Male APPswe/PSEN1dE9 double transgenic mice overexpressing the mutated genes for human amyloid precursor protein and presenilin 1 (APP/PS1; Figure 1A), as well as age-matched male wild-type (WT) mice were purchased from Beijing HFK Bioscience Co., China. The mice were provided food and water ad libitum and housed under controlled conditions (22 ± 2°C and 12 h/12 h light–dark cycle). All the animal experiments were conducted in accordance with the “Guidelines for the Care and Use of Mammals in Neuroscience and Behavioral Research,” and were approved by the Laboratory Animal Ethical and Welfare Committee of Hebei Medical University (IACUC-Hebmu-2023054). The experimental mice were divided into following four groups: WT group, APP/PS1-Sham group, gonadectomized APP/PS1 (APP/PS1-GDX) group, and gonadectomized APP/PS1 with T propionate (TP) supplementation (APP/PS1-GDX-TP) group, where mice were subjected to either bilateral orchiectomy or sham operation. For APP/PS1-GDX-TP mice, male APP/PS1 mice were castrated, and subcutaneously injected immediately with TP at 1 mg/kg per day for 28 days. This dosage was adopted based on a previous study (Qiu et al., 2016). All the mice were sacrificed at the age of 4 months. A previous study illustrated that APP/PS1 mice show mild cognitive deficits at this age (Yu et al., 2019). The experimental procedures were briefly illustrated in Figure 1A.

On the 8th day before sacrifice, mice underwent habituation in a 40 cm by 40 cm soundproof square open field for 5 min. The next day, a novelty recognition test was conducted using the same apparatus. The test consisted of a familiarization trial, novel object localization (NOL) test, and novel object recognition (NOR) test. The experimental procedure followed a previously established method, with minor modifications from the study (Jacobs et al., 2021). In brief, the mice were exposed to two identical wooden objects during the 10-min familiarization trial. After a 1-h interval, the NOL test began, where one object was relocated. Another 1-h interval followed, and the mice were subjected to the NOR test, where one object was replaced with a novel hemisphere-shaped wooden object. The DigBehv Animal Behavior Analysis Software was used to record the time mice spent exploring each object within 2 cm and calculate exploration ratios and recognition indices. The objects and arena were cleaned with 70% ethanol between each trial.

During the last 6 days prior to their sacrifice, the mice were subjected to water maze test to evaluate their spatial learning and memory abilities. These tests were conducted in a controlled environment within a soundproof room. The main apparatus used for the experiment was a circular plastic tank with a diameter of 120 cm and height of 50 cm, which was partially filled with opaque water. A platform with a diameter of 9 cm was concealed 1 cm beneath the water’s surface, located in the first quadrant and served as an escape point for the mice during the trials. The training sessions were conducted over the first 5 days. Each daily session consisted of four trials, with an hour interval between each trial. In each trial, the mouse was placed in the water maze starting from different quadrants each time, and was given 60 s to navigate the maze and locate the hidden platform. The time taken to reach the platform, known as the escape latency, was recorded. On the sixth day, a probe session was conducted. During this session, the escape platform was removed from the maze. Each mouse was given 60 s to freely swim in the maze. The time spent swimming in the first quadrant, as well as the number of crossings over the platform’s previous position, were recorded. To accurately analyze the movement patterns of the mice during the trials, the DigBehv Animal Behavior Analysis Software (Shanghai Jiliang Software Technology Co., China) was utilized.

For biochemical or molecular biological assays, mice were sacrificed by decapitation and brains were removed quickly. Bilateral hippocampi were dissected, which were immediately used for biochemical assays or frozen in liquid nitrogen and stored at −80°C for later use based on the study aims. After decapitation, seminal vesicles were dissected, blotted and weighed. Trunk blood was collected and serum was harvested by centrifugation and stored at −80°C for measuring T levels. For immunohistochemistry (IHC) staining studies, mice were anesthetized and perfused transcardially with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). For mitochondrial ultrastructure analysis, mice were perfused transcardially with fixative containing 2% paraformaldehyde and 1.25% glutaraldehyde. Brains were postfixed in the corresponding fixative for 24 h at 4°C.

