Heterozygous 5xFAD colony founders were bought from Jackson Laboratory (MMRRC_034848-JAX, B6.Cg-Tg (APPSwFlLon,PSEN1*M146L*L286V) 6799Vas/Mmjax). The colony was expanded in-house by breeding heterozygous progeny on C57BL/6J congenic background. For genotyping, the primer sequences 5′-CGG GCC TCT TCG CTA TTA C-3′ (mutant reverse), 5′-ACC CCC ATG TCA GAG TTC CT-3′ (common) 5′-TAT ACA ACC TTG GGG GAT GG-3′ (wildtype reverse) were used28. All experiments were performed on homozygous 5xFAD mice of both sexes from the age of 4 weeks to 32 weeks, in compliance with the Swiss Veterinary Law Guidelines and the ARRIVE guidelines, and approved by the ethics committee of the Cantonal Veterinary Office of Geneva. Throughout their life span, mice were group housed ranging from two to five animals per cage with food and water ad libitum. Transparent individual ventilated plexiglass cages were maintained on a 12 h dark/light cycle at 22 °C in a temperature-regulated room and protected from exterior pathogens.
Animals were assigned randomly to one of three sex-balanced groups: sham treatment (n = 19), low power PBM treatment (6 mW/cm2, n = 20) and high power PBM treatment (600 mW/cm2, n = 21). Sham treatment consisted in handling and immobilization of the mouse under the PBM device similarly to experimental group animals, but with no illumination produced by the device. The mice were subjected to a battery of behavioral tests at baseline, prior to any treatments, when they were 4.5 (± 0.5) weeks old and again after the end of thetreatment (5 months later).
Everyone involved in the experiments and analyses of their outcomes was blinded. To ensure equal treatment of the sham mice, the person operating the PBM device only selected a mouse ID code on a custom graphical user interface, without knowing what light power (or no light) would be emitted. The group assignment was known to our quality assurance manager and remained confidential until data collection and processing had been completed. The entire analyses methodology was made ready in the form of automated scripts which were not changed after the unblinding.
The behavioral tests, except the Y-maze, were based on Monteiro et al.29.
On the first day the mice were habituated to an open field arena (44 × 44 cm with 30 cm high gray non-reflective walls; Ugo Basile S.l.r., Italy). Each mouse was placed in the arena facing the wall and left to explore it for 30 min.
Novel object recognition test (NORT)
Recognition memory was evaluated during the 2 days following the open field habituation, in the same arena. On the first day, two identical objects were symmetrically placed in the arena, at a given distance from the walls. Mice were allowed to freely explore both the objects for 10 min, after which they were returned to their home cages. Twenty-four hours later, the probe session was carried out: one of the objects was replaced by a novel one (similar size and texture but different color and shape). Each mouse was again given 10 min to explore the objects (Fig. 1A). The discrimination index, i.e., time exploring the novel object minus time exploring the old object over the total object exploration time, was calculated. The animal was considered to be exploring an object whenever it was facing the object with its nose within 5 cm from the object’s center.
Short-term spatial memory was measured 24 h after the NORT, by placing the mouse at the end of one of the arms of a radial maze with three arms (A, B and C; arm length 35 cm, width 5 cm, wall height 15 cm; Ugo Basile S.l.r., Italy), facing the center and letting it explore the maze for 10 min before being returned to its home cage (Fig. 1B). The ratio of spontaneous arm alternations (e.g. ABC, BCA, CAB, BAC) to overall number of arm entries minus two, gives an indication regarding working memory, as a healthy mouse is expected to remember the arm of the maze that it previously emerged from and will show a tendency to enter the less recently visited arm.
Morris water maze
To assess spatial reference memory, mice were placed in a white circular pool (120 cm diameter, 60 cm high, Ugo Basile S.l.r., Italy) filled with water at 25 ± 1 °C, made opaque with non-toxic odorless white dye (Guage Eco by Caran d’Ache SA, Switzerland). Spatial cues were placed on the walls around the pool (signs displaying a square, stripes, triangle and a cross printed in A4 format). The pool was divided into four imaginary quadrants and a transparent escape platform (10 cm diameter, 30 cm high, Ugo Basil) was placed in one of the quadrants submerged 1 cm under the water’s surface. The mice had to learn the position of a hidden platform over a period of 4 days with four trials on each day. At the beginning of each trial the mouse was placed facing the wall of the pool in a different quadrant in a pseudorandom order that varied from day to day. Each trial was completed whenever the mouse reached the platform (see Fig. 1C) or when a 120-s timeout period elapsed. On the fifth day, the platform was removed, and a single trial of 60 s was performed (probe trial) during which the accuracy of the mice’s platform seeking was quantified.
Mice were filmed during behavioral experiments using a Zelux CS165MU/M (Thorlabs, Inc., USA) monochrome camera with a varifocal lens (model ACLV0412IR3H, Aico Electronics Limited, China). Automatic mice tracking was performed using DeepLabCut30 with a ResNet50 neural network. For each behavioral maze, a ResNet50 was trained on manually annotated data and then refined and validated on unseen data. All further behavioral data preprocessing and analysis was performed in Matlab (The MathWorks, Inc., USA) and Prism (Dotmatics, USA).
