We obtained approval from the Taipei Medical University Institutional Animal Care and Use Committee. We used B6C3 hybrid background (C57BL/6 X C3H/HeN), double transgenic (Tg) APPswe/PSEN1dE9 (APP/PSEN1) mice (Jankowsky et al., 2001), and littermate wild type (WT) mice which were purchased from National Applied Research Laboratory, Taiwan (N = 66, n = 12 for each wild type group and n = 11 for each AD group). Mice were genotyped in-house with primers for APP (transgene F-AGGACTGACCACTCGACCAG, R-CGGGGGTCTAGTTCTGCAT; internal positive control F-CAAATGTTGCTTGTCTGGTG, R-GTCAGTCGAGTGCACAGTTT) and PSEN1 (transgene F-AATAAGAACGGCAGGAGCA, R-gCCATGAGGGCACTAATCAT; internal positive control F-CTAGGCCACAGAATTGAAAGATCT, R-GTAGGTGGAAATTCTAGCATCATCC). Only male mice were used in this study to avoid the confounding factor of mouse estrous cycle. All mice, regardless of their WT or APP/PSEN1 status and antidepressant treatment, were randomly allocated to cages. Between two and five mice were housed in each cage, depending on the supply conditions. Animals were fed with food and water ad libitum in 12 h light/dark cycle with constant temperature and humidity. We inspected the mice daily and recorded body weight weekly to assess their health. During euthanisation, an overdose of anaesthetic agents was used at study termination.

Sertraline (CenZoft concentrate solution 20 mg/mL, Center) or paroxetine (Seroxat 15 mg, GSK) was administered in drinking water initially at a dose of 25 mg/kg/day and 10 mg/kg/day respectively. Treatment was initiated at 3 months of age, and was continued for a duration of 12 months. However, we observed unexpected sudden mortalities in APP/PSEN1 transgenic mice treated with initial dose of paroxetine (10 mg/kg/day), and no significant abnormal gross finding was confirmed by veterinarian. Therefore, we decreased the dosage to 5 mg/kg/day four and a half months following the initial treatment.

To identify the sertraline and paroxetine concentrations in serum and brain, LC-MS/MS analysis, by using the mass spectrometer (Agilent triple quadrupole 6470), was performed on mouse samples after 12 months treatment of antidepressants based on the previous modified methods (Peng et al., 2008; Olesen et al., 2016). During sacrification, blood samples were collected by cardiac puncture and centrifuged for 10 min at 1,500 × rpm. The supernatant was stored at −80°C freezer for further analysis. 50 μL of serum was added with 5 μL of internal control (fluoxetine; concentration: 1,000 ng/mL), 50 μL sodium hydroxide (concentration: 0.1 M) and 0.5 mL ethyl acetate. The mixture was vortexed thoroughly for 2 min. The supernatant of the organic phase was collected and centrifuged at 13,000 × g for 5 min. For cerebellums, they were stored at −80°C after perfusion fixation by formaldehyde solution. Mouse cerebellums were added with 600 μL ethyl acetate and 5 μL of internal control (fluoxetine; concentration: 1,000 ng/mL), then homogenized by tissue homogeniser (Kurabo Sh-100) for 90 s at 1,600 × rpm, and then centrifuged at 2,500 × rpm for 10 min at 4°C. A volume of 300 μL top organic layer was further transferred to eppendorf tubes and centrifuged at 13,000 × rpm for 5 min. Both supernatant from processing of plasma and cerebellum was filtered by PVDF (0.22 μm) followed by LC-MS/MS analysis. Two MRM transitions were monitored for sertraline, paroxetine and fluoxetine (sertraline: 306.3–275.2; paroxetine: 330.1–192.2; fluoxetine: 310.1–148.1; Supplementary Figure S1). Calibration was performed by linear calibration based on nine-points and the range was 1–640 ng/mL.

