All experiments were conducted in strict conformity with the Policies of the French Ethics Committee. Animal surgery and experimentations conducted in this study were authorized by the French Direction of Veterinary Services (APAFIS#12311-2017071816217821 v3). Swiss and C57BL/6 male mice (age 5 weeks; 30–35 g) were purchased from Janvier Labs (Le Genest-Saint-Isle, France) and Envigo (Gannat, France), respectively. Sprague–Dawley (8-week-old) male rats were purchased from Janvier Labs. During all of the experimental period, rodents were group-housed (six per cage) and maintained under controlled environmental conditions (12-h light/12-h dark cycle; temperature of 23°C ± 2°C; humidity of 50 ± 10%) with food and water ad libitum.

Mice were anesthetized with isoflurane, placed in a stereotaxic frame and bilaterally implanted with probes in both left and right hippocampi. The stereotaxic coordinates for the bilateral injections into the hippocampus were anteroposterior: −2.4, lateral: ±2.7, and ventral: −3 (in mm from bregma, according to Franklin and Paxinos (2007). Microdialysis experiments were conducted with artificial cerebrospinal fluid (aCSF) in freely moving mice 24 h after surgery as previously described (Guiard et al., 2005). Dialysate samples were collected each 20 min and analyzed for ACh contents using an ultrahigh-performance liquid chromatography (UHPLC) method (see below). Basal values were determined for each mouse, and changes in ACh outflow induced by acute administration of vehicle, DPZ alone, or in combination with MEF were expressed as percentage of these basal values.

Concentrations of ACh in the microdialysis samples were analyzed by the Pronexus Analytical AB company (Stockholm, Sweden) using UHPLC tandem mass spectrometry (MS/MS). Briefly, the UHPLC-MS/MS system included a PAL autosampler (CTC Analytics, Zwingen, Switzerland), an Advance UHPLC pump, and an EVOQ Elite triple quadrupole mass spectrometer (Bruker Daltonics, Fremont, CA, United States) equipped with electrospray ionization source operating in a positive mode at +4,500 V. The source parameters were as follows: probe gas flow: 50, nebulizer gas flow: 60, probe temperature: +300°C, cone gas flow: 30, cone temperature: +200°C, CID gas: Ar 1.5 mTorr. A Titan PFP column (100 × 2.1 mm, 1.9 μm, 120 Å pore size) purchased from Sigma-Aldrich (Sweden) was used for UHPLC. Deuterated ACh (acetylcholine-1,1,2,2-d4 bromide; ACh-d4), used as internal standard, was purchased from C/D/N Isotopes (QMX Laboratories, Dunmow CM6 2PY, United Kingdom). Typically, 10 μL of aCSF or hippocampal microdialysate was mixed with 20 μL of ACh-d4 internal standard (50 nM, in aCSF), and 5 μL of the mixture was injected into the column. The mobile phase A was 0.1% (vol/vol) formic acid in water; the mobile phase B was acetonitrile with 0.1% formic acid. The linear elution gradient was as follows: 0–0.5 min: 5% B; 3.5 min: 55% B; 3.6 to 4.0 min: 70% B; 4.1 min, 5% B. The flow rate was 400 μL/min, and the total run-to-run time was 4.8 min. The calibration curve was linear in the range of 0.1 to 409.6 nM ACh (R² = 0.999); the estimated limit of detection was 0.05 nM, and the lower limit of quantification was 0.15 nM.

Acetylcholinesterase activity was quantified using a cholinesterase activity assay kit according to manufacturer’s recommendations (MAK119; Sigma, St. Louis, MO, United States). Briefly, mouse hippocampus was homogenized (10 μL/mg) in 0.1 M Na/K phosphate buffer (PB) (pH 7.5) at 4°C followed by centrifugation at 30,000 × g for 5 min. Aliquots of the cleared supernatants were added to the kit reaction mixture containing various concentrations of DPZ (1–50 nM) and/or MEF (0.5–4 μM), and the reaction proceeded at room temperature. Absorbance at 415 nm was quantified using VICTOR plate reader (Perkin–Elmer Inc., Waltham, Massachusetts, United States) at 2-min intervals from time 0 (tissue sample addition) up to 10 min. Acetylcholinesterase activity was expressed as units/L. One unit of AChE is the amount of enzyme that catalyzes the production of 1.0 μmol of thiocholine per minute at room temperature at pH 7.5.

These experiments had to be performed in rats because blood volume in mice was too low to allow collection of serial serum samples for studies over 24 h after drug administration.