PBM was performed using a custom optical system consisting of an M810L4 LED (Thorlabs, Inc., USA) collimated into a homogenous ~ 1 cm diameter beam using a f20.1 mm aspheric lens (model ACL-DG6-B, Thorlabs, USA) placed approximately 15 mm from the LED surface and a f25.4 mm bi-convex lens (model LB1761-B, Thorlabs, Inc., USA) placed approximately 13 mm from the first lens. A black cone with a 1 cm opening was 3D printed to further ensure that the light was only applied to the mouse head and to help center the mouse head in the NIR light beam which is invisible to humans. Mice were immobilized in a custom-designed restraining cylinder. The peak power of light emitted by the device was set to 470 mW (600 mw/cm2, high power PBM condition) or 5 mW (~ 6 mw/cm2, low power PBM condition) based on the parameters used by Oueslati et al.31 and by De Taboada et al.15, respectively. The light was pulsed at 100 Hz with a 20% duty cycle.
Immunohistochemistry was performed on mouse brain sections. Mice were first anesthetized by an intraperitoneal injection of pentobarbital (50 mg/kg body weight) and perfused intracardially with 0.9% saline solution for 1 min and then with 4% paraformaldehyde in 0.15 M sodium phosphate buffer (PFA 4%) for 4 min. Brains were dissected, post-fixed overnight in PFA 4% followed by a cryoprotection into 30% sucrose in PBS for 24 h and then frozen within isopentane at − 55 °C cooled in a SnapFrost® (Excilone, France). Coronal sections (25 µm) were cut on a Leica CM3050S Cryostat (Leica Microsystems GmbH, Germany) at − 20 °C. Sections were collected from the prefrontal cortex to the beginning of the cerebellum and stored in an antifreeze solution (0.2 M sodium phosphate buffer, glycerol 25%, ethylene glycol 30%) in 96-well plates at − 20 °C.
Prefrontal cortex and hippocampus sections from all mice were processed for multiple staining following protocol described below. Brain sections (25 µm thickness) were incubated in a multi-well plate with constant agitation. Plaques were detected by either antibodies or by Methoxy-X04 (Bio-techne, 4920/50); microglia and neurons were detected by antibodies in a multi-well plate as described in Table 1.
Briefly, the sections were rinsed at room temperature (RT) in TBS (10 mM Tris pH 7.6, 0.9% NaCl) (3 × 10 min) and then incubated in a blocking solution of TBS with 0.1% Triton and 5% bovine serum albumin (BSA) for 1 h. Sections were incubated overnight at 4 °C with primary antibodies listed in Table 2.
After three washes, sections were incubated with the secondary antibody diluted 1/1000 for 2 h at RT. After washing, some sections (see Table 1) were incubated 15 min at RT with Methoxy-X04. Sections were mounted on slides in VECTASHIELD® Vibrance Antifade Mounting Medium (Reactolab, H-1700-10). Brain sections were mounted on slides in VECTASHIELD® Vibrance Antifade Mounting Medium with DAPI (Reactolab, H-1800-10).
Slides were loaded into a customized version of a digital slide scanner TissueScope LE120 from Huron Digital Pathology, consisting of a dual camera setup for brightfield and fluorescence imaging. Emitted fluorescence was collected by a 20×, 0.75NA WD 1mm objective and imaged on a Teledyne Photometrics–Kynetics camera. Raw data were down-sampled by a factor 2 yielding an isotropic pixel size of 0.6 μm. Settings for acquisitions and experimental parameters were kept constants for all sections stained with the specific antibody to avoid biases in fluorescence intensity.
Histology data analysis
Stained sections were automatically analyzed using a custom pipeline written in Python in two distinct brain regions: prefrontal cortex and dorsal hippocampus. These regions were manually annotated by an expert using a custom Napari plugin. Damaged regions or regions exhibiting artifacts were excluded from further analysis. Plaques were segmented on methoxy-X04 stained images using iLastik (with all features included) that was previously trained and validated on manually annotated data. Neurons were segmented on NeuN stained images using Stardist algorithm32. For quantifying the microglial response to plaques, the ratio between the pixel intensity (for Iba-1 staining) close to the plaque and far from the plaque was computed. To do so, the plaque centroid and radius was extracted from the methoxy-X04 channel. The pixel intensity on Iba-1 channel was then averaged in the annulus between the plaque edge and 12 μm away from the plaque edge (i.e. close to the plaque), and in the annulus between 18 to 60 μm from the plaque edge (i.e. far from the plaque). The normalized ratio between the average Iba-1 intensity close to, and far from, the plaque was then calculated to estimate the microglial response to plaques. If microglia density was higher close to the plaque, then the ratio should be higher than 1 and if the density is similar then the ratio should be close to 1.
Statistical analyses were performed with Matlab (The Mathworks, Inc., USA) and Prism (Dotmatics, USA). For variables with more than two levels (groups), such as the main inter-group results (sham vs. low vs. high power PBM), first the normality of data set was confirmed using the Lilliefors test. Given data being normally distributed, one-way ANOVA (analysis of variance) was performed. For variables with two levels (such as baseline vs. endpoint behavioral comparison) either paired Student’s t-test or—for non-normal data according to the Lilliefors test–Wilcoxon’s signed rank test was used. The n values can be found in the figure legends and correspond to the number of mice analyzed. Results are presented as mean ± SEM (standard error of the mean) unless stated otherwise. Statistical differences were considered significant at p < 0.05; denoted by one asterisk in figures; two of three asterisks denote p < 0.01 or p < 0.001, respectively.