The serum samples were prepared as the same procedure described in the previous section for the quantification of sertraline and paroxetine. The serotonin and BDNF levels in the serum was quantified using the ELISA kits for ST/5-HT (5-hydroxytryptamine; FineTest EU0253) and for BDNF (Cloud clone SEA011Mu). Firstly, the plates were washed, and then 50 μL of standards, samples, or blanks were added to the pre-coated plate, along with 50 μL of biotin-labeled antibody. The plate was sealed, gently tapped, and incubated at 37°C for 45 min. Subsequently, a washing step was carried out, followed by the addition of 100 μL of HRP-Streptavidin Conjugate. After another washing step, 90 μL of 3,3′,5,5′-Tetramethylbenzidine (TMB) substrate was added, followed by 50 μL of Stop Solution. The absorbance of the plates at 450 nm was measured using the Bio-Tek µQuant Universal Microplate Spectrophotometer.

The mice were sacrificed under a general anaesthesia overdose, followed by transcardial perfusion with ice-cold PBS and 10% formaldehyde. Subsequently, the brains were collected and kept in 10% formaldehyde for 24 h, followed by preservation in 70% ethanol at 4°C. The tissues were trimmed, and dehydrated with serial alcohol solution. They were then embedded in paraffin wax, and were cut to a thickness of 3 μm. Coronal brain tissue sections, intended for histopathological assessment, were obtained with the specific aim of capturing cuts that encompassed a sufficiently large area of the hippocampus. These sections were positioned approximately 2.3 mm caudal to the bregma, in accordance with the mouse brain atlas by Paxinos and Franklin (Paxinos and Franklin, 2001). This location corresponded to approximately 8.3 mm from the cranial apex.

For H&E staining, the paraffin section slides underwent deparaffinization with xylene and hydration using serial alcohol solutions. The sections were stained with hematoxylin solution (BioTnA TA01NB) for 1 min, followed by a 5-min wash with running tap water. Subsequently, the slides were counterstained with Eosin (BioTnA TA01ES) for 1 min. Dehydration with alcohol, clearance in xylene, and mounting with mounting medium followed.

Brain lesions in H&E-stained sections were assessed and graded by a veterinary histopathologist based on the INHAND (International Harmonization of Nomenclature and Diagnostic Criteria) (Shackelford et al., 2002). Evaluated areas and parameters included acidophilic material (characterized by strong eosinophilic deposits in the centre surrounded by pale-stained structures in a radial pattern), atrophy (decreased volume/cell counts), and gliosis (characterized by densely stained and small-sized elements) of the isocortex and hippocampus. Grading was performed as follows: Grade 0 - no remarkable histopathological changes (<1%); Grade 1 - minimal (1%–5%); Grade 2 - slight (5%–25%); Grade 3 - moderate (26%–50%); Grade 4 - moderately severe (51%–75%); Grade 5 - severe/high (>75%).

For thioflavin-S staining, the slides were deparrafinized and hydrated with xylene and serial alcohol solutions. Stains were performed with 1% thioflavin-S solution (Sigma-Aldrich) for 10 min, and then the slides were washed in running tap water for at least 10 min. The slides were then dehydrated with alcohol, cleared in xylene, and mounted with mounting medium.

Thioflavin-S staining was quantified both automatically, involving the calculation of the positive staining area relative to the entire brain, and manually, by counting the number of positive spots in the cortex and hippocampus. These assessments were conducted in a blinded fashion. Additionally, we engaged a veterinarian pathologist in identifying true positive signals to perform these assessments in a blinded manner.

Congo red staining was performed by using the stain kit (BIOTnA Biotech TASS12) to identify amyloids. After deparaffinization and hydration, the slides were stained by congo red solution for 15–20 min followed by rinsing in distilled water. They were then immersed in alkaline alcohol solution for differentiation and hematoxylin for counterstaining. Following dehydration, the slides were cleared in xylene and mounted with resinous mounting medium. We captured the images by using a compound optical microscope (Olympus, Japan) and analyzed the Congo red staining spots using ImageJ.

