals were anesthetized, measured from anaesthesia induction to the righting reflex (i.e. when mice were back on all four paws). In order to reduce the number of animals for experimentation and to get sufficient group sizes, the experiments were performed three times with 2 recovery days in between. Each animal did not receive the same condition more than once, i.e per test day the drug treatments were balanced over the animals so each condition was tested in every animal eventually. Normal distribution of data was evaluated by Kolmogorov-Smirnov and one-way ANOVA was used to compare all conditions. Electrophysiology recordings. Electrophysiological recordings have been carried out as previously described 48 . Briefly, hippocampal slices (400 μ M thick) were obtained from aged WT and Tg2576 mice, chronically treated either with vehicle or with GEBR-32a (0.03 mg/kg s.c.), by means of a tissue chopper and subsequently were maintained under perfusion (1-3 ml/min) with saline solution (in mM: 124 NaCl, 4.4 KCl, 1.0 Na 2 HPO 4 , 25.0 NaHCO 3 , 2.0 CaCl 2 , 2.0 MgSO 4 ,10 glucose; continuously aerated with 95% O2 and 5% CO2) in an interface chamber at 29 °C for 90 min prior to recording. A concentric bipolar platinum-iridium stimulation electrode and a low-resistance glass recording microelectrode filled with saline solution (5 mΩ resistance) were placed in CA1 stratum radiatum to record the extracellular field excitatory postsynaptic potential (fEPSP). Basal synaptic transmission was assayed by plotting the stimulus voltages against slopes of fEPSP. For LTP experiments, a 15-min baseline was recorded every min at an intensity that evoked a response ~35% of the maximum response. LTP was induced using a θ -burst stimulation, consisting of 4 pulses at 100 Hz, with the bursts repeated at 5 Hz and each tetanus including three 10-burst trains separated by 15 s [33]. Normal distribution of data was evaluated by Kolmogorov-Smirnov test and two-way ANOVA was used to determine the rescuing effect of GEBR-32a on synaptic dysfunction.