The postfixed brain tissues were dehydrated in graded ethanol, cleared in xylene, and embedded in paraffin wax. Wax-embedded brain blocks were sliced into 5 μm coronal sections. The sections were dewaxed using xylene and gradient ethanol for deparaffinization and rehydration. The brain sections were blocked in 5% normal goat serum after antigen retrieval, and then incubated with rabbit anti-Aβ_1–42 (Biolegend, specifically recognizing extracellular Aβ_1–42) or mouse anti-3-NT antibody (Santa Cruz Biotechnology) overnight at 4°C. After three washes in phosphate buffer, the sections were incubated with goat anti-rabbit IgM-HRP or goat anti-mouse IgG-HRP for 2 h at room temperature. All sections were stained for 5 min in liquor containing 0.05% diaminobenzidine and 0.03% H₂O₂ in 0.05 M Tris–HCl buffer (pH 7.6). The photograph of stained sections was taken by light microscope (Olympus). The area/number of the Aβ_1–42 plaques and the average optical density of 3-NT were measured by Image Pro Plus 6.0.

For spectrophotometry analysis, hippocampal tissue blocks were weighed and homogenized immediately with 0.01 M ice-cold phosphate buffer saline (PBS, pH 7.4) and centrifuged based on the study purposes. Levels of MDA, the ratio of GSH/oxidized GSH (GSH/GSSG), the activities of CuZn-SOD and Mn-SOD, ATP levels, as well as the activities of mitochondrial complexes and citrate synthase (CS) were measured spectrophotometrically according to the instructions of the corresponding detection kits (MDA: Code A003-1; GSH/GSSG: Code A061-1; CuZn-SOD and Mn-SOD: Code A001-2, Nanjing Jiancheng Bioengineering Institute, China; ATP: AKOP004M; Complex I: AKOP005M; Complex II: AKOP006M; Complex III: AKOP007M; Complex IV: AKOP008M; Complex V: AKOP009M; and CS: AKAC007M, Boxbio, China). Serum T levels were measured using a T ELISA kit (Item No. 582701, Cayman) based on the kit instruction. For spectrophotometry analysis, SH-SY5Y cells seeded in 96-well plates (4 × 10⁴ cells/well) were processed by different treatments. The cell viability after different treatments was detected with MTT method by adding 50 μl DMEM/F12 and 50 μl MTT reagents to the wells. After incubation for 3 h at 37°C, 150 μl of DMSO was added to the wells and their absorbance was read at a wavelength of 590 nm. The cell GSH/GSSG ratio and ATP levels were respectively measured using the GSH and GSSG detection kits (S0053, Beyotime Institute of Biotechnology, China) and the ATP levels detection kit (A095-1-1, Nanjing Jiancheng Bioengineering Institute, China), with the absorbance read at 412 nm for GSH and GSSG and at 636 nm for ATP levels. The levels were normalized to the total protein amount of the sample.

Mitochondrial membrane potential was estimated in dissociated hippocampal cells using JC-1 dye that has two forms of fluorescence (Cossarizza et al., 2019). The ratio of aggregate form (exhibiting red fluorescence) to monomer form (exhibiting green fluorescence) reflects the level of mitochondrial membrane potential. Tissue blocks were grinded in PBS solution and filtered using a nylon mesh screen. The harvested cells were incubated in JC-1 solution (10 μg/ml, Cayman) at 37°C for 30 min, washed, resuspended in 500 μl PBS, and immediately analyzed by flow cytometry. The red fluorescence intensity was measured at Ex/Em: 525/590 nm. The green fluorescence intensity was measured at Ex/Em: 490/530 nm. The ratio of JC-1 aggregates to JC-1 monomers was calculated.

The hippocampal tissue blocks were dissected and washed in phosphate buffer, they were postfixed with 1% osmium tetroxide for 2 h, dehydrated in acetone, and embedded in araldite. Ultrathin sections (70 nm) were obtained with a microtome. After staining with uranyl acetate and lead citrate, the sections were examined by a transmission electron microscope (Hitachi HT7800) at 80 kV for mitochondrial ultrastructural alteration. The abnormal mitochondria rate was calculated by the number of abnormal mitochondria/total number of mitochondria (Mansueto et al., 2017).