Mice were deeply anesthetized and transcardially perfused with a 4% paraformaldehyde solution in Na/K PB (0.4 M, pH 7.4). Brains were removed, and series of one in twelve 30-μm-thick coronal sections were cut at hippocampal level, rinsed in PBS, and then incubated 1 h in a blocking solution containing 10% of normal donkey serum (Bio-Rad, Marnes-la-Coquette, France) and 0.1% BSA (Euromedex, Souffelweyersheim, France) in PBS supplemented with 0.25% Triton-X100 (PBST). Primary antibodies were goat anti-GFAP (1:500, Abcam, Paris, France), mouse anti-Cx43 (1:250, BD transduction, Le Pont de Claix, France), and rabbit anti-Cx30 (1:250, ThermoScientific, Les Ulis, France) and diluted in blocking solution and applied overnight at 4°C. Sections at the very level of injection sites and displaying signs of lesions were discarded. After several rinses in PBST, sections were incubated for 120 min at room temperature in a mixture of Alexa Fluor^® 488–conjugated highly cross-adsorbed donkey anti-rabbit, Alexa Fluor^® 555–conjugated donkey anti-mouse, and Alexa Fluor^® 647–conjugated donkey anti-goat antibodies (all at 1:500; Life Technologies, ThermoScientific) in blocking solution. Sections were then rinsed extensively and finally mounted onto glass slides, cover slipped using Mowiol^® + Hoechst (1:10,000), and stored at 4°C.

Immunolabeled hippocampi from LV-injected mice were pictured using an Olympus BX-51 microscope equipped with Mercator Software (Explora Nova, La Rochelle, France). The area of interest was framed at ×20 magnification lens for each injection site and visualized on DAPI-labeled sections. Quantifications of DAPI, GFAP, Cx30, and Cx43 were conducted with the FIJI software (ImageJ 2; National Institutes of Health, Bethesda, MD, United States). Gray values for the GFAP, Cx30, and Cx43 immunolabeling in the area of interest were determined in each series of hippocampus sections.

All data are expressed as mean ± SEM, and normality of data distribution was evaluated using the Shapiro–Wilk normality test. Alternation performance and object exploration time were converted to percentage of alternation or preference index, respectively. Spontaneous alternation results were analyzed by two-way ANOVA followed by Tukey multiple-comparisons test. Novel object recognition or location results were analyzed using one-sample t-test vs chance level. Morris water maze (MWM) task was analyzed using two-way repeated-measures ANOVA followed by Tukey multiple-comparisons test (place-learning) and one-sample t-test vs chance level (probe test). Hippocampal ACh outflow was analyzed by two-way ANOVA followed by Fisher least significant difference post hoc test. Serum and brain levels of DPZ and MEF were analyzed by two-way ANOVA followed by Bonferroni post hoc test and Kruskal–Wallis test followed by Dunn post hoc test, respectively. Connexin 30 and Cx43 expressions were analyzed by two-way ANOVA followed by Tukey multiple-comparisons test. Significance level was set at 0.05.

Different cohorts of mice were tested using the Y-maze alternation task after i.p., or p.o., administration of DPZ either alone or in combination with MEF (Figure 1A, left panel). Regardless of the route of administration, scopolamine significantly impaired spontaneous alternation compared to vehicle-treated mice (P < 0.0001; Figures 1B,C). Whereas DPZ at 0.1 mg/kg (DPZ0.1) i.p., exerted no effect (P = 0.69; Figure 1B), DPZ 0.5 mg/kg (DPZ0.5) significantly reversed the scopolamine-induced deficit (P < 0.0001; Figure 1B). DPZ0.1, when combined with MEF at 1 or 3 mg/kg (DPZ0.1 + MEF1 and DPZ0.1 + MEF3), restored spontaneous alternation up to the control group (Figure 1B). Similar observations were made in mice that received treatment of DPZ0.25 p.o., which did not reverse scopolamine-induced deficit (P = 0.69; Figure 1C), but the same low dose of DPZ combined with MEF at 3 or 10 mg/kg p.o., significantly reversed the alternation deficit (both at P < 0.0001; Figure 1C), up to that noted in vehicle-treated controls, suggesting a potential synergistic effect of MEF and DPZ on working memory after a scopolamine challenge.

When subjected to the NOR task (Figure 1A, right panel), all mice spent the same amount of time exploring both identical objects during the familiarization session (all, P > 0.05; Figure 1D). In contrast, during the NOR test session, scopolamine administration abolished the increase in exploration time of the novel object observed in control group. Neither MEF1 i.p., nor DPZ0.25 i.p., prevented the scopolamine effect (P = 0.57 and P = 0.52, Figure 1E). However, the combination of drugs (DPZ0.25 + MEF1) was as efficient as DPZ at 1 mg/kg to restore exploration preference of the novel object in scopolamine-treated mice (Figure 1E).

Mice were i.c.v., injected with Aβ_25–35 peptide or Sc.Aβ on day 0 and chronically treated with DPZ and/or MEF from days 1 to 16 (Figure 2A). Familiarization phase demonstrated that these treatments had no effect on the time spent in exploring identical objects (all, P > 0.05; Figure 2B). In contrast during the NOR test phase, Aβ administration abolished the increase in exploration time of the novel object (P = 0.25; Figure 2C). While MEF1, DPZ0.25, or the combination DPZ0.25 + MEF0.1 did not reverse the Aβ-induced impairment, mice treated with DPZ0.25 + MEF0.3 or DPZ0.25 + MEF1 demonstrated a marked improvement in preference index (P = 0.021 and P = 0.0017, respectively) up to that obtained with DPZ at 1 mg/kg.