For immunofluorescent staining of anti-glial fibrillary protein (GFAP), following deparaffinization and hydration, the slides were blocked with 10% BSA in PBS and incubated with the primary antibody against GFAP (1:600; Genetex GTX108711) at 4°C overnight. Subsequently, the slides were incubated with the secondary antibody Alexa Fluor 488 (1:750; Abcam ab150077) for 1 hour at room temperature. Both primary and secondary antibodies were diluted in 3% BSA in PBS. The slides were mounted with a fluorescence mounting medium containing DAPI (Origene). Immunofluorescence was visualized and captured using a confocal microscope (Stellaris 8 confocal microscope, Leica).

For immunohistochemistry staining, the primary antibodies and their dilutions were anti-phospho-mixed lineage kinase domain-like protein Ser345 (pMLKL; 1:200; NovusBio NBP2-66953) and anti-ionised calcium-binding adapter molecule 1 (IBA1; 1:200; Bioworld BS90680). The slides were scanned using a MoticEasyScan Pro 6 at ×40 magnification.

For both staining techniques, positive-stained cells and total cell counts were manually calculated at ×40 magnification.

To assess the long-term memory, mice were subjected to the novel object recognition test (NORT) based on modified methods from a previous study (Zhang et al., 2012). The novel object recognition tests were conducted at 3 and 12 months post antidepressants in a rectangular white arena (60 cm × 80 cm, surrounded by 60 cm-high walls). The entire assessment consisted of 4 phases: pre-habituation, habituation, training, and testing. On the first day, animals were brought to the testing room to freely explore the box in the absence of objects for 5 min. On the second and third day, mice were habituated to the empty box for 20 min per day. On the fourth day, each mouse took a training trial with exploration of two identical objects for 10 min and followed by a testing trial after an inter-session interval of 1 h. During the testing, animals were placed back to the same box, where one of the two familiar objects was switched to a novel one, to start a 5 min testing phase. Recording videos were analyzed by using ACTUAL TRACK software. Object exploration time was defined as the length of time the subject spent sniffing, pawing, or directed its nose within 2 cm of the object. Sitting or standing on the object was not recognized as exploration. The exploration time was analyzed manually using 2 stop watches. In the training session, the location preference in the training phase and recognition index (RI) in the testing phase were calculated using the following formula:

Recognition index (RI) = Time exploring novel object/(Time exploring novel object + Time exploring familiar object)×100%

The contextual fear conditioning was done at 15 months of age to evaluate by measurement of the tendency of freezing behavior. On training day, mice were given 3 min of exploration in a chamber with current-regulated shocker (Coulbourn Instruments) and followed by 2 s of electric footshock (0.75 Ma). Mice were then removed from the chamber 30 s later. The testing was performed 24 h following the training day, mice were returned to the chamber and freezing behaviour was recorded for 3 min. Freezing was defined as the absence of any movement with treshold of more than 1 s. Freezing was measured using the FreezeScan video tracking system and software (CleverSys).

The behavioral tests were carried out in a blinded manner, with the individuals conducting the tests unaware of the group assignments.

All of the data are presented as mean ± SD. All statistic analyses were performed using GraphPad Prism (GraphPad Software, Inc., San Diego, CA). The Log-rank test was used for survival analysis, and one-way ANOVA followed by Dunnett’s multiple comparisons test was employed for other comparisons. Significance was set at 0.05.

Both WT and APP/PSEN1 mouse model were used to study the effect of antidepressants on initiation and progression of AD. Sertraline and paroxetine were administered in drinking water for a period of 12 months to study the long term effects of these medications. The daily dose of sertraline was 25 mg/kg/day, while paroxetine was initially given at a dosage of 10 mg/kg/day for the first four and a half months, and then reduced to 5 mg/kg/day for the remainder of the study. These initial dosages were chosen based on the recommendations provided in the guide for treating neuropsychiatric symptoms of AD patients (Lleo et al., 2006) and the dosages employed in animal models (Taler et al., 2013). Unexpected sudden mortalities happened in paroxetine group, which was limited to APP/PSEN1 mice (p < 0.001; Figure 1A). For both WT (p = 0.004) and AD (p < 0.001) mice, the groups treated with paroxetine demonstrated more weight gain compared to control groups (Figure 1B).

To determine the antidepressant levels in systemic circulation and central nervous system, we performed LC-MS/MS analysis by using serum and cerebellum tissues. Data confirmed the existence of sertraline in our samples with acquisition time of 7.964 min and MRM transition of 306.3 to 275.2 (Supplementary Figure S1A), and acquisition time of 7.609 min and MRM transition of 330.0 to 192.0 for paroxetine (Supplementary Figure S1B). After pooling data from WT and APP/PSEN1 mice, sertraline was shown to be 8.6 nmol/L (n = 17) in serum and 289.8 ng/g (n = 20) in cerebellum, while paroxetine was 120.0 nmol/L (n = 13) in serum and 792.9 ng/g (n = 14) in cerebellum (Figure 1C). These results were in consistent with the serum concentrations of antidepressants from patients’ samples (Reis et al., 2009).

Sertraline and paroxetine belong to the class of SSRIs, which elevate extracellular serotonin levels by inhibiting its reabsorption by presynaptic cells. To determine whether the SSRI concentration was sufficient to impact the physiology of mice in this model, we quantified serum serotonin levels to assess the impact of the SSRI concentration on mouse physiology in this model. We observed a significant increase in serum serotonin levels in WT mice treated with sertraline (p = 0.01; Figure 1D), and a non-significant trend towards elevated serum serotonin levels in the paroxetine group. However, these effects were not observed in the APP/PSEN1 mice. As for BDNF, a protein crucial for stimulating the growth and preservation of neurons, no significant changes were observed among all groups (see Supplementary Figure S2).

By using thioflavin-S staining to evaluate the β-amyloid deposition, we observed minimal thioflavin-S deposit in WT mice across all treatment groups (Figure 2A left panel and Figure 2B). However, sertraline (p < 0.001) significantly increased the total area of thioflavin-S deposit compared to control group in APP/PSEN1 mice, as quantified automatically using ImageJ software (Figure 2A right panel and Figure 2B). This finding aligns with the result measured by a veterinary pathologist (p = 0.001; Supplementary Figure S3A). To further investigate the brain areas affected, we manually counted the thioflavin-S deposits in a blinded manner and observed a significantly higher number of deposits scattered across the isocortex (p < 0.001) and hippocampus (p < 0.001; Supplementary Figure S3B). These findings were supported by Congo red staining, which revealed that sertraline increased deposits in both the isocortex (p = 0.003) and hippocampus (p < 0.001) of APP/PSEN1 mice (Figure 3). We also found that paroxetine exacerbated the deposition in the hippocampus of APP/PSEN1 mice (p = 0.004).

In the APP/PSEN1 mice, elevated levels of acidophilic material and gliosis were observed in both the isocortex (p = 0.020 and = 0.004) and hippocampus (p < 0.001 and = 0.004) of the sertraline-treated group when compared to the control group (Figures 4A, B). It is noteworthy that the acidophilic materials were surrounded by glial cells (Supplementary Figure S4). Additionally, there was a non-significant trend indicating an increased presence of Iba-1⁺ cells, a microglial marker, in the sertraline-treated APP/PSEN1 mice (Supplementary Figure S5). In contrast, paroxetine significantly exacerbated the gliosis (p = 0.020) and also led to hippocampal atrophy (p = 0.010) in the APP/PSEN1 mice (Figure 4A, middle and right panels). This treatment was also associated with a trend of increased GFAP⁺ cells, an astrocyte marker, in both WT and APP/PSEN1 mice (Supplementary Figure S5).

Novel object recognition test was performed to evaluate the cognitive function of mice with or without antidepressants (Figure 5A). The results showed sertraline (p = 0.009) and paroxetine (p = 0.04) significantly reduced recognition memory of WT mice 3 months after antidepressants administration (Figure 5B left), while the cognition of APP/PSEN1 mice in sertraline (p = 0.03) and paroxetine (p < 0.001) group were decreased in 12 months (Figure 5B right). However, no significant difference among groups was found in the contextual fear conditioning (Supplementary Figure S6). The result implied that sertraline and paroxetine impair memory function after long-term antidepressant medication.