Human neuroblastoma SH-SY5Y cells were used in in vitro studies (ATCC, Manassas, VA, USA). They were grown in DMEM/F12 medium containing 10% fetal bovine serum and 1% penicillin and streptomycin at 37°C in a humidified incubator with 5% CO₂ and 95% air flow. For cell viability experiments, the cells were first treated by T (1, 10, and 100 nM) for 24 h and then by H₂O₂ (0, 50, 100, 200, and 300 μM) for 24 h (Grimm et al., 2014). Flutamide (F, AR receptor antagonist) pretreatment for 1 h was chosen to test the role of AR in the effects of T on cell viability in exposure to H₂O₂. For the experiment about the effects of AR inhibitor and T on the oxidative damage, SH-SY5Y cells were treated with F for 1 h, and treated with T for 24 h, followed with H₂O₂ treatment for 24 h. In this part, four groups were included, that were, control group, H₂O₂ group, T + H₂O₂ group, F + T + H₂O₂ group. Stock solutions of T and F were prepared in DMSO (final concentration of DMSO was 0.001%). Stock solution of H₂O₂ was prepared in DMEM/F12. The control group treated with vehicle solutions (DMSO or DMEM/F12). Cells were plated for treatment when they reached 70%–80% confluence. After treatment, cells were washed twice with cold PBS, then were subjected to corresponding detection.

Oxygen consumption rate (OCR) of cells was examined using a Seahorse Bioscience XF24 Analyzer. Cells were seeded in XF24 cell culture microplate at a density of 2.5 × 10⁴ cells/well. After the cells adhered to the wall, culture medium was replaced with medium containing 1 mM sodium pyruvate, 2 mM glutamine, and 10 mM glucose, Then, the microplate was placed in a CO₂-free incubator at 37°C for 1 h. The OCR was recorded simultaneously before and after the sequential injection of oligomycin (0.5 μM), FCCP (2 μM), and rotenone and antimycin A (0.5 μM). Data were extracted from Seahorse XF24 software, and were normalized to cell counts.

Cells were cultured on glass slides. After treatment, JC-1 working solution (Item No. 15003, Cayman, USA, 10 μg/ml), MitoTracker Red CMXRos (M7512, Invitrogen, USA), and Hoechst 33258 (Beyotime Institute of Biotechnology, China) were added in the culture medium respectively according to different experimental purposes. After incubation in dark at 37°C for 10 min, the cells were washed in warm PBS gently. Fluorescent images were acquired using a two-photon microscope (FV1200MPE, OLYMPUS). JC-1 monomers were detected at Ex/Em 490/530 nm, JC-1 aggregates were detected at Ex/Em 525/590 nm, mitochondria stained with MitoTracker Red CMXRos were detected at Ex/Em 579/599 nm, and cells stained with Hoechst 33258 were detected at Ex/Em 530/461 nm. Ten fields were randomly selected and at least 100 cells were analyzed in each sample. The mean fluorescence intensity (integrated density/area of view) of each cell was measured using ImageJ software (Pediconi et al., 2023). Mitochondrial membrane potential was analyzed by calculating the ratio of red to green fluorescence. The mitochondrial mass was expressed by the mean fluorescence intensity of mitochondria stained with MitoTracker Red (Sghaier et al., 2019). Punctiform mitochondria were considered fragmented mitochondria, mitochondria with clear networks were considered tubular mitochondria. The number of cells was counted when 80% of mitochondria in the cell are under mitochondrial fragmentation.

After extraction with TRIzol from hippocampal tissue blocks or cultured cells, 1 μg of total RNA was reverse-transcribed using random primers to obtain the first-strand cDNA template. Then, quantitative real-time PCR (qPCR) was performed with 2 μl of cDNA (diluted 1:10), 2 μl of each specific primer, and 2 × All-in-One™ qPCR Mix (GeneCopoeia Inc.) in a final volume of 20 μl. PCR was performed as follows: an initial cycle at 95°C for 15 min, followed by 40 cycles at 95°C for 10 s, 60°C for 20 s, and 72°C for 20 s. The melting curves of the PCR products were analyzed to confirm the specificity of amplification. Gene expression was analyzed using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as internal control. For all samples, qPCR was performed in triplicate. Relative quantification was performed using the 2^–ΔΔCt method. The sets of primers were as follows: for mouse genes, SYP (5′-cag ttccgggtggtcaagg-3′ and 5′-actctccgtcttgttggcac-3′), PSD-95 (5′-taccaaagaccgtgccaacg-3′ and 5′-cggcattggctgagacatca-3′), Mn-SOD (5′-tgaacaacctcaacgccac-3′ and 5′-gaaggtagtaagcgtg ctc-3′), PGC-1α (5′-gaaagggccaaacagagaga-3′ and 5′-gta aatcacacggcgctctt-3′), NRF-1 (5′-tggagtccaagatgctaatg-3′ and 5′-agagctccatgctactgttc-3′), TFAM (5′-caggaggcaaaggatgattc-3′ and 5′-ccaagacttcatttcattgtcg-3′), Drp1 (5′-caggaattgttacggttcccta-3′ and 5′-cctgaattaacttgtcccgtg-3′), Mfn1 (5′-aacttgatcgaatagcatccga-3′ and 5′-gcattgcattgatgacagag-3′), OPA1 (5′-gatgacacgctctccagtg-3′ and 5′-tcggggctaacagtacaa-3′), GAPDH (5′-actcttccaccttcgatgcc-3′ and 5′-tcttgctcagtgtccttgct-3′); for human genes, PGC-1α (5′-ctg caggcctaactccaccca-3′ and 5′-actcggattgctccggccct-3′), NRF-1 (5′-ctactcgtgtgggacagcaa-3′ and 5′-aattccgtcgatggtgagag-3′), TFAM (5′-ctactcgtgtgggacagcaa-3′ and 5′-aattccgtcgatggtgagag-3′), Drp1 (5′-aatcgtcgtagtgggaacgc-3′ and 5′-tccaccccattttcttctcct-3′), Mfn1 (5′-gcctcctctccgcctttaactt-3′ and 5′-gccttcttagccagcacaaag-3′), OPA1 (5′-cgggaacttgaccggaatga-3′ and 5′-cgcagctggaaggtagatgt-3′), and GAPDH (5′-ctcctccacctttgacgctg-3′ and 5′-ccacc ctgttgctgtagcca-3′).

For mice experiment, total DNA was extracted from tissue blocks using an animal tissue genomic DNA kit (ZP307-2, ZOMANBIO), while it was extracted from SH-SY5Y cells by cell genomic DNA extraction kit (ZP308, ZOMANBIO). Mitochondrial DNA (mtDNA) copy number was determined by quantifying mitochondrial gene (16S rRNA for mice and D-Loop for human cells) and nuclear-encoded gene (HK2 for mice and Actin for human cells) expression via qPCR. The sets of primers were as follows: for mouse genes, 16S rRNA (5′-ccgcaagggaaagatgaaagac-3′ and 5′-tcgtttggtttcggggtttc-3′), HK2 (5′-gccagcctctcctgattttagtgt-3′ and 5′-gggaacacaaaagacctcttctgg-3′); for human genes, D-Loop (5′-accacccaagtattgactcaccca-3′ and 5′-ccgtacaatattcatggtggctggc-3′), Actin (5′-cccctggcggcctaaggact-3′ and 5′-acatgccggagccgttgtcg-3′). qPCR was performed in triplicate. Relative mtDNA copy number was calculated by the ratio of mitochondrial gene to nuclear-encoded gene using the 2^–ΔΔCt method.

The proteins from hippocampal tissue blocks or SH-SY5Y cells were homogenized in ice-cold RIPA buffer (No. R0010, Solarbio, China) containing 1% PMSF. An immunoblotting analysis was performed following a previous method (Yan et al., 2021). The presence of particular proteins was examined using rabbit anti-synaptophysin (SYP, Abways), rabbit anti-postsynaptic density protein 95 (PSD-95, Abways), rabbit anti-Mn-SOD (HuaBio), rabbit anti-peroxisome proliferator-activated receptor-γ coactivator 1αantibody (PGC-1α, ABclonal), rabbit anti- nuclear respiratory factor 1 antibody (NRF-1, ABclonal), rabbit anti- mitochondrial transcription factor antibody (TFAM, GeneTex), rabbit anti-mitochondrial fission protein dynamin-related protein (Drp1, Cell Signaling Technology), rabbit anti-mitochondrial fusion protein mitofusin 1 (Mfn1, arigo), rabbit anti-optic atrophy protein 1 (OPA1, GeneTex), or mouse anti-3-NT antibody (Santa Cruz Biotechnology). After washing three times, the membrane was incubated for 2 h in IRDye^® 800-conjugated goat anti-rabbit secondary antibody (Rockland) or goat anti-mouse secondary antibody (Rockland) at room temperature. The relative band densities were measured by an Odyssey infrared scanner (LICOR Biosciences, USA). The expression of target proteins was normalized to β-actin that was used as endogenous control (rabbit anti-β-actin antibody, Santa Cruz Biotechnology).

Data are shown as the mean ± standard deviation (SD). Grubb’s test was applied to remove possible outliers. Levene’s test was applied to test the homogeneity of variance. If homogeneity of variance is equal, statistical analysis was applied by one-way analysis of variance (ANOVA, F-statistic) followed by Tukey’s multiple comparison test. If the variance was unequal, Welch’s F test in one-way ANOVA (F′-statistic) was used and post hoc analyses were done using the Games-Howell procedure. For the Morris maze, escape latency was analyzed using two-way ANOVA with repeated measures, and the number of platform crossings was tested using Chi-square test. Statistical analyses were performed using the SPSS 21 software. P < 0.05 was considered statistically significant.

In male AD model mice, gonadectomy was performed to induce androgen depletion, which is verified by significantly reduced seminal vesicle weight (seminal vesicle weight/body weight: 0.6 × 10^–3 ± 0.4 × 10^–3, a sensitive bioassay of androgen levels; Yamane et al., 1986) and serum T levels (0.1 ± 0 ng/ml), compared with sham-operated APP/PS1 mice (seminal vesicle weight/body weight: 5.1 × 10^–3 ± 0.8 × 10^–3, serum T levels: 7.4 ± 1.0 ng/ml). As a control for APP/PS1-GDX mice, the cognitive, pathological, and mitochondrial changes of APP/PS1-Sham mice were firstly compared with WT mice fully. To explore effects of T deficiency on the cognitive performance of the 4-month-old APP/PS1 mice, NOL (Figures 1B–D), NOR (Figures 1E–G), and Morris water maze (MWM) (Figures 1H–K) tests were performed. NOL and NOR test are sensitive to detect impaired recognition memory (Leger et al., 2013). Considering the GDX-induced apathy and motor problem may affect the interactions of experimental animals with the objects in NOL and NOR tests, MWM test were performed following NOL and NOR test. Morris water maze test is common methods to evaluate hippocampal-dependent learning, acquisition of spatial memory and long-term spatial memory (Jacobs et al., 2021). In NOL test, all the mice spent the same time in exploring the old and the novel localization of the familiar objects, except for the APP/PS1-GDX mice that spent less time on exploration in the new location than that in the old location (Figure 1B). APP/PS1-GDX mice showed significantly low recognition index compared with WT mice (Figure 1C). In NOR test, the mice in WT, APP/PS1-Sham, and APP/PS1-GDX-TP groups exhibited more time exploring the novel object and APP/PS1-GDX mice spent similar time exploring either the new or the old object (Figure 1E). In Morris water maze test, during the 5 training days, the escape latency of WT mice progressively decreased, while APP/PS1 mice with or without castration exhibited longer escape latency than WT mice in the 2nd, 4th and 5th day (Figure 1H). An obviously decrease of escape latency was found in APP/PS1-GDX-TP mice compare with APP/PS1-GDX mice in the 3rd training day, and a slight but non-significant reduction of escape latency was observed in APP/PS1-GDX-TP mice, when compared with APP/PS1-Sham mice and APP/PS1-GDX mice in the last 2 days. In the probe test, group difference was found in the time spent in the target quadrant and in the number of platform crossing (Figures 1I–K). APP/PS1-Sham mice spent less time in the target quadrant than WT mice and orchiectomy further decreased time spent of APP/PS1 mice in the target quadrant (Figure 1I). TP supplementation to APP/PS1-GDX mice restored the time spent in the target quadrant to the level of APP/PS1-Sham mice as well as WT mice. Relative to WT mice, APP/PS1-GDX mice showed the reduced number of platform crossing (Figure 1J). Testosterone deficiency further compromises the cognitive deficits of male APP/PS1 mice.

The pathological changes of hippocampus in experimental animals were revealed by observing the altered expression of synapse-related proteins (PSD-95 and SYP) as well as Aβ deposition. Significantly decreased expression of PSD-95 and SYP at mRNA and protein levels were detected in the hippocampus of APP/PS1-Sham mice relative to WT mice, which were further reduced in APP/PS1-GDX mice (Figures 2A–D). Significantly increased area and number of Aβ_1–42 plaques were observed in APP/PS1-GDX mice, compared with mice in other three groups (Figures 2E–G). Testosterone deficiency further increases the pathological alterations of the hippocampus in APP/PS1 mice.

Increased oxidative stress is common in normal aging and AD brain. Oxidative stress is caused by imbalance between production of ROS and antioxidative defense. To assess the extent of oxidative stress, we measured MDA, 3-NT, and GSH/GSSG ratio, as well as CuZn-SOD and Mn-SOD activity/levels. MDA is an important indicator of lipid peroxidation caused by ROS damage to cytoplasmic membrane (Liu et al., 2021). 3-NT is another oxidative damage marker formed due to nitration of protein-bound and free tyrosine residues by reactive peroxynitrite molecules. GSH and GSSG are two different forms of glutathione. As a cofactor for antioxidative enzymes, GSH can be oxidized to GSSG. CuZn-SOD functions in the nucleus and cytoplasm to eliminate ROS, while Mn-SOD works within the mitochondria to clear ROS (Liu et al., 2021). The MDA levels in the hippocampus were significantly higher in male APP/PS1-Sham mice than in WT mice. Testosterone deficiency further increased MDA levels and 3-NT immunoreactive intensity in the hippocampus of APP/PS1 mice (Figures 3A–C). The ratio of GSH/GSSG, as well as the activity and levels of Mn-SOD enzymes in the hippocampus was less in APP/PS1-Sham mice than WT mice (Figures 3D–G). Castration to male APP/PS1 mice further decreased the ratio of GSH/GSSG, as well as the activity and levels of Mn-SOD enzymes in the hippocampus. There was no obvious difference in CuZn-SOD activity between APP/PS1-GDX mice and APP/PS1-Sham mice (Data not shown), though its activity in APP/PS1-Sham mice was significantly low relative to WT mice. Testosterone deficiency aggravates oxidative stress in the hippocampus of male APP/PS1 mice.

Mitochondria are the main source of energy and ROS production. Mitochondrial dysfunction can lead to ROS overproduction and subsequent oxidative stress (Guo et al., 2022). In AD, impairment of mitochondrial functions has been implicated as the primary cause of ROS generation (Tobore, 2019) and is one of the early events of AD. Thus, mitochondrial functions index, such as mitochondrial membrane potential (revealed by the ratio of JC-1 aggregates to monomers, Figures 4A, B), ATP levels (Figure 4C) and the mitochondrial complex activities (Figure 4D) were analyzed, to elucidate whether T deficiency aggravated the mitochondrial dysfunction in the hippocampus of male APP/PS1 mice. Group difference among WT, APP/PS1-Sham, APP/PS1-GDX, and APP/PS1-GDX-TP mice was found in the hippocampal JC-1 ratio and ATP levels, as well as activities of mitochondrial complexes IV and V. APP/PS1-Sham mice showed decreased JC-1 ratio and activities of mitochondrial complexes IV and V than WT mice (Figures 4A, B, D). Testosterone deficiency significantly reduced JC-1 ratio, ATP levels and mitochondrial complex IV activity in the hippocampus of male APP/PS1 mice. There is no group difference in ATP levels between WT mice and APP/PS1-Sham mice. No altered mitochondrial complex activities were observed in complexes I, II, and III among WT, APP/PS1-Sham, APP/PS1-GDX, and APP/PS1-GDX-TP mice, as well as in complex V among APP/PS1-Sham, APP/PS1-GDX, and APP/PS1-GDX-TP mice. Testosterone deficiency exacerbates mitochondrial dysfunction in the hippocampus of male APP/PS1 mice.

Mitochondria are energy source of cells. Local energy deficits are correlated with the reduced content of mitochondria with normal function in afflicted brain area. Mitochondrial biogenesis is a predominant strategy to maintain healthy mitochondrial content. Thus, relevant parameters reflecting mitochondrial content, such as CS activity (a mitochondrial matrix enzyme) and mtDNA copy number, as well as the key inducer and effectors regulating mitochondrial biogenesis, namely PGC-1α, NRF-1, and TFAM (Kozhukhar and Alexeyev, 2022), were analyzed to evaluate the influence of T deficiency upon mitochondrial biogenesis in hippocampus of male APP/PS1 mice. CS activity and mtDNA copy number were significantly reduced in the hippocampus of APP/PS1-GDX mice relative to APP/PS1-Sham mice, as well as WT mice (Figures 5A, B). Decreased CS and mtDNA copy number were also present in APP/PS1-Sham mice compared with WT Mice. Testosterone supplementation to APP/PS1-GDX mice showed amelioration to relevant parameters observed above in hippocampus. In addition, the expression levels of PGC-1α, NRF-1, and TFAM were significantly decreased in APP/PS1-Sham mice relative to WT mice and gonadectomy to male APP/PS1 mice induced further reduction of them (except PGC-1α protein level), which was reversed by T supplementation to APP/PS1-Sham level (Figures 5C–E). These findings suggested T deficiency further attenuates mitochondrial biogenesis in the hippocampus of male APP/PS1 mice.

Mitochondrial dynamics are also one of the approaches maintaining normal mitochondrial content. Mitochondria, as dynamic cell organelles, switch morphology between fragmented and tubular shapes via mitochondrial fission and fusion depending on different situations (Piquereau et al., 2013). Mitochondrial dynamics are regulated by the fission proteins Drp1 and Fis1, as well as the fusion proteins Mfn1, Mfn2, and OPA1 (Piquereau et al., 2013). Abnormal mitochondrial dynamics are involved in age-related neurodegenerative conditions. Therefore, the effects of T deficiency on the mitochondrial dynamics in the hippocampus of APP/PS1 mice were analyzed by detecting the altered expression levels of Drp1, Mfn1, and OPA1, which are involved in regulating mitochondrial dynamics. Relative to WT mice, Drp1 expression levels of male APP/PS1-Sham mice were significantly increased in the hippocampus (Figures 6A–C). Orchiectomy to APP/PS1 mice resulted in the further increased Drp1 expression levels in the hippocampus and TP supplementation restored Drp1 expression levels of APP/PS1-GDX mice to Sham mice level. In contrast, Mfn1 and OPA1 expression levels were significantly decreased in the hippocampus of APP/PS1-Sham mice compared with WT mice (Figures 6A–C). Castration to male APP/PS1 mice further reduced the Mfn1 and OPA1 expression levels in the hippocampus. In addition, mitochondrial ultrastructure displayed striking differences among the experimental groups (Figures 6D, E). Compared with mitochondria from the hippocampus of WT mice, which presented a normal mitochondrial structure with clear cristae, some mitochondria from the APP/PS1-Sham mice showed slightly disorganized cristae. Orchiectomy aggravated disorganized cristae of mitochondria from hippocampus of APP/PS1 mice and increased abnormal mitochondrial number. Testosterone supplementation to APP/PS1-GDX mice improved the ultrastructural alterations of mitochondrial cristae. The data above demonstrated that testosterone deficiency aggravates the imbalance of mitochondrial dynamics of male APP/PS1 mice, escalating mitochondrial fission in the hippocampus.

To study how T exerted its effects on neural cells, the alteration of oxidative stress was first investigated in human neuroblastoma SH-SY5Y cells due to critical involvement of oxidative stress in AD process. SH-SY5Y cells were exposed to 0–300 μM H₂O₂ for 24 h following 24 h pretreatment of 0–100 nM T. Appropriate concentration of T and H₂O₂ was chosen based on cell viability. Cell viability was significantly higher in 10 nM T + 200 μM H₂O₂ group than in 200 μM H₂O₂ group as well as 100 nM T + 200 μM H₂O₂ group (Figure 7A). Thus, 10 nM T and 200 μM H₂O₂ were used for later experiment. In addition to decreased cell viability, 200 μM H₂O₂ reduced GSH/GSSG ratio and increased 3-NT levels of SH-SY5Y cells. To validate whether the protective effects of T were mediated through androgen pathway, a specific AR antagonist flutamide was utilized. Administration of 10 μM AR antagonist flutamide for 1 h before T pretreatment inhibited the beneficial effects of 10 nM T on SH-SY5Y in cell viability, GSH/GSSG ratio and 3-NT levels following 200 μM H₂O₂ exposure (Figures 7B–D). Testosterone ameliorates H₂O₂-induced oxidative damage via androgen receptor.

The mitochondrion is the main source of ROS and its dysfunction leads to oxidative stress. Thus, the following study was focused on mitochondrial function. Mitochondrial membrane potential and ATP levels in H₂O₂ group were significantly lower than that in control group and decreased mitochondrial membrane potential and ATP levels in H₂O₂ group were prevented via T pretreatment. Administration of flutamide before T pretreatment inhibited the above effects of T pretreatment on mitochondrial membrane potential and ATP levels of SH-SY5Y cells following H₂O₂ exposure (Figures 8A–C). By measuring the OCR (Figure 8D), it was discovered that the basal respiratory function, ATP production, maximal respiration and spare respiration capacity were decreased in H₂O₂ group relative to control group. T + H₂O₂ group showed higher basal respiratory function, ATP production, maximal respiration and spare respiration capacity than H₂O₂ group (Figures 8E–H). There is no significant difference between F + T + H₂O₂ group and H₂O₂ group in the basal respiratory function, ATP production, maximal respiration, and spare respiration capacity. Testosterone pretreatment alleviates H₂O₂-induced mitochondrial dysfunction of SH-SY5Y cells via androgen receptor.

Some important parameters related to mitochondrial biogenesis were analyzed to evaluate the effects of T pretreatment on mitochondrial biogenesis under the condition of oxidative stress. It is found that decreased the mitochondrial copy number, as well as PGC-1α, NRF-1, and TFAM at mRNA and protein level was present in SH-SY5Y cells of H₂O₂ group compared with control group, while increased the mitochondrial copy number, as well as PGC-1α, NRF-1, and TFAM in SH-SY5Y cells was shown in T + H₂O₂ group relative to H₂O₂ group. No obvious difference in the mitochondrial copy number, as well as the expression of PGC-1α, NRF-1, and TFAM was detected in SH-SY5Y cells between F + T + H₂O₂ group and H₂O₂ group (Figures 9A–H). Moreover, mitochondrial mass (a parameter reflecting mitochondrial biogenesis to some extent) was analyzed via quantifying fluorescence intensity of MitoTracker Red by laser confocal microscopy. The mean fluorescence intensity in H₂O₂ group was significantly weaker than that in control group. T + H₂O₂ group presented strong mean fluorescence intensity relative to H₂O₂ group. There was no significant difference in mean fluorescence intensity between F + T + H₂O₂ group and H₂O₂ group (Figures 9I, J). Testosterone reduces H₂O₂-damage to mitochondrial biogenesis of SH-SY5Y cells via androgen receptor.

To determine the effects of T pretreatment on the balance of mitochondrial dynamics under the condition of oxidative stress, the altered mRNA levels of Drp1, Mfn1, and OPA1 regulating mitochondrial dynamics were detected and the mitochondrial fragmentation extent was analyzed by MitoTracker Red CMXRos staining. Relative to control group, the significantly increased Drp1, as well as decreased Mfn1 and OPA1 was found in SH-SY5Y cells of H₂O₂ group (Figures 10A–C). On the contrary, compared with SH-SY5Y cells of H₂O₂ group, SH-SY5Y cells of T + H₂O₂ group showed the decreased Drp1, as well as the increased Mfn1 and OPA1. For Drp1, SH-SY5Y cells demonstrated higher mRNA levels in F + T + H₂O₂ group than T + H₂O₂ group and lower mRNA levels in F + T + H₂O₂ group than H₂O₂ group. There is no significant difference of mRNA levels between F + T + H₂O₂ group and T + H₂O₂ group for Mfn1, as well as between F + T + H₂O₂ group and H₂O₂ group for OPA1. Furthermore, H₂O₂ group showed more SH-SY5Y cells with short punctiform mitochondria labeled by MitoTracker Red CMXRos dye than control group (Figures 10D, E). SH-SY5Y cells of T + H₂O₂ group had less short punctiform mitochondria and more long tubular mitochondria than SH-SY5Y cells of H₂O₂ group. No significant difference was found in the numbers of SH-SY5Y cells with short punctiform mitochondria between F + T + H₂O₂ group and H₂O₂ group, though the reduced trend in the numbers of SH-SY5Y cells with short punctiform mitochondria was detected in F + T + H₂O₂ group relative to H₂O₂ group. It was demonstrated that testosterone pretreatment reduces H₂O₂-induced mitochondrial fission, alleviating mitochondrial fragmentation of SH-SY5Y cells.