From days 11 to 16, mice were challenged with the MWM test. Amyloid β i.c.v., injection induced a significant memory deficit (from learning days 3–5) in comparison with Sc.Aβ injection (for all days, P < 0.0001; Figure 2D). DPZ0.25 or MEF1 showed no effect, whereas DPZ1 significantly reversed the Aβ-induced memory deficits from learning days 2 to 5 (P < 0.001 and P < 0.0001 for day 2 and days 3–5, respectively; vs Aβ/VEH; Figure 2D). In addition, DPZ0.25 combined with MEF0.3 or MEF1 showed a significantly superior learning memory activity (from learning days 2–5) when compared to DPZ0.25 alone.

On experimental day 16 (probe phase), the time spent in each of the four quadrants of the circular pool was measured during a 60-s session (Figure 2F). Interestingly, data highlighted the significantly superior cognitive activity of DPZ1 and DPZ0.25 + MEF1 groups (P = 0.013 and P = 0.0063. respectively; Figure 2F) as those treatments restored performance up to the level noted in Sc.Aβ-injected control mice (P = 0.011).

The effects of DPZ alone or combined with MEF on ACh outflow in the hippocampus were assessed using intracerebral microdialysis in awake freely moving mice. Intraperitoneal administration of DPZ1 alone or DPZ0.25 in combination with MEF1 caused a transient ACh overflow (P < 0.01, Figure 3A). Twenty minutes after injection, hippocampal ACh levels were increased by up to 2.5-fold in DPZ1 and DPZ0.25 + MEF1 groups and were significantly higher than in mice treated with the vehicle (both P < 0.01, Figure 3A) or DPZ0.25 alone (both, P < 0.05, Figure 3A). Administration of DPZ0.25 or MEF1 alone did not significantly modify ACh outflow.

In vitro assays were used to quantify the inhibitory effect of DPZ alone or combined with MEF on AChE activity in brain extracts. The addition of MEF (0, 0.5, 2, and 4 μM) did not modify the potency of DPZ to inhibit AChE activity (Figure 3B).

Donepezil (0.25 or 1 mg/kg) alone or in combination with MEF (1 mg/kg) was administered p.o., daily for 14 days, and the drugs were quantified in serum using an LC-MS method. The daily coadministration of MEF1 did not change the serum concentration of DPZ determined after 14-day treatment with 0.25 mg/kg p.o., daily of the latter drug (P > 0.05; Figure 4A). Brain concentration of DPZ was dose-dependent (DPZ0.25: 13.00 ± 2.17 ng/g, DPZ1: 78.1 ± 11.5 ng/g, P = 0.0036; Figure 4B). No significant difference was found between the brain concentrations of DPZ comparing DPZ0.25 + MEF1 (26.20 ± 5.26 ng/g) and DPZ0.25 groups (P = 0.33; Figure 4B). Mefloquine concentration in serum ranged between 78 and 113 ng/mL all along the 24 h after the last drug administration on the 14th treatment day (Figure 4C). In brain, MEF concentration reached 229.0 ± 19.5 ng/g 24 h after the last drug administration (Figure 4D).

Lentiviral vectors expressing shRNA against Cx30 or Cx43 (or control) were bilaterally injected into the hippocampus. Mice were tested in the NOL task after a 14-day recovery period (Figure 5A).

DAPI labeling and GFAP immunolabeling showed that the hippocampal density of all cell types, and especially astrocytes, did not differ between the groups of LV-injected mice (Figures 5B,C). Quantification of brain sections confirmed that Cx30 and Cx43 hippocampal expression was significantly decreased in sh-Cx30– and sh-Cx43–injected mice (P = 0.0007), respectively, compared to sh-GFP–injected group (Figure 5C).

During the familiarization phase of the NOL task, LV-injected mice examined the two identical objects for a similar exploration time (all groups, P > 0.05 vs chance level; Figure 5D). The duration of exploration of the displaced object was similarly enhanced in each LV-injected group given vehicle (P < 0.0001 for sh-GFP and sh-Cx30; P = 0.034 for sh-Cx43; Figure 5E). In contrast, all the LV-injected mice treated with scopolamine presented a preference index close to 50% showing clear-cut memory impairment. In all groups, a DPZ1 was able to prevent NOL memory deficits induced by scopolamine (P = 0.0027 for sh-GFP; P = 0.0066 for sh-Cx30 and P = 0.0001 for sh-Cx43 mice; Figure 5E). Interestingly, although DPZ0.25 was unable to prevent the memory deficit caused by scopolamine in sh-GFP injected mice (P = 0.28), this dose significantly reduced the scopolamine effect in both sh-Cx30 and sh-Cx43 groups (P = 0.0033 and P = 0.019, respectively; Figure